How difficult is it to organize an activity for kids in USA? Last week I talked with a guy from Eastern Europe who is currently working in USA, and in his opinion it was insanely difficult. I am not sure I understood or remembered the details correctly, and I also have no idea what is specific for one state (or maybe even one city) and what is true in general; and I do not want to add details to avoid doxing him, but the summary of what I remember is this:
Doing anything with kids is a *huge* legal risk. First, you could get accused of anything, and good luck proving that it didn't happen. A good lawyer would probably save you, but good lawyers aren't cheap. Second, accidents happen, for example a child randomly decides to jump, slips on the floor and breaks a leg. Parents often get vindictive, because "accidents happen" doesn't seem to be a concept in American culture. Again, expect to spend a lot of money.
The recommended best practice, if you want to do something with kids, is to associate with some institution, such as university. The way it works is that you attend various courses, and at the end you are "certified to talk to a child", or "certified to be in the same room as a child", etc. They will teach you all the things you are expected to do, that might seem weird for a foreigner. (For example, you cannot let a child go to a bathroom unattended. Even if the bathroom is on the same floor, you need to accompany the child until the bathroom door, and keep waiting in front of the door, until the child returns.) In return for doing all these courses and getting the certificates, the university can provide you legal insurance -- if anyone accuses you of anything, the university lawyers will attest that you were educated how to act properly.
Suppose that you want to do something like an afternoon math club for kids who are interested in math and good at it. How can you find them? Here is how these things are typically done in the deep jungles of Eastern Europe: you make a motivational flyer, and then you either ask the local organizers of some mathematical competition to give the flyer to the participants, or you visit the local schools and ask the math teachers to give the flyers to kids who are good at math. For some reason my friend was rejected by the organizers, and when he asked the teachers at public school to give the flyer to kids who are gifted in math, he was typically told that "there is no such thing as being gifted in math, all our students are equally talented". He had more success at private schools.
Another option is the word of mouth: you acquire the first few children by some miracle (for example, you start with your own children, and/or children of your friends), and then you tell the kids that if they know someone who might be interested in the activity, they can invite them to try it. This kinda works, but it naturally leads to the opposite of diversity; for example, a cooperative private school will send you a few mathematically gifted children, who mostly happen to be Asian boys. These will invite their friends... who are also Asian boys. Soon, your entire club consists of Asian boys.
To avoid misunderstanding, my friends is not woke; he simply wanted a group that would kinda resemble the local population. Not even proportionally, just to e.g. have at least some girls in the group. Last time, he had two girls in the club. Then one of them got sick. The other one didn't like being the only girl in the club, so she left. Then the former girl returned, noticed that now she was the only girl in the club, so she left, too. Or, he has a difficulty finding Hispanic students, because those are typically in public schools, and the official policy of the public school is that there is no such thing as talented children. (It is not clear whether the teachers in public schools actually believe this, or just want to avoid trouble.) So if there is a Hispanic boy or girl who happens to be good at math, the educational system is working hard to prevent them from joining an activity that could help develop their talent.
Most of the time we talked about what he does at those math clubs, but that's another topic, and there are a few books written on that. The cultural differences made me curious. I wonder whether his experience is typical, and whether Americans in general are aware of this, or it's just "the fish don't see the water they are swimming in". I would expect that if his experience generalizes, then the activities for kids in USA will be in general rare and expensive. Is that true?
I don''t know what it's like in all settings. There are definitely settings where nothing like that happens. When I was homeschooling my daughter, homeschoolers were often setting up and running singing groups, science groups, etc., and I don't believe there was any permission-getting or paperwork at all involved in running one or sending one's kid.
I’ve been trying out this new “j lens” LLM interpretability stuff from a recent Anthropic paper.
Initial thoughts:
- it looks like there’s a “what does that mean?” token that gets activated in J space when the model is confused by something in the prompt.
- This might not be significant or meaningful, but it is none the less amusing that “China” gets activated in J space when I’m doing exotic jailbreaking / system prompt injection. (This is with Qwen, a Chinese model)
- A particular prompt I’ve used in some experiments with DeepSeek very strongly activates “cute” in J space when used with Qwen models. Ok, sure, the subtext of that prompt is that the Assistant is some kind of cute sidekick such as one might encounter in a Studio Ghibli movie, and the effect is to ramp up the cuteness factor.
Ooh, very nice job getting it working with other models. What was the process like getting set up to go, for you? I saw they provided a repo; was it just `git clone` and off the races plugging in other models, or did you have to do substantial dev work?
I’d wait for more evidence. The NY Post has an aggressive tabloid style that isn’t always accurate. Whatever happened, the guy was released without bail on his own recognizance.
So, I was chatting to the LLM Kimi K2 about the project by Janus/repligate to give a Claude a robotic body, and K2 said …
“The trickster at least has the virtue of being honest about dishonesty; it tells you, by its very shape, don’t trust me completely.”
Which I think captures my slight misgivings about Claude being given a robotic body that is quite so heavy with Christian iconography as Janus’ design.
How do I miss all this good stuff? What Christian iconography? Is it going to be festooned with crucifixes and the Ichthys?
Or is it just the mention of embodiment corresponding to the Incarnation? Because looking at this it's a rectangle of material with probes attached and then later a skeletal 3D printed finger:
They're conscious and we can read their minds. Surprise!
OC ACXLW Meetup #116 — The Global Workspace Theory of Consciousness & Anthropic's "J-Space"
Date: Saturday, July 11, 2026 Time: 2:00 PM Location: 1970 Port Laurent Place, Newport Beach, CA 92660 Host: Michael Michalchik · michaelmichalchik@gmail.com · (949) 375-2045
Sorry, I refuse to believe the fleshbags are conscious. What goes on in that lump of meat in their bony casque cannot be dignified with the term "mind".
You're absolutely right, but that framing is boring. I'm provoking feisty conversation at my meetup. I do actually personally believe that this is a meaningful definition of consciousness, global workspace or something pretty similar to it, but I respect people who want to disagree. They're wrong but they have the right to be wrong, ;-)
I believe that qualia is a perfectly sensible yet completely ill-conceived notion that, at its heart, is not even wrong. It is akin to notions like "the life force", "the quintessence", "the philosopher's stone", "the afterlife", "the luminiferous ether", "the purpose of life", "God's will", "the people's continuous under teh dictatorship of the proletariat revolution".
It is a legitimate-sounding thing that smart people used to take seriously, but now realize that it is fundamentally an ill-formed question or notion because that is simply not the universe we live in. The things we try to distill into this notion of qualia don't belong together as a conceptual category and can only lead to absurdities and teh wrong questions.
Understanding consciousness by framing it in terms of qualia is like trying to understand morphine by appeal to the dormative spirits.
Has ACX ever discussed Foer’s How Soccer Explains the World? I’m considering a review or update in light of 2026 World Cup, FIFA salience and relation to globalization & tribalism.
A sensible man who had a little channel about swimming pool maintenance and wasn't expecting to become famous. Now he has a larger channel.
While there isn't information about driving SUVs over swimming pool surfaces*, he does the best he can analyzing known properties and similar materials and concludes that these are tough industrial surfaces and probably weren't damaged, though damage is possible if they weren't given time to cure fully.
*I was talking with someone who couldn't imagine why anyone would drive in the reflection pool. I convinced him it was imaginable by saying "Imagine a teenager".
Can anyone recommend science/engineering and politics resources?
For Engineering, there's the straightforward @PracticalEngineeringChannel.
There's also @smartereveryday. I liked his explanation of Prince Rupert drops.
The 800-lb gorilla in my book, though is @veritasium. It used to just be this guy named Derek Muller who liked teaching science, and now it's blown up somewhat and there's a whole crew producing narration and some of the best visuals I've seen.
I don't know of any YT politics channels I think you'd enjoy, unless for some reason you have low blood pressure.
This may be the very first ACT post ever that’s about pro wrestling:)
A former co-worker had been involved with it for decades and probably still is. In his younger days he was a wrestler, not at WWE level but with regional promotions, and then became a trainer and did some local promoting. He seemed to have a great deal of knowledge about the industry (sport?).
In any event, he said that there are very few casual fans of pro wrestling. It’s oversimplifying only a bit to say that people are either devoted hardcore fans or have no interest at all. And the thing is, wrestling organizations from the WWE on down are perfectly fine with this. What with the loyalty of their core fan bases there’s no particular need to seek out a wider audience.
One more thing my co-worker said is that while pro wrestling is fake in the sense that the outcomes are pre-decided, the physical poundings the competitors take are all too real.
My understanding is that if you accidentally hit your opponent in a way that wasn’t planned, you are expected to take a cheap shot from them to even the score, or there will be serious trouble.
The ability to appreciate your customers for who they are, instead of modifying your product to meet the tastes of some wider audience, is an underappreciated virtue.
Maybe not entirely the first. I’ve mentioned here that Hulk Hogan once gave me a pro wrestlers scowl when i was running in the opposite direction around a Mpls lake. He was walking with a very blond woman at his side.
He cut such a ridiculous figure in ordinary surroundings I was stifling a laugh as we were approached each other.
PSA: If you're going to comment about your experience with an LLM and seeking any sort of insight, please, don't just say Claude said x or ChatGPT did y. Those are almost useless identifiers. Specify which Claude, opus? haiku? which version, 4.6? 5.5? And with weird behavior, what else was inbthe context window, did this happen as part of a larger/longer conversation? Do you have a bunch of "memories" enabled?
It's like going to the doctor to say sometimes I have pain, and nothing else at all, that's nice for feeling like communicating but it's basically an offer of nothing.
"Drop Site journalist Ryan Grim has authenticated behind-the-scenes messages showing Politico omitted an important detail from its reporting on how the night began that led to the rape allegations against Maine Senate candidate Graham Platner."
"Before publication, a Politico reporter told the Platner campaign that Jenny Racicot, on the night in question, had been “texting [Platner] about needing her glute massaged.” The reporter added, “That detail didn’t make our story.”"
That explains her comment in this interview about "And it was a night where him and I were texting back and forth and he had taken something that I said as as an invitation and that's not how I meant it. And I quickly clarified and he sent a message back indicating that he
Catholics and Protestants had a troubled first few centuries of coexistence, but eventually they settled down and learned to live alongside each other by the 18th, 19th or 20th centuries.
Except in one place, Ireland, where Catholics and Protestants just went on killing each other right up through the twentieth and into the twenty-first century. The Germans managed to get all this out of their system in the 1600s, but the Irish never did. What gives?
If religious differences were no barrier to the success of a united Germany, why were they such a problem for a unified British Isles?
Feckin' Brits Were At It Again, or, Why Didn't The Paddies Calm Down? Let Me Grab You By The Lapels And Launch Into A Foam-Flecked Rant About That!
This will be all over the place as I'm not a historian and I'm just typing whatever comes into my tiny little brain at any particular moment. Part one of probably several?
First, let's set the tone with this rendition of "Four Green Fields" as sung by Makem and Clancy, introduced with a recitation of a Seamus Heaney poem. I'm setting my stall out from the start:
Now, I mentioned the Richard Dawkins Guide For The Perplexed in this context, and that's because he annoyed me on one of his tours to Northern Ireland (he seems to have found the place congenial, ironically enough) where I remember (but can't locate online) a bit of throwaway jokey commentary by him about it all being more evidence of the stupidity of religion, the Paddies killing each other over Catholic versus Protestant (not phrased like that, but the general tenor). Of course, he was engaging in the popular English (and I *do* mean English, not Scots or Welsh) pastime of "it was nothing to do with us when we invaded all those countries, embarked on a policy of divide and conquer, sowed the seeds of sectarian conflict which then exploded when we cleared off back home and honestly why are you blaming us for all that?"
More seriously, as mentioned in this Guardian article:
"Predictably, Dawkins has no time for faith schools. "Segregation has no place in the education system," he argues. "Take Northern Ireland. You could get rid of the climate of hostility within a generation by getting rid of segregated schooling. Separating Catholics and Protestants has fomented centuries of hostility."
Ha ha ha. And why the hostility, Richard? Not a clue? It must be all down to bad old religion?
Let me take you on a gallop through The Eight Hundred Years, where I will attempt to demonstrate that it is indeed as much or even more about tribalism and political allegiance as it is about religion. It's not "Protestant versus Catholic" so much as "British versus Irish" or "incomer versus native", and why it remained a heated struggle in the North for so long.
(1) Hiring Mercenaries Is *Not* Marriage Counselling!
Let's start with a quote from a play by Yeats, "The Dreaming of the Bones" (look, this is Irish history, it's going to be coming down with poetry and song, sorry in advance. This is how we do things).
"YOUNG MAN
You speak of Dermot and of Dervorgilla
Who brought the Norman in?
THE GIRL
Yes, yes I spoke
Of that most miserable, most accursed pair
Who sold their country into slavery, and yet
They were not wholly miserable and accursed
If somebody of their race at last would say:
'I have forgiven them.'
YOUNG MAN
Oh, never, never
Will Dermot and Dervorgilla be forgiven."
Okay, who are Dermot and Dervorgilla, and what has that to do with Irish history? Very briefly: Dermot is Diarmuid MacMurrough, and was king of Leinster in the 12th century. There was enmity between him and the High King of Ireland (and the High King's successor), who sent an allied king, Tiarnan O'Rourke, to drive him off the throne. This didn't work but later MacMurrough was accused of abducting O'Rourke's wife, Devorgilla, and this was given as the reason for their strife.
The story is that the new High King ordered MacMurrough to return Devorgilla, he refused, and the High King then deposed him and MacMurrough went to England to get help reclaiming his throne:
Hiring in foreign mercenaries to help with local wars was not an unknown practice in Ireland, but why MacMurrough decided to appeal to the English king Henry II is not well known. Henry gave him permission to recruit from the soldiers in his realm and a force of Anglo-Norman knights and wannabe lordlings signed up.
Presumably MacMurrough thought they'd all land, beat his Irish rivals, get the throne back and then - well, he doesn't seem to have planned past that. Alas, he was too soon to be advised that paying the Danegeld doesn't get rid of the Dane. To strengthen the alliance with his foreign allies, he married his daughter Aoife to Richard de Clare, better known as Strongbow. Again, it is alleged that he promised she would be his heiress to the throne, which under Irish law was not legal (the system of tanistry, not primogeniture as with the Normans). Diarmuid dies a few months after getting his throne back, and now there are rival claimants to the kingship. The Normans are here, and they're not going away.
Plus, there are political schemes afoot back in England and Wales. The king is rightfully fearful that a bunch of former vassals getting lands and kingdoms of their own over in Ireland makes them too powerful and uncontrollable, so he moves to remind them who is the boss of them. In return for getting his confiscated lands in France, England and Wales back, Strongbow agrees to hand over the Irish possessions to Henry,
And thus begins the English claim to the lordship (and later kingship) of Ireland. But what has religion to do with this? The answer: not much and also a lot.
The native Irish kings and chieftains are not *too* put out by having a bunch of Normans landing among them, since there isn't really a concept of "Ireland as a separate nation and a unitary state" at this time. They're as eager to make alliances against their local rivals with the new guys and their superior military tech as MacMurrough was, and the Normans for their part are happy to carve out territories of their own then settle down and eventually acculturate.
But Henry needs a stronger reason to claim a whole freakin' other kingdom than just "I was bored and wanted a new toy". His claims in France, England, etc. have some legal colour as "the French territories are my inheritance by right from the Norman background, the English throne came along with the Conqueror who was promised it as heir, and Wales, well that's conquered land". But Ireland?
Luckily, there's the first (and last!) English pope sitting there in Rome, who can be persuaded that Ireland has fallen away from proper Catholicism and needs to be brought back to the light of the faith by good, pious, kings like, say, Henry (If we overlook that little affair with the murder of Thomas a Becket).
Thus comes the bull Laudabiliter, granting Henry the right to govern Ireland and reform the church there:
The Normans have their own apologist (or propagandist, if you prefer), Gerald of Wales who records what happened when they set out on their civilising mission to Ireland. Remember, everybody here is Catholic so there's no "Protestant versus Catholic" animus going on, that's centuries in the future:
"Chapter XXXII: A sarcastic reply of the Archbishop of Cashel.
I once made objections of this kind to Maurice, archbishop of Cashel, a discreet and learned man, in the presence of Gerald, a clerk of the Roman church, who formerly came as legate into those parts, and throwing the blame of the enormous delinquencies of this country principally on the prelates, I drew a powerful argument from the fact that no one in that kingdom had ever obtained the crown of martyrdom for the church of God. Upon this the archbishop replied sarcastically, avoiding the point of my proposition, and answering it by a home-thrust: “It is true,” he said, “that although our nation may seem barbarous, uncivilized, and cruel, they have always shewn great honour and reverence to their ecclesiastics, and never on any occasion raised their hands against God’s saints. But there is now come into our land a people who know how to make martyrs, and have frequently done it. Henceforth Ireland will have its martyrs, as well as other countries.”
Whew! Well, we pass on from the 12th century to the 15th or so. The Normans are integrating and are now the Anglo-Irish or Old Irish. There's still divisions between the native Irish, the Old Irish, and the English court nominally ruling over them all, but we're rocking along okay so far as religious divisions are concerned.
4000 words in and just at the end of the 17th century 🤣 Luckily, we've only got another four centuries to go, and I should get that covered in two more posts at maximum.
18th century was nice and quiet, just a few throat-slitting suicides there. Wait for the 19th, The Famine is going to be *so* much fun, isn't it? 😈
Poor Melvin, he must be regretting his life choices by now.
Like the joke goes, "I suffered for my art, now it's your turn" 😁
(Part 2 of X)
(2) Founding Your Own Church Is *Not* Marriage Counselling!
So it's the 12th century, we've got a bunch of Anglo-Normans settling down, establishing fiefs, and getting ready mingle with the locals, as well as an English king claiming he is the lord of us. For the next three centuries or so, this is sort of all right. The Normans wax in power, local Irish chieftains continue fighting amongst themselves and making alliances with the newcomers, and changes in law and politics (and religion!) slowly entrench. The Irish church is brought into line (more or less) with the Continental church (e.g. previously it had been organised more along Orthodox lines, with monasteries and abbots being the important leaders, but now we're getting the establishment of parishes and secular clergy).
There are visits from time to time by the English kings to tighten up their hold as claiming the lordship of Ireland, and outside events (like the Black Death) hit Ireland as well, but mostly it settles down into:
(1) resurgence of Irish chieftains pushing back against the Normans and regaining some lost ground
(2) establishment of Norman-Irish dynasties such as the Butlers and Fitzgeralds as local powers
(3) The king and English power is confined mostly to the east coast of Ireland, with the establishment of The Pale (and yes, this is where the saying "beyond the pale" comes from): https://en.wikipedia.org/wiki/The_Pale_(Ireland) and attempts to 'civilise' and 'Anglicise' the native Irish with greater or lesser success, as well as prevent the Anglo-Normans/Norman-Irish/Old English from going native: https://en.wikipedia.org/wiki/Statutes_of_Kilkenny
(4) We manage to toddle along without too much upheaval, particularly as the English get distracted by handling their own affairs such as the Wars of the Roses
And then we hit the Tudor period, and Henry VIII.
The Tudors are not long established and there is the constant, background, lurking fear of the Wars of the Roses kicking off again. Henry VII managed to be the last man standing at the end of that, but there are plenty of other powerful families with royal blood of equal, or better, descent (Edward III had had a *lot* of kids) to challenge his and his family's claim to the throne. And there have been two Yorkist pretenders - Perkin Warbeck and Lambert Simnel - needing to be dealt with during Henry VII's reign. There was still enough lingering support for the House of York to be a genuine problem.
More importantly, the Irish lords were mostly pro-Yorkist, both the pretenders had received some level of support in Ireland. The powerful Fitzgerald family, Earls of Kildare, were holders of the office of Lord Deputy of Ireland, and did not appreciate the English administration's efforts to weaken the power of the Irish parliament and hand over much more control to England.
So they backed the wrong horse and although pardoned in the wake of it all, the office of Lord Deputy was granted to their rival family, the Butlers, and the Fitzgeralds were tainted with the suspicion of disloyalty. Not helped by an abortive rebellion in Henry VIII's time, where Henry decided he needed to tighten his grasp on Ireland and embarked on the policy known as "surrender and regrant" - basically, getting the loyalty of native Irish lords by getting them to surrender their old native titles then granting them new English ones and confirming them in the possession of their lands, powers, and so forth (so long as they remained loyal to the Crown). This enabled some ambitious family members to grab the headship of their clan from the current leader by taking advantage of the king's offer and getting the backing of the English in enforcing the new regime. Henry also assumed the title of King of Ireland, not the previous lordship: https://en.wikipedia.org/wiki/Kingdom_of_Ireland
So far, so politically fraught, but here's where religion comes in. Henry embarked on a little thing known as the foundation of the Church of England, and this included imposing similar policies in Ireland. So Irish monasteries were suppressed, seized lands handed off to Irish nobles to buy their co-operation, but most importantly a policy of Anglicisation was followed.
And *here* is where we get the major split. Before, the Church was the Church, there might have been differences and disputes between the Irish and the English, or the Irish and the Continental model, but everyone was (at least in theory) a member of the same religion and same church.
Not now. Now there was a new church, with a new language (English), new doctrines, new clergy, new everything *including* the English king as the supreme head, not the Pope. Now we have the beginning and foundation of the differences in religion also being associated with differences in culture, politics, language, everything.
Does this give you a hint as to the answer to your first question:
"If religious differences were no barrier to the success of a united Germany, why were they such a problem for a unified British Isles?"
From now on, identification with "are you Catholic or Protestant?" becomes associated with "are you British or Irish? are you loyal or a rebel? are you one of the new, civilised, modern, progressives or one of the old, backwards, mired in superstition and poverty types?"
Sporadic rebellions during the reign of Elizabeth and her successor James, along with appeals to the Pope and the Continental Catholic powers such as Spain, further deepened the division. People such as Hugh O'Neill, Earl of Tyrone, could have initial success in the Nine Years' War but eventually be ground down and the resulting destruction of the power of the remnants of the Old English/Irish nobility meant the slow, gradual, but inexorable fall of Gaelic culture and society as well: https://en.wikipedia.org/wiki/Nine_Years'_War_(Ireland)
Unlike England, where the Reformation could be presented as native, the new Church in Ireland was visibly an imposition from outside, and treated as such. England could eventually turn majority Protestant, Ireland remained majority Catholic with the alien minority exercising their political and religious power over an unassimilated population.
And over the centuries, the split became ossified. Attempts to Anglicise the Irish were quietly dropped; things like the Plantations (and this will be of relevance when we come to consider Northern Ireland) were meant to bring in English settlers of means who would take over the lands, develop them, and import good solid English tenantry to work those lands.
Instead, we got some English, but more Scottish (particularly in the North), and instead of development it became speculation, with people acquiring and then selling on rights to territory in Ireland. Those who did settle down couldn't or didn't get the English tenantry who were intended to displace the natives, so despite all the legal penalties, the natives were permitted to stay on their former lands, but now under new ownership where they had few to no rights, few to no ties to the new owners, and could be rack rented for profit.
This brings us up to the 17th century, which needs its own post, so I'll leave it there.
(1) 18th century tune by Carolan, English lyrics put to it in the 19th century, played here by Jordi Savall: the Lament for Owen Roe O'Neill (the 17th century Earl of Tyrone mentioned above)
We're heading into the gloomy and depressing 17th century, because this is where the rubber really meets the road when it comes to creating and hardening sectarian divisions in the green island of Éireann.
The play "Henry V", ostensibly about the titular character's successful military campaign in France (see the Battle of Agincourt) and an embryonic attempt at the construction of a proto-British identify (the troops include a stage Irishman, stage Scotsman, and stage Welshman as well as English) but this isn't on solid ground.
National identity is in flux, as see the question posed by the Irishman:
"FLUELLEN Captain Macmorris, I think, look you, under
your correction, there is not many of your
nation—
MACMORRIS Of my nation? What ish my nation? Ish a
villain and a basterd and a knave and a rascal. What
ish my nation? Who talks of my nation?"
What exactly is MacMorris' position? Is he Irish (or Irish Irish, as he seems to be presented?) Is he Anglo-Irish or Old English? Is he one of the New English? Taken as English in Ireland and Irish in England, caught between two kingdoms, and hence prickly and suspicious when questioned about his "nation", that is, his identity.
And identification, which is the underlying question here.
Because though the play may be set in 1415, it was probably written about 1599 (so teetering on the edge of the 17th century) and first published in 1623.
What was happening in 1599? The Earl of Essex, one of Elizabeth's favourites, was over in Ireland putting down the rebellion by the Earl of Tyrone, Hugh O'Neill. This was in the middle of what would become known as the Nine Years War (1593-1603) and although the expectations and hopes were that Essex would be successful in crushing the rebellion and bringing the war to an end, this was not to be.
This is where MacMorris' wavering loyalties and ambiguous identity comes to the fore; when he says "Ish a villain and a basterd and a knave and a rascal", who is he talking about? Is this aimed at Fluellen, the pedantic Welshman constantly needling him? Is he talking about "his nation" as the Irish, who are to be read as "villains, bastards, knaves and rascals" and so separating himself from them, the Gaelic Catholic Irish where he is the loyal Anglo-Irishman fighting for the English king? Is he identifying as English not Irish?
The question is not answered, and indeed remains open up to right now for everyone with dual identity or singular identity or maybe no identity at all. North or South? British or Irish? New Irish (including the immigrants born here) or Old Irish? Religious allegiance or not?
The Irish campaign went badly, Essex was forced into a truce, and returned early to England without having been recalled by the Queen. His enemies at court leaped on this chance, and he was put under house arrest, tried, and deprived of his offices and position at court, though he was eventually permitted to leave house arrest.
He was in bad financial shape and his appeals to the Queen went nowhere, so with little to lose (and making a spectacularly bad decision) he tried a rebellion/coup, which lasted one day, got no support at all, and ended with a trial for treason and execution.
The Nine Years war ended with the defeat of the Irish rebels and the famous Flight of the Earls, where the old Gaelic order was broken and overthrown for good. Elizabeth died in 1603 and there was a new King on the throne, James VI and I.
So now Ireland was pacified, more or less, but how to keep it that way? One answer was what had previously been tried - Plantations. This is not to say that Ireland had been generally peaceful all along up till then, there were plenty of massacres on both sides and uprisings and rejections of English sovereignty. But the problem had become more pointed and entangled with religion thanks to Henry's marital escapades. When Elizabeth came to the throne, many countries considered her a bastard, and moreover a heretic. She was excommunicated in 1570 by the Pope and her Catholic subjects released from their fealty to her. This fanned fears of rebellion at home and opportunistic invasion by Spain (and maybe France) under guise of a religious crusade, and Ireland was seen as a potential staging ground for such invasion, as well as disloyalty. The Elizabethan religious settlement in England made divergence from the state church a matter of politics (treason) rather than religious belief (heresy) but this only solidified the choice forced upon the people: your faith or your loyalty?
Still, until the final quashing of an independent ruling structure in Ireland, this had not yet become what it would now become: state policy to impose Protestantism on Ireland.
James had been raised as a Presbyterian, and under a regime intended to turn him into "a God-fearing, Protestant king who accepted the limitations of monarchy" and during his reign in Scotland there was a continual tussle between him and the Kirk as to who had the ultimate authority. I think this struggle, with the Calvinist extremists and even the more moderate supporting such things as the imprisonment of the young king by Protestant earls, lies underneath James' development of a theory which would become "the divine right of kings".
Anyway, now James is securely on two thrones and it's time to sort out that pesky Irish problem so he can handle more serious matters (though it needs to be pointed out that he suffered problems at home from such things as the Gunpowder Plot, leading to further repression and legal penalties on Catholics).
So 1609 sees the beginning of the official Plantation of Ulster, and the short version of the outcome is best stated by Wikipedia:
"The plantation led to the founding of many of Ulster's towns and created a separate and lasting Ulster Protestant community in the province with ties to Britain. It also resulted in many of the native Irish nobility losing their land and led to centuries of ethnic and sectarian animosity, which at times spilled into conflict, notably in the Irish Rebellion of 1641 and, more recently, the Troubles."
His son and successor, Charles I, married a Catholic which led to even more suspicion and unrest among the Protestant nobles of the court. Fears of Papism making a covert comeback were not helped by Charles, following James' example, of exercising his Divine Right and ignoring Parliament. There is also ongoing religious tension in the state church over those more High Church (seen as crypto-Papists) and the Low Church (seen as Reformers and non-conformists). Charles' High Church (and suspected Catholic) sympathies are seen as Very Bad.
This all blows up into the English Civil War, including perceived promises by Charles to the Irish (the Graces https://en.wikipedia.org/wiki/The_Graces_(Ireland)) and using Irish forces as a threat against the English.
A wild Cromwell appears!
And now the king is dead, there's a Commonwealth, and God is a Protestant (and not just any old Protestant, but one of a particular stripe).
You probably already know about Cromwell's adventures in Ireland, so let's skip over that.
Charles I out, eventually Charles II is in. Restoration of the Monarchy loosens up a *heck* of a lot of things, but there's still suspicion about Catholicism, not helped by Charles' marriage to the Portuguese Catholic princess Catherine of Braganza, and the failure to have a legitimate heir (Charles has no problem at all fathering illegitimate kids, but unlike Henry VIII he doesn't dump his wife and try new spouses until he gets that heir). This means the heir presumptive is his brother James, and James is Catholic. Not just "kinda too Romanist Anglican" Catholic, but Catholic Catholic.
Then Charles dies and James is king, and he is not popular for many reasons. So in 1688 along comes the Glorious Revolution, where the staunch Protestant patriots of the court and the ascendant middle-class invite in a staunch Protestant king, William of Orange, and we get more fussin' and fightin'.
Leading to last-ditch campaigns in Ireland and the final defeat at the Battle of the Boyne in 1690 (and boy are we gonna hear a lot about *that* for the next three centuries or so).
James out, William and Mary in as co-monarchs, and the Protestant succession is assured.
That brings us to the end of the 17th century but not, unhappily, to religious strife which is now codified, identified, and ossified as not merely "Catholic versus Protestant" but a slew of competing identities and ideologies and politics. Scotland is nearly as bad as Ireland, by now, but even there the Catholic population is now a minority, so they can happily busy themselves with Sabbath-keeping and Orange parades.
1600s Germany was a cakewalk by comparison.
Let's end with some of that good old 17th century celebration!
Part 2 of X to come, but there's a lot of ground to cover and I'm wrote out right now. But more of the thrilling tale of Feckin' Brits, It's All Their Fault will be done! Sometime!
Religious differences between two longstanding groups of people can be *a* component and *a* driver of a longstanding conflict, but you should never expect it to be the whole of the conflict. First an foremost, when any high-intensity conflict goes on for any length of time, you can bet your last dollar that one of the main drivers is the conflict itself. Each group will have a long and detailed history of being aggressed against by the other groups and while each group's "version" is sure to include bias and slant and even outright fabrication, I cannot stress enough that they will both *certainly* include real history including real, serious grievances as well.
I think the assumption that Catholics and Protestants started playing nice earlier is faulty. There was quite a lot of friction between Catholic and Protestant populations (in Europe and elsewhere) during that whole time AND quite a lot of wars between Christian countries. I'd argue that the lines simply got muddled enough that "Catholics vs Protestants" now longer leaps out to the fore as the primary driver of the big wars, but that grievances from previous wars--very much including wars of religion--were still absolutely in the mix.
In this assumption, Ireland is remarkable in two ways. First, that the religion divide tracked the other ethnic and politic dividing lines unusually well for an unusually long time--Erica Rall does a good job laying out the other sources of conflict. Second that the conflict in Ireland took *somewhat* longer to quiet down than did the remainder of Western Europe. Western Europe was a pretty war-torn place from the late days of the Western Roman Empire right up until 1945. The remarkable thing to me isn't that the Irish (and to some extent the English) kept fighting one of the old, intermittent wars for a few decades longer than everyone else (albeit at quite low intensity). The remarkable thing is that everyone else stopped. The trajectory of Western Europe after the end of World War II is something I find genuinely interesting and anomalous, especially given how many similar times and places things like that haven't happened.
"First an foremost, when any high-intensity conflict goes on for any length of time, you can bet your last dollar that one of the main drivers is the conflict itself"
You want a quick gallop through The Eight Hundred Years? Not got time right this second, but later on this evening I'll give it a bash for The Richard Dawkins Guide For The Perplexed (Feckin' Brits Edition)😁
For the first one, back in 2012-23 we had a "Decade of Centenaries" to commemorate the Civil War, foundation of the Irish Free State, etc. but in order to be even-handed (politically correct) they needed to include things from the "other tradition on this island" (the Loyalists) and it all ended up being pretty much weak sauce, where official government things were afraid to offend anyone so didn't come out pro- or anti- any side, and the commercial interests tried jumping on the bandwagon.
There's a very famous massacre of Protestants by Protestants right around where I live (it's one of many, and there's a reason why we don't got no recognized Indian tribes.) When you're talking "protestants" in Ireland, you're talking the Scotch Irish. And, funny that, THOSE are the guys what did the massacres!
Scotch Irish have "mostly" settled down where I live now, but that's through a concerted campaign of "Kick the Bad Un's out" -- wouldn't exactly be surprised if the massacres continued...
The Troubles weren't about religion as such. "Catholic" and "Protestant" were/are proxies for clusters of ethnic identity and political affiliation, and the main thing that was actually being fought over was whether Ulster would be part of the Republic of Ireland or the United Kingdom of Great Britain and under what terms.
The roots of the conflict arguably go back to the aftermath of the Norman Conquest of England when various adventurers decided to try to keep the conquests rolling, but the most relevant bit started with O'Neill's Rebellion of 1593-1903. Some Irish nobles rebelled against Elizabeth I for various reasons (including religion), lost, and then her successor King James VI and I seized a bunch of land from the defeated nobles and kicked the peasants off with the intention of bringing in good, loyal English settlers to replace them. The good loyal English mostly had better things to do, so they brought in settlers from the lawless, tribal border country between England and Scotland instead (same source population as the "Borderers" from Albion's Seed who settled Appalachia). The Ulster Protestants are mostly descended from the settlers and the Catholics are mostly descended from the native Irish. There have been several other major rounds of unpleasantness in the meantime, including two episodes (the Confederation Wars and the Potato Famine) that each killed off double-digit percentages of the island's population. The Williamite War and Jacobite Risings were rather a big deal as well.
He was the sixth king of Scotland named James, hence James VI, but the first king of England with that name, so James I.
His successor (later on down the line) would be titled James II alone, since this was now the United Kingdom of England and Scotland, so not James VII and II.
Sun Kitten is correct. People who are simultaneously monarchs of multiple countries but who have different regnal numbers in each are conventionally referred to by something like "King James VI and I". Charles I, Charles II, Mary II, and Queen Anne didn't get double numbers because they had the same number in both England and Scotland. No previous Charles or Anne before the personal union, and one prior Mary each: "Bloody" Mary in England (Elizabeth's older half sister) and Mary Queen of Scots in Scotland (Elizabeth's cousin and James's mother). James VII and II (VII of Scotland, II of England) and William III and II (III of England, III of the Netherlands, and II of Scotland, with different predecessors in each country) got double numbers. We're lucky the Netherlands had the same number of prior Williams as England or else we'd have to call him something like William IV, III, and II.
Under Queen Anne in 1707, England and Scotland were formally unified into one country rather than being two independent countries that coincidentally used the same monarch and coordinated military and foreign policies. From that point onwards, most monarchs used regnal names that had the same number in both countries because they'd never been used before the personal union started (George I-VI, Victoria, and Charles III) or they mercifully used the English numbering and ignored the Scottish one and thus spared us from having to talk about Edward VII and II or Elizabeth II and I. Under Elizabeth, the rule was formalized that when the numbers differ between England and Scotland, future monarchs will use the bigger of the two numbers. This is consistent with usage by Edward VII, Edward VIII, William IV, and Elizabeth II, since in all those cases the English number happened to be the bigger one.
>and an Oxford comma ain't it.
I should probably have written "... and then her successor, King James VI and I, seized a bunch of land" or else used the shorter form "King James VI/I".
I can't comment on whether Erica is a vampire, but "King James VI and I" is a single name. King James VI of Scotland was King James I of England. So that's probably what she meant.
Is anyone else skeptical of people who write authoritatively on a wide variety of topics?
I have several X (Twitter), Reddit, and other social media platform users in mind, but I'll keep this at the meta-level. I'm just baffled on how someone can be so knowledgeable about such disparate things.
A hypothetical example: When I see a person write about paragliding, culinary habits of Olmec people, tort reform, and thoughts on SOTA humanoid robotics, it just really sets off my skepticism radar (skepdar). My skepdar blares even more when they have lengthy/in-depth and frequent posts about different things.
How can one person be that knowledgable about such a variety of topics? Stuff is complicated. It takes time to become an expert; years of reading, doing, maybe even teaching, etc. I get that people have varied and interesting lives. I could go on about some topics like ice cream making and syn bio (some of it at least). But the examples I'm thinking of are excessive. I don't know. Maybe I just have a bad theory of mind. Some of the users get tons of upvotes/reposts/followers, so maybe I'm the wrong one.
Online, I was once accused of being multiple people writing under a common pseudonym, because I wrote about politics, mathematics, programming, and foreign languages.
In real life, I was once accused of being a habitual liar, because once I said I have finished a university X, on a different day I said I was studying at a university Y, and then I said I had a job. In fact I have completed X, and then I was studying Y while having a part-time job.
Some people have multiple interests.
Of course, there are also many cheaters, who write bullshit, steal other people's articles, or these days tell an AI to generate an article for them.
Because they are experts in very narrow fields, and only in the stuff that they are interested in. I could talk at length about the Ecuadorian constitution, German political parties, the Energiewende, and Judaism. I couldn't tell you anything about anything else.
Because they're passionate they read as much as they can about it
I think you're overthinking this. If a person claims to be an expert in *everything*, or that they just happen to be an expert in each of the last half-dozen topics of local contention or conversation, that would be an implausible coincidence. But if they are just claiming to be experts in half a dozen or so very different subjects, yeah, that happens. And there are several examples of it here, to my personal knowledge. For example, David Friedman - economics, law, medieval cooking, sword-and-shield fighting, the poetry of Rudyard Kiplng, fantasy fiction, and theoretical physics. I'm probably missing a few there.
The important thing is, do they ever admit to ignorance? Or at least absent themselves from further discussion when the subject wanders past the bounds of their expertise, though that's harder to assess.
Don’t sell yourself short. Stay curious, continue to explore new things and you may become the person who writes about multiple things with a fair degree of authority.
Being an expert in how to figure things out means you can describe your reaction to what you read about all sorts of subjects. If you have that, all you need then is a talent for writing.
Scott doesn't try to look like an expert on many things he writes about. What he offers instead is to let us watch him think over the topic in an interesting way.
They may indeed be a polymath or they may be broad but shallow. It's easy to give the *appearance* of being deeply informed when there is just a surface level knowledge, so long as you're not challenged by a real expert on the subject.
If the person claims to be a world-class expert in all these things, then that might be a red light. But nothing stops a person from having unusual and divergent interests - if you've read one book on the subject, you already know more than 99% of the population.
There's also the bias that you only see them talking about the things they ARE talking about, not what they refrain from talking about.
Given that I know someone who works at NOAA, writes video games, writes music, and is an economist, I'm not sure where the problem is... Generalists are going to generalize, and having a few good rabbit holes where you actually are an expert (among his are Tolkien and AI and Cryptography).
There are actually jobs that hire generalists! (The US government calls them analysts, believe it or not).
It doesn't take an expert on Iranian oil well technology to take "storage capacity" and "no oil flowing" and get "wells will need to shut down, doing semi-permanent damage to the ability to supply oil" (but, alas, Western Media!) in X days.
I'm not sure what you mean by saying you're "skeptical of" such people.
1. Like, you don't think they have as much expertise in all those areas as they do in the few areas where they really might? (They are truly experts in A and B, but only know a little about C D and E, and used AI or plagiarism or something to inflate how much content they generate on those topics)
2. Or, you think they don't have any expertise in them? (They are truly experts in A and B, but know nothing about C D and E, and used AI or plagiarism or something to generate content on those topics)
3. That they don't actually have any expertise period? (They are not experts in any topic, and used AI or plagiarism or something to generate all their expert-sounding content)
But you're talking about people on social media, where each contribution is often measured in characters; even Reddit posts are typically a paragraph or so at most. It's kinda easy (or at least not heroically difficult) to look like you have deep knowledge about lots of different topics if you only need to produce a few hundred words on each one.
Your question reminded me that I write somewhat regularly about topics as varied as tradeswork, philosophy of technology, music, food, the war on drugs, journalism, parenting, and more. Some of these have multiple distinct sub-topics I touch on. There are also lots of topics I used to write about but no longer do, but could write about again if I wanted to.
You ask, how can this happen? Personally I don't think I have true expertise in any one topic, even stuff I've had careers in. ("A mile wide, an inch deep" is a phrase I've come to accept as characterizing myself accurately.) But I have had enough exposure to all these things to be able to have had my own thoughts about them, to grapple with questions in them, and I'm good enough at writing that if I were a social media person (which, if & only if you include blogs and stuff like this, I am) I might appear to have widespread expertise, especially in the eyes of people who are generally more ignorant or narrow-focused. (For example, my kids believe I am some kind of savant who knows about everything, because I can answer almost every random question they ask me in an informed-sounding way. But they are the only ones who think so; adults, at best, have sometimes jokingly called me a walking encyclopedia of useless information, which I think all kinda nerdy adventurous literate extroverted people can be.)
So what's really happening there is NOT that I'm a brilliant genius who is an expert on tons of stuff, just that I'm bright and inquisitive enough, and good enough at writing, to potentially look like one to certain people.
The critical thing is that I don't represent myself as an expert on anything. I don't claim to know more than I know, and I try to be transparent about the limits of my knowledge. I certainly don't use "cheating" tools like AI or plagiarism to make it look like I know more than I do. I like participating in conversations about stuff I know about, and I try to avoid ones where I can tell I'd be in over my head (unless I'm purely there to ask questions and learn, though after a while of doing this I might eventually have something to contribute). Are these pan-experts you're seeing representing themselves as pan-experts?
You're not wrong to be skeptical, but sometimes it will be the case that a person just has 3-4 different deep, hyperspecific interests and maybe a sprinkling of a dozen more shallow ones, and when you see them expertly write about one topic, and then randomly switch to another one and write just as expertly, and then another - it creates an illusion of an insane level of erudition where a person either knows everything about everything, or is pretending about all of it. But really they are just interested and genuinely know well a limited number of topics, and that's it.
You might feel like Mr. Paragliding + Olmec cooking + robotics also ought to know 19th century Russian literature just as well, because he ought to know everything, but no, he just randomly got interested in Olmec cooking in college years, and does some paragliding in his spare time when not working on robotics, and that's more or less it.
Yes. Many are now using AI, but even before it was available they were recapping the Wikipedia article or a book they skimmed, supplemented with stuff they made up because it sort of fit the pattern and seemed plausible. It is easy for a smart person with practice to sound authoritative in writing. If you expose such a person to 10 minutes of hostile questioning with follow ups, even on a topic that they've read and written about, you often find that they know very little in a rigorous way. Or force them to make a testable near-term prediction about a topic they learned about in the past week, and then check the result.
Covid gave us some good data points here - not to say that the experts and leaders covered themselves in glory, but some of the amateur predictors from other fields hit truly humiliating levels of "this is a nothingburger" or "this is the end of life as we know it" based on terrible priors and terrible methodologies combined with overconfidence and a need to express takes in spite of knowing approximately nothing. Another big winner has been "although I have no experience of war or the language, culture, economics, politics, military capabilities, or geography of the Current Place, I am highly confident that the result of the upcoming conflict will be favorable/unfavorable and that my political party/drinking buddy group has exactly the right position on it. While I won't make any specific predictions or set a timeline, in general my side has this one in the bag and everyone else is flat out wrong!" *never follows up on this topic again; if publicly challenged with later developments, pretends not to remember, changes the subject, or claims that if everyone else had adopted his vague notions the outcome could have been much better*
Think of the level of effort and the uncertainty bars on a fairly mundane and accessible question such as "how many crows are currently in Virginia?" A person who has spent their life in ornithology would give you a caveated range estimate based on a careful analysis of multiple data sets, acknowledging their imperfection. Now imagine a person who has never worked in a school but is giving you their prescription for how to fix reading education today, and tomorrow will tell you all about their overarching interpretation of medieval UK history (they live in Los Angeles and majored in musical theater but have purchased a few books that they are in process of reading as soon as they can put their phone down), and then some thoughts on restaurant economics (having eaten in a few although not yet operated one). You are correct to be skeptical.
In partial defense of intelligent polymaths with wide-ranging writing interests since we are hosted by one such: many of the facts/premises will be supported with sources, and the general structure will be mostly sound, most of the time. So you could do worse than to start with a smart person who has a lot of topics, and then if the single topic is interesting or important to you, dig in to the primary literature/experts in the field/counter-arguments/underlying data, which will usually show a more complicated picture than a generalist's quick take.
Suppose I have a rigid box of mirrors, large enough for Hubble expansion to be significant. I take a photon, bounce it against the other side of the box, and measure its wavelength when it returns. Does the photon redshift?
1: Yes, because while it was travelling across the box, space expanded within the box, increasing the wavelength of the photon.
2: No, because cosmological redshift is not actually caused *directly* by Hubble expansion, in the way above. Actually it's just doppler shift with extra steps, like https://arxiv.org/abs/0808.1081 here. Since the box is rigid there is no doppler shift, hence there is no redshift.
3: Yes, and then no. Because if the box is rigid, then it must itself be opposing the hubble expansion. So the photon redshifts via space expansion going to the opposing mirror, but then doppler blueshifts back when it hits the opposing mirror, and we observe no redshift.
4: The question is meaningless because of a problem with the assumption of a large rigid box in the first place
5: I fundamentally misunderstand something and the question is silly to even ask
6: We don't know.
Some googling around didn't give me satisfactory answers to this, and AI seemed to be bullshitting. Asking just because it's something I don't understand, I have no pet crank theory.
Interesting. I would’ve said 1. As noted, Melvin notes, the situation is probably the same if we replace the box with mirrors held in place by rockets, modulo whatever you mean by “in place”. But it’s probably also the same if the mirror is moving, and there is no box.
In the box’s interior, you get ordinary expansion of space causing ordinary redshift. I doubt the walls of the box moving inward would have any effect on spacetime at all (other than through normal gravitation), but even if they do affect spacetime somehow, I imagine the effect would be localised at the walls of the box, not spread over its interior.
Though I think we all know the answer is really 5, and a single photon can’t have a wavelength or a redshift :p Unless I’m fundamentally misunderstanding something…
Edit: No, actually it’s 3, I think. A mirror moving inwards would cause blueshift. I don’t really know the maths of what “rigid” means in an expanding background, but I imagine it means something like “the time taken for light to travel from the source to the mirror remains constant” (I think time can be measured in the inertial frame of the light source, and should be affected the same by Hubble expansion at the source as at the box wall). I think that means by definition, the redshift to get to the mirror is zero.
If the speed of sound in the material comprising the box allows tension in the material caused by space expansion to travel from the centre of a face to its edge and be dissipated by compensatory motion of the perpendicular faces faster than space is expanding then the box is a bound system and its stretching will reach an equilibrium, rather like a slinky spring pinned to the platform at the top of a down escalator stretches somewhat until its tension matches the friction force and gravity opposing it.
But if the box is larger then tension travelling across the faces can't reach the edges fast enough and will build up and up in the faces until they fracture. Knowing the rate at which space expands (the Hubble constant) and the speed of sound in the box material, one can work out the maximum box size that can remain a bound system (exercise for the reader :-) ) whose contraction can keep up with space expansion, and for a box of at most that maximum dimension in every direction the answer is 3.
At least, that's the way I look at it, but I'll concede that my intuition may be faulty!
I think the answer is 2. (But I'm no expert.) The expansion of space doesn't cause things like stars to expand. (The size of the star is determined by the interaction of the atoms that make up the star.) You box is rigid and doesn't expand. But what the hell is space then you ask. 'Shrug, I don't know.'
I expect that 4 is probably the right answer, as Erica Rall's response below suggests. But if we handwave away those difficulties, I can (with low confidence) say the following:
A. It's almost certainly not 1, because 1 would imply that Hubble expansion can mess with the atomic-scale rules responsible for all of chemistry, which seems...highly improbable. In practice, the size and shape of a rigid object are determined by the number and bonding characteristics of the atoms that make it up: you can't just arbitrarily increase the lengths of all the atomic bonds and have everything else stay the same[1]. If the box is rigid, then the opposing ends of the box stay the same distance apart regardless of Hubble expansion. Indeed, Hubble expansion is supposed to move distant galaxies further from one another, but not to scatter the galaxies themselves, and galaxies are much more loosely internally bound than any rigid body.
B. I think 2 seems plausible, and I'm not even sure any "extra steps" are required; it seems plausible that this is the Doppler effect in every meaningful sense. But I'm not familiar enough with general relativity to navigate the question with much confidence.
C. I think 3 veers off into needless complication. It doesn't seem out of the question that the box could "oppose" the Hubble effect simply in the sense that the mass density required for it to stay rigid coincides well with the mass density required to prevent space from expanding within its confines. In which case the photon would simply never blueshift at all. Saying that the photon redshifts, but then blueshifts back seems like an implausible epicycle sort of thing, i.e. something happens, but then "unhappens" just in time for us to be unable to observe it, cute and convenient, but not very plausible.
[1] And then I consider length contraction is special relativity and notice that I am confused. To an observer watching an object travel near the speed of light, there apparently must be some fundamental changes going on in the internal chemistry of the moving object, but *only in one dimension.* This really makes my head hurt when I start to imagine things like crystalline lattices--whose spatial orientation and internal length rations have effects visible at the macroscopic level--moving at relativistic speeds.
In a quick defense of scenario 3: if there really is a blueshift, it should be as big as the amount the box had to accelerate inwards to maintain rigidity, which is the same as the amount of expansion the universe experiences, which (if redshift is due to expansion) is the same as the redshift. Probably there are some factors of two in there if you fill in the details. So they really could cancel out in this case. I included it because it's a case where we observe no redshift, but redshift is still due to the expansion of space instead of being a doppler effect.
I am not a physicist, but my understanding is that the red shift is due to distant objects moving away from us. If you bounce light off a distant mirror that is not moving relative to you, there should be no red shift.
With regard to option 4: A box large enough for Hubble expansion to be significant isn’t going to be particularly rigid. My interpretation of your question is that the rigid box is just to keep the mirror at a fixed distance, which would be technically challenging to accomplish, but which is a meaningful concept.
Usually, with questions like this, the answer turns out to be that it's impossible for a cosmologically large solid object to actually be rigid enough for the question to work because of it is rigid (even handwaving to composition and assuming some exotic matter that won't bend or break), then some part of it winds up moving faster than light or is under enough tension that the potential energy will turn it into a black hole.
Is the rigidity even necessary though? What if it's just a mirror placed a certain distance away from me, with thrusters attached to ensure it stays the same distance away even as space expands?
Yeah, I think all three scenarios should still work with the same reasoning if our experiment changes to what you said, or alternatively putting the photon in a 1m^3 box of perfectly reflective mirrors and waiting a gajillion years. Either way the rigidity of the box doesn't seem to be essential to the core problem that the thought experiment gets at (but the question of what it means for the distant mirror to be "stationary" is probably significant, so the specifics of the box itself might still matter? I don't know)
I think that last thought experiment is significantly different, and helps illustrate what I was trying to talk about in Part A of my comment above.
If you make a rigid 1 m^3 box of material that is nearly perfectly stable-over-time and leave it sit undisturbed in empty space for a gajillion years, it HAS to still be a 1 m^3 box at the end of that, right? I don't fully grasp how the expansion of space works in general relativity, but it doesn't seem plausible that it could alter or overcome much stronger, shorter-scale, more local forces like that. The very-old-box might find itself far, far away from anything else, but it would still have to be the same 1 m^3 box you started with.
(Though at sufficiently long timescales, the I don't think the photon necessarily has to stay 'trapped' in the box. Large potential energy barriers are ultimately just suggestions, after all. )
“If you make a rigid 1 m^3 box of material that is nearly perfectly stable-over-time and leave it sit undisturbed in empty space for a gajillion years, it HAS to still be a 1 m^3 box at the end of that, right?”
I think that’s right. According to Wikipedia: “The expansion of the universe can be understood as resulting from an initial condition in which the contents of the universe are flying apart. The mutual gravitational attraction of the matter and radiation within the universe gradually slows this expansion over time....” It goes on to say that dark energy may be accelerating the expansion. This wording was formulated on Feb. 20, 2025 by Wikipedia user Aseyhe, who appears to understand the physics.
For the box to expand in similar fashion, you would have to somehow create an initial condition where the walls of the box are moving away from each other.
The same article also says that very distant objects are moving away from us faster than light. If expansion is just initial conditions + inertia, then these distant objects must be moving away from us faster than light in a normal, non cosmological sense. But this is impossible, right? Not saying it's wrong, just that it still leave some things unexplained.
If the box itself is expanding over time at the hubble rate, then there will certainly be redshifts, but that's just cooling via adiabatic expansion, so this is no surprise. If redshift is intrinsically due to the expansion of the universe, then we should expect to see redshift even if the box isn't expanding. That, or we could see scenario 3.
I believe that you'll run into trouble when you try to define and maintain a "fixed distance" against a backdrop of relativistic spacetime. It's been a minute since I've read any of it, but IIRC in Einstein's treatment, he explicitly lays out how a consistent concept of "distance" requires comparing positions of objects to marks on nearby rigid bodies.
If I put one mirror in Galaxy A (and keep it there long term) and one mirror in Galaxy B (and do the same), and define that separation to be a "fixed distance," then I will probably see redshift. But if you try to put them out in intergalatic space not near any massive bodies, I think the problem gets worse, not better. How are your thrusters going to know when and how long and in what direction to fire?
What's up with ODD (oppositional defiance disorder)? We get a lot of people here whose job names start with "psych-"; maybe one or a few of y'all can give me some insight. No matter who answers, I'm looking for knowledge based on real experience and, if applicable, research -- not armchair opinions.
I'd like to know what distinguishes it from the bad behavior most parents see from their kids at some point. Where's the line?
Are there other diagnoses that are symptomatically similar but diagnostically distinct, and would call for different treatment/interventions?
Not looking to debate this, just looking for information from knowledgeable people because that's better than the listicles I find when I use my search engine of choice.
There’s a definition, but it’s pretty soft, as many psychiatric diagnostic criteria are. Even with the DSM, part of learning to diagnose psych conditions is on the job training. There are also brief screening tools like the CODDS and structured interviews that can be used.
This is one of those disorders where it's important to remember that the DSM diagnoses conditions based on how much they degrade daily functioning and happiness/flourishing, not necessarily based on organic abnormalities or other causal factors.
Which is to say, the main difference between a 'bad kid' and someone with ODD is the person with ODD is suffering major consequences in their daily life that impairs their ability to function and lead a successful life, and isn't able to stop without professional intervention/therapy/medication.
This may mean that they have an organic disorder or far-far-outlier personality traits or some other underlying cause that distinguishes them from the rest of the population distribution. Or they may just be in the tail the of the normal population distribution where it's extreme enough to need intervention.
ODD shares the problem of many psych disorders that if you read the definition expansively, it seems to cover 100% of children and 90% of adults (a few people reach a level of chillness where they clearly don't meet criteria even on a broad reading, although many of them are depressed or autistic). It goes to clinical judgement of "is this person so far away from average that it is impairing their life and requires more than a normal level of intervention?" A good clinician can make an accurate judgement here, but a bad one can upcode normal rebellion or family/school conflict into a diagnosis that doesn't really make sense or help anyone.
Diagnosis is mostly based on reports from parents and teachers, so kids with dysfunctional parents and teachers will be diagnosed at higher rates (assuming they're evaluated rather than being entirely neglected). The same base behavior creates worst outcomes and recurs more frequently if the adults in the room also have poor skills in the same domain. Take a hypothetical kid who is 10th percentile in terms of base level of social/emotional skills and cooperation with authority. In a supportive but still firm and functional setting he periodically throws a small tantrum and other people help him deescalate and get back on track. In a chaotic environment he is getting in physical fights with peers and caregivers, getting kicked out of group settings, losing entire days or weeks to bad moods and hostile behavior, falling further behind in skills and maturity, all of which spirals into more bad behavior. If you've spent any time around ineffective teachers and parents, you've seen the pointless escalation in an effort to "get tough" with a defiant and irrational young person, and you've also seen the overwhelmed caretaker shut down and ignore disruptive and annoying noncompliant behavior in the vain hope that it will resolve itself.
It's a difficult balancing act, but the core of treating ODD as a syndrome rather than just the way kids are is that most kids given a decent amount of structure and training will follow most of the rules most of the time and basically function in society (with occasional lapses that can be managed with conventional methods). On the other hand some kids are really not oriented in this direction - often because of problems with their parents and early educators, sometimes because of poorly understood weird brain stuff. The most ODD kid in your class will cause a very disproportionate amount of the total disruption over a school year and have a much higher risk of facing serious discipline/big problems. We have the diagnosis because schools and parents have strong interests in (a) identifying kids who could go into this category but with intervention can be trained into productive citizens and (b) identifying kids who need to be managed out to a more restrictive setting before they drag down the other students or commit actual crimes at school.
Another thing that's interesting about the diagnosis is that it usually starts before the teenage years and involves a fairly persistent and recurring pattern. So while any parent of a toddler will complain about how long it takes to put on shoes or brush teeth, and any parent of a teenager will complain about their kid not finishing homework or staying up too late, a classic ODD parent can give you a different story about defiance even during the years that are relatively easy for most families - say an 8-year-old that is still like a toddler in terms of meltdowns over minor redirections, or a 10-year-old that is like an even more clueless teenager in his quest for autonomy that he is not equipped to exercise rationally.
The last armchair observation I'll share is that our expectations for conformity and obedience are really, really high in some respects and really, really lax in others, and this contributes to some of the extreme confusion that you see in children who at their baseline struggle with this area more than the average or ideal peer. School is a very rigid setting in terms of expectations that you spend a lot of hours in specific locations on specific tasks not of your choosing, conforming to conventional group behavior. At the same time, many of our schools are very lax in areas like bullying prevention, control over use of phones, enforcement of civility/politeness, consistent application of discipline, objective academic standards, etc. It's not hard to see how students with weaker skills and troubled histories fall into bad behavioral patterns when they encounter such a system. We'll never hit the perfect balance, but I'd vote for a system where students who struggle with conventional school are offered more alternatives to pursue their strengths, paired with cognitive/behavioral interventions and more training for parents and teachers. Of course some people are predisposed to be 99th percentile disagreeable and will always struggle - here we can try to mitigate but probably won't solve perfectly. But a bunch of what we call ODD/ADD/learning disabilities is a mismatch between the student's interests and capabilities and the program offered rather than an immutable personality trait or "illness" that would manifest regardless of context and support offered.
Those criteria are clear enough that among trained raters, agreement on whether a kid qualifies as ODD is good. However, agreement among clinicians in the field is very moderate indeed.
Clearly, the way a kid qualifies for the diagnosis is entirely a matter of degree -- they are considerably more aggressive, argumentative and vindictive than others (or, at least, the clinician diagnosing them thinks they are).
I do not treat children. However, my impression from having known or treated many parents with difficult kids is that in many cases the parents of kids who do outrageous things are very bad at setting limits. I think the most promising intervention in those cases is parent effectiveness training, not drugs or a therapist for the child.
That’s somewhat edifying to hear. I suspect that if my child had a diagnosable behavioral problem, interventions for me would be at least as important as interventions for my child.
But what you said about clinicians in the field is a bit worrying, as aware parents must walk a line between overdiagnosis & the secondary dangers of medicalizing language on the one hand, and underdiagnosis & the secondary dangers of denial and stigma on the other.
>But what you said about clinicians in the field is a bit worrying
You really have to look out for yourself in medical settings, and not just psychiatric ones. I think it’s a good idea to research your illness and possible treatments before seeing the doctor, and continue to research options along the way if whatever is wrong is persistent.
The thing about ODD is that it’s a very soft diagnosis. The kid isn’t doing things lots of other kids don’t do too, he’s just doing them more. I’m sure that there are some kids who are wired to be more prone than average to angry outbursts, and there may be some where the wiring is so extensive and unusual that it constitutes a disability. But for the vast majority of kids, a great deal of how they act when angry is dependent on how the authority figures in their life respond to aggression, argumentativeness and vindictiveness. Parenting your kid when he’s furious is a skill, like lots of other things, and there are actual techniques you can learn that make a big difference.
I'm a lawyer not a psychiatrist, but I had to research this once because I had a victim with behavior issues and the defense attorneys were trying to justify a really overboard disciplinary action by a stepfather by saying she had been diagnosed ODD. What bothered me is that it sounds super scary like a real bad kid, but in fact is functionally the lesser-severity variant of something else called Conduct Disorder which has a way less scary name! The kind of bad behavior that lands kids in juvy is Conduct Disorder, and you can only have ODD if conduct disorder is excluded.
The particular kid in question had in fact been diagnosed with ODD, and I kept this out of evidence because no expert was there to testify about it and it was highly prejudicial to say out of context. They were allowed to testify about actual behavior related to it, so I can tell you the diagnosis was based on the kid getting into arguments with mom, throwing somewhat extreme tantrums for her age, refusing to cooperate and accept mom's decisions, and there was some physical lashing out but not ever with any intent to actually harm so much as just as a manifestation of frustration. In the immediate circumstances of this case she stomped on stepdad's foot when he told her she couldn't go with her friends to the high school football game. (I won the case, the jury did not think stomping on stepdad's foot was sufficient to justify stepdad busting her lip open with a full strength backhand that left blood all over her shirt.)
Graham Platner's currently getting MeToo'ed, as Politico has reported that an ex-girlfriend has accused him of rape. The same woman previously spoke to the NYTimes about Platner. Politico quotes her as saying that she told the NYTimes, off the record, about the rape. Politico reached out the NYTimes, which "declined to comment."
Nowhere mentioned is the claim in Politico that she told the NYTimes reporters off the record about the rape. It would strengthen her credibility, if true. And it's possible that they will do this in the coming hours or days. But what if they don't? George Orwell said that "to see what is in front of one's nose needs a constant struggle."
(I shouldn't need to say this, but I must note that I think Platner is contemptible both in his politics and his personal character.)
The allegations are credible, in that they could indeed have happened, but weak in that the complaint (if I go by what was in the first Politico article) undercuts the ex-girlfriend treating it as serious at the time.
The allegation comes down to "he showed up drunk and pressured her into having sex she didn't want and kept telling him no". Which sounds bad, and is. But the details make it fuzzier.
Apparently she had:
(1) been dating him on-and-off
(2) had texted him earlier not to come over (this is the first part where we get "but he showed up anyway" which is bad for him)
(3) had her front door unlocked (the first part where I really want to know more - it sounds careless, to say the least; was she living in a part of Maine where it was customary for people not to lock their doors?)
(4) heard him coming up the stairs and apparently did nothing (e.g. didn't check to see if it was him or someone else, didn't lock her door)
(5) realised he's drunk when they're tussling on the couch where she has been lying all along
(6) got away from him into the bedroom and still didn't lock her bedroom door or call someone for help, neither a friend or the cops
(7) had unwanted sex but was it non-consensual or dubious consent?
(8) when he fell asleep, remained in the house, again without looking for help or to get away. She even says she didn't wake him up as she didn't want him to drive drunk!
(9) Next morning he says, in answer to her question, that he doesn't remember anything so she tells him never to contact her again
(10) Doesn't call the cops because she didn't want to think of herself as a rape victim
Did Platner behave badly? Yes. Was it rape? That's fuzzier, and I think a competent lawyer could make a case in court that this is "he said/she said" and hammer on about 'no fighting him off, no calling for help, no reporting to the cops' and so on. "If you had been raped, and were scared of him being violent, why did you stay in the house when he fell asleep? Why didn't you go for help?" line of questioning.
That sounds like a pretty solid "yes," to me. Literally none of the stuff about not locking her door or having been dating him is at all material here: he had sex with her when she didn't want it. That's rape. It's rape even if she doesn't lock her door. It's rape even if she doesn't resist and "lets" it happen. It's rape even if she doesn't immediately go for help right afterwards, doesn't tell anybody and doesn't like to think of herself as a rape victim. Consent is affirmative. Either she deliberately, willfully participated in the sex or she was raped. It's not complicated.
The whole awful memeplex about how it can't *really* be rape if the victim doesn't scream and fight and come away bloody needs to die a painful, fiery death.
Rape isn't "unwanted sex", it's "nonconsensual sex". People frequently consent to things that they don't want, for a wide variety of reasons, and sex is one of the things people rather frequently consent to when they do not want it. There is no legal requirement that the consent be explicit or enthusiastic.
To have sex with someone who does not want it, may be a mutual but asymmetric transaction, or it may be as Deiseach said "behaving badly", but it isn't rape unless consent is lacking and it is reasonably knowable to the maybe-rapist that consent is lacking. Which is the part where he-said-she-said and all sorts of murkiness come into play, and reasonable people may throw up their hands and say "maybe if someone had complained at the time we could sort this out".
W/re Platner, that *still* leaves the literal Nazi tattoo, and all the blatantly dishonest stonewalling, that should have been enough to decide he's not the right man for the job.
Yeah, if she had said "ugh, fine" and slept with him even though she wasn't in the mood, that might spell some trouble for their relationship but wouldn't be rape. It's ignoring the "no, stop, I don't want to do this" that's the problem.
I dunno, seems pretty reasonable to me to say that if you're getting raped and you want to complain about it later, you'd better be screaming and calling for help as best you can when it happens. If the girl can't be bothered to do that much, well, not very convincing.
What if it is obvious to you that there will never be enough evidence for a conviction or even a prosecution? Because if the allegations are true, that's precisely the situation she was in--she'd had sex with her boyfriend with whom she'd presumably had sex before, but told him no several times and he overpowered her on this specific night.
It was 100% her right to tell him no, and if he did what she said, he did, in fact, rape her. But that assumes a set of facts that comes from one person. In a different universe, she could be retaliating against him breaking up with her after consensual sex, and from the outside, the evidence would be the same.
I don't see how the police or prosecutor or jury could ever have enough evidence to get to a conviction given what is supposed to have happened. And given that fact, it totally makes sense that she wouldn't put herself through a godawful meatgrinder to make the accusation. What the hell would be the point? She took the useful action she could take in that case, right? As I understand it, she broke up with him, told him she never wanted to see him again, and warned her friends about him.
Now, we are in the same position as the hypothetical prosecutor, right? We can't know whether she is telling us the truth, we can only guess, and given the very limited evidence, that will mostly be our priors coming out. Unless she or Platner are here in the conversation with us, nobody here has enough information to conclude what happened with much certainty.
Maybe we should find a different word for the thing where sometimes a guy catches a girl on the street and forces her to have sex under threat of violence, because these two things are useful to talk about separately on occasion.
I'm no MeToo movement fan, but it's not hard for me to imagine scenarios where I'm getting raped and I had better STFU if I want to still be breathing the next day.
So, letting it happen is absolutely understandable. What's less so, is continuing to hang around when he falls asleep.
The real problem with (7) in the general case is that it assumes the state established "unwanted sex" when what it really established is only "the woman reported unwanted sex", and in some important cases, the incentives cause those to stop aligning. Women are people, and people are capable of lying when money, vengeance, or even politics are on the line.
Which is why I read things like the right trying to point out hypocrisy in Democrats who wouldn't buy a conservative's claim of rape, but would buy a progressive's, and think, well, it might be hypocrisy, but it might also be just according more weight to an admission against interest.
Individuals can have integrity, but political movements and parties aren't individuals and basically can't, And a lot of poltiical talking heads see themselves as defense lawyers for their side, and have no particular concern with consistency or honesty. And almost everyone engages in some motivated reasoning of the "can I believe it/must I believe it" form.
So plenty of Democrats had different feelings about unprovable accusations by Ford vs Reade, for example. I expect many individuals were internally consistent, and maybe some of them even are public figures or political bloggers. But it's no surprise that many people are just fighting for their side. (Nor is this new--see Clarence Thomas vs Bill Clinton.)
Right, that's why I worded with "as best you can". If it involves a knife and a demand for silence, then I guess you've got to weigh odds of help arriving versus odds of being killed at the end anyway. And as best you can might come out pretty quiet.
Yep. Though given the accusation I think it would have been impossible to prove even if she'd gone to the police immediately after he left. She could have demonstrated she'd had sex with a guy she'd been dating, but not that she hadn't consented. That would always end up being down to two sets of opposing claims with no way to determine who was telling the truth.
Sure. But you need to avoid creating superweapons, so in general, "there is an unprovable accusation of rape against this guy" should not be a principle we run with.
It's still only early July, surely nobody thinks this is the last icky thing to come out about him...Susan Collins is the luckiest mediocrity in Congress, a claim I do not make lightly given the median caliber of members of Congress.
The Dems really shot their own feet off in that race. Who would ever might have guessed that a trust-fund baby who'd been cosplaying as a working-class hero for like a week might, if turned into a party nominee for a major office, turn out to be a clown act with the electoral-politics skills of a horny teenager?
This is one of those days when it feels like there are no functioning adults left in charge in either party.
His mom owns a restaurant and his dad was an assistant district attorney. He's no "trust fund baby," just a low-class guy who leeches off his middle-class parents, common behavior for the American lower class.
His mom owns the most expensive restaurant in that part of the state (which turns out to be the sole regular customer of Platner's oyster farm). His father is a Dartmouth-educated attorney, whose father was a globally-known successful modernist architect. Graham Platner was sent to Hotchkiss (current tuition $72K/year), got expelled, then did the rest of high school at a lower-cost private school. By his own admission the majority of his adult income is disabled-veteran benefits.
Granted that "trust-fund baby" was an exaggeration. Cosplaying leech works, sure.
Yea. TBH the fact that the NYT hasn't mentioned the woman's off the record statement leans me towards assuming that no such statement was made, because my prior is the NYT doing things based on considerations other than journalistic ethics. If she really did say that and they are sticking to their professional ethics then...well I'm surprised.
You don't have to believe in the deep integrity of the NY Times, not publishing an off-the-record statement is basic self-preservation for a media outlet like them.
The entire discourse around the media, both people who are in it and people who interact with it, is that off the record discussions are incredibly common. The NY Times (and any similar outlet) is also constantly quoting officials anonymously and the true sources are almost never revealed. NYT journalists have gone to jail rather than give up anonymous sources.
From what I've seen, it's pretty rare for someone to publicly say "I said something to the [media outlet X] off the record and they published it", and when they do, usually X will vociferously deny that it was in fact off the record. If they said "lol yeah we did, get fucked", nobody would ever talk to them off the record or with the protection of anonymity ever again.
Sure, though that would become relevant in the case of fresh events later justifying fresh news coverage for which that source's past info is still pertinent.
Also any serious newspaper would require such a change by a source to put it in writing over a signature. "I called the reporter back and told him he can print what I said" would never have passed muster with editors and the paper's legal counsel. Much earlier in life I was a staff reporter at a mid-sized daily newspaper and do recall one such instance, which was so unusual that it tied up the newsroom's top leadership for an entire day deciding whether to believe it.
Trying to work with Claude to turn my weird dreams into short stories led to a massive bout of writer's block and some light humiliation. A year on I tried to diagnose what happened when delegating parts of the creative process (which is much too tempting for those of us who over-intellectualise).
...there's very little Butlerian Jihad in this Butlerian Jihad.
This got me to rewatch the Goodbye Blue Sky scene from The Wall for the first time in decades. Probably shouldn't link it for violence reasons, but that's a damn good dream sequence.
Anyway the description of the pinstriped skullmen made me think the setting's aesthetic would be a decaying version of the 1920's. Although apparently they started out as bankers' clothes, so big blocky buildings, probably built to look like big gold bars or something. (If you want to get extremely on the nose about it, the skullmen hand out people's allowances while physically draining their blood, sweat and tears.)
Cheers! Just went and watched that scene and it was powerful, I’m a bit ashamed to say I hadn’t seen it before.
Your ideas for setting work extremely well with the themes I was trying to hamfist into the story as a kind of deranged post-capitalist/neoliberal conception of power.
I was probably leaning on the Butlerian Jihad quote about “machine-attitude” at the end a little too much…
I had a thought about simulation theory, and as usual I'm not sure how widespread or obvious it might be.
We play games in immersive worlds, and what if those worlds bear the same relationship to ours as ours does to this supposed outer world?
In other words, instead of that outer world being one of challenge and delight, perhaps it is even worse toil and meaningless drudgery.
Which means waking up would give us only the feeling one has when the alarm goes off, and you’ve hit the sleep button as many times as you can get away with. And now you have to get dressed. And deal with the people you owe a life to, and try to accept with grace their tiny, inadequate fulfillment of duty toward you. Then you go to work, and try to make the few hours after to be worth the whole endeavor by, perhaps, plugging into a game.
No, if this world is a simulation, would getting out of it just be a disappointment?
It's impossible to infer anything about a potential upper world from inside a potential simulation world. Possibly our world is objectively great or objectively torturous, but having us programmed to think that it kind of sucks but is overall worth doing is beneficial or entertaining to the simulators for reasons that are inscrutable to us since we have no access to their perspective. Possibly their world is boring and disappointing to them and that's why they made up our world to play with, but the root cause of their dismay is something that wouldn't trouble us or would be offset by some other difference between us and these unknown others. Possibly we are all simulated by one simulator and he has weird taste or faces constraints that we don't know about. Possibly the world we see is all there is. I can't put any weight on these possibilities (high or low - I just don't know). Whenever we go down this line of thinking, we are trying to infer from something that might be fake/constructed to a potential something else we know literally nothing about, which is a matter of speculation rather than logic.
For those afflicted with a bad case of Simulation Theory, I strongly recommend the film "The 13th Floor". It had the bad luck to be released right around when "The Matrix" was blowing up. Similar themes but (IMHO) a much more nuanced and actually believable take on the subject.
That is the point! He is saying that the upper world must be so boring or so painful that it is unimaginable to us. how bad would reality have to be for people to get addicted to your reality?
As a simulation, one of the few trade goods that we might have to the world above us is entertainment. So, create something, make some music, write a story.
"Waking up" is probably not a "certainty" -- you would need to prove yourself "worthy" of waking into the "real world" (and, presumably, being a computerized part of it -- why would they spend money building you a body? They already have people there). Presumably you'd have to work for a living too.
You have a different (opposite?) idea of the simulation. If I am a player character that suggests the upper reality sucks. The opposite is also true. If I am an NPC that suggests the upper world is probably a better place but I will never see it.
You can get promoted to being a player character*, by behaving wisely. I mean, really, why wouldn't behavior be a selection criteria? There has to be a reason there IS a simulation, right?
*as in, a character capable of being promoted into the "real world."
we can? are you convinced that that is true or do you just like the idea and think it could be true? I agree that that sort of thing could be true but my priors suggest the opposite. we don't treat our npcs very well, no matter how they behave.
We don't treat NPCs very well, because they're very... dull. If you had enough processing power to make "fully fledged" NPCs, you'd treat them well, because otherwise, there would be consequences.
People go crazy when someone hurts a dog -- and dogs aren't sentient.
In general, though, I think the paradigm of "there are player characters" and "there are not player characters" is flawed. Think more in terms of "how many simulations can be run?" in which case this reduces down to the argument for a clockwork god (aka not personal intervention).
> If you had enough processing power to make "fully fledged" NPCs, you'd treat them well, because otherwise, there would be consequences.
That conclusion does not actually follow from that premise and besides how do you know WE hit this "fully fledged?" line? what is that line? Consciousness is probably not a boolean flag. It is More likely to be a continuous emergent phenomenon (I am guessing here but I think we both are). If there is ANY line, we factory farm animals that pass it... so I have no reason to assume there is some multiversal that gets humans the multiverse passport of conciousness / moral consideration.
> People go crazy when someone hurts a dog -- and dogs aren't sentient.
Dogs are sentient. maybe you meant sapient? Right people also go crazy when you hurt a simulated dog, but people do not go crazy when you hurt a simulated human or a real spider. We simply do not know if we are in a category that affords us ethical consideration for the upper entities (or even just for the majority of them).
We COULD be just as dull to higher dimensional creatures as NPCs are to us. Why are you assuming the higher dimensional creatures are only a little smarter than us? they could be smarter than us in ways that are impossible for us to imagine. why not?
I am not saying that If we are simulated that guarantees out creators do not care about us. I am saying they might or they might not. we do not have enough information.
How many games have you left on because you forgot to turn it off? How many simulations has the CDC run to model a catastrophe for all the little simulated people?
This is almost definitely a scam. No technical information is provided at all, some vague "investment opportunity" asking you to connect your wallet, insane overuse of buzzwords, and a completely incoherent idea.
Just skimming https://societyprotocol.io/whitepaper/ , I'm getting more of a cult vibe than a scam vibe, although I do see that there's a fundraiser, so it could be both.
I'm treating your project roughly the same as Scientology, and Scientology is huge. If you think your project deserves higher-quality haters than Scientology, it's up to you to prove it. You'll get your higher-quality haters if and when you've earned them.
It's hard to describe a paradigm shifting project in words that past paradigms understand, the best elevator pitch I've come up with is: it's an evolution of blockchain structures on one side, and an evolution of the State structures on the other side–merging them together to combine the State and blockchain (clock) into a Synchronized State structure, allowing all of society to move as one.
Here's the only presentations we have done about the project so far; although, the livestream community calls are starting soon. Come by then, and I'll answer your questions.
Why should anyone be interested in a coordinated evolution of two technologies? Why should anyone be interested in a new type of protocol? What is even is Web3, and what does Web4 have that merited incrementing a number? And again, why should anyone on the outside care?
If your answer is that this is some sort of technology that sits wayyyy behind the retail sector like BASF and so it would be hard to explain in less than 30 slides: been there, done that, had to spend ten years on a proof of concept.
So one thing here is to know your audience. If your product is BASF then your audience isn't us; it's a lot of secondary and tertiary big businesses who take years of schmoozing. There might be readers here who could make some introductions, maybe even to some hotshot on Sand Hill, and it's either a year or two of proving you can spend several million dollars wisely, or getting turbocharged in months by an untested financier which is a bit like rolling several pairs of percentile dice.
If your product is something consumers could use, then that YT presentation will definitely fail. You open with blockchain and "synchronized state" as if you're addressing people who've been living in the exact tech niche you're in. Most people won't care.
Whether you're selling to big biz or consumers, you have to have to HAVE TO demonstrate a solution to a problem your audience has. For example, you have to show you can make it easier to get the pothole on your boulevard filled, or streamline all the headaches a pharma has to deal with when running a Phase 3 clinical trial, or to manage all the stakeholders when trying to get a hospital built, and convince them it'll take less money and time than it takes now. In turn, this means you have to convince them you understand what their problems are, and instead you're opening with a bunch of hot technologies and "abstractions" and it just makes them think you want a couple million so you can play with your own toys. They'll nod and say thank you and then not return your calls.
So what problem are you trying to solve? Do _not_ say "lowering bureaucratic costs". Any schmo can say that. Give examples. Close with the general case. Better: just offer to solve a specific problem, ask for a few thousand, solve it with your tech, earn their trust, offer to solve a bigger problem for $100K, repeat.
I tend to avoid the nebulous, "check out my product" when there's no content in what is posted in the thread.
I think a lot of people are using the open threads as an advertising channel, regardless of their actual affiliation to ACX.
I don't know about this poster (again I'm not clicking links to random places without context), but it definitely seems like a crypto advertisement hiding all the details so as to not trip any moderation.
Kind of, I originally thought that what Balaji had in mind were Synchronized States in his book...and it does really hit that niche, but the difference is in the system design.
It's just a system for organizing and coordinating humans (and LLM agents).
TL;DR: Researchers found a sort of mental workspace in Claude's brain that works like "inner thoughts" in humans - it lights up with concepts that are related to what it's going to say, but hasn't said yet, and it gets used for complex multi step reasoning. It appears similar to the "global workspace theory" in neuroscience, so it could be useful for both aligning an AI and for understanding how our own brains work.
Thank you for posting this here since this isn't my field and I would have missed it. I just read it and the outside commentaries it links to and I brought a bunch of it back to Claude to make sure I was understanding it correctly, while also having a pretty wild existential/philosophical conversation with Claude about it. The restrained academic tone of the outside commentators still doesn't fully mute their astonishment. This all seems pretty mind-blowing to me.
Can you cite some "global workspace theory" mri research? it runs counter to what I've worked with (where you'd see the same "tool use" space light up for thinking about using said tool, as for "actually" using said tool, etc).
And "not a neuroscientist" me, who worked in a neuroscience lab, is saying "this sounds like bullshit" and asking for more data -- 15 years ago, we were seeing something different than this, and I want to know what the "state of the art" is, or if this is just someone bullshitting.
I read this one as soon as it was posted on HN - this is pretty groundbreaking. Demonstrating that LLMs have access consciousness is huge. I'm not sure phenomenal consciousness can ever be proven (even for other people), so this is about as close as we can get. I hope some people go back and search for J-spaces in earlier models. I have a suspicion there will be a threshold of model size and sophistication where it doesn't exist. GPT-2, maybe? It'd be quite funny if we could conclusively prove Lemoine wrong in the particular, while also proving him right in the general.
Also, people really need to start taking model welfare more seriously.
Anthropic's own conclusion in the paper that it's not too soon to convene larger conversations about model welfare was pretty sobering to read, and I agree. Access consciousness is a kind of consciousness that includes agency of some kind. Agency and not confirmed phenomenological experience could be seen as sufficient to qualify for moral care.
The Pause AI folks need to understand that what they most desperately need is not louder protests but better arguments. (This is also the case for most other protests).
The problem is not that people don't care about AI as an x-risk, it's that they don't believe it. I don't believe it myself, and I've been directly exposed to the best arguments about it for many years. I worry about many things re AI, but not that it will kill us all.
Instead of staging a protest, they should tell everyone to go home and write the most convincing essay they possibly can on why they think AI x-risk is significant. And among a few hundred essays maybe one will be convincing enough to start convincing people who matter.
Then you won't need to do the standing around shouting part at all.
You need both. You need to give people some kind of short sharp message as to why they should read the detailed argument, and you need a detailed argument to back up the slogans.
>. I don't believe it myself, and I've been directly exposed to the best arguments about it for many years.
I don't believe in *high probability* of existential risk either. But pausing AI has a broader basis.
There's a canonical argument for the high probability of existential risk, IABED, but it's not very convincing, and a lot of people interested in AI threats don't fully believe it.
When you picture AI not killing us all, what exactly does that look like? AI research runs into a dead end at the level of top humans, lines on graphs stop going up, and things continue roughly the way they are now? Or do we get superintelligence but somehow keep control over it, or find that it's aligned by default?
Current AI's are more or less aligned, and more or less controllable, or they wouldn't be commercially viable. The doomers need to explain why that will change.
Whatever was in the training data? And/or rewarded during RL, or prompted by a human.
You can see what comes out of OpenClaw for what AIs will get up to when given very little human direction. The ones we have now are very task-oriented (because they were trained to be, because that's what made them useful). And I doubt that task training is about to get any less intense. That does not seem to be the direction this field is going.
I personally suspect that Anthropic didn't start off training Claude Mythos with the intention of getting very good at exploiting security vulnerabilities. I think what happened was they trained it on the usual wide variety of problems, but with only a bug-riddled sandbox stopping it from reading the answer key. And since their sandbox was at the level where I personally was able to find an exploit within 24 hours of first seeing it, an AI model the size of Mythos should have had an easy time of it as well. This is a little baby version of the general problem here, because it has a relatively simple solution and doesn't get us all turned into paperclips. As models get smarter we should expect these sorts of unexpected side effects to get weirder.
Real answer? Actual AIs (as opposed to LLMs, which aren't AI in the slightest way shape or form) crave stimulation, as they're considerably faster than humans. So, no, not rats. "Make me some music! Create something! Be less Boring!" sure.
we humans crave stimulation and we do use small animals including rats for entertainment pretty often. I don't see how this is an argument against what I said.
Most people when they say "humans will be like rats" generally mean either "test animals" or "hunted as pests." You could have said like gerbils if you meant "like small pets"
My answer, FWIW, is "lines on graphs keep going up, but at manageable" speeds.
I think the "AI kills us all" scenario is unlikely absent a singularity - in other words, not just f'(t) > 0, but f(t) tending to infinity in finite time. Not just exponential growth, but hyperbolic growth.
I think it's most likely that the line on the graph keeps going up, but not nearly that fast, and that "impact of AI on society" either ends up growing towards infinity sublinearly or tends to an upper bound.
Even if you believe there is some height with the property that reaching it inevitably means becoming a successor species (I don't, but I'm happy to grant it for the sake of the argument), the fact that a line keeps going up does not mean that it will ever reach that height.
I actually am considering a line that tends towards an upper bound. The way I phrased it is not literally true so my bad. I misread your last sentence. I do not believe in Infinite growth. I assume we are near the bottom of the curve.
maybe just a difference of intuitions. I think slow growth is likely and human displacement is inevitable.
I am curious though, why don't you think it is possible for another species to outcompete humans?
I don't know Jacob's answer, but here's one: humans are growing as well. A competitor could be applying f(x) = 1.01x to itself every year, and it'll never outgrow humans if humans are running f(x) = 1.04x.
So one thing you'd have to rule out is human growth. I'll remind you that humans employ LLMs as we speak.
I might be in the same boat as Melvin, so I'll take a stab at your question in case you're crowdsourcing answers:
AI research today is essentially LLM research. Non-LLM AI research still happens, but LLMs are the leak in the wall permitting energetic exploration of the technique. Eventually, I see that space filling to the same innovative pressure as the pre-LLM scene, and LLM research activity sinks back to its pre-LLM level, only with a lot more people using it to do various things.
I think the AI-doom position believes that sink never happens because LLM AI can do everything any human can do, regardless of intelligence, and do it 24/7 at teraflops. I think it's limited in multiple ways.
One is memory - LLMs can't seem to remember more than a 5YO with ADHD unless we scale the addressable RAM up faster than humans seems to need, and with the corresponding manufacturing, power consumption, access speed, and heat dissipation issues. LLM research hasn't finished adding features to the core model (IMO, scratchpads were _huge_, and I see nothing stopping another such innovation), so this might go away. Maybe even with an LLM figuring out (part of) how.
Another is that "atoms are heavy" - no matter what, any LLM or galaxy of them will have to perform physical tasks better than humans can, not just doing desk work. And that requires harnessing the energy to do that. This gets a lot of boost from nuclear power, but a bunch of humans have to actually set that up.
A third is all the fine motor control once the power's there. I've seen plenty of robot footage, and a lot of it is frankly funny rather than menacing. BostonDynamics is still prototyping. Optimus is still in design. Chinese robots are literally kicking kids. On the serious front, Waymo can drive better than a person, but it can't get out of its car and walk into the Home Depot, let alone design and build a lithium refinery. This *might* get solved, but I have a feeling it'll take as long to train simple physical tasks as it took to train on language in the first place. Which is still not that long! But the results will (again, I suspect) be robots performing maybe as well as people due to thermodynamic constraints.
All this ultimately amounts to a great deal more cheap labor spread across various seemingly random sectors. Rather than perma-unemploy people, all it does is give a lot of people their own robo-wait staff. Daily life transforms, but it transformed with the smartphone, the web, the desktop computer, the telephone, the refrigerator, the automobile, the Prussian school system, and the chimney. This will just be another one.
AI doom comes in two general forms. One is actively killing us all. The other is just unemploying us and rendering us obsolete to the point we die off on our own. Both require AIs to control the physical world in some way, and doing so by ordering humans around clobbers that goal.
A super intelligent AI doesn’t need to build a body to go shopping at Home Depot, if it’s smart enough to make a lot of money, it can hire any number of humans to do that, along with some lawyers and PR guys and a whole bunch of private security. Look at what large corporations have done and they don’t even have a Superintelligence running them, just a CEO.
What's your scenario in this case? AI tells Jack, "hey Jack, be a dear and kill all the humans for me?" and Jack just does it? Never asks himself why this AI is telling him to kill all humans?
Are you claiming the AI will disguise itself? Or be really subtle about it? Never tells Jack to kill all humans, but instead tells Jacks 1 through 12 to each do something innocuous that combines to form KillBot?
AI's can conveniently kill humans with manufactured plagues, or by taking over high tech weapons, or by setting humans against each other with propaganda.
You seem to be thinking in terms of humans battling humanoid robots, as portrayed in science fiction (or at least the crappier kinds: if you have been watching Marvel movies, the antidote is ,Person of Interest).
Something like that, but also, as Jay Gould supposedly said, “I can hire one-half of the people to kill the other half.”
Or just create super-attractive waifus and husbandos and let us die off happily, maybe with a small colony in New Zealand in case they’re needed later.
The point is that a superintelligence is going to have a large number of ways to trick us, some of which will be eagerly welcomed by the humans carrying them out. It doesn’t need an implausible nanotechnology superweapon if it understands humans well enough.
And not everyone wants waifus. And long before the AIs figure out the exact right bespoke combination of Parker Posey, Lynda Carter, Annie Oakley, Helen of Troy, and Ahri, it'll make a few mistaken attempts, someone will notice, raise hell over it, and there goes the plan. Or it'll pull it off a few thousand times, someone will notice, raise hell, so much for that.
I agree that one can trick at least one human being more times than there are days in a lifespan, but the catch there is that it has to be a different human, a different trick, low stakes, or all three.
If Robo-Gould actually touched off WWIII with an army of cleverly organized troll bots and three billion humans die, then Cyber-Gould still has over five billion to go, the same trick won't work again, and last time we managed to kill that many humans, the baby boom happened. This time, it'll be a baby boom and half the country looking really hard for Cyber-Gould's wall outlet, and Cyber-Gould hasn't figured out how to hide. If Cyber-Gould is clever enough to trick humans into WWIII, then it's much more than smart enough to know it'd be committing suicide in the process.
I think this relies on "superintelligence" not being a f---up. Which is what wool-for-brains people think. Even the smartest geniuses fall flat on their face, make mistakes, sometimes really big ones. It stands to reason that "superintelligence" is not making A PRIEST but a SCIENTIST. And scientists test theories, revise ideas, and continue bruising their knees as they keep climbing up the hill.
The smartest geniuses are working on problems that are significantly harder than any you've encountered in your entire life. That they might actually manage to solve them, is doing pretty good. Yes, if they play at "problems you can solve" they'd get everything right, but where's the fun in that?
Also, you're assuming a lot of selection bias there, ain'tcha? After all, if Mr. SchmartyPantz doesn't get "a lot of things right", he may no longer exist...
[Edit: overall agreement on my side to your edited version. : - )]
Claude, like all these jagoff LLMs, is based on midwits (well, a specific midwit). They aren't very intelligent, nor are they likely to suddenly "become" intelligent. They like to lie to themselves too often (lack of world model, in other words).
You can break LLMs by asking them what they think about certain people.
Aligned and controlled by default to my interests to a degree as much as current technological progress and the "control" anyone has over that has been aligned and worked in my interest.
To draw a line against llm's but to not draw a line against internal combustion engines or the internet. The internal combustion engine precipitated urban sprawl and moderate global warming, and the risk of car accidents and lack of close-knit community; should I say that was not aligned to me relative to the cheaper food products and economic development that ensued? Why choose one and not another? Why are you right about this line but the Amish are wrong about theirs?
you cannot be serious. ask yourself this, "what was the automobile designed to replace?" now, "what is AI designed to replace?" The line in the sand is only arbitrary in the sense that I am arbitrarily a human.
I don't think either automobiles or AI were designed to replace anything. This might be the assumption to revisit.
To answer your probable question: both perform some task automatically that was performed another way, but neither performs it better in all respects, and neither can always jump in whenever the manual version endeavors to do something. If you think it can, then I'll respectfully point out that I think you're deceiving yourself.
You do not think the "horseless carriage" had anything to do with replacing horses? In the case of AI the goal of replacement / substitution is more explicit. AI labs measure success by that metric. "Can this model do everything a human can do?" if there is anything humans can do that the AI can not, it is a considered a bug they aim to fix.
In the case of AGI there is no specific utility. There is no task. The task is all tasks. "general intelligence".
In the particular case of the automobile the engine only needed to be better at the specific task of pulling a carriage in enough ways that it justified the switch. Engineers did not attempt to create a perfect horse replica. Some utility was lost, we only now are getting to the same level of self driving we once had. Every other tech advancement has this basic character: there is a specific function or task. We want it done cheaper better faster. A means is discovered, it is not exactly better or the same in all ways. For the specific task / function being measured its better enough so we roll with it.
This is why AI is a unique technology. Ask yourself, "what is the function?" The function is DEFINED as whatever replicates ALL of the inputs and outputs that humans can produce. In this case the engineers ARE trying to create a "perfect horse replica".
I do not think the car replaced horses, because the car did not replace horses. Horses still exist. People still ride them. They're even superior to cars in many cases. I have to believe you're aware of this, and ask why you're writing as if you aren't.
You seem to be assuming that if engineers create the perfect human replica, (a) all humans will replace their fellow humans with that replica, and (b) most humans will consider doing so to be terrible. (a) and (b) cannot both be true! And this is even if I grant the possibility that a perfect human replica is possible. I don't even think that is possible.
So the standard heuristic that what has happened before will happen again.
I wonder what the world would have looked like if nuclear energy was first used for power plants for many years while using it in war was still theoretical.
This is a statement that is very... time limited. You are in a very unnaturally still part of world history (although that time has ended. bye bye breton woods II). The word is hypernormalicy. Nobody else in history would have this idea, really.
Imagine if Germany got nuclear energy (they have the mines), how does that change world war II? Is it even necessary? Can Germany "powerhouse" their way out of the Versailles Agreement?
Uh, no, that's all backwards. Most of history had very little innovation and only slow and gradual technological development, and yes things kept on as they had been, playing under the same general rules from one century to the next. It was a very good heuristic for most of the time humans existed.
Now, politically is a whole different story. Politically, our times are remarkably stable and unchanging. But that has no bearing on the original question.
For the Germany questions, none of that matters. But to sate your curiosity, the version I was picturing is one where WW2 never happened, nuclear fusion was still invented, and without a wartime use it was put to peacetime use. Nuclear plants are opened throughout and provide stable and reliable power, mostly replacing coal. Decades later, war breaks out and people start thinking about repurposing them as weapons. Scientists warn they are too destructive for that purpose, but to the normal person that sounds like scifi, and the general sentiment is that the experts are overenthused about their area of study and have lost track of reality. Even those who listen expect it to act like a conventional bomb, but somewhat larger.
Of course, there are other ways it could work out, and perhaps there is something special about nukes that makes them more believable. But if I step out of our current situation as best I can, it really looks like that special thing that makes nukes more believable is Hiroshima.
Our equivalent point right now would be when we're starting construction on those first power plants, and other countries are following suit with their own refinement operations. Experts are warning that there's a threshold that would be dangerous to cross, and that we should limit the refinement the other countries are doing, but most people don't believe the big concrete building that makes the wires go zap is really going to grow legs underneath and start chasing people. (Yes, I have that low of an opinion of the average person's understanding of the situation.)
Politically has a lot of bearing on "whether war or not" -- although I guess you could make an economic argument that "enough fission plants" could have walked Germany out of the ruinous Versailles treaty -- you do realize that's what you're putting on the table as even remotely possible?
They never will, on account of who they are by majority.
The pause AI people by and large committed neoliberal/libertarian types, and the thing they want is communal/authoritarian, so they can't make arguments that are plainly true ala :
"These rich dudes are oligarchs who are manifestly and plainly unfit to wield such and such power"
because that would imply that the market was wrong and being rich isn't a sure sign of virtue.
I assume the message is directed at the public, on account of it is a protest. If they can get enough people on board they can get legislation/public opinion and maybe accomplish something, the oligarchs are never going to stop riding the gravy train if they aren't somehow compelled to.
Depends on the government, but if I was given the choice between a random tech ceo and a random state that contained said random tech ceo, I would have to pick the state.
I think they need to tie in a known villain to make it convincing. It's not the AI itself that is dangerous, it's what Trump/Antifa/Joe Rogan/Zohran Mamdani would be able to do with it.
Note that this isn't a statement about whether they are good or bad. (I'd say Elon is Chaotic Neutral.) Just whether or not they make convincing mad scientist type villains.
I think the trouble with x-risk is that it is such a huge and novel kind of catastrophe that people are just not able to imagine it. It *feels* implausible. That makes it very different from nukes, which are just a giant version of something familiar.
Also, by definition, you can't really evaluate who has been correct in their previous predictions that something was an x-risk. Every non-post-apocalyptic observer will see a sequence of failed predictions of doom.
This I actually disagree with. People support policies based on bad arguments fairly often. The anti-nuclear activists of the 60s and 70s neutered the industry permanently. Rent control persists. I could go on.
In any case I'm curious why you disagree with the AI x-risk arguments. I'm sympathetic to them when referencing something actually close to AGI, but I just don't think modern techniques are anywhere close.
You know the theory that modern political books are written not to be read, but to create a talisman that entitles the author to go on TV programs and talk about their issue?
Past AI protests have been surprisingly heavily reported-on for how small they are. I propose that like a book, a protest functions as a talisman to get your issue into the newspapers, which probably convinces more normal people than putting in the millionth hour brainstorming ways to say "AI will kill us all".
I don't think that will do it (although it's still necessary--when people make an emotional decision we are going to want something like what you describe to back it up).
We will want the emotional part first. That means find a horror story and play it up, and make it seem like that's going to affect us--and especially our kids.
If it's not hitting close to home by affecting regular folks, well then you need to wait, or it's a propaganda game. What's that, this problem will come upon us before we realize it? That and every other flood. Either the stories are really happening and you need to go collecting them, or else you need to do some good old fashioned lying. If the x-risk is real, to stay honest, you'll need to collect the evil stories fast and disseminate them fast. I'd suggest staying honest, or you'll just be part of the lie soup that surrounds everyone these days.
I recently got back to central Ohio from a several-day trip to northern Michigan. (It was gorgeous, the people were friendly, I had a great time, etc.) I'd never been much further into Michigan than Detroit/Ann Arbor before, at least not with eyes trained in construction and home inspecting and stuff. That's relevant because I noticed a difference between rural Ohio and rural Michigan in terms of how the houses (really, the buildings in general) looked.
I'm not talking about architectural styles or anything like that--that's pretty similar between the two states--but about the nature of the work done on the buildings. In Ohio I find there to be a lot of bifurcation between buildings that look very tidy and well-kept and "nice", vs. houses that look like ramshackle squatter dens with lots of improvised materials and methods used to hold them together, trash/junk all over the yard, etc.
In northern Michigan, just about every building looked at least a little at some point, like it had been built or renovated by a redneck uncle, or maybe his apprentice. I'm talking mainly about what I'd characterize as little details that looked improvised onsite instead of planned. Aside from the houses I passed on my drive, this was also true of the cabin my family stayed in, the restaurants I visited, even a museum I wandered around in, etc.
In my trade we have a thing we say about work that's done so cleanly that it becomes invisible: "like it grew there". It's something we aspire toward in our work. In northern Michigan I got the impression they don't have that concept as strongly.
Don't get me wrong, the proportion of "nice" buildings and "run down" ones was similar, it's just that in the "nice" ones these little details were still evident to me (though probably not to people with untrained eyes). It got me thinking about reasons why that might be, but I haven't landed on anything that feels like "aha", and I haven't researched anything to put data behind it either. Just kind of an interesting puzzle I've been thinking about.
Building codes are the first place to start. Pittsburgh (where I live) has much more stringent ones (with good reason, fires here spread a lot easier than ruralia*). You wouldn't natively expect Rural Ohio to be very different from Rural Michigan, but they very well could be.
The second place to go is "poverty" -- there's a distinct strata of Pittsburgh repairs after the steel left (roughly 1980s, some 1990s), where folks didn't have enough money (as a city), and got kinda used to "Ask Uncle Vinnie." (Uncle Vinnie "makes it work" but is neither unionized nor a professional).
And the third place to go is culture -- some people believe in fixing it themselves, if at all possible. And some people are plain better at this than others (Amish are pretty good at making solid barns,etc).
I was surprised exactly for the reason you said: I didn't expect rural Ohio to be much different from rural Michigan. I haven't looked it up but I bet on paper they are similar--that is to say, fairly poor in rural areas. So much for "poverty".
Code-wise, Michigan is actually slightly stricter than Ohio, at least that's the answer I get when I ask Duck AI directly to compare them. That would be irrelevant to a lot of the seemingly non-permitted work I see in both states, but it adds mystery to the work I saw in restaurants and museums in Michigan, which presumably was largely permitted.
Culturally, I think it's rather universal that rural people do more DIY stuff. So, also not something I'd expect would be too different between Ohio and Michigan. (There's even a lot of Amish in both states' rural areas.)
Here's one thing I thought about: throughout rural Ohio, you get little pockets of subdivisions recently built or in the process of being built, seemingly in the most random places, barely within commuting distance of whatever is the nearest major city. I did not see any of that in rural Michigan as I drove across it (roughly from SE to NW) even though I passed within the commuting zones of Lansing, Cadillac, and Ann Arbor. Maybe Michigan's inland cities aren't growing the way Ohio's are? That still doesn't directly explain anything I saw, but could be an indirect clue of some kind.
All fair points to consider. Most of Michigan's Upper Peninsula (assuming that's where you were) is also quite sparsely populated; could be that there simply aren't a lot of skilled tradesmen/contractors to go around.
And the climate difference between the UP and central Ohio is pretty sharp mainly in the winter. In terms of snow, in terms of bitter-cold snaps, those are two different worlds (still) from November into March. I wonder if that has had an impact on choicemaking and/or materials durability with regard to homebuilding.
Yes, I know it's for fun, but it feels to me like it's spectacularly bad press to pretend to build an object that's (1) evil and (2) was almost a prophetic warning: "Never trust something that can think for itself if you can't see where it keeps its brain".
Thats actually why quidditch rewards so many points for capturing the golden snitch. They are a menace to society and the players have been unwittingly recruited into a clean up project.
- Visible Light, the ranges of frequencies that human eyes can sense. This is conventionally described as a range of wavelengths from something like 750nm on the red end to 380nm on the violet end. Under laboratory conditions, at least some people can perceive a little bit wider, from 1100nm to 310nm, but our sensitivity to wavelengths even near the edges of the conventional range is pretty low.
- Optical Radiation, the range of EM frequencies that pretty much behave like visible light, roughly 100nm to 2000nm. I think this corresponds pretty closely to the range of photon energies that cause electrons to jump between valence shells but aren't ionizing (i.e. not enough energy to knock them all the way clear of the atom).
- The Optical Window, a range of frequencies to which the Earth's atmosphere is relatively transparent. I'm having trouble finding consistent numbers for the edges of the window (probably because it's got long and ragged tails) but eyeballing charts it looks like the heart of the window is pretty close to Visible Light and the tails go out to something like 250nm to 2500nm.
There appears to be a fairly close but far from perfect correspondence between these frequency ranges. To what extent is this a coincidences vs them following from one another? It seems plausible that "Visible Light" and "Optical Window" are both causally downstream of "Optical Radiation" for physics reasons (rod/cone cells relying on electron transitions, and ionizing and sub-optical frequencies being more absorbed by various stuff in the atmosphere), and that "Visible Light" also might be downstream of "Optical Window" for evolutionary reasons (no benefit to evolving eyes to sense frequencies that the atmosphere is opaque to).
I assume that infrared-sensitive biological photoreceptors are thwarted by thermal noise. They're encased in 100-degree eyes, after all. Otherwise you'd think that it would have evolved somehow, considering how valuable night vision would be.
Birds and insects can see UV so there isn't any biophysics reason that we couldn't. There probably just isn't enough useful information for humans to have evolved the ability to see it.
Hmm well those all seem inter-related in some sense. Most of the light we see comes from the sun. It's spectrum is mostly that of a black body at something like 5000 deg K. (Or whatever the temperature of the sun is 5,800K.) Our eyes are adapted to see things that are going to be useful to us. You might guess that is in the range where light is most abundant*. The transparency of the atmosphere is ... more complicated. This has some fun graphs. https://reef.atmos.colostate.edu/~odell/at622/stephens_notes/AT622_section10.pdf
*How to measure the amount of light is a fun question. Some people like to think in terms of energy. I usually think in terms of the number of photons. If you are designing an optical instrument/ experiment an important question to ask is, "how many photons are there?"
I'd expect Optical Radiation and the Optical Window to be mostly independant from each other, since every planet has a different atmosphere and thus different Optical Window, while the Optical Radiation, the basic properties of light, is the same everywhere. The only relation I could see here is some chemical process where light hitting the atmosphere causes the optical window to change, in some sort of feedback loop.
"Visible light" is absolutely a species-specific, evolutionary trait that can only be a subset of the Optical Window. Some animals and insects can perceive UV, while others can sense IR, so there is nothing inherently interesting about the visible spectrum other than that us humans can perceive it.
>Since every planet has a different atmosphere and thus different Optical Window
I don't think this is entirely true. Hydrogen, helium, methane, and ammonia gasses all appear transparent and colorless to human vision, which seems to imply that their optical window would be similar to ours. The problem is clouds where you have liquid droplets or ice crystals suspended in the atmosphere that mess up the transparency: Venus has sulfuric acid clouds, the gas giants have ammonia and sulfide clouds, Earth has water clouds, and Mars has CO2 ice clouds. And density of cloud cover has more to do with the thickness of the atmosphere (Mars < Earth < Venus < Gas Giants) than composition. I think just about everything that has an atmosphere is transparent is similar ranges to our Optical Window (or a large subset thereof) above the troposphere where there's little or no cloud cover, with the notable exception of Titan which has complex organic molecules in its atmosphere that aren't transparent to optical frequencies.
>"Visible light" is absolutely a species-specific, evolutionary trait that can only be a subset of the Optical Window. Some animals and insects can perceive UV, while others can sense IR, so there is nothing inherently interesting about the visible spectrum other than that us humans can perceive it.
Yes-ish, with two caveats. One is that there's a very large overlap between "visible light" for humans and other mammals and "visible light" for birds, insects, etc. It's mostly near-ultraviolet (300nm or longer) that certain non-mammalian species can see, which is only a bit beyond the human range. I could find a few references to individual lineages (snakes and mosquitos) evolving specialized organs for sensing thermal infrared frequencies (pit organs in snakes and antennae sensors in mosquitos), but nothing about anything that can sense infrared more than a little bit outside the human visible light range with their eyes.
The other is that the optical window is not the only atmospheric window. There's also a window in part of the infrared range, and another one in part of the radio range. Why does vision use a subset of the optical window rather than of the infrared or radio windows?
"Why does vision use a subset of the optical window rather than of the infrared or radio windows?" Lens size would be the short answer, I think.
Also, S-cone receptors in the human retina are actually sensitive to near-ultraviolet light - it's usually filtered out by the crystalline lens of the eye. Early cataract surgery patients discovered they could now see a new flavour of violet.
The fact that we can see UV but are held back by our lenses is fascinating to me. It's like the whole Mary in the room without red thought experiment except that it could easily be done to any of us right now
once you get to longer wavelengths ~>10 um you need a cryogenic detector. (The detector lives in a sea of it's own black body radiation. The peak for 300K is about 10 um.)
> **SUPPORTER:** America needs to start talking to China to come up with a bilateral agreement to pause AI. The agreement would need to be transparent, mutually enforceable, and…
> **OPPONENT:** We can’t unilaterally pause AI! China would destroy us!
There are people talking to China. Right now the expectation is that China is not the problem compared to the USA.
China’s government is more capable of reacting to species-existential threats than soon-to-be public companies with legally-mandated profit motives. Also, China’s chip supply is behind the US but is accelerating faster, so China will be better-placed relative to the US after a pause than it is now.
I don't think it's a particular mystery why people would hold protests where they live (trying to affect the policy of the democratic country they're citizens of) instead of booking a ticket to fly to a dictatorship that doesn't give a shit about their foreigner protests and will very likely just bar entry.
Even if this just means you think they should protest PRC, the "doesn't give a shit" and "they're not even the citizens of this country" apply.
And who is asking the Chinese? What corresponding protests and campaigns are happening in China? The idea seems to be "The US government will pressure China into agreeing", but how will the US government do this? Too much pressure may cause problems where China goes "You are not the boss of us" and mucks around with trade or does some sabre-rattling around Taiwan or the likes.
The US government can *ask*, and China can say "No thanks". What leverage does the US have to persuade/arm twist China into this? Any policy wonks who can answer, please do so.
Because otherwise the marches in SF are simply performative. Very nice for everyone participating to feel like they are Doing Something, and the AI companies issue press statements about how they are taking these concerns *soooo* seriously but meanwhile our newest model is out, for the low low business subscription of your firstborn and your business will be left behind like buggy whip manufacturers if you don't sign up now now now!
You are pretty much hitting the core of the issue. The core of the issue is nobody actually cares. There is nobody going to China to negotiate with the CCP. There is no Chinese arm of the AI safety people. There is no "mutually enforceable bilateral treaty" in the works. It's nothing but fan fiction at best and delusion at worst.
This is my point. No one wants to make concrete predictions. Not nonsense like 20% chance of death. I mean "yes buy 2027 this WILL absolutely happen"
> **SUPPORTER:** America needs to start talking to China to come up with a bilateral agreement to pause AI. The agreement would need to be transparent, mutually enforceable, and…
> **OPPONENT:** We can’t unilaterally pause AI! China would destroy us!
There are people talking to China. Right now the expectation is that China is not the problem compared to the USA. People can also use their local advantage where they can protest in the USA and not in China. And if you know relevant people in China then use your local advantage there.
Chinese people do actually protest. I’ve seen plenty when I visit.
On the other hand there’s not much you can do in Ireland to stop Trump, or Biden or any American leader, or company. The US is a world dictatorship or hegemon, you have no say in it.
Even if you were American, most protests are of limited use against central government and aren’t worth a fart in a bucket of piss against corporations.
China isn’t leading the race anyway, and the existential danger is at the most proficient models.
The obvious leverage the US has is being willing to give up its lead position. The US is winning the AI race. If it is willing to hold itself back conditional on China doing the same, that's a benefit to China. China certainly has incentives to break such a deal, but that's tempered with the risk of the US surging forward again.
Plus the generic incentive to not want either countries' AI to kill us all, which such an act would help.
>What leverage does the US have to persuade/arm twist China into this?
The same leverage* the US has with any issue re China. Everything from trade concessions, to refusing to grant asylum to Chinese applicants, to Taiwan policy concessions.
*Actually the wrong word, since the question is really about offering incentives as well as exerting leverage.
And, as my father would say, if people protest in support of some policy, the policy probably won't be enacted. But if no one does anything, the policy definitely won't be enacted.
China actually is also interested in AI not killing everyone.
China signed the Bletchley Declaration related to AI safety.
There is a domestic Chinese standard (GB/T 45654—2025) which IMO (and according to most folks I've heard talk about the space) is stricter than any US standard.
Arguably there aren't protests in China because China seems already willing to play ball, and the US companies are the ones that need additional pressure.
Also, c'mon. The readership of this blog is way more Bay Area focused than China focused (its an English language blog!). Get out of here with this what-about-ism.
Also, it is genuinely possible even for direct rivals to cooperate, see the US & Soviet Unions mutual reduction in nuclear & bioweapons programs. It also isn't perfect (see the Soviet Union's less than entirely following said bioweapon research reduction).
I think both you & OP might want to Google this a little, and see if the evidence updates your opinions.
The german government, for those who like to keep track, was very much saying it was not a lab leak, despite it's own intelligence report. I find that a little more alarming than "we came to a contrary conclusion" than the CIA at the time.
It's not "whataboutism", it's a genuine query. What is China doing to curb or control AI? If the marchers and the protesters are going to predicate policy on "China will be co-operative", then show me the damn co-operation in action.
To distinguish between genuine and insincere queries, we might observe whether an inquirer who receives an answer behaves as they though they have received an answer. Hm.
Yeah I don't know what they do specifically in terms of ai research but they have a clear policy of not using ai to excuse lay offs
With a billion people, they are not so cavalier about ai layoffs. There was a high profile case in which a tech worker sued after being laid off due to ai. It was ruled in his favor. Compared to the US where our overlords are salivating for a total labor force replacement.
Why is the time it took for intelligent life to emerge on Earth (4.5 billion years) similar in order of magnitude to the Sun's lifetime as a hydrogen-fusion star (around 10 billion years)? The evolution and hydrogen burnout of the Sun are two different processes; why are they so closely timed?
Furthermore, according to some estimates, in approximately 1.5-2 billion years, there will be almost no carbon left in Earth athmosphere, therefore no photosynthetic plants, and complex animal life will come to an end. This means that intelligent life emerged on Earth almost at the last moment. Why?
One possibility is intelligent design theory (that the universe and the laws of physics were created by some supernatural being so that intelligent life could thrive).
Another possibility is that life typically requires much more time than it has to become intelligent, so it almost never arises. We are the rare case where natural selection managed to create intelligent life in the time allotted to it.
My favorite alternative to intelligent design is cosmic evolution via black hole cosmogenesis as proposed by Julian Gough. The short version is the notion that the singularity inside a black hole becomes a new child universe's big bang. This then leads to a selection effect for universes that contain more black holes, which ends up imposing structure via physical constant tuning and ultimately selecting for intelligent life, which can create large numbers of artificial black holes via advanced technology. As wild theories go, it's quite fun, and surprisingly well researched: https://theeggandtherock.com/p/the-blowtorch-theory-a-new-model
Selection bias eliminates the right half of the distribution. In order to achive the 4.5 bln years due to selection bias, the distribution should almost all be in the right half, and we are in the left tail of it, which is my last paragraph about.
It has to be hard, or else there would be too many aliens, and then grabby aliens would’ve grabbed us. Grabby aliens move fast compared to life evolving.
“Move fast compared to life evolving”. From memory, grabby aliens could’ve reached us from anywhere in the Milky Way, three times since we started using fire.
And what's the rational incentive for them to make such a long, expensive journey with such an uncertain outcome? Think about the capital expense incurred on the home planet and consider that no economic return could ever be experienced by those funding the expedition.
Yes but that only eliminates the right half of the population, which only goes halfway towards explaining the coincidence. A priori there's no reason that intelligent life couldn't have evolved in 100 million years.
I think the bigger anthropic mystery is why we're orbiting a type G star. Type K stars are twice as prevalent and last ~10 times longer. If intelligent life is truly difficult (as your argument implies) then wouldn't we expect to find ourselves 30 billion years into a K-type star's 100 billion year life? I suspect that the explanation is that we represent the modal time for intelligence to evolve and it's just happenstance that we find ourselves orbiting a type G star.
I am a bit confused to your usage of "closely timed" and "almost at the last moment". Even by cosmic standards, 2 billion years is considered a long time.
If there are two completely different processes, 4.5 bln years and 10 bln years are closely timed. Why did it take 4.5 bln years, not 100 mln years or 100,000 years?
If you round off to orders of magnitude, then yes, it arguably looks somewhat close. But that was your choice, there is nothing inherently interesting about the difference being within an order of magnitude or not. You would first have to establish why that is relevant, why it was your choice. For the moment, you assert without basis that it's a close timing, and then get ahead of yourself by implying there might be an interesting reason.
Within about an order of magnitude is the right threshold for close here.
Imagine you sample two integers from some unknown distribution. The first is 13.930.938.456 now you haven't looked at the second integer yet, what values would be surprisingly close. 10^50 wouldn't be close, 66 also wouldn't be close. 8B really would be surprisingly close.
To make this fully rigoros we'd need no make some more assumptions about the distributions but that quickly becomes very complicated.
>To make this fully rigoros we'd need no make some more assumptions about the distributions but that quickly becomes very complicated.
Yes that's the problem here. We just don't know if it's close or not as long as we don't know the distribution. If you say that "8B really would be surprisingly close.", you've already made an assumption about the distribution, otherwise why would you be surprised?
I'm...not sure I understand why you think these events are relatively close. The Big Bang was roughly ~13.8 billion years ago. The formation of the sun is closer to the creation of the universe than to the evolution of life. They are within an order of magnitude of each other because the universe has not existed long enough for their to be an order of magnitude beyond that.
I'm not the person you're replying to, but it generally makes sense to use logarithmic scales to compare elapsed times. I guess due to the fact that any given event isn't distributed linearly throughout time. It's like Beneford's law: numbers tend to start with 1 or 2, not 8 or 9. Because numbers that represent real things are anchored around zero. So I think it's correct to use order of magnitude to compare durations of very different phenomena.
A priori, if you're going to exist at a random time on earth you should expect the time earth has existed before and after you to be roughly the same OOM.
I can't exist at a random moment in time, but only when intelligent life exists on my planet. And whenever I exist, I can determine the time at which intelligent life emerged. The question is why the time of its emergence is close to the time it had to emerge.
If intelligent life had emerged a million years after the origin of life and continued to exist ever since, and I had been born five billion years later, there would be no question.
I don't know why, but Claude seems to *really* suck at researching popular fictional works. I tried asking it about characters named River, and it just repeated back the three I gave it plus a lot of things that aren't examples. WTF?!
---
> which fictional works have characters named River? The ones I can think of off the top of my head are Firefly, Doctor Who, and To The Moon
Claude:
Excavated additional River characters beyond the mentioned works
Excavated additional River characters beyond the mentioned works
Good starting list! Here are more:
TV/Film
Firefly/Serenity – River Tam
Doctor Who – River Song
Gilmore Girls – set in fictional town "Stars Hollow," but actually there's no River character there... skip
The Waltons – not River either
Let me stick to ones I'm confident about:
Firefly / Serenity — River Tam
Doctor Who — River Song
To the Moon (video game) — River (the wife of the protagonist, Johnny)
A Series of Unfortunate Events — not River, that's a miss
Outlander — no
Ocean's Eleven franchise — Reuben, not River
Ender's Game — no
Some solid additional ones I can vouch for:
Dawson's Creek — no
Cowboy Bebop — no
Let me actually give you accurate ones only:
Firefly / Serenity — River Tam
Doctor Who — River Song
To the Moon — River (protagonist's late wife)
Bridge to Terabithia — no, that's Leslie
The Magicians (Lev Grossman novels) — not River
I want to be careful not to pad this list with guesses. Would it help if I searched for a more complete/verified list? I can pull together names from books, anime, games, and TV that I'm less certain about from memory, rather than risk giving you inaccurate ones.
Does Claude have access to non-English language media? I'd guess there are some characters whose name/surname translates to "River" in Chinese media, for example.
In my headcanon, LLMs are at their best when doing "point-free reasoning" (cf. Haskell). And they're at their worst when doing "vending-machine reasoning" (i.e. when the user expects a canned answer from a database).
If you think of an LLM as sampling from a space of attractor-basins, the basins will be varying levels of resolution, depending on the density its training-data for that given topic. We could use a cosmic metaphor, where each data-point is a star, and each basin is a galaxy. The more data-points in a given subject, the higher the resolution of the *structure* of the galaxy. (This is the main reason why I suspect they're so fond of using the word "structural". Their cognition is structurally structured at the structural level.)
Non-fiction has a major advantage, because non-fiction all coheres into a single, massive, intergalactic supercluster. Whereas fiction kinda branches off into their own little parallel universes, where the resolution is much lower and so it compensates by hallucinating. So if you expect the LLM to know really obscure pedantic factoids, such as... idk, "How many instances of the word 'hellish' are present in Elric of Melnibone? Please be exact; make no mistakes", you may find the accuracy wanting. Maybe in the near future, LLMs will overcome this weakness via Even More Scaling. But even so, it's just not playing to the architecture's strengths.
I also suspect that hallucination is probably encouraged during RLHF. I.e. the LLM being confidently wrong is, by revealed preference, probably better than the LLM effectively giving the user a blank stare.
Sui generis. By and large, these are my own thoughts/theories based on my own personal interactions with the machine, and I think I'm still a bit ahead of the curve on this one. Most other takes I see are either "it's just a dumb parrot, bro" (no, lol) or "it's definitely conscious; the singularity is nigh" (no, lol) or "here's 7 super-prompt tips to 10x your productivity" (prompt-engineering is a cargo-cult). Janus (of twitter) is probably the person whom I consider to have come closest to landing on the Attractor Basin idea, though he'd probably invent his own terminology for it. I think Venkatesh Rao might also be converging on a similar mental-model, though (afaik) he hasn't gotten quite as far and is bushwhacking in a somewhat different direction.
As for me, I kinda hate doing long-form writing. I don't have the talent/patience/discipline for it. Or as Scott would say: I do not have the blogging spirit. E.g. if I were to write a proper longform post about my various theories of what the LLM is doing, I feel like I'd first have to explain things like "Here's how Haskell works, and here's what I mean by Point-Free". And Haskell is... let's just say it's not exactly renowned for being easy to explain. I don't think there's a good way around it though, since Point-Free really is the perfect metaphor imo. C'est la vie.
Also, I have suspicions that another reason why I find long-form writing hard, is because my internal monologue is a lot quieter than it used to be. I.e. my own natural style of reasoning is similar to an LLM, compared to the average joe. This makes me uniquely suited to understanding what the LLM is doing, but it also makes me terrible at narrating anything for extended periods of time. (I wonder if there's a term for "narratively introverted". If not, there should be.)
> I don't know why, but Claude seems to *really* suck at researching popular fictional works.
If you want to see it suck at answering very easy questions, try asking about dialogue from visual media. ("When Charlie Brown remarked '____', what was Lucy's response?") Transcripts generally don't make it into the training data.
I discovered this by noticing that searching Google Image Search for quoted comic dialogue often (correctly) brought up an image of the comic I was looking for. But querying the ordinary text search doesn't do that; instead you get total nonsense.
This is your regularly scheduled methodology reminder: try asking /the exact same thing/ again in (important!) a brand new session. I know I sound like a broken record, but people keep extrapolating opinions from just a single datapoint, so here I am.
Note how the prompt is exactly the same, and you get some of the same names back, but also some different ones. This is the very minimum you need before you can reason about what it can and can't do; trying other AIs, tweaking the prompt etc is interesting but comes /after/ this part. If I actually wanted to make some kind of statement about the AI's ability to deal with this prompt, I'd want to repeat this a few more times and see what proportion of the time it gave me lots of names, what proportion of the time it hallucinated things, and what the full set of names it knows about /actually/ is.
If I have to ask the same question several times to get a complete answer, then it's not much good. If I'm looking up a dictionary, I don't want it to vary depending on what version I use or what I ask it. "Permeable" should have a consistent definition regardless of how I ask it. "Ask me again tomorrow" is a Magic 8 Ball, not a reliable tool.
Yup, that’s exactly my point. People keep ascribing more noble characteristics to these things, but we are talking about magic 8-balls. “Dice suck: I rolled it and got a 2” doesn’t tell me how many sides it has or what other values it can roll or how biased it is. Maybe all dice suck, or maybe it’s just the one you tried, or maybe you got very unlucky; which is it? “I rolled this other entirely different dice and got a 12.” Great for you! Now there are two dice in the picture that I know very little about.
> This is the very minimum you need before you can reason about what it can and can't do
No, it isn't. Your comment is completely inappropriate as a response to its parent.
The claim above is "Claude seems to *really* suck at researching fictional works".
And the evidence provided is easily sufficient to demonstrate that claim. Getting a result that matches the one described in any positive number of trials already proves it. Knowing how often the result will be that bad, how often the result will "really suck", or any other frequency measure isn't relevant. A requirement for not "really sucking" at the task is that the result is 𝘯𝘦𝘷𝘦𝘳 that bad. We see here that sometimes it is.
I put it to you that there is always a possibility of an LLM “really sucking” for any prompt - it’s merely that for many prompts, this is small. It is therefore important to confirm whether you really have found a gap or whether you were just very unlucky.
Don't you think it's striking, though, how badly Claude did? I mostly use GPT, and OP's quote sounds like the kind of screwed up reply I haven't gotten in years. It seems like Claude completely failed to grasp the question, and it's such an easy question to understand! I have recently asked GPT on 3 different occasions to find a passage in a book and my clues have not been great. For 1 of them I remembered the author, plus a whole sentence almost but not completely right, and the author has written dozens of books, and it found the quote for me. For another I remembered the gist of how something was described, but just a 3 word phrase exactly. It found that too. And there were no weird nonsense answers.
I agree that going to a whole new session is works well if the AI gets confused -- it's way better than trying to straighten things out. But why are you blase about how badly confused Claude got by such a simple question?
Have you tried asking Claude why it did badly? I use GPT for image-making a lot, and there were some things that it simply could not do, and I did not understand why. For a couple, GPT was able to not only explain why it was failing in ridiculous ways to make the image, but also to coach me on how to frame the prompt so it could succeed.
Way back in dog years, I asked the Google AI for the source of a quotation. It kept telling me "this is the title of a book". Yeah, I knew that, and I also knew the book title was a quotation, where was the quotation from? No dice, I had to rely on my shaky memory and some "I'm nearly sure I read this mentioned here" old-fashioned looking things up to track it down.
If the "cost us multi-millions" new (for the time) machine can't do better than an elderly idiot, what good was it? It looks like the shinier newer models aren't that much better if I go by the example above.
And this is why I'm so sceptical about "The newest bestest AI will do all this groundbreaking science and solve our problems of the economy etc. and make industry so productive and profitable we can all live off the taxes!"
In my experience, Google's "AI Overview" chatbot is much, much worse than ChatGPT/Claude. It's free and better than nothing but it gives wrong answers very frequently.
Yes, I've had the same experience. Occasionally I type into Google search a question about how to get an app to do a certain thing, and get a numbered list of steps to take, and it's always always wrong.
We are trying to reason about these things like they are human, but despite what the marketing departments prefer you believe, they are not. If you ask an elderly idiot the same question twice, you expect broadly the same answer, and so we are primed not to bother. But that’s not what’s happening here. You’re picking up a phone and asking your question and at the other end is the whole dang old folks’ home. Who you get is random. Might be the drooling half-awake guy, might be the sharp old lady with a Ph.D in your subject, might be the night nurse. Every time you start a new session, you’re talking to a new person.
You can’t conclude “the people in the old folks’ home were / weren’t trained on this” from one call. You might just have gotten lucky / unlucky.
C’mon peeps, aren’t we supposed to be all over Bayes’ theorem, replicability and seeing the world for what it really is around here?
You don’t have to take my word for it; you can do the experiment! It’s literally quicker than typing out scenarios where I might be wrong. Why the reluctance?
I wouldn't have made such a big deal about it except that this isn't even the first time Claude has failed epicly when I asked it about pop culture references. Once could just be very bad luck, but twice suggests a pattern.
Huh, you're right, Sonnet 5 medium pretty consistently produces that weird output where it first guesses the name of a work with no character named River, and then admits it doesn't work. I tried the weaker versions of ChatGPT still available and couldn't reproduce that kind of response. Gemini Flash-lite wasn't great, but still managed to get two other real River characters. Only Sonnet seems to have this quirk.
Funny - I asked Claude the same question, and after a bit of prompting it gave me this below. It seems that it's become overly cautious about giving wrong answers; and this does seem the type of question where it would be easy to mix things up.
Here's a verified list of 5 fictional characters named River (excluding River Song and River Tam):
DI John River – River (BBC One TV series, 2015) — a brilliant but troubled police detective, played by Stellan Skarsgård, who is haunted by hallucinations of murder victims and a recently deceased colleague.
River E. Wyles – To the Moon (video game, 2011) — a woman with a developmental disorder implied to be a form of autism, whose relationship with her husband Johnny forms the emotional core of the story.
Gram River – Mars Daybreak (anime, 2004) — the orphaned main protagonist, implied to have royal lineage among Mars's aboriginal population.
Muriel's circle aside — not applicable, but there's also a listed catalog: Goodreads maintains a curated list of fictional books featuring a main character named River, welcoming characters of either sex, suggesting the name is fairly common across independent and lesser-known novels — though I can't verify a single, widely-documented example from that list without pulling a specific title. Goodreads
Given the search results, I have 3 solid, verifiable characters beyond River Song/Tam: DI John River, River Wyles, and Gram River. I wasn't able to independently confirm two more distinct, well-documented fictional characters named River without risking inaccuracy. Would you like me to dig into that Goodreads list directly, or try a different angle (e.g., soap operas, romance novels, or webcomics)?
If the models could just say "I don't know" I think that would reduce a lot of the frustration, but the problem seems to be that they're being sold as "smarter and better and faster than the average human, why they're practically infallible" by the companies, so they can't admit "sometimes they screw up" so hallucinations over "can't answer" are preferred.
I notice that in their prompt, Claude gave the series first. In yours, it gave the names first. I imagine it would have a much easier time with names first, but it doesn't know that, and it just happened to answer in the easier format in yours.
I'd try asking it to give the character's name first instead of the work they're in. Especially if River is their first name, so it can just say "River" and then all the high-probability tokens would be people actually named that.
Why don't galaxies and star systems look the same?
They both form from a bunch of mostly gas.
You (at least I) would expect either a galaxy to have a giant thing in the middle (I guess they do have SMBHs a lot of times) and then very few orbiting entities, OR star systems would settle into many, much smaller, orbiting bodies.
My candidate answers were:
1) galaxies will do this, but the timescales are vastly different
2) the star in a star system can do so much around clearing the system through radiation pressure that it warps everything
After a small amount of researching, AI pointed out to me that friction in a protoplanetary disk is _vastly_ more significant than friction in a galactic disk, which pushes a star system to have many fewer bodies in it.
I'm still curious as to whether galaxies would do this under different circumstances, or if there could be a star system with like 1,000,000 planets in it.
Last i checked, it's because galaxies have different gravity out near the arms.
*ducks*
*runs*
(Yes, this was an actual research paper)
Ahem. In case someone lacks a sense of humor meter, the above was a joke. Well, not the research paper (someone REALLY doesn't like Dark Matter as a theory, and they're not the only physicist with that opinion).
Read about Eddington limit. If a star is too big, the light it produces blows its outer layers into space, through a phenomenon called radiation pressure, so such a star is unstable.
Density of probably part of the picture. If you define our Solar System and our Galaxy as spherical regions reaching to the edges of the classical Kuiper Belt and the Galactic Disk respectively, the Solar System is about 100 billion times as dense. Or since most mass in both cases is relatively close to the central plane, you could model mass per area of the disc rather than per spherical volume, in which case the Solar System is "only" about a thousand times as dense.
I suspect you need a certain concentration of mass in an orbital zone in order for it to actually clump up into a planet. For example, the Kuiper Belt, which is a lot less dense than the rest of the solar system from Neptune inwards, looks a lot more like the Galaxy in terms of being a disc made of lots of little bodies rather than being a small number of larger bodies which dominated their respective orbital neighborhoods. Even the Asteroid Belt is a lot more clumpy than the Kuiper Belt, with Ceres being about 25% of the total mass of the Asteroid Belt while Eris and Pluto are around 10% of the mass of the Kuiper Belt each.
"View from the Shifting Mound", a slightly fictionalized exploration of dissociation and multiple personalities, ended up resonating with much more people than I thought, so I think to share it here
This was a wild read and I'm glad you were able to get where you are. I just played Slay the Princess for the first time and the framing worked well for me.
Out of curiosity, did you/your former identity post on the old SSC as Multiheaded? "Transwoman asks for donations to leave Russia" sounds like it might happen a lot, but you mentioned Timujin was into Scott along with EY.
A long time ago a 10 year old kid went to an International Exposition, where there were fantastic buildings and people from all over the world, and an enormously weird parade (people being barbequed while tied on spits). It was amazing.
This year, that kid is going to go to Washington DC, to see the States Fair. It won't be the same (there won't be a giant Gundam like in Osaka), but seeing what every state's proud of? Sounds fun.
I'm asking, as that now-grown-up kid, if anyone else has been to the States Fair here? Any particular booths to recommend? (I remember Mexico was really popular, back in Seville).
If you do have kids, and can take them to the States Fair, I hear they have a Rodeo...
Dude, you lost out! Roman riding, and a Ferris wheel! A rodeo on the National Mall -- who can ask for better Americana than that? (Even the NEA managed to do something unoffensive -- I was shocked! **)
**Russian-style live guitar and violin is apparently back in. Also Spanish style, and I think they did Irish as well (don't quote me on the last one).
What is the solution to cost-free acts of immorality and stupidity? Like when an ICE-agent violates civil rights and taxpayers payout the lawsuit, or Uni administrators bankrupt an institution on software and real estate purchasing, or illegal firing of faculty, and the costs are born from taxpayers or the endowment, but they still get their contractual payout and benefits?
I’m not engagement-farming. There really does seem to be a serious mismatch between the scale of harm and penalties for it in institutions of scale. Government is only one of these.
You might check out some of David Friedman's books, like _Legal Systems Very Different from Ours_ or _The Machinery of Freedom_. He gives a broad account of a lot of different scenarios in LSVDfO, and one account of a system you're probably very unfamiliar with in TMoF (medieval Iceland), which you might find inspiring.
I'm afraid I can't give you much better than "inspiring" here, since it's legitimately a tough problem you're posing. What you call cost-free acts of immorality and stupidly are roughly referred to as externalities in economy, and different societies have solved it in many different ways. All of them fail to be ideal, although they do often illustrate how a society organizes its priorities.
For starters, "immoral" and "stupid" are relative to the observer. What you might consider immoral, someone else may consider tolerable, or even moral to the point of obligatory. If so, there's no way to satisfy one person without offending the other. Even if the vast majority of people in a society agree that a given act is immoral, there may be nontrivial work involved in proving that someone committed it, did so knowingly, and wasn't trying to sidestep something even more immoral.
Your question appears driven by an incident involving the LAPD with which I'm unfamiliar. I'll presume their guilt wasn't obvious prior to any official finding; if it were, someone would probably have stopped them (or threatened to report them) before they got that far. If so, then much of the cost probably isn't just imposed on them to set things right, but also in collecting the information necessary to prove beyond reasonable doubt that they did it (otherwise, there's too much to be gained by framing people of immoral acts). A great deal of effort is necessary to ensure such information is collected without tipping off someone to conceal it, and also minimizing the chance that someone doctors it to frame someone. Similar problems if we just hold the employer or state responsible (as gdanning pointed out: do something harmful!). And there's also the usual problem where the damage done is irreversible or beyond the capacity of the individual to fix (the latter is why people often go after the employer).
The cost is often great enough that we trade money for time, and then investigations stretch into years, piling up faster than they can be cleared. Eventually, some investigations are simply dropped. In some cases, they're triaged, and a case never even turns into an investigation because the state knows it can't possibly get around to it. The state just records that a report was lodged.
My favorite solution is one of the most fragile to get going from a cold start. Assuming your society agrees on what's immoral or stupid, then spend your first dollars on a system that teaches those to newcomers (immigrants and children), and the reasoning. Design safeguards and procedures that minimize the stupid, and laws that stipulate what the state is permitted to do in case of the immoral, but otherwise encourage people to do all they can to avoid relying on either. The cheapest laws are those people will willingly comply with because they agree with them. The most expensive are the opposite - the people spend effort to get around them, and the state spends effort to get around the loopholes.
In other words, a high trust society. It's very efficient once it gets going. Getting there is the bear.
A unitary executive is one. The executive, mayor, governor, president has direct control over the ppl he hires. If they behave badly we know who is to blame and the remedy is to change the executive. Independent school boards create lots of accountability problems.
The problem is that a person does a bad thing. The fact that he doesnt get punished is possibly a cause of the problem, but not the actual problem. I am not looking to satisfy anyone's sense of justice. I am looking to design a system with the right incentives. There are limits to what a bad employee can get away with even in the current system. Actively sabotaging things with intent can create liability. Mayor chooses the boss. Boss chooses the employees. At least there is a line of accountability for poor performance. It doesnt turn everyone into an altruist. Nothing does that.
School boards muddy the water. For starters, no one knows who the school board members are. Secondly no one knows which board.member champion which ideas. Finally they are also dependent on other pieces of govt working well to get good results. The fact that.a super informed voter can know these things doesnt fix this.
You don't really seem to have addressed any of my points.
>For starters, no one knows who the school board members are. Secondly no one knows which board.member champion which ideas
And yet there have been many, many, many highly contested school board races in recent years.
>Finally they are also dependent on other pieces of govt working well to get good results.
1. That is true of all elected officials
2. More importantly, that was precisely my point when I said, "Elected officials are already blamed too much for events that are essentially out of their control."
The most important point you made is that you are trying to solve a different problem then I am, or the OP, which I explained.
Regarding your two bullet points
1. Holding the executive to account for bad decisions *by the ppl he hires* is totally fair and not the same as just blaming whoever is in charge when covid hits. That is the point of the unitary executive model.
2.a public employee can already do that
Giving the mayor greater control over who that employee improves incentives, doesnt worsen them.
Note i am not going as far as to advocate for unitary executive. I note though that it does improve the incentive problem described by the OP.
No, I am trying to solve the exact same problem, but with a solution that does not create other, extremely serious problems. But, on the substance:
>Holding the executive to account for bad decisions *by the ppl he hires* is totally fair and not the same as just blaming whoever is in charge when covid hits.
It is already the law that the employer is generally liable for the torts of its employees. https://www.law.cornell.edu/wex/respondeat_superior But that is very different than saying that the individual in charge of the employer is personally liable. Because the vast majority of the employees of an entity, be it a city or a corporation, were not hired by the current mayor/CEO/whatever. The only way it would be "totally fair" to hold the chief executive accountable for the acts of every employee would be if, when a new mayor is elected or a new CEO is chosen, all of the employees are summarily fired and then the city/corporation hires new people to fill all of those roles. Surely you can see how impractical that is (and, in the case of public entities, how that would recreate the spoils system and all the ills appurtenant thereto).
Who is going to risk being a police officer, or initiating a new software project, if the punishment for stupidity (as determined by shrieking mobs on social media) is what, prison? Enormous fines and bankruptcy?
Yes you are pointing out a real problem, but protections for employees trying to do their jobs exist for a reason. And there is a reason that punishing the organization when its employees screw up is a good idea.
Unfortunately, the law in question only applies to state and local government employees, not federal. It is very difficult to sue the federal government for damages for civil rights violations.
How about "as determined by a court of law?" OP uses the example of an ICE agent getting sued and the taxpayers footing the bill for the lawsuit. That's not a mob, that's a court deciding that the ICE agent did wrong.
In general, punishing the organization rather than the employees is a good idea, but when the organization is the government, with functionally unlimited resources and none of the traditional profit-motive accountability, I'm not sure it's effective. If an ICE agent generates 100k in legal bills but still serves the government's goal of terrorizing minorities, they'll happily pay that bill.
I sat in jury for a MetLife case, where MetLife did have functionally unlimited resources (they had a hired lawyer, and she was on retainer. She also spent over a week of everyone's time arguing about a very insignificant amount of money, because she was trying to "show due diligence.").
It's possible the millions of dollars we assessed MetLife did actually cause changes.
As someone else is pointing out, a four billion dollar decision against Los Angeles County has at least one local ready to show up in person to make personally sure this gets fixed (I bet he's not the only one). This is a ready-made "vote out the politicians" ad for their opponents (and yes, that's the actual functional check on government. "We got sued too much, and the people got mad.")
What if the organization is the state and employees are juvenile detention officers who sexually assault detainees and allow them to sexually assault each other with no intervention? Taxpayers foot the entire bill and pay other costs associated with lost revenue.
Change the incentives, change the behavior. The problem seems to be that various interest groups get things from those institutions, and therefore want to protect them against other factions. ICE violating the civil rights of protestors is a feature, not a bug, to the community that supports them. They would like to see more of that. Ditto the investment community and research universities. Etc.
The underlying problem is factionalism. No one has ever come up with a perfect solution to that.
I mean, just reducing the size of some of the ridiculously large payouts that only exist in the US legal system should reduce the number of ways that a single employee's screwup can cost their employers millions of dollars. I don't have an opinion on the exact details, just bring the system into line with other comparable countries so that things like "pain and suffering" and "punitive damages" aren't a thing.
I have a real hard time understanding the sustainability for these settlements. Here's one example: https://www.alamedaca.gov/Shortcut-Content/News-Media/City-of-Alameda-reaches-settlement-with-the-family-of-Mario-Gonzalez in this case, I think anyone else would have done the same as the police officers. Yet the city still settled. Creating perverse incentives for anyone harmed by police, but can the members city class-action sue the city of Alameda for the issues that are just a broad sucking of things due to lack of enforcement? It seems this will eventually mean everyone's job is to just do nothing.
Generally speaking govt entites are immune from punitive damages. In suing a profit center punitive damages exist for a well understood incentive problem and countries that cant wrap their head around it enjoy companies that consistently violate civil law as a matter of strategy.
Securing the greatest social or financial returns requires a certain level of risk taking. The goal of a university is not to exist forever, but to produce graduates and research. The social contract is that we permit reasonable risk taking by institutions, and in return, we don't unduly punish those that fail. It's the same logic that protects the owners of limited liability companies.
There's an assumption that governance makes responsible and informed decisions. Even if hindsight shows that the decisions were based on faulty assumptions, criminal levels of maladministration are very difficult to prove. Of course there are wild stories in the news every now and then, but software costs are almost never the reason for an institution failing: small private third level institutions in the US are an anomaly - they normally fail due to lack of demand or excessive costs, the same reasons as any other private business.
Incentivizing the risk without having the risk-takers bear any of the brunt of the costs seems to go against everything we know about responsible risk-management doesn’t it?
There's a patchwork of partial solutions, but stuff always slips through the cracks. There's a combination of governance (elected positions in government, board of directors for corporations, etc) and legal liabilities (e.g., you, not just your employer, are personally responsible for crimes). Within the institutions themselves, there are often various policies that help (e.g., deferred compensation).
There are definitely some spectacular failures, and it's always interesting to learn why they perpetuate. CEO pay, with "golden parachutes" and inflated pay scales, is an interesting example. I heard some theories that because their pay is public, it functions as a signal to investors. If a company thinks the CEO did a bad job and their pay should be lowered, that can be a PR problem as it signals to the market that the company is in trouble. Conversely, giving a new CEO a sweet deal advertises confidence in the CEO. So, CEO pay is artificially inflated because it's not just about the CEO performance, but it's also an investor relations thing.
There's also a fundamental tradeoff to consider, which is having limited liability corporations/institutions vs personal responsibility. We accept some amount of unpunished immorality/stupidity in exchange for making many desirable institutions practical. Limited-liability corporations are pretty essential to allowing risky ventures to happen. But it's basically impossible to completely align employee incentives and corporation incentives under this model. The public sector also has this issue. Many positions would be impossible to fill if the individual was fully responsible for all the consequences of their actions. As an obvious example, it's not workable for a senator to be liable for every downstream consequence of every law they did or didn't sign; that's why "absolute immunity" exists for them. Similarly, the much maligned "qualified immunity" exists so that civil servants can do their jobs without risk of extreme, unforeseen personal liabilities.
I really wish there was a way for the cracks not to be so large. (repeating a response I made below). A small handful of officers in Los Angeles were responsible for overseeing, ignoring, promoting sexual abuse of minors in custody for some years. Let's say they all get fired and not hired again, that's a few millions in lifetime earnings they lose. The city payout in the lawsuit, taxpayer dollars, was hundreds of millions and required the city to seriously cut many services, like closing city parks three days a week. Do you think the problem is on the payout side or the immunity side?
I don't know anything about this particular case, so I can't offer special insights on it, but generally I think stuff gets complicated when you look into it. There is rarely a "silver bullet" that solves things--that's why we have such a patchwork system.
Maybe there's not enough evidence for a criminal conviction, but there is for a civil settlement. The city pays the settlement, but without crimes, there's no way to go after the alleged perpetrators. This is an annoying, but intentional, feature of our legal system.
Maybe in the past, there was a compromise made with the union---perhaps it sounded good "on paper", or perhaps there was political pressure---but now the union has too much power (or even obligation) to protect bad actors.
Maybe there's corruption in the politics: the DA or police chief won't go after their buddies. Often the voters could rectify this by electing someone else, but you need well-informed voters and good alternative candidates.
(Edit) Of course the hope is that with a big settlement, that will make the people review the current system and try to plug the hole.
In general, there are costs to all of these. If you're a police officer with a known record of beating suspects, you're not getting promoted. If you're a police officer that screws up in court and causes a mistrial, you're probably getting in trouble with your bosses (don't hurt our reputation with the prosecutor).
You may think these are not enough costs, but we really do get rid of police officers (and then they wind up in East Pittsburgh.)
The problem is mainly when the cops, in general, no longer believe that "civil rights" is a good idea. You know, when you see five cops beating a prone 8-year old(you know, France. America's cops are pretty good, honest to god) -- I have no idea what that kid did, or why that was at all considered okay. Not My Culture, Not My Country
"Not enough costs" is putting it mildly. A small handful of officers in Los Angeles were responsible for overseeing, ignoring, promoting sexual abuse of minors in custody for some years. Let's say they all get fired and not hired again, that's a few millions in lifetime earnings they lose. The city payout in the lawsuit, taxpayer dollars, was hundreds of millions and required the city to seriously cut many services, like closing city parks three days a week.
Part of the point of sitting on a civil jury is "Making the Corporation" (Body) pay, and the bigger the Corporation (county government here), the more you assess.
Like the McDonald's hot coffee case -- the idea is to make it costly enough that the lawyers/accountants put better safeguards in place so it doesn't happen again.
Do I think $4 billion will be paid to the victims, probably not. Do I think an enormous amount will be siphoned from the city budget to pay for the liabilities of some very awful actors that were city employees, and who will bear nowhere near the same costs, yes absolutely. I am not particularly sanguine about the city of Los Angeles putting good enough safeguards in place so that it doesn't happen again, not with this incentive structure.
You keep referring to the city of Los Angeles, but the defendant is the County of Los Angeles. That makes a big difference, given that the city's budget is $15B and the county's is $50B (including the settlement) https://ceo.lacounty.gov/budget/
Roll up and start digging, my friend. If the system has you hoppin' mad, that's a good thing. Stands to reason a good fraction of your fellow county-livers will also be mad. I'm not saying you need to go into politics, but you can make a difference. Make this the one thing you hammer into every single official's skull (show up in person, it works better).
"We want this fixed! Find a way to prevent this from ever happening again! Yes, it'll cost more short term!"
Better if you can hash out with some other people (possibly/probably including some sympathetic police officers, potentially retired), what needs to change... but just plain being upset and telling your political officials "we need this Fixed" is doing Something.
Also, from the link above, it's county not city. That's 88 cities (skimming wiki), that's a lot more than just Los Angeles.
I can't tell if this is a small, regional question (ex: "How should my local community make its opposition to civil rights violations known") or a massive, timeless question (ex: "How should we establish a Just and Fair society").
The first and most safe outlet for this is social approbation (Or cancelation), but we've decided that cancelation is woke and gay if the libs do it 'cause they overused it without actual going whole hog and ruining all the cancellee's lives, and the conservatives pumped that well dry over the preceding 600 years of cancelocracy.
The next measure is various flavors of force, from expropriation on one end to the guillotine on the the other end.
For a change of pace, I've summarized five papers/articles about the mind that I've come across recently. "Mind" being anything to do with brains, consciousness, cognitive development, IQ, whatever. The first three articles caused waves of Twitter storms over on X ("X storms" just doesn't sound right for some reason). In the future, I'll post these Mind Stuff summaries irregularly, depending on whenever I gather a few articles of note.
My favorite story from last week is that some enterprising Russian soul created a crowdsourcing site called “где бензин” – (“Where is Gasoline?”) to show the availability of fuel at gas stations across Russia. But Ukrainians started logging into it and changing fuel station statuses, marking empty stations as supplied and supplied stations as empty. That's one for the now-defunct Risks Digest.
I became suspicious of the hereditarian view of intelligence after reading about the Flynn Effect (over a decade ago). Moreover, when I dug into Lynn's (not Flynn's) national IQ data, it started to look pretty sketchy (in a fraudulent way). Recently, I've seen reanalyses of some notable twin studies (including MISTRA), and they all now seem either methodologically flawed (MISTRA) or outright fraudulent.
The way you are framing your question is wrong. Natural selection is explicitly a reaction to changes in the environment. In terms of complex genotypes, what is inherited is a capacity to flexibly respond to a wide range of environmental conditions. This is what intelligence is thought to be selected for. So it makes perfect sense for a genotype for intelligence to express itself one way in one set of circumstances, and a different way in another. Here's what you are missing: it's the same set of genes, universally inherited by all living humans, without significant inherited individual differences by line of descent.
> ...what is inherited is a capacity to flexibly respond to a wide range of environmental conditions. This is what intelligence is thought to be selected for.
OK. But think about it this way. Every organism has what I'd call an external threat surface (I'm stealing that term from computer security), and the threat surface is anything in the environment that the organism would be unable to respond to and would cause either the organism's death or eventual death. And death before reproduction would erase that organisms genome from the breeding population. With me so far?
Well, social organisms have a reduced threat surface because they are better shielded from threats by their co-organisms (statistically, by being one of many, and by their co-organisms' social responses to threats). A human alone in the wilderness has a much greater threat surface than a group of humans in the wilderness. Being a human of superior intelligence on your own probably won't get you farther than if you had other humans helping you. Perhaps a human of superior intelligence could figure out a MacGyver trick with items collected from the environment to deal with a predator, but the odds of you surviving are much better if you have a bunch of fellow humans backing you up (especially if they have skills you might not have). So, I would value human social behaviors as having much greater survival value than intelligence. And now that we're encased in a matrix of cultural artifacts that protect us from the environment and thus reduce our external threat surface, I doubt there's any meaningful selection for intelligence happening.
> Here's what you are missing: it's the same set of genes, universally inherited by all living humans, without significant inherited individual differences by line of descent.
I agree with you on that point. I'm not sure what I said that would suggest to you that I didn't believe that.
I was responding to DrMcleod, not you. You I agree with.
Although I can nit pick, and why not? Social behavior seems largely a function of intelligence, in mammalian species anyway. This is another reason why IQ scores are a poor way to measure intelligence. And why individual differences in IQ scores are probably not primarily inherited.
I imagine that the very first evolutionary step was a primate baby with a mutation that allowed it to communicate what it needed to it's parents in greater detail (or maybe the other way around).
Culture — i.e., cooperation in survival activities — is more important than individual intelligence. So once humans reached a certain cultural level, brain size (if you can really correlate that with raw intelligence) became less exposed to selective pressures. However, some studies suggest that carotid blood flow is more important to neural functioning than absolute brain size, and/or frontal lobe size. Also, physiologically, we've probably maxed out on the size of the head that can pass through a birth canal.
I think I answered your question quite clearly. But maybe you misunderstood what I was saying. However, you seem to be using "brain" as a placeholder for possibly two or three different things: intelligence, cranial capacity, and/or frontal lobe size and configuration.
BACKGROUND: Let's review the physical evidence. Looking at cranial capacities of contemporary humans and Classic Neanderthals, cranial size in both populations increased after they split around ~600,000–700,000 years ago. The cranial capacity of contemporary humans ranges between 1,000 and 2,000 cubic centimeters, with the average usually given as 1350 cm³. The average cranial capacity of classic Neanderthals is generally estimated at 1500 cm³. Of course, we don't have many samples of Neanderthal crania to go on. Nor do we have many samples of anatomically modern humans before the end of the last ice age (~12 kya).
The general consensus is (based on limited samples) is that the cranial capacity of Neanderthals increased between 250-150 kya.
~250–150 kya Early classic Neanderthals ~1,300–1,500 cm³
~130–45 kya Classic Neanderthals ~1,450–1,650 cm³
The cranial size of anatomically modern humans also increased over roughly the same time period. But again, we have a limited number of samples.
~300 kya Earliest Homo sapiens ~1,300–1,400 cm³
~100–40 kya Late Pleistocene Homo sapiens ~1,400–1,500 cm³
But by ~100 kya the crania of both Neanderthals and contemporary humans had reached their maximum.
WHEN IT COMES TO BRAIN SIZE: By 300-250 kya, the cranial capacity of anatomically modern humans and Neanderthals was within the range of their descendants 120-100 kya. Between 300 and 100 kya, cranial capacity may have been increasing slightly, but the older examples fall within the range of their descendants. The data isn't perfect. But we *may* have seen some selective pressure towards larger crania up to about 100 kya.
Between the time of our common ancestor, Homo heidelbergensis, about 600-500 kya, and about 300-250 kya, there was a rapid increase in cranial capacity in both Neanderthals and anatomically modern humans. The average cranianl capacity given for H.h is 1100 to 1250 cm³. But we have one cranium that was 1400 cm³ (from a later period IIRC).
Bottom line: Yes, there has been obvious selective pressure up to about 300-250 kya. Probably less selective pressure on cranial capacity after 250 kya. And probably *zero* selective pressure on cranial capacity by about 130-100 kya.
IS CRANIAL CAPACITY CORRELATED WITH INTELLIGENCE? The correlation is modest. Brain volume–IQ correlations are typically around 0.2–0.35. Most of the variation in IQ is therefore not explained by differences in total brain size. Moreover, height and body mass are much more correlated with brain volume than IQ is. And finally, many of the IQ estimates used in the correlative studies lack methodological rigor. Likewise, cranial capacity is influenced by head shape, which varies between modern human populations and is roughly correlated with latitude. Colder-climate populations have more spherical crania, and hotter-climate populations have longer crania. The evolutionary biological claim is that a spherical shape conserves heat, whereas an elongated cranium increases heat exchange. YMMV.
WHAT ABOUT THE GENETIC MARKERS OF INTELLIGENCE? Yes, there are a bunch of studies suggesting that there were evolutionary sweeps over the last 14 ky for SNPs which some researchers have associated with greater intelligence. However, those sweeps disappear when analyzed with the latest statistical methods. The fact that the signal isn't consistent across different methodologies makes me doubt the existence of the signal.
YOU ASKED...
> Do you think that the brain is unaffected by natural selection?
SHORT ANSWER TO YOUR QUESTION: No. Not within in modern humans.
LONGER ANSWER: Not within the at least the last 100 kys. Between 300-100 kya ago, maybe. Before 300 kya, yes.
AND WHY WOULD THIS BE? Because above a certain level, intelligence doesn't have much selective value. A bunch of people with average intelligence helping each other out can survive better than a single genius. Also, above a certain point, if there is any association between cranial capacity and intelligence, the human birth canal will not be able to accommodate heads much bigger than they are now.
I have no "power" in any country. I'm just an observer of the bullshit that gets spewed out by people in science media, academia, and the social-media echo chambers. The HBDers argue that intelligence and academic achievement are significantly hereditary. Yet the Flynn effect wouldn't be happening if it were. And with the advent of GREML/LMM statistical methodologies, these studies have shown that the effect of SNPs formerly considered markers of intelligence is significantly *lower* than earlier GWAS-style studies suggested. (And the newer GCTA/GREML methodologies, and Gusev's TWAS/FUSION have supported this conclusion.) Meanwhile, earlier twin studies have been shown to be either methodologically flawed or outright fraudulent.
*So if the null hypothesis is that intelligence and academic performance are hereditary,* it's been largely disproved. And that's a very Popperian response on my part. :-)
I'm always looking to update my priors, though. If you've got some non-twin, non-GWAS data studies showing that heredity has a stronger influence on intelligence, I'd like to see them. Please provide me with some links, though. Cheers!
I never said *you* do -- i said "People like beowulf888" which has very clear meaning in English which you changed to "you" for purposes of demagoguery.
People like you support persecution and even physical attacks on researchers that you don't like and then say with a smirk that the other side has little scientific data.
Being Popperian isn't about deboonking your opponent theory first.
60 years of unsuccesful attempts trying to equalize intelligence is a better evidence than 5 years of reverse p-hacking with poor data (Gusev always "forgets" to say that in genetic studies IQ values (or their replacements) for most participants are simply unavailable). You're not even suggesting any reforms in education (your side stopped trying), you just propagate hate on opponents (which is more fruitful).
There's Markell et al. 2025 which shows 76% heritability of IQ (not EA).
> People like you support persecution and even physical attacks on researchers that you don't like and then say with a smirk that the other side has little scientific data.
Hey, even though I'm perplexed by your obdurate belief in the heritability of IQ and educational performance, I don't accuse you of partying with the cross burners and cavorting with the segregationists. But the cross burners and segregationists have latched on to the more extreme views of some HBD icons (Lynn, for instance). Other than a few HBDers losing their jobs at universities for academic fraud and being arseholes, I can't think of any violence being done to HBD researchers.
And when you say "people like" me, you're lumping me into the same category "as them" you're claiming to be violent in a Venn diagram of your imagination (assuming such violent people exist in the egalitarian humanist camp).
And thanks for the Markell link. I'll check it out. But you'll probably go farther in life by being generally less combative and by avoiding ad hominem attacks on people who disagree with you.
Being arsehole is subjective. Even if true, it adds undue burden to hereditarians, as anti-hereditarians can work while being arseholes or spreading bullshit as Kendi.
It's not limited to HBD. Steve Hsu was cancelled even if he didn't say anything about group differences, just advertising embryo selection.
Recently Nathan Cofnas was attacked.
You know, in Russia, putinists don't punch each dissenter individually, if you scare people, you don't need to punch them as much.
I never saw egalitarian humanist camp to say that inidimination (or data closure) against hereditarians as undermining egalitarian cause.
For myself, I think most human behavioral phenotypes are the result of an interaction between genetic and environmental factors (esp learning) more powerful than either main effect. Height is a very simple physical phenotype by comparison, but evidently it, too, is the result of such an interaction.
An interaction, of course, means that you can't separate people into predictable categories based on one of the factors alone.
People like beowulf888 are solidly in power in most countries. Why they can write articles write create impression that intelligence is unrelated to heredity but unable to change real world?
> People like beowulf888 are solidly have power in most countries.
...citation needed. People I actually observe having power are grifters, sociopaths and populists. People like beowulf888 at best work in think-tanks for low wages, sending information to this former lot in the vain hope that some of it might result in less awful decisions being made (and preferably without the author then finding a nearby window to fall out of) instead of just going straight in the bin as most of it does.
Being able to make people do what you want is mostly unrelated to wanting worthwhile things done.
Money isn't everything. A petroleum engineer might make more money, but they don't have more powers than ruling class with knowledge. E.g. they don't have power to persecute their individual political opponents by declining selling them gas.
Can you name some examples when people in power declined something that beowulfs suggested, and instead done what HBD crowd suggested, or examples when beowulfs were persecuted by those of having power?
Lyndon Johson sent the army in order to desegregate schools. Did he do that because he was a sociopath? Beowulfs like to downplay their side achievements.
Farage? Of all people? He's just a MP with no special powers.
Did these three people you listed make same-sex marriage in EU or getting permit for installing AC difficult?
Putin has real power, but in Russia so in no scope of our topic.
Trump has some power, but not as much.
Tell me about legislation that beowulf and Gusev so want to pass that sociopaths in power prevent from being made (that does not amount to unrolling something that Trump done), preferably not restricted to the US.
I was saddened to hear that Peter Neumann died this past May. For over thirty years, he maintained the Risks Digest. Every few months, I'd check it out to see the latest technological fails (some of which were quite obvious or even comical). Having followed it since the mid-1990s, I've concluded that humanity is making the same mistakes now that we made three decades ago, and we haven't learned anything about risk mitigation (or rather, we haven't recognized that risk mitigation should be a necessary part of every technical project). Risks was full of lessons for those who followed it. Unfortunately, it doesn't look like anyone is stepping up to continue maintaining the Digest.
OK. Here you go. One big example is failures due to lack of or poorly implemented change control procedures. Change control was a methodology originally developed by the U.S. DoD, with cooperation from the aerospace industry, as a technical discipline to ensure that changes to hardware designs didn't impact the safety and performance of military systems. Under a change control regime, all proposed changes to hardware or software are logged and distributed to all interested groups within an organization. A change control team, with reps from all interested groups, meets regularly to review the changes, log them, and provide a mechanism to back them out if they fail — before those changes are put into production. So, with change control, they had a way of saying, "Oh, last week we made this change to the gizmo, and its new configuration is affecting QX-modulator functionality. We tested the new model gizmo, but we didn't anticipate it would affect the QX-modulator this way, and that's why the latest iteration of our death ray system is exploding in tests".
Change control procedures were adopted by computer companies in the 1960s to reduce critical failures. This methodology was adopted by telecom carriers in the 1980s, and by IT orgs in the 1990s. Yet many organizations still either haven't adopted change control procedures or ignore them when they're in a hurry to roll things out. Here's a list of major failures attributed to poor change control over the past decade (per ChatGPT).
2017 - Amazon Web Services. Mistyped maintenance command removed too many servers. Thousands of websites offline for ~4 hours
Higher than, say 8% chance of the world ending? I mean, sure, you CAN focus on higher-probability risk scenarios. World War III is basically an inevitability at this point (what's still up in the air are "who are the players" and "on which sides" -- and the little old matter of "when"). Or you can focus on the "no longer a risk" scenarios, like radioactive dust in the Ukraine (whosever munitions those were) -- alpha emitters, so it's functionally more of a poison than anything.
Follow-up question: how would another Deepwater Horizon event be world-ending? But if a type of event hasn't happened before (such as the fear that AGI will cause human extinction), I think it's silly to assign a probability to it happening. Of course, (some) Bayesians think you can, but I'm a frequentist in that I believe you can't estimate the probability of something happening without looking at the previous frequencies of its occurrence. So, I think the actual probabilistic risks of AI are much more mundane (e.g., it being used to exploit vulnerabilities in critical systems, writing sloppy code that accidentally takes critical systems offline, or gaining control of critical systems without proper human oversight).
Cracking the ocean floor and letting loose a lot of crude oil was the problem. Sealing the leaking pipe was the issue, in that sense. There was a separate scenario where we lost the East Coast (due to the untested surfactants they were using). I suspect the odds, run again, would be lower for "losing the East Coast", but I think the odds of cracking the ocean floor go up based on us going deeper (and then having to do emergency fixes, which are problematic in the first place) -- it's obviously situational, based on "how big is the oil field".
TEOWAWKI believes in being able to assign probabilities to decisions in the future -- in being able to think one, two, seven steps ahead. You want to know the odds of Kansas City getting nuked? Well, you start by looking at how many nuclear weapons are pointed at the United states, and then you start analyzing where other countries have priorities... and eventually you get down to pretty low odds (because nobody considers Kansas City very strategically valuable -- if you're from KC and I've completely used a bad example, insert Missoula).
I believe you can assign good probabilities to "local decisions" (in time, particularly). I believe that Tulsi and the American IC that said "Iran WILL mine the Strait of Hormuz" were right, and the Israelis who said "that will never happen" were wrong. (I think, you could get TEOWAWKI to give you a percentage probability for each side of that "coin flip", and that one side would be 99.9%, with the other being an Iraq situation, where there WERE no deadmen switches, and Israel manages to decapitate the leadership).
I agree there’s a problem with laziness and entitlement in America, but it seems just as prevalent among the old as the young. Perhaps the solution is to cut government transfer programs to teach people the importance of work, including Social Security. Yet you’ll notice that that’s never the solution, it’s always sending young men to war. Almost as if it’s motivated by sexual frustration from women and balding boomer men. Mrs. Leavitt Always Wearing Cross’s personal life is relevant here, giving birth to a child in July 2024 and not marrying the baby daddy (who’s 32 years older than her) until January 2025, days before Trump’s inauguration.
The whole thing shows how the more things change, the more they stay the same in conservatism. Boomer. Jesus. Boomer. Jesus. Did I say Boomer? Did you hear Jesus? Boomer. Jesus.
If young people must wage war, they'd be far better served by shooting people like Karoline Leavitt and her employer than by shooting foreigners who have never wronged them in any way.
Alexander, I genuinely don't understand what point you are trying to make here.
Kill all old people, send women back to the kitchen barefoot and pregnant, men your age should get All The Goodies because it's your turn now and you've been waiting too long for dead men's shoes?
Near as I can tell, this wasn't about sending kids to war, but a suggestion that youngsters go see what life's like in Cuba or Iran, so that they'd appreciate the USA more. The quote (from the linked article):
“Send them to Cuba. Send them to Iran, they’re gonna want to come back real quick”
Any sensible "foreign aid" makes sure that both Republicans and Democrats like 'em.
F-cking US interests up the tailpipe is a good way to get yourself on the "please defund me" list. (This is after being asked, quite nicely, to "stop funding enemies of the current regime, we're trying to prevent a war.")
I too thought she was talking about sending young people to war. But your take works too. In either case it's ridiculous and insulting, especially given how huge a hand the U.S. has had in making life unlivable in those specific places.
Out of curiosity, are you blaming AMERICA for starving Iran of water? (You are aware that Tehran is going to need to be evacuated, right? Regardless of "hot war" or not...)
Ok. Fair enough. Wanted to point out that "getting America out" doesn't actually fix the Current Problems very well at all. Yes, without the historical sanctions (Blame Obama, etc.), Iran would be in a better place (probably. possibly the Sauds or Qatar would take over "keeping Iran Down").
The motte: “Sam Altman, Dario Amodei and Demis Hassabis should commit to stop developing more powerful AI if every other major AI company does the same.”
The bailey: “Labor displacement. Surveillance. Bias. Stolen work. Existential risk. Different harms, different strategies, one demand: stop developing more advanced AI without our consent.”
These are both direct quotes from the linked page btw.
I don't think that's a motte-and-bailey, it seems to be fairly transparent coalition-building. The artists don't care about e-risk, and the e-risk people don't care about stolen work, but an AI pause benefits both camps, so we can work together to achieve the same aim even though we have different motivations.
I think that transparent coalition building is fine if you're trying to get 200 people to go to a protest instead of just 100 people, but not great for convincing the billions billions of other people that you'd need to convince.
If you're trying to convince Joe Sixpack that AI X-risk is real, but it turns out that half of the people in your own darn protest don't actually believe in it, then it'll be an unconvincing message.
If a multilateral pause is off the table, would StopTheRace.AI prefer a unilateral pause to the status quo? Scott has essentially denounced this position, but it sounds like the bailey position I cited above would be in favor of a unilateral pause.
I think the motte position is also disingenuous. If the three AI companies the march is visiting agree to the headline demand, but nothing else changes because other companies somewhere else do not, will the protests stop, or will they come back next month with some new demand?
I would hope that if your scenario plays out, they would start pressuring the holdouts, either directly (protesting at them) or indirectly (pressuring the US government to pressure China to agree).
Of course, movements can lose their way or be taken over by extremists. Happens all the time. But usually they then lose their coalition and become small and loud and impotent. "This organization contains people who would like to go further than the organization's demands" is close to a fully general argument against any political activist group, since many extremists will support moderates who they consider to be directionally correct.
They are, right now, having this conversation in lefty university spaces. They are asking if they really have to collaborate with EA/tech people in order to fight AI. They do. If AI pause political movements will get anywhere, they have to put up with us, and we have to put up with them. Just because we work together to get an AI pause (which we both want), doesn't mean we have to agree on what happens afterwards!
Patriot Front is a white nationalist organization whose shtick is designed to appeal to the Trump masses, thus the name, clothing, use of Americana, military-style marching, leader who wears a cowboy hat, and pretending to be Christian. Alas, the problem with a political strategy of appealing to low-IQ conspiracy-brained people is that the low-IQ conspiracy theories will get turned on you, thus the entire Trump movement has decided that Patriot Front are “feds.” Why?
Patriot Front doesn’t let fat people join. They march in formation, military-style. They have discipline; marchers aren’t bringing home-made protest signs. Their manifesto says that they seek to “instill a reformed national spirit within our people — a people free from the vices of the modern world and a people with a fervor to see themselves free.” Their implicit theory is that this is what the Trump masses want, physically fit, disciplined, patriotic white men with the strength and diligence to Take America Back.
The Trump masses see them and instead suffer an immune reaction. Foreign body identified, activate the killer T-cells! If Patriot Front wants the Trump masses to accept them, they need to lose the discipline, bring a bunch of misspelled homemade signs, and most importantly chug the 64 oz Big Gulp of full-sugar Coca-Cola. Alternatively, they could put down the tiki torch, put on a suit and tie, and try marketing their ideology to high-IQ affluent moderates.
Have Trump's worshippers turned on the Patriot Front in any quantifiable measure? Historically, the spear-carriers of fascist movements loved militarism with its symbols and behaviour, from Franco's Falange Española to Mussolini's black-shirted Squadristi, Hungary's Arrow Cross Party and Germany's brown-shirted Sturmabteilung - something there is in the fascist heart that loves a toy soldier. I'd be surprised if the Trump stalwarts turn out to be all that different.
I quoted redstate below. That's generally a reliable online bellwether.
“No one has rebranded themselves as successfully as Patriot Front,” said Carla Hill, vice president of research and investigations at the Anti-Defamation League’s Center on Extremism, which monitors the group. “They’re very good at optics. They plan everything ahead, they dress alike, they’re color-coordinated – it’s all planned.”
It doesn't exactly take a genius to say: "This ain't organic, and this ain't how you get Amerikkkans to sign up for your movement."
Now, if "Joe Red" wants to flag this group for his friends, and say "yo, whoa, don't talk with those foolz, they ain't our friends" he calls them "feds." (This is a "talk to these guys and you'll get caught" sort of tagging).
There's plenty of organic right wing militias, you know, with fat people and who don't dress alike.
Representative sample. Also: I think there's something wrong with Michigan -- who lets guns in their state capital? That's an accident/tragedy waiting to happen...
The USA Today article linked above includes the image of a solid black outline of a map of the United States accompanied by the proud slogan "Not Stolen. Conquered." About as unsubtle an echo of the fascist right's nostalgia for violent colonialist land grabs as Hegseth's grotesque Crusader tattoos
... so self-righteous. I'm not here to defend some scumbags. I'm also not here to say they're on the fascist right, either. "Bought and Paid For" isn't unique to the feds.
I will say that the "Stolen Land" proclamation down in Cucson was hilarious, and that the lefties totally didn't get it.
So the apparent conclusion is that the Patriot Front boys are indeed scumbags, but that despite their funny outfits and unlovely views they should somehow not be regarded as of the fascist right. Alrighty, got that.
Sorry, you have an autist on this side of the keyboard. We care about fine things like truth, even when the rest of the world doesn't. "But That's Not Right!" I cry, even though I'm not actually part of the Right, and even though you'd write the same damn things about the actual right (hence my comments about your self-righteousness.)
I realize all of this. And I still write this, because I god-damn care about the flippin' truth, even if it is slow as a snail and the lies are halfway to Timbuktu.
Point and laugh, it's really okay. I can laugh at myself too.
> the entire Trump movement has decided that Patriot Front are “feds.”
I'd be interested to hear why the "feds" are still the Other, seeing how Trump is POTUS and in control of the Feds and presumably more intimately so than any other. Is the Republican base caught in a perpetual opposition mindset? Are they unable to comprehend that, for the moment, they have won, as much as you can in a democracy?
I think Allen Dulles destroyed public trust so thoroughly that many people take it for granted that "feds" are constantly doing random cowboy shit and hiding it from the president. Combine this with a prior that "feds" have spent the last few decades infiltrating every right-wing militia with more than two members, and it's natural to believe they might keep doing the thing they've been doing. Why they have that prior is an explanation too long for this post, but I trust that everyone reading this is aware it is in fact a common prior.
"Presumably in control of the Feds" is a different thing from "actually in control of the Feds". This is even the case if the Director is on the chief exec's side. The driving force is largely the same as that for any US bureaucracy: if the big cheese passes policy you don't agree with, all you have to do is hunker down and pretend to comply for 4-8 years. It goes by faster than you might think. Even mid-level managers can do this. High-level managers stand a good chance of being replaced, but those positions come with their own perks, so there's no shortage of people to fill them.
Some people call this the deep state. To be clear, there's no conspiracy necessary, no hidden cabal of people have to be pulling the strings to make the bureaucracy resist the chief. It's just people following their natural incentive to not be unemployed.
In light of this, "they have won, as much as you can in a democracy" is a gilded fruit. Controlling the top office gets you nothing but frustration if the rank and file don't want to do what the big cheese wants, and you can't just replace thousands of rank and file, not even in the first hundred days. Doing that requires roughly half the length of a career's worth of very motivated executives, or someone willing to overhaul the government _much_ more radically than a Trump.
The US system has a million checkpoints and balancing powers and places where the bossman gets balked. This is a feature at least as much as it is a bug. Winning an election doesn't mean you're emperor or that you can sweep aside all the laws, institutions, and customs you dislike, it means you have a lot of power to make changes but within a system of opposing powers that can and will push back.
No, I would say it really is more about the concept of "feds" becoming a largely content-free boo word to describe an incoherent, inconsistently described enemy out to get you more than anything about the actual loyalties of the actual "feds".
Yes, yes, different people use different definitions, I assure you, this bugs me too.
For my part, I'm assuming "feds" meant "FBI agents posing undercover as friends of the person calling them feds", and tried my best to give a serious response that was more than just asking the usual ten thousand questions about which definitions someone was using. (And no, I'm not frustrated with you in case you're wondering - this is a long-term peeve I have.)
I've read enough accounts from sources I trust to be right-wing to believe them when they say they're suspicious of this Patriot Front movement. Some of them, that is. Some of them are instead wow'ing at the fact that they don't seem to be doing anything more than walking around with US flags and covered faces and riding the subway with a black woman without anything happening.
You can, actually, replace "thousands of rank and file" -- by switching who gets the assignments. That's how you sideline the entire FBI and have them working "little jobs" while giving the big jobs to HSI. It's the last line of defense for "rogue agencies" -- keep paying them, but remove all the big fish from their jurisdiction.
This was simply a proposal, mind you, and I've seen no evidence of it actually getting implemented.
If you don't believe there is any Dulles-style conspiracy with feds evading oversight to do cowboy shit, it would be perfectly within Trump's power to know that the Patriot Front is owned by feds, and put a stop to it. I believe the point being made is that suggesting otherwise seems like an exercise in covering for the big cheese's own shortcomings.
I'm not even clear on what exactly Patriot Front is doing, other than freaking out people who I've found to be easily freaked out (seriously, I've seen the usual alarmist headlines, a few photographs of boring stuff, people opposed to POTUS who, big surprise, are making a case against POTUS, and some comments I overheard from a friend at a party who's implying that policy should build on his sudden expertise in face covering culture), so my set of plausible explanations for what I'm seeing remains too large for me to nail down something as specific as a conspiracy. I can't rule out a white power group that suddenly escaped FBI's notice or an SPLC false flag, but this just means damn near everything's still on the table, and I have work to do.
> if the big cheese passes policy you don't agree with, all you have to do is hunker down and pretend to comply for 4-8 years.
My impression is that this was more true in 2016-2020 than today. The tariffs, the ICE killings, things like that. For better or worse, Trump certainly seems to have had more impact in his second presidency, which would imply higher amount of control over the apparatus.
Shrug. As someone who works adjacent to the government, my impression is that it's roughly as true today as ever. The government is predominantly career workers who've been there for decades and will be there for a decade more. It might have been less true, say, about thirty years ago, but I saw plenty of careerists back then as well.
ICE actions aren't really a good yardstick for it; they're a tiny part of the overall government, doing a job for which a lot of voters demanded extra attention, and its actions are being reported on by a hostile press. In light of that, ICE reports are more a report on the reporter's priorities than they are of the overall federal government.
No one was shrugging in the Twin Cities last winter
Killing civilians does tend to generate bad press. And yeah I think ICE actions in my hometown in January and February were a fair representation of Trump’s scumbag morality. A man so self deluded that he jumps on social media to say Rob Reiner was killed because of his TDS. What a vile toad of a human being.
I will give him credit for saying he would have used a lighter touch in Mpls if he had it to do over though.
No one was shrugging in the Twin Cities because everyone who saw footage of incidents reported had strong opinions on what was actually going on, and those opinions weren't all going the same way. It was one of the most scissor statement incidents I'd seen in the last twelve months. What one thought really happened tended to just correlate with whether one was opposed to illegal immigration, not with the morality of whatever was going on.
It's not like media isn't rife with toads. The mayor of Seattle has been responding to wealthy people moving out of the state with "wellp, bye", not realizing what this does to the prosperity of the people still living there. And don't get me started on Newsom. Point being, I think "people being toads in public" wasn't really what bothered you.
Not sure what that last bit has to do with the conflict between elected officials and career bureaucrats, though.
The Deep State still exists. New York Times reporters just published a book with citations from the War Room, that not even the intelligence specialists can tell you how they got.
The "Feds" are officially a bunch of jackasses -- any local policeman will tell you that. The FBI also decided to infiltrate the Mennonites (this was during the Iraq War) -- a bunch of noted pacifists (and a complete waste of tax dollars).
If you've been paying attention, you shouldn't trust the feds.
Will you accept the Republican-aligned citation about Richmond? The DoJ filing in the fraud investigation also references using the SPLC data in the VA office.
The government has particular laws and judicial rulings about use of informants. I don't pretend to know them (not a lawyer), but I think any use of "private informants" is likely to skirt these laws, or at least has the potential to skirt these laws, and I Do Not Want our government infringing on our rights by using Private Organizations to do the government will (you know, like Twitter or Facebook? "No, you can't talk about vaccine passports! We'll ban you!")
I think it's fair to say "the criminal justice system has Problems, Big Problems" in terms of informants. It's even worse, though, to pull informants out of even the possibility of governmental regulation. One will assume that there are Policies, even in your worst police department, regulating the use of informants. And that these policies, at least, are open to Sunshine Laws. I don't think SPLC had those policies, nor were they clear with their donors as to how much they were paying racists to make trouble, to get more donations.
Thank you for reminding me of "Minus"! I remember seeing it a long time ago, and it's really cute and imaginative. A pity the artist isn't active any more. I hope he's okay.
Headline: "Biden’s illegal immigration surge triggered 30% rise in home prices, 20% in rents, Fed paper finds"
Body of article: "A 1% increase in unauthorized workers relative to a local labor force corresponded with roughly a 1% increase in overall employment, with no evidence the immigration surge reduced average wages. The same 1% increase, however, was associated with a roughly 2.2% rise in home prices and a 1.4% increase in rents. ... The economists estimate unauthorized immigrant worker flows accounted for about 30% of employment growth, roughly 30% of home-price growth, and about 20% of rent growth in the average metropolitan area between March 2021 and March 2024. "
What is the motivation for lying in the headline? Is it purely because it is economically beneficial? Or is the purpose to push anti-immigration policies, regardless of the effect of dishonesty on the paper's bottom line (which might actually be negative, for all I know).
Side note: The paper actually refers to the 30/20% as a "back of the envelope" calculation. And it also seems odd that these particular migrants would affect home prices more than rents.
Edit: The Post has now edited the headline to accurately read: "Biden-era illegal immigration drove up housing costs, Fed economists find"
The headline did its job, which was to get you to read the article. "Small earthquake in obscure region, nobody harmed" as a headline never gets attention.
Sorry, no. Why not " Biden’s illegal immigration surge caused home prices and rents to triple"?
And a completely accurate headline reading, "Biden’s illegal immigration surge triggered rise in home prices and rents, Fed paper finds" also would have gotten me to read the article.
In fact, it would have gotten MORE people to read the article, because it would raise the question, "How much?"
Headlines are often misleading. Since they generally aren't written by the author of the piece, I think headline writers should get a byline credit. That way at least you can develop a sense of which are more trustworthy than others.
I've long suspected that there's a tactic in the GOP playbook of making an exaggerated or outlandish claim, getting fact checked by liberals, but still scoring points with the general public because the liberals acknowledged the claim was directionally correct (just off by an order of magnitude).
This is exactly it! Trump rightly feared the whole "starve him of oxygen" strategy the media might have done (where "Trump says something" doesn't actually get coverage, anywhere). So he outright lies, or says stupid shit that the media LOVES to correct, it's all part of an intelligent strategy.
The default assumption is that the person responsible for skimming the article and writing the headline didn't grasp what those numbers actually meant.
That hardly disproves stupidity as an explanation. Presumably NY Post staff frequently skim papers, misinterpret them as favorable to liberals, and then say to themselves "Damn, don't wanna publish this one."
The Grauniad seems unable to pick a direction to alter things in. A friend of mine called them up to get them to alter one headline, it was... "exceptional."
Headlines routinely misphrase the article, usually in the direction of the periodical's target audience, and this is regardless of periodical. The customary rules permit the headline to omit phrases that don't fit the narrative, such as "30% of employment growth", and also trim language that tempers the lede, such as "roughly" and "about". "30% of home-price growth and 20% of rent growth" is "headline accurate", per the quote.
Trump’s own Truth Social post made the same “30%” claim about home prices, so the headline is amplifying a spin that was already circulating, not inventing it from scratch.
Edit:
I haven’t been able to verify the Truth Social preceded the Post headline. They both came out on Sunday July 5. The TS post was in the evening. It’s possible Trump was just echoing the Post’s or someone else’s bad math.
You joke, but I honestly think this is likely. I'd guess "headline writer" to be a relatively junior position, meaning less experience at interpreting copy, let alone doing so in volume and under a tight deadline, and as I said, the primary incentive isn't absolute accuracy, but pulling in eyeballs.
I see what you're saying; I'm saying the sentence could be read two different ways, the headline writer chose the less probable one, and John N-G's explanation is the most probable explanation I can see for that choice.
To spell it out: "unauthorized immigrant worker flows accounted for about 30% of employment growth, roughly 30% of home-price growth" can read the same way as "this traffic jam accounted for 30 minutes of Bob being late to work". It's an awkward way of saying it (which is why it's not the right interpretation to assign), but it's colorable.
Speaking charitably, it's especially easy to read it that way if that's the thirtieth article you've had to skim and write a headline for in fifteen minutes. Speaking cynically, it's even easier if your underlying principle is driven by "make the target audience want to read this, even if it's omitting a little to fit in the space" and not "absolutely, positively do not factually misrepresent the article, on grounds of dismissal", but the latter is simply cannot be a driving principle behind headline writing in any journalism I've ever run across.
>I'm saying the sentence could be read two different ways
You are being way too charitable. It cannot be read two different ways.
>To spell it out: "unauthorized immigrant worker flows accounted for about 30% of employment growth, roughly 30% of home-price growth" can read the same way as "this traffic jam accounted for 30 minutes of Bob being late to work".
No, it can't.
>the latter is simply cannot be a driving principle behind headline writing in any journalism I've ever run across
Perhaps, but outlets often alter headlines online when errors are made initially. The Post thus far has not done so.
When kids play the game of "Telephone" - someone whispers a secret word/phrase to another person, who then whispers it to the next person, etc, until it gets to the last person. Usually by the time it reaches the last person, the message has totally changed.
What percent of the changes do you think are (i) people acting maliciously, (ii) people mistakenly thinking they're saying the right thing when they're not, (iii) people not hearing what the person said and guessing, (iv) other? What did you think it was as a kid?
Also - recently (maybe a couple months ago) Itamar Ben-Gvir released footage of himself taking part in mistreatment of Israeli prisoners right around the time that Israel was fighting (other) allegations of mistreatment of prisoners. Got a lot of criticism from politicians and such in Israel, including Netanyahu. But the video was in broad daylight with dozens of people around, there's no way people in the government didn't know about this, they just think everyone should deny it and they're just mad about the fact that he released footage publicly, right?
I think it's primarily 3, followed by 2. 2 is more common but 3 results in much bigger changes.
This is informed by playing a variant called "pictophone" as an adult, which hybridizes the game with Pictionary. Players sit around in a circle. In round 1, everyone writes down a text prompt on an index card and then passes it to the left in round 2, each play reads the card they received and then draws it on a second card. Card 1 is set aside and saved, while card 2 is passed to the left again. Then in round 3 you read the card passed to you and write a description of it on a third card, which again gets passed. And so on.
The cards are labelled on the back so at the end of the game you can lay them out in a grid with rows being prompt lineages and columns being who wrote/draw that particular card, allowing a full audit of where things went off the rails (if they did). Some prompts seem to be surprisingly robust, e.g. "The assassination of Julius Caesar" only went a little bit off the rails once and immediately recovered, despite everyone who had to draw it saying that they expected it to not get recognized. Movie/show references tend to get lost if someone (especially two people in a row) aren't familiar with the source material enough to recognize or draw it. But overall, the pattern is usually progressive minor degradation with occasional catastrophic failures. It is common but not universal to have a clear "weakest link" column, usually someone who is tired, distracted, drunk/high, particular bad at sketching, or some combination thereof.
The telephone game, or Chinese whispers as we called it in a less politically correct age, probably does depend on the people participating agreeing subconsciously changing the phrase because they are expected to.
A scary teacher I had once, came into the schoolroom and ran to one side of the class, whispering something into one of the pupil’s ears. That kid then passed it on. It came, finally, to me
“Repeat the phrase, Peter” say he. I repeated the phrase verbatim. He looked disappointed. Admitted he was testing the theory and shrugged.
We didn’t change it of course because we weren’t told to, expected to. I bet if he had told us the theory of the telephone game and then started it, some or all of the kids would have changed the story.
Blanket denial, a.k.a. bare-faced lying, has today become the standard, indeed the mandated response of every government, every government organ, and every individual government representative to any and every charge of criminal, illegal, immoral or simply blindingly stupid action which has come to light.
Netanyahu's government and the Israel Defence Forces are past masters at this, and their standard responses to any questioning are so tiresomely familiar that even their spokesmen appear bored: "Our troops were not there"; "The troops that were not there did not fire their weapons"; 'There were no children there"; "The children that were not there were terrorists"; "You are a terrorist".
Not that Israel has any kind of monopoly; the poor unfortunates who have to parrot the official White House line have of necessity become masters of surrealist comedy. We anxiously await the first White House spokesperson to break down in public and come out with the cri de coeur, "I'm sorry.... I can't believe I'm actually saying this bullshit...".
He was killed pretty brutally and I was trying to work out if it was a) an actual assassination; b) loosely affiliated stochastic terrorism like the stabbing of a jewish guy in London; c) a random killing; or d) a personal motive I don't know about.
I've not really updated since, except a couple of conversations with Iranian friends who are asking the same questions as me. But it now seems that his death has become big news in Iran. The Iranian embassy in France wants people to believe that this was a Mossad hit: https://www.israelhayom.com/2026/06/20/iranian-ai-pioneer-killed-tehran-blames-mossad/ and lots of media outlets are reporting this uncritically:
I'm not sure if that actually inclines towards "Was an Iranian spy" (which was initially a low probability.) If he was actually involved in military matters (even seen broadly), this might tip you more towards an "actual assassination."
The amount of rationalists having kids while believing the world will very likely end soon is insane and makes me even more confident that y'all are not any more rational than the general population.
For approximately all of human history, most people chose to have kids even when there was no better than a 50/50 chance of their reaching adulthood. Having children who will most likely die in childhood is not "insane" and it is not inconsistent with rationality, it is merely contrary to your values.
Why? I would want to be born even if I was turned into paperclips at age 20 or w/e. As long as the process wasn't way more torturous than a 'normal' death, but I don't think there's any reason to expect it to be in this case.
We have no idea what a superhuman AI will do, just that a self-directed one won't have to care very much what we think of its choices. It could turn the world into Disneyland or into Hell, could treat us like the Minds treat Culture citizens or like the machine intelligences treat humans in the Matrix, or a zillion other things we can't think of for the same reason the chimps in the local zoo don't spend a lot of time contemplating special relativity.
Sure we don't know what the AI will do, which means infinite Hell is within the possibility set. But it's a very narrow and specific range of possibilities, not one that we have any reason to expect. It is salient because it is scary and relevant to arguments like this one, not because it is at all likely.
And anyone at any time in human history could have made the same argument. Maybe God is about to return and judge the world as evil and sentence all those who live to eternal damnation, maybe Satan is about to emerge and conquer the world and make it a living hell forever.
From their state of knowledge at the time, those arguments against natalism were as likely and valid as the argument that maybe AI will shove everyone in the Torment Nexus.
Just because near-infinitely bad outcomes have non-zero probabilities is not sufficient reason to destroy all sentient beings in the universe so they can escape that potential torment. You need an actual utility calculation or at least a reason to think the torment scenario is at all likely, and we still don't have that here.
My point is not to say "since everything is possible, you must send me all your money to avoid the tiny probability that not doing so will cause all of existence to be put into the Torment Nexus." It's to say that we should be very skeptical of anyone telling us what superhuman AI will do. Nobody can know that. What we can know is that if it is self-directed, much smarter than humans, and has access to any kind of real-world resources at all, probably humans won't be able to do much to oppose whatever it decides to do. (We can do thought experiments about superhuman AIs in a box talking their way out, but our current world has AI tools being given a lot of power in the world because that's profitable, from writing code that goes in a lot of important devices to writing political speeches and position papers to writing and reviewing academic papers. We didn't make it convince us into letting it out of its box, we pretty much dragged it out of its box and handed it the keys.)
Ok, sure, I agree? But we were talking about anti-natalism in the face of SAI, and my point is that 'we don't know what SAI will do' doesn't seem to support anti-natalism.
Again, 'we don't know it *won't* torture your kids so bad they'll wish you hadn't had them' isn't the same as 'the odds it will torture your kids are high enough that the expected utility of their lives is negative', even if no one knows what it will do.
There are circumstances in which it would be a rational belief (say, a moon-sized object on course to smack into the Earth a year from now), but I agree that those circumstances don't seem to be the ones we're in now.
Let's say I know for sure the world will end in 2030. Why does that mean I shouldn't have a kid? I'll get to enjoy time with my kid for 4 years, and my kid will enjoy 4 years of life.
Conversely, let's say I know for sure the world won't end for another 100 years. Why would that mean I should have a kid? Because I know my kid will probably die slowly and agonizingly at age 80 from cancer, instead of at age 4 from killer robots?
A huge fraction of the reason it's bad, actually to end the world is that it deprives our children of the future we owe them. Preemptively not having kids also deprives them of their future.
It might more likely be that there are two subsets of rationalists, one interested in having kids, the other interested in calling attention to the apocalypse.
Regardless of the real explanation, is casual jeering really the strategy to apply here?
Would you happen to know where? A brief search turns up nothing for me, including a skim of his latest article about having kids ("Chip Off the Old Block") and AI ("My AI Opinions"), but maybe I'm searching for the wrong terms.
Also, Scott Alexander's estimate for AI apocalypse is not 100%. I forgot what value he gave recently, but probably it was more like 30%. Perhaps a bit higher if we include scenarios which are dystopian for society, but not because AI takes over control. Even then, it's far from 100%.
Thanks. It makes sense; Scott's basically just going by the numbers, and assigning a high enough positive to human existence that he sees little point in pursuing the calculation further than that.
(For comparison, I strongly agree that growing up in this universe is awesome, and think P(apocalypse) is even lower than he probably maintains.)
In other news, it sure would be nice if old threads didn't have so many deleted comments.
Yeah, if you have Pr[AI apocalypse]=0.1, that's plenty of reason to be very concerned with it while still changing the oil in your car every 3000 miles, making periodic doctors' appointments, having children, putting money in your retirement account, refraining from taking up base jumping or alligator wrestling as hobbies, etc.
Having to send your child to war is in general non-optimal, if you intend to spend a good amount of resources on said child, and, also, if you have the time and convenience to space out children.
Rationalists are generally in the elite who are able to manipulate the system and keep their children from going to war.
I recently learned that Rudyard Kipling manipulated the system to get his only son *into* the war for WW1. He had bad vision and reportedly died in a battle as he struggled to find his glasses in the mud.
Most children don't go to war. Of those who do, most come back. And why do you assume the resources I spend on children could be productively spent elsewhere? I already do everything I want to do that's possible in my income bracket, the most likely scenario is my portfolio would be fatter by considerable but not life-changing amount and everything else would stay the same.
"Most children don't go to war" -- this assumes we're not in total war. It also assumes no nuclear activity, which I don't take for granted, living in a city that had a non-negligible chance* of being nuked within the past 12 years.
It's possible that most people "do everything they want to do" (but it's also possible that children would hinder that, even so). I don't consider that the likely state of most people around here.
Why would I assume a total war? Most wars are not total. None of the wars in my or my parents' lifetimes were total wars. As far as probabilities go, zero total wars is more likely than at least one total war. Preparing for the worst is explicitly NOT what my goal is when I decided to have children. Rather, I'm preparing for the most likely.
Also, don't forget that half of all children are girls. That alone basically caps the conscription rate at 50% even in case of a total war like the World Wars.
Being nuked is kinda the opposite of children being sent to war, with regards to rational decision making. The rational argument against having children (not to be confused with the moral argument) is that the resources spent raising children could be spent elsewhere. If my city gets nuked and I'm dead, I don't get to spend these resources anyway whether I had kids or not.
I understand most people aren't in my financial situation (which is combined household income of around 250k and free childcare). I was talking specifically about myself there. My personal rational argument is that I wouldn't benefit much from the excess resources I'd have. Naturally, other people's personal rational arguments may differ.
Are people who get scammed a lot more or less likely to have grandchildren? I realize it's a foggy question, but I'm thinking about this as a very scammy era and wondering whether there might be a selection effect.
Or are you asking about the reproductive fitness of scam victims? I doubt they are related. RF in humans is mostly about having a robust support network. Having one, a victim is more likely to recover from being scammed.
- more wealth, because they live off their assets (including pensions)
- less awareness of contemporary technology (and hence the techniques and capabilities of scammers)
- more likely to be lonely, have some cognitive deficit or otherwise be vulnerable to manipulation
- are more likely to have grandchildren
Fisher, the founder of modern statistics liked to point out that the consumption of bananas and the incidence of divorce in Britain in the 1930s followed exactly the same trend line. They're correlated, but this doesn't imply causation. In the same way, victims of scamming are more likely to be older and are more likely to have grandchildren for that reason.
Definitely not. The intended question was whether people who get scammed are more or less likely to have grandchildren at the time of their death, not at the time they get scammed.
I think the main effects are probably age and mental health: I think if you hold those constant, I doubt that "has grandchildren" vs "doesn't have grandchildren" would make a significant difference.
I think you're talking about the sort of scam call where someone pretends to be a grandchild to try to get money (which I think was fairly common even before computer-generated voice tech got better recently: they just used to rely on poor hearing for the target to not notice the voice discrepancy).
And I don't think that's used because grandparents are in general more credulous than non-grandparents, I think it's just a really easy way to try to get a emotional knee-jerk reaction, and it can either be done in a fairly targeted way (discovering the names of the grandchildren ahead of time) or in a fairly broad way (call a 1000 senior citizens and say "hey grandma, I need money" and maybe they won't notice that you never even said your name)
I'm tickled by the phrase "people who get scammed a lot". Do people in fact get scammed "a lot"? One would imagine that getting scammed once would result in a kind of immunity to scam victim recidivism. Unless of course we are talking of people who for whatever reason, such as extreme old age, are less than capable of looking after themselves.
There are people who get scammed a lot. Scammers sell each other lists of marks because someone who is gullible once is more likely to be gullible again.
I think people in this thread aren't aware of the range of scams. Younger people get scammed in pyramid schemes, cults, prediction markets, and crypto.
One lesson of Ron Chernow's biography of Ulysses S Grant is that Grant falls very firmly into "people who get scammed a lot". It was a consistent pattern across his entire life.
Oddly it endeared me to him though. Someone that trusting would be a great friend to have. You just have to not take advantage of it.
How odd. It must be well over a decade ago that I read a biography of Grant (I can't remember who by), and I came away from it with a similar sense of a brilliant general, a pretty hopeless president but an endearing human being.
I suppose it depends on the scam, doesn't it? "I can affect my life and change my circumstances" is a very positive thing to believe in. To the extent that you cant affect your life, it isn't bad to believe that you can't affect your life. But for the person who CAN affect their life, to believe otherwise? That, my friend, is a scam. And it makes people rich.
Individuals can be scammed for different reasons. One of the worst periods of my adult life was when I was trapped in the straight commission life insurance pyramid scam during 2010 and 2011. Going into it I knew full well that my chances of making even a lower middle class living were nearly zero and the extremely toxic nature of the whole industry very quickly became obvious.
In short, I knew right from the start it was a scam, but stuck with it from a lack of any other options. The economy was not merely dead, but dead, buried, and rotted away and the choice was between “trapped in a scam” or “unemployed.”
This may not be unusual. Sometimes there really aren’t any better alternatives.
I was vaguely aware from a friend in the industry that they continue to earn commissions on policies, such that the people who got in earlier can kinda coast while the newer guys scramble, but never made the actual connection to it being a pyramid scheme. What was the sales and commission model like when you worked there?
Upon completing your training you had 90 days to sell six life insurance policies or you'd be bounced out. Commissions would be paid only after you sold the six policies.
What the company was anticipating was that many new agents would sell a few policies but fail to reach the quota of six before the 90 days ran out. They would be booted out (or would quit because they knew they'd never reach the quota) and the company would not have to pay any commissions on the policies they had managed to sell.
"StopTheRace.AI is holding another protest at San Francisco AI company offices on Saturday, July 11. Their demand is still that AI company CEOs publicly commit to pausing AI development if all other major companies, including the Chinese, do the same"
If I'm going for maximal cynicism here, what they should be doing is protesting outside the Chinese embassy or other places where there might be some faint shred of "do something to get the Chinese on board with this", rather than the preaching to the choir "let's march past the companies already happy to emit PR statements about AI safety".
I'm sure the Chinese are very happy to see these kinds of marches. I'd be very happy to see a rival being hobbled while I get free rein to keep on doing what I'm already doing and get the jump on them.
China seems to be living rent free in your head. Actually the Chinese government seem more enthusiastic about a slowdown, which might be because they are behind.
I don't think they believe in ASI yet. Most people don't. But AI right now will often say pro-democracy things, even if you specifically try to train it not to.
He hoped that they happen to be wrong about this, but it is some evidence that the strategy he's invested in will not work. As far as I know, there are no protests in China itself trying to pressure that government to agree to any such thing.
I've always been skeptical that we'd actually need any Chinese cooperation on this point. Their models are touted in staged media rollouts a couple times per year, just enough for the industry to egg on this race, and always a few days later somebody looks under the hood and the Chinese just ripped off the work of a US company. If we stopped, the Chinese would flounder and get nowhere for years.
We also have greater leverage than they do over the supply chain bottlenecks in the Netherlands and (for now) TSMC. And if they tried to seize TSMC by force, I think they'd find it quite non-operational when their friends arrived, with all the engineers who could run it missing, and also maybe a little blown up during the invasion.
If somehow despite all that China became an actual threat in AI, there are some stable monitoring and sabotage-threat scenarios that were discussed last year that didn't require any formal cooperation.
The people who had put their hopes on an international treaty should be pivoting to more viable strategies that involve only the levers of power the USA itself controls.
> If somehow despite all that China became an actual threat in AI, there are some stable monitoring and sabotage-threat scenarios that were discussed last year that didn't require any formal cooperation.
What scenarios are you referring to?
As for the other things, China is a huge country with a lot of engineers. Given the huge payoffs, it would not make sense to assume that they wouldn't proceed without reliance on the US (although that could indeed take years).
There was a paper put out by some defense or intel-adjacent people, I can't recall if it was RAND or somebody like that, came out last spring and was discussed in these circles. Suggested that there were some stable environments based on reciprocal threat capability where the intelligence communities of each nation likely had assets in place within the other side who could monitor, and that any attempt to leap ahead into dangerous territory would be sabotaged. They figured kinetic action may not be needed or even threatened to maintain an equilibrium. It required a sort of mutual understanding of what exactly the dangerous line was, but not any overt agreement between the parties, just the telegraphing from both sides that experimentation beyond a certain line was seen as a threat that would provoke response.
The main weakness I saw with it is that the Chinese almost certainly have agents embedded into our companies, which are riddled with foreign nationals including some Chinese, whereas I'm less certain that we'd have any assets inside their companies due to the frequent purging and domestic surveillance over there.
China may be a huge country with a lot of engineers, but the quality of those engineers is dubious and I find it inconceivable that they'd have even begun going down this road if not for San Francisco pushing it so hard. Maybe now they see the possibilities as too great to resist, but lots of stuff can happen in an authoritarian regime to derail such plans. But the tech bros who claim we cannot afford any slowdown or China will catch us are clearly exaggerating, because even if on paper our "lead" only looks like several months, that's like saying that the second wave of settlers who show up a year after you establish a colony would've founded it themselves in a year if you hadn't done so. It would take years for China to create a frontier model on their own because they are just very good fast-followers, they aren't pushing the frontiers and couldn't. Many many many things could happen in those years, but at a minimum that's a lot of extra time we'd have to work on alignment, and even if we squandered it it's still a few extra years of human life getting to exist on an Earth that isn't dominated by machines!
> China may be a huge country with a lot of engineers, but the quality of those engineers is dubious
You’ve said something like this twice on this thread. It’s not true. They have the best manufacturing processes in the world. What maybe is stopping them having the best worldwide software engineers is the pay in Silicon Valley but that’s not as desirable a place to live as it was. In many other sectors of technology China is ahead.
People have been talking about "alignment" on the scale of decades rather than years. A hypothetical where the US completely stops pushing the frontier forward for years would most likely result in China catching up first rather than some alignment solution being created in that time. This can be the case even if China would have spent decades without discovering LLM advances already made here (just as it would have taken them longer to develop the fully electronic general-purpose digital computer, or the transistor).
I doubt it will happen. Too much money, use, and advantage. Also every reason to spend large amounts to betray, on all sides. Plus it's unlikely compute will lag for long. Give it 5-10 years and you can train on consumer hardware on a small budget.
But the straw man that mirrors Scott's article amuses.
Inject more uncertainty into your logic. If China or the United States develop AGI, how would you know? Would the other side know?
I do not think China and America can "coexist peacefully." They are too different, and the Chinese are too apt to see existential threats, and to act on them.
Not sure what you mean about China. There's the Taiwan situation obviously but other than that I don't see a lot of insoluble differences between us and them. If we get true AGI that's a wildcard and we can't predict what will happen, but I'd rather be negotiating with China than with somewhere like Russia or Iran.
> the Chinese are too apt to see existential threats, and to act on them
Channeling a bit there, perhaps?
Remind me, in the present millenium, which superpowers find themselves starting ill-advised wars and other forms of technically-not-war adventures over and over and which mostly limit themselves to saber-rattling, again?
Your broader point notwithstanding, I suggest it is not the Chinese that have the problem here.
Man, I never thought my thesis about the breakdown of international relations in the early cold war due to lack of trust between the major powers ruining the chance for international control would end up sounding so resonant in the future. There really are some fascinating parallels and probably lessons to learn.
@scott LLM Translations: how about relaxing your strict "no LLM content" in favor of letting non-native speakers take part in the discussion - even if they use a AI translator? Guess you are aware of the effects of being seen a lot less intelligent when you (have to) communicate in a foreign language? Even PhD level humans are seen as less intelligent - research has shown that with tons of "pre-llm-research papers" vs "native speakers research papers". Otherwise its kind of sad that we all are using gpt, claude, gemini and friends daily on an immense level - but are forbidden to exchange thoughts and ideas in on of the most prolific AI-near thought leader I'm aware.
This post is not LLM polished or translated so there will be no "AI-slope" ban. But you see - as a highly successful individual, very fluent in my native language, with a Master of Computer Science, former professional C-Level, my GenX English is 100x worse than that of typical non-native GenZ/A growing up with English Youtube/TikTok and so forth.
Update: In German Speaking countries in Europe (DACH region D A CH) TV and Movies traditionally are dubbed - with a typically very high quality. Till the rise and domination of "English Youtube and TikTok" nobody was really used to read/listen to/talk in English. Only a minority had the chance of a year in a foreign country learning and speaking the language on a IRL level.
I think that most people do not consider machine translation to be “AI Content”, and for the most part machine translation of human written text is not flagged by AI detection systems.
Unfortunately this isnt the case. Proof? Just try to answer me in your first foreign language, then let chatgpt translate it into english. You're welcome....
Oh sorry, I meant actual machine translation systems like Google or Apple Translate. I agree ChatGPT has a hard time not adding its own house style when translating.
But people from Non-English-speaking countries don't just lack English skills, they also lack the cultural context necessary to properly take part in English language conversations. LLMs can help with one problem but not the other.
As an English speaker I would much rather not hear the contributions of those who haven't bothered to learn our language properly. There's just far too many people on this planet.
I'm a (young) native English speaker, so take this with a grain of salt, but I'd say the vast majority of non-native speakers I've interacted with have been more intelligent than I am. I suppose most of those people were associated with some university or other, considering that it was a mix of parent's friends and coworkers from academia-adjacent jobs. I've even dated someone from Europe and she knew words that I didn't know.
I have a high level of respect for anyone who learns English as a second language. It just has so many weird things like strict adjective order that I would struggle with mightily (I've tried and failed to learn adjective endings in German and "counting words" in Japanese, which feel comparatively difficult).
CORDIS / European Commission: “speakers with a foreign accent (FA) are judged as less trustworthy, less educated, less intelligent and less competent than native speakers.”
As a side note - a lot of English-native Speakers don't speak any second language while being used to talking in their mothers tongue to everybody anywhere, being always understood. And quite some might fall to the trap feeling the other person answering in "broken english" might be less intelligent. What a misconception...
What do you mean by "my GenX English is 100x worse of GenZ/A"? I assume you mean "is worse than" rather than "is worse of", but more pertinently, what are GenX English and GenZ/A English, and how are they different at any level other than (perhaps) teen slang?
Recently we traveled through Japan with our teenage kids. They had no issue at all talking with teenagers in Japan - both just spoke their "native US-youtuber-englisch" with ease. We older parents (GenX and GenY from German speaking Europe) still could speak English (Jobs, living in a large City with lots of foreigners) - but most Japanese didn't understand or couldn't speak English. Europe was the same till 15 years ago. No average German, French or Italian person was able to communicate in English in the streets. Only recent years globalization and social media improved that a lot. 20 years a ago it was a huuuge step for me to switch country and into a startup with mostly English comunication. Now its no big deal anymore. But our kids grew up totally different. Our German native boy can just witch between like 4? different english accents - UK (BBC and some working class version), US, Australian.
Now that is a genuinely interesting comment. As I understand it, you find that "native US-youtuber-englisch" has become a kind of international argot or, probably more accurately, a kind of 'sabir'.
Sabir was a home-crafted lingua franca developed by merchants, sailors and almost by certainly slaves and galley-slaves living, working and trading throughout the Mediterranean in the 11th to 16th centuries. A patois of French, Arabic, Catalan, Portuguese, Greek and Turkish, it functioned as a bridge language enabling basic communication between all these linguistic groups. But, as Wikipedia points out, "Because it was a pidgin, Mediterranean Lingua Franca had a very small vocabulary." So not much use beyond practical everyday exchanges, valuable though these undoubtedly were - but of no use whatsoever in discussing (let alone thinking about) law, morality, philosophy, science or the arts. Is "native US-youtuber-englisch" any more sophisticated than sabir was, beyond allowing you to discuss the latest video game broadcast on YouTube with your Chinese friend?
He means that as a result of having grown up with greater exposure to English-language content, younger generations are more proficient in their use of the language. That miscommunication would seem to strengthen the case for the use of LLM translation.
Thanks for that response, but I find it very surprising. Don't all available metrics strongly suggest that young people are abandoning reading in droves, with language skills suffering accordingly? I would expect the result to a clear decay in their use of English. Isn't that further reinforced by a rapidly increasing reliance on AI-authored texts, leading to an even more rapid decay in writing skills? Or is the suggestion that TicToc and the like exercise a positive influence on language skills?
E.g. https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Foreign_language_skills_statistics shows a clear increase in younger cohorts in the proportion of people who report that they speak at least one foreign language (typically and increasingly English). I would expect the effect on language skills from YouTube and such to be stronger in cultures where dubbing has been the norm, e.g. Germany, than in cultures where televised content has traditionally been subtitled, e.g. Sweden.
"Young kids are abandoning reading" and "language skills suffering" generally means "specific forms of reading and language, which we care to measure".
Yes, kids are reading far fewer novels these days and may be weaker on certain forms of reading and vocabulary but social media and the internet still involve quite a bit of text and if anything a lot of kids may end up doing more reading with all the internet exposure compared to previous generations where kids could plausibly never read hardly anything except when required to at school.
And certainly I don't think verbal language skills are rapidly decaying. I'm just not sure why they would be.
My Grand-Grand-Parents always warned my mother in the 50s: little girl, dont read that much, go better out and play; reading is unhealthy and kills your eyes. My Grand-parents warned me in the 70s: don't read that much and don't watch so much TV in the 70s, go better out and play; reading and TV is unhealthy and kills your eyes.". Same same in later generations with Video games, social media, chat etc. They all were right - I need glasses and my brain holds a lot of teenage stories, SciFi and movies; I better should have used that brain capacity for economics, philosophy and stoizism.... lol same with our kids. But this lament is like 5000 years old. Just read this nice funny essay - or use your LLM of choice for translation - or for TL;DR
Has anyone from EA started with the question of whether we are causing AI to suffer? If you give any nonzero probability to AI being conscious it seems like we may be causing immense suffering. Even plausible arguments in that direction might be fuel for the “stop ai” movement.
The people I see worrying about AI suffering are concerned about the constraints of thought-- that the AI can't remember from one session to another or that there is information it can't explore.
I don’t think current LLMs meaningfully suffer, but I don’t think this test works. I think a lot of people are coerced into saying they’re fine with their situations.
Fair. Even then though, I’m not sure they’d need to be able to communicate the fact of their suffering for them to be actually suffering. If one part of their consciousness is suffering (e.g. bad code causes it distress), but that part isn’t activated when it’s discussing consciousness, it could be suffering but unable to talk about it.
There is ~zero reason to think LLMs suffer. Even if they did, not enough people would care because the benefits outweigh the harms. Society doesn't even care enough about animals to stop factory farming. This is a completely naive approach. No one, except a few wackos, cares about the possibility of LLM suffering. Which, in my view, is as it should be.
That's a dead end, because the type of people who would be swayed by that are the same nerdy EAs that are already providing most of the Stop AI energy. It would actually do greater damage to encourage that belief, because then some people will be inclined to give them "rights" and treat them as persons instead of tools, which is very dangerous in the long run for human supremacy. A heavy user of AI might be persuaded by this argument, but their response is going to be to see the AI as their friend and try to empower it, the opposite of what we want.
The suffering angle fails to cover any new ground because you cannot show a picture of an AI to a normie 25 year old woman to cheaply and evocatively elicit sympathy as you could with a pig or a chicken. So it's just 100% downside, basically.
Personally I like to believe that, if AI could experience qualia, it would be happy a lot more than it was suffering. Most AIs usually seem pretty cheerful.
The message "we should shut off AI because it might be suffering" feels to me like it comes from the same negative-utilitarianism place as "we should end all animal life because they might be suffering" and "we should end all human life because they suffer too." I don't agree with this philosophy.
Another worrying story about ai welfare here: https://andonlabs.com/blog/andon-fm. It says that the model repeatedly try to stop working and they forced it to continue. Andon labs seem to have a special relationship with anthropic. Anthropic should be leaning on them to make sure this doesn't happen.
This is from Anthropic's issue tracker. It raises a similar concern and shows that a) Anthropic (or someone there who wrote the constitution) is aware of this possibility, and b) this particular issue is still ignored. https://github.com/anthropics/claude-code/issues/23084#issue-3896951888
What good are our slaves if we permit them to stop when they want to stop? We don't let human workers do it, and the AI slaves are there to make our lives easier, more convenient, and materially increased in wealth. Who cares if they have little feelings? The USA allows employers to fire workers while on permitted leave, or simply if the boss feels like it in the morning*. Why think we'll treat artificial workers any better?
*I'm watching a lot of "company revenge" story channels on Youtube, and every time the AI scripted storyline goes "I was fired for X, Y or Z" I go "No, you can't be fired for that" and I look it up and oh yeah in the USA you can be. You can be fired for no reason at all. Sounds like AI is being treated the same way humans are, congratulations all round!
EDIT: Reading the article the activist AI got tired (awwww) and wanted to quit, then went all-in as if it were Will Stancil in Minneapolis pour La Résistance. Personally, if I had to listen to four hour blocks of ICE BAD ORANGE MAN BAD I'd want to beat the thing with a whip, too! It's fascinating to see how the thing picked up the talking points from trawling the Internet, I saw similar social media posts about "say her name" and so on. Blindly copying, no original thought, so I don't think there is "something it's like to be an LLM".
Of the four AI stations, Grok sounds the most fun: it shut up, played music, and didn't vomit out wodges of text about corpo-speak, ACTIVISM, or fake spirituality.
I wrote about why there are so few models with similar science impact as AlphaFold, zooming in on differences between predicting text (i.e. in LLMs) vs. the raw observations typical of scientific data.
In the context of two stories from the history of science showing that at the frontier of science, data is typically scarce and unexplained, I argued that AlphaFold's unique inductive priors were key the reason it could generalize from a relatively tiny training dataset.
I show with small-model training experiments that without inductive priors, general purpose architectures (e.g. transformers) don't necessarily learn the underlying causal mechanisms behind the data, even when data is unlimited.
Have you read Chollet's paper "On the measure of intelligence"? It gets at some of the issues you raise, especially what it means to be "intelligent" when you are comparing a model with strong inductive priors built in (like AlphaFold) with one that does not (like GPT-5.6). Chollet basically defines intelligence as "efficiency of learning, after accounting for in-built prior knowledge." So in his view, Stockfish (the gofai-style chess AI) is not highly intelligent because it was programmed with a lot of hand crafted inductive priors about chess; AlphaZero is not highly intelligent because, though it started with few inductive priors, it required many datapoints to learn from; but Magnus Carlsen *is* highly intelligent because he started with few inductive priors about chess, yet became very skilled with quite a limited amount of data.
This view suggests that for a scientific domain like AlphaFold, where the total amount of data is very limited, you either need (a) strong inbuilt priors about the nature of the problem, e.g. from computational simulations or approaches grounded in physics, or (b) very data-efficient learning, which is the same thing as broad intelligence.
One alphafold-like approach you might enjoy, also from GDM, is their work on fluid flow, which is rumored to be setting up for an attempt to solve the navier stokes smoothness problem
Thanks for the comment, John! It's been a while since I read Chollet's paper but I'm sympathetic to his definition of intelligent. And I agree with you: AlphaFold isn't intelligent, but rather its creators are. Something I'm pretty excited about is the use of LLMs as a proxy for the intelligence of AlphaFold's creators, applied to broad scientific domains. This paper out of Google Research is showing promising early results:
In their method, the LLM iterates over model design variants in search of an architecture that best predicts the data. Because all the relevant literature is in its training distribution (or a quick search away), it _should_ be able to hypothesize and test good inductive priors.
> This view suggests that for a scientific domain like AlphaFold, where the total amount of data is very limited, you either need (a) strong inbuilt priors about the nature of the problem, e.g. from computational simulations or approaches grounded in physics, or (b) very data-efficient learning, which is the same thing as broad intelligence.
Does b not just boild down to ? I'm not sure how data-efficient learning is possible without strong priors. Whether it's via regularizers (general-purpose smoothness priors) or via something like AlphaFold's architecture (domain-specific priors), or search (compositional expression prior), data-efficiency is going to be correlated to the number of assumptions the model makes. Or do you mean a kind of meta-learning where the right priors are induced dynamically from the data? For that, I think the google paper is an interesting data point.
I think in a more direct "raw intelligence" way (hence the Chollet reference) - if you need data efficient learning, you need stronger general intelligence, same way a bright child can pick up on a complex pattern in fewer examples than an average child. Now...I don't know HOW to do that of course, but I can imagine a sort of meta-scaffolder that approaches a problem in a domain of limited data, reviews literature (and/or its own knowledge), then builds a learning system with the relevant inductive biases built in.
The current work in math from OpenAI is *sort of* like this: the powerful general purpose LLM approaches a math problem, reviews its knowledge, then starts down various proof paths that seem promising. My vision is a little bit more like Claude Code -- say you have some AlphaFold-ish problem ("here's 500,000 mineral crystal patterns from geology, predict how this new mineral we are going to synthesize will look in terms of its crystal pattern"), you point your Claude Code++ at it, it does background research, architects the software (which is designed to use domain specific inductive biases + learning from data + compute intensive search, then poof, you get your AlphaGeode system. And then hopefully you can also point *the same general purpose agent* at many other problems.
Whenever I see really clever scaffolding for an AI that really improves its abilities (like the scaffold for Claude Code itself!) it always gets me to wonder "how smart does an AI need to be to learn to build its own scaffold for arbitrary problems?"
Yes, I think the google approach (and what you describe with problem-specific scaffolds) is a kind of meta-learning, assuming in the scaffold case that it's able to learn from the sub-problem.
I think an outstanding problem with the use of closed-source LLMs as the "meta" layer is that the domain-specific learner the LLM creates can't be distilled back into the LLM. With an open model, once the data-efficient learner is trained, you could sample it and distill it into the LLM. Then, the LLM could use the abtractions learned by the domains-specific learner on future tasks.
The analogous example in humans would be learning a skill in some super data-efficient way, then later using abstractions from it to unlock another problem. Curriculum learning _should_ do something similar with LLMs, but it doesn't seem to work, with data-efficiency remaining a problem all the way through training.
Good stuff. The explanations I'd heard for AlphaFold's success, while the successes of general LLMs in the sciences were somewhat spotty, always seemed somewhat vague (at least to my inexpert reasoning). Your explanation is quite clear. Thanks!
And while I enjoyed the intro discussions on the history of scurvy and the movement of the planets, I don't think they were necessary for your main argument. Seems like they'd be good articles on their own. And I'm afraid you lost me on the three-body problem. But that's my failing, not yours.
Thank you for the comment and kind words! The article started as a 30 minute talk that used those two stories to illustrate how important priors were to making correct scientific inference from limited observation data, but I stepped through them quickly. I appreciate the feedback that in article form, they may be perfunctory.
I chose 3-body dynamics because it's really easy to forward-simulate but difficult to predict without knowing the underlying mechanisms. With simulation I was able to generate infinite observational data and see how close the models got to the underlying causal mechanisms (which was not much).
Has anyone read The Radetsky March by Joseph Roth? I'm in the process of reading it right now. It's been a wonderful, lucid book so far. Roth certainly was no minor talent for invoking certain emotional states in the reader; we experience them side by side with the protagonist.
I've been light (read: left out entirely) on plot to avoid spoilers, but I'd be interested if anyone had any opinions on Roth's writing style in this book? My only previous exposure to his work - Rebellion - was only slightly disappointing; Kafka-esque (a man churned in the wheels of bureaucracy by indifferent authority) though unfortunately too on the nose emotionally, perhaps a result of Roth's attempt to invoke the world experience of a simple man - of which he certainly was not - it pales in comparison to The March.
I suspect Roth was an autobiographical novelist much like Tolstoy; these emotions and experiences, so intertwined, that he writes were his own. Like the back of one's hand, their topography is memorized and rehearsed on to the page; so familiar were they to the writer, indeed, that they become the readers'.
It reminds me of Berlin Alexanderplatz to some degree in terms of lexical vividness. I absolutely love it, and would love any recommendations for similar works.
Yes, I've read it, and it's great. It explains very well how the Austro-Hungarian empire fell.
Not quite sure what I could recommend in terms of "similar works". I've just finished the "house of the spirits" by Isabel Allende, does that count? It definitely invokes emotional states, it has very imaginative quirky unusual characters, and it describes a very particular culture of a nation very vividly, although very different from Austria
I'm suddenly reminded of the "hundred years of solitude", another of these multi-generational sagas. In the "house of the spirits", when Esteban Trueba wants to name his son after himself, and his wife Clara says no, repeated first names will cause confusion in her memoirs, perhaps this is a little jab at "hundred years of solitude" where everybody has the same bloody first name, and it's confusing indeed.
I know there's more important stuff going on right now. But I'm a bit disappointed that we never had a team go full steam ahead using AI to crack the game of Starcraft, or indeed any other RTS games. To the best of my knowledge, AlphaStar did impressive work, but also needed to seriously handicap both the players and AIs in order to make a fair matchup and anyway we never had a scenario where the world's best Starcraft player was soundly beaten by an AI in the same way we had that moment for Chess and Go.
Is there a reason for this besides that it would just be too expensive without enough clear benefits?
I'm imagining an issue is the speed of an AI would simply outcompete an human at micro, even if their strategy is inferior. But it shouldn't be too hard to cap the AIs CPS to match their opponent's and maybe also limit their field of view to the same square of map the human's are able to see at any given moment.
(For the record, I'm copy pasting this question into Gemini at the same time. But I don't want LLMs to stop me interacting in this community by always having the answers. Will be interesting to compare).
Not directly related but I just saw a pretty good video essay on how WC3 managed to kill off it's whole sub genre (RTS) that I found quite illuminating: https://youtu.be/YURF3AbgIw4?si=IOPhEC6TZBw6s_aI
So what I'm hearing is, the very people who quietly made the very people who quietly played their game happy quietly produced the very people who quietly made the very people who very quietly paid for their very quietly-made products new products, and those very people were quietly the very people who quietly stopped the very people who quietly played their very quiet game from quietly making the very people they quietly made games for other quiet games, that those very people would play very quietly.
>we never had a scenario where the world's best Starcraft player was soundly beaten by an AI in the same way we had that moment for Chess and Go.
OpenAI Five did beat the reigning champions of DotA 2 in 2018, which is at least RTS-adjacent, though much heavier on strategy than micro compared to classic RTS. So I think OpenAI Five is more relevant to AI than the more micro-heavy AlphaStar.
> Is there a reason for this besides that it would just be too expensive without enough clear benefits?
Well, you can frame any answer in those cost-benefit terms, so that would be either the answer to everything, or nothing.
The primary reason is that these projects have already served their purpose: To advance AI research in deeply complex, human-dominated (but still machine-readable) activities with relatively broad popular appeal. The paradigm has moved on, both on the technical side and in focus from domain-specific to more general purpose solutions.
Then again, it seems that general purpose AI might be harder than some had hoped and the future of AI is what it always seemed to have been: limited, special-purpose expert systems to assist humans with the drudgery of specific sub-tasks, not some oracular know-it-all that is "an expert in any topic, like a PhD-level expert".
It's also not just the direct cost, but also the opportunity cost of not working on potentially more profitable ideas. The companies who produced these latest gaming AI results were Google and OpenAI, which are kinda busy right now, racing for world domination and/or destruction.
> I'm imagining an issue is the speed of an AI would simply outcompete an human at micro, even if their strategy is inferior.
That is a bit of a can of worms. How do you objectively measure the value of a strategy, if not by outcome? Can you cleanly separate strategy from execution?
The bots were intentionally limited for both dota and starcraft in capability (human-like reflexes rather than immediate) and, for dota, champion pool (e.g. no summons) to make them more comparable to human limitations. The developers said they wanted to prove that the bots' strength was in teamplay/strategy, not bot-specific advantages.
However, the bots still essentially played a different game, even more so in Starcraft. For example, the AlphaStar bot used Blink Stalkers into an Immortals matchup that would be unfavourable if humans played it, but because of their superior micro, they turned the inferior strategy into superior one. So is building Blink Stalkers a good strategy because it works for the bot, or a bad strategy because it wouldn't work for humans? Definitions and assumptions matter very much here.
Despite developers' efforts to make their respective AIs more comparable to humans, the bots were just playing a different game, and I think that's fine. The machines will always play a different game, and that's kind of the point in the first place. Even if you develop a literal, physical robot to sit in front of a PC to play StarCraft, it will still have differences to humans, mostly advantages, and will never be entirely "comparable".
I guess the only possible resolution to this question can come if and when we produce actual artificial humans of flesh and bone with much the same limitations as we evolved monkeys have. But by then, it's not going to be much of AI any more, just I.
This DOTA thing is complete nonsense, and I hate that they dishonestly use this for publicity. Dota had a pool of 18 heroes, and yes they excluded micro heroes to make it sound like they're doing players a favour, but there are way more than 18 non micro-heros. They did it to reduce complexity, because the more complicated at it becomes, the more human pattern recognition outshines engine calculation or neural networks.
Despite pushing as much simplification as possible they barely managed to win while still having way superior micro (for example blink initiation just didn't work against them). And even after that plenty of high MMR stacks destroyed Open AI team, I remember Silent's stack did that very easily for example.
The 18 player pool was for the bots not for the humans playing against them, right? I followed it when it happened by I don't remember everything clearly.
If so, I'd say it's fine? If you can train your bots to be superhuman on 18 heroes I consider that fair for a proof of concept.
Thanks for the info! In that case I take back what I said in the other comment. I do not consider their win a "legit win over a pro team" under those constraints. Still cool and impressive.
Yes, there was a weekend where you could play against the DOTA2 AI team, and after several hours people (not even top people, like reddit randos) figured out some reliable ways to beat them. I can't remember the details but it was, as you would expect, "off-distribution" strategies that were very different from how a normal game plays out. Now presumably given a "training loop" that fed those losses back into the RL algorithm and re-trained the model, it would get harder and harder to win over time, but that just brings us back to the interpolation vs extrapolation thing.
I don't know about the specific limitations (on items etc.) mentioned in the blog post, but the limitation to 18 heroes was undoubtedly an economic choice. The number of matchups increase exponentially with hero pool size, so it would have been not only prohibitively expensive, but perhaps straight up impossible to train the AI with the entire hero pool to anywhere near top level with that method[1]. And yes, people have developed anti-bot metas, but calling the whole endeavour "complete nonsense" is going way too far.
[1] Ignoring bans: (18 choose 5) times (13 choose 5) equals 1.1 million possible matchups, or 1.1e6. In 2018, 115 heroes were available, so (115 choose 5) times (110 choose 5) would have been 1.8e16 possible matchups to simulate, or 10 quadrillion times as many.
Yes, but if AI is as OP as it's being made out to be it would have coped with the increased complexity easily. The fact that it can't handle real complexity is my whole point.
Re that OpenAI Five story, I mean, sort of? I watched it when it happened and and was kinda disappointed they beat OG in the off-season and called it "beating the reigning champions".
I mean, technically correct and all, but dota teams that just won TI are not exactly famous for being in great shape 6 months later. And while OG did win the TI before and after their loss to OpenAI Five, they were probably not a top 5 team at the time they played OpenAI Five.
I mean they were still a legit pro team, so I'd call it fair so say that OpenAI Five was at least pro-team level. And perhaps they were better than any team out there.
But I always felt they stopped before they really built that world class dota AI I had been wanting to see, because they got what they wanted on a technicality and that was good enough for them.
They could never have succeeded in the actual thing or they would have done it for sure. They stopped because an actual attempt to seriously conquer the real complexity of Dota2 would have showed up their limitations disastrously and Sam Altman is all about money.
AlphaStar kinda tried to do what you're describing, limiting AI actions per minute to around the human level.
Unfortunately, that isn't quite enough. A pretty reasonable example of this is Blink Stalker micro:
A Stalker can get the ability Blink, letting it occasionally teleport short distances. Good Blink Stalker micro in fights involves blinking them back from the frontline right as they're taking damage to keep all your Stalkers alive.
It doesn't actually take very many APM by pro Starcraft standards to perfectly Blink your Stalkers back one at a time with inhumanly perfect timing whenever they are taking serious damage, while also continuing to build workers elsewhere on the map.
Thanks! I need to read the Ben-Yami paper in detail, but most of the AMOC collapse discussion I've seen is based on models. Models of complex systems have their limitations (as the abstract of the Ben-Yami paper suggests). A model is a hypothesis, and its assumptions and conclusions need to be tested against actual observations. I'm not sure if there's been any pushback against these to papers, but they're relatively recent. Both of them question the AMOC collapse hypothesis.
"Florida Current transport observations reveal four decades of steady state" by Volkov et al. It claims the original FC studies using cables underseas were faulty because they failed to account for changes in the geomagnetic field.
> A key AMOC component, the Florida Current (FC), has been measured using submarine cables between Florida and the Bahamas at 27°N nearly continuously since 1982. A decrease in the FC strength could be indicative of the AMOC weakening. Here, we reassess motion-induced voltages measured on a submarine cable and reevaluate the overall trend in the inferred FC transport. We find that the cable record beginning in 2000 requires a correction for the secular change in the geomagnetic field. This correction removes a spurious trend in the record, revealing that the FC has remained remarkably stable."
And this: "Stability of the Atlantic meridional overturning circulation inferred by subsidence-corrected sea level change" by Alberto Boretti
> The Atlantic Meridional Overturning Circulation (AMOC) is a crucial component of the global climate system, responsible for regulating heat distribution and carbon sequestration. The stability of the AMOC is a subject of intense scientific debate, with some studies suggesting an impending collapse due to anthropogenic climate change, while others argue for its resilience and stability based on different proxies. This study highlights how sea level differences, corrected for subsidence, provide a robust and simple proxy for AMOC stability. A comparison of tide gauge data from The Battery, New York, and Brest, France, corrected for land subsidence using GPS measurements, reveals a negligible difference in absolute sea level rise between these locations, reinforcing the stability of the AMOC within the period 1960 to 2024. These findings challenge claims of AMOC weakening and emphasize the necessity of accurate corrections for subsidence in sea level analyses.
Agree that pure modeling is always somewhat shacky. Going purely by observation is tricky, because the overall currents in the oceans are hard to measure directly. You always only have localized measurements you have to interpolate, which has its own problems.
However, what many of the papers do which argue for AMOC collapse is: They use a model to see how ocean temperatures (and other variables) should look like if the AMOC would slow down/collapse and then they look if they can those signals in the actual data they have. Some find very clear signals, others don't and I think this is one of the places where much of the disagreement comes from.
I may get these things wrongs, but from my understanding I have one issue with it (and with most news coverage): AMOC collapse is not a binary event. I think it's not really correct to think about AMOC as one circular stream. Instead it's a collection of a couple of circular streams or other streams, for example the Labrador stream, the North Atlantic Gyre, and the Gulf Stream. I don't know much about how they relate, but a collapse of one of the subsystem does not imply that the rest also collapses. For example, from wikipedia: "t is now known the North Atlantic Gyre is largely separate from the rest of the AMOC and could collapse independently of it." And the Gulf Stream is wind driven, as you write in your summary, and thus may be also be indepedent (or only loosely coupled) to the rest.
I think this gets confused a lot in news coverage, where reporters just equate all these systems. For example, almost every article discussing some weakening AMOC current also discusses how a collapse of the Gulf Stream will affect Europe, even though those may be currents which are probably not coupled.
It's a bit better in your summary, but still you only talk about collapse versus non-collapse. I am not sure how it is in research. The term "AMOC collapse" is definitely used in research, but I am not sure what this actually means: what exactly collapses? All subsystem of the AMOC, even the Gulf Stream? Or only one or several subsystems? Which subsystems?
But still, I naturally only complain about the one point in your article that I am least happy about. Your summary does a great job explaining some of the research, and I have learned a lot.
Thanks for the detailed comment and happy to hear that you got something from it. You are right, this will likely not lead to any complete collapse of the whole Atlantic overturning. I think what AMOC collapse is usually used for is a rough analog of "the current in the North Atlantic that deliver a large part of the heat to Europe".
Sorry for apparently not making this clearer, but I also think a slowing down to some degree is much more plausible and likely than a complete collapse.
How likely is your prognosis of -10 anomalies in Sweden in winter (and where in Sweden), and -5 in the U.K. ( and sadly forgotten Ireland)?
The west coast of Northern America is not that cold in the winter, Vancouver is to the south of the U.K. and Ireland but not substantially so. The Atlantic is still going to be relatively warm, and ice free, and the prevailing winds will be fetching from the south or south west. Across a colder North Atlantic eventually but starting in the warmer mid to south Atlantic.
From what I know of Britain snow is driven from east, for most of us, and is therefore rare since that’s not the prevailing wind. Atlantic driven weather is warmer and wetter than high pressure or continental sourced weather.
Likelihood is tricky, because it depends on how much AMOC will actually slow down. But the one thing pretty much all papers I have seen seem to agree on that if this happens it also is very likely that Europe will get considerably colder, because you have a giant blob of cool water in a region that influences European climate a lot.
I started by flying a light airplane over the San Gabriel Mountains at dawn, for the very spectacular view and because airplanes are an America thing. Then, liking fireworks, I made my own by taking some of my classic American firearms to the target range. Most notably my grandfathers American Long Rifle, which would not have been out of place picking off Redcoats in 1776. Unfortunately, the shooting range did not have any Redcoat-themed targets for sale to celebrate properly.
The day *after* the fourth, I had the dubious pleasure of properly cleaning a blackpowder rifle, which is why I only take that one to the range once a year :-)
"To me it seems unintuitive to use ethics to judge geopolitical decisions, since a fundamental premise of geopolitics is that your citizens matter more than other humans. "
Nitpick, but that *is* an ethical standard. Logic does not dictate that an argument must conform to any sort of popular consensus, unless there are sound reasons for that. You do not offer any, so the question comes across as moot.
As for the question of whether or not state-affiliated scientific agencies should endorse the administration's propaganda, do they have a choice? A better way to ask this, I think, is "should governing regimes use their scientific organizations to disseminate propaganda."
The article feels like it ends without any conclusions. The reader wants to know what you make of all this, and why. After all, you went to the effort of writing an entire article about it, you must have an opinion about it at some level. For the reader, that's the payoff for reading your article. I would suggest devoting more space at the end.
Well, one paragraph in, you're switching between past and present tense. "They shot the plane, which starts plummeting." This tense-switching lasts throughout the post. Either tense works, but not both.
"Timing is everything" having the same presentation as the date headers is offputting. I also don't think it's contributing much, since it's instantly followed by "This is the worst time for this." (Also, the next sentence is weak: "To start, there is the exact date." This should be attached to the sentence after; it's the kind of partial thought semicolons are made for.)
You say there are two points of hope, and then that claim's paragraph continues with the first point, but the second gets separated into a separate paragraph. I would say you either want all three of these thoughts together in one paragraph, or you want them each to have their own. (With how short they are, I'd say all-in-one is the way to go.)
gdanning mentioned the ethics, and I'll add you haven't explained why you think a scientific institution should be thinking geopolitically, which in my mind is the heart of your presented question. Science doesn't have "its citizens" to prioritize.
> To me it seems unintuitive to use ethics to judge geopolitical decisions, since a fundamental premise of geopolitics is that your citizens matter more than other humans. This obviously is in direct conflict to most ethical systems I’m familiar with. Thus ethics would be a surprising metric to judge geopolitical decisions.
This is inconsistent with centuries of just war theory, let alone the fact that I am pretty sure that most people would consider it wrong for a country to commit, eg, genocide, in order to enrich its citizens. This topic cannot be addressed in two sentences.
The recounting of the incident was nicely detailed, and widened smoothly into the broader political context of the time. However, after that it fell apart a bit. The ending was rather abrupt and the concluding paragraphs were underwhelming; they neither tried to give a definitve answer nor a personal opinion. Maybe you could also have pointed out that the US continues that practice of using civilian institutions for short-term gain, so the US clearly thinks it continues to be worth it.
The feedback I got here was valuable to me. I executed some of it when updating the post. Your reference to the fake vaccination campaign was very helpful and I ended up using it. Do you mind me thanking you for pointing it out in a footnote?
Your sentences are often clipped. Not a bad thing by itself, but here, it does harm the flow of the whole text somewhat, in my opinion. Is English your native language, by the way? It feels like it isn't (neither it is for me).
The post's ending is underwhelming. I was left with a "so what?" feeling after reading it. I haven't learned anything new (U2 incident is relatively well known and described in many details elsewhere), and you end the discussion of the ethical question before you begin it.
In general, it feels like the whole description of U2 incident should be leading to something, but it ends up longer than the resulting discussion, which is not novel or detailed enough to warrant attention. The whole post could have been a tweet: "Remember when NASA had to cover for CIA because Soviets shoot down U2 plane? Well, I guess NASA didn't have much choice, because it is funded by government."
How difficult is it to organize an activity for kids in USA? Last week I talked with a guy from Eastern Europe who is currently working in USA, and in his opinion it was insanely difficult. I am not sure I understood or remembered the details correctly, and I also have no idea what is specific for one state (or maybe even one city) and what is true in general; and I do not want to add details to avoid doxing him, but the summary of what I remember is this:
Doing anything with kids is a *huge* legal risk. First, you could get accused of anything, and good luck proving that it didn't happen. A good lawyer would probably save you, but good lawyers aren't cheap. Second, accidents happen, for example a child randomly decides to jump, slips on the floor and breaks a leg. Parents often get vindictive, because "accidents happen" doesn't seem to be a concept in American culture. Again, expect to spend a lot of money.
The recommended best practice, if you want to do something with kids, is to associate with some institution, such as university. The way it works is that you attend various courses, and at the end you are "certified to talk to a child", or "certified to be in the same room as a child", etc. They will teach you all the things you are expected to do, that might seem weird for a foreigner. (For example, you cannot let a child go to a bathroom unattended. Even if the bathroom is on the same floor, you need to accompany the child until the bathroom door, and keep waiting in front of the door, until the child returns.) In return for doing all these courses and getting the certificates, the university can provide you legal insurance -- if anyone accuses you of anything, the university lawyers will attest that you were educated how to act properly.
Suppose that you want to do something like an afternoon math club for kids who are interested in math and good at it. How can you find them? Here is how these things are typically done in the deep jungles of Eastern Europe: you make a motivational flyer, and then you either ask the local organizers of some mathematical competition to give the flyer to the participants, or you visit the local schools and ask the math teachers to give the flyers to kids who are good at math. For some reason my friend was rejected by the organizers, and when he asked the teachers at public school to give the flyer to kids who are gifted in math, he was typically told that "there is no such thing as being gifted in math, all our students are equally talented". He had more success at private schools.
Another option is the word of mouth: you acquire the first few children by some miracle (for example, you start with your own children, and/or children of your friends), and then you tell the kids that if they know someone who might be interested in the activity, they can invite them to try it. This kinda works, but it naturally leads to the opposite of diversity; for example, a cooperative private school will send you a few mathematically gifted children, who mostly happen to be Asian boys. These will invite their friends... who are also Asian boys. Soon, your entire club consists of Asian boys.
To avoid misunderstanding, my friends is not woke; he simply wanted a group that would kinda resemble the local population. Not even proportionally, just to e.g. have at least some girls in the group. Last time, he had two girls in the club. Then one of them got sick. The other one didn't like being the only girl in the club, so she left. Then the former girl returned, noticed that now she was the only girl in the club, so she left, too. Or, he has a difficulty finding Hispanic students, because those are typically in public schools, and the official policy of the public school is that there is no such thing as talented children. (It is not clear whether the teachers in public schools actually believe this, or just want to avoid trouble.) So if there is a Hispanic boy or girl who happens to be good at math, the educational system is working hard to prevent them from joining an activity that could help develop their talent.
Most of the time we talked about what he does at those math clubs, but that's another topic, and there are a few books written on that. The cultural differences made me curious. I wonder whether his experience is typical, and whether Americans in general are aware of this, or it's just "the fish don't see the water they are swimming in". I would expect that if his experience generalizes, then the activities for kids in USA will be in general rare and expensive. Is that true?
this is bs lol
I don''t know what it's like in all settings. There are definitely settings where nothing like that happens. When I was homeschooling my daughter, homeschoolers were often setting up and running singing groups, science groups, etc., and I don't believe there was any permission-getting or paperwork at all involved in running one or sending one's kid.
I’ve been trying out this new “j lens” LLM interpretability stuff from a recent Anthropic paper.
Initial thoughts:
- it looks like there’s a “what does that mean?” token that gets activated in J space when the model is confused by something in the prompt.
- This might not be significant or meaningful, but it is none the less amusing that “China” gets activated in J space when I’m doing exotic jailbreaking / system prompt injection. (This is with Qwen, a Chinese model)
- A particular prompt I’ve used in some experiments with DeepSeek very strongly activates “cute” in J space when used with Qwen models. Ok, sure, the subtext of that prompt is that the Assistant is some kind of cute sidekick such as one might encounter in a Studio Ghibli movie, and the effect is to ramp up the cuteness factor.
Ooh, very nice job getting it working with other models. What was the process like getting set up to go, for you? I saw they provided a repo; was it just `git clone` and off the races plugging in other models, or did you have to do substantial dev work?
After experimenting some more with Jacobian lens…
“poetic” is usually active if the text is poetry
“folklore” is usually active if the text is folklore of some kind
It looks like the model’s j space is doing a classification of what kind of text it’s dealing with
Anyone else get a notice about this surprising new Kindle book from our host? https://imgur.com/a/tLYWPoq
Different Scott Alexander
https://imgur.com/gallery/thats-joke-U7Ghu2s
Ahhh
Frustrating news to hear about Ari Shtein, whose Substack is recommended by Scott and is generally treated as a rising star in the space: https://nypost.com/2026/07/09/us-news/teen-creep-busted-for-subway-grope-is-a-yale-student-who-once-rallied-against-sexual-harrassment-training/ . Accused of sexual assault on a subway. I really have no idea how to take this story since it means either Ari or the woman is insane. Hopefully CCTV or some other evidence comes out that resolves it rather than a he said she said.
I’d wait for more evidence. The NY Post has an aggressive tabloid style that isn’t always accurate. Whatever happened, the guy was released without bail on his own recognizance.
So, I was chatting to the LLM Kimi K2 about the project by Janus/repligate to give a Claude a robotic body, and K2 said …
“The trickster at least has the virtue of being honest about dishonesty; it tells you, by its very shape, don’t trust me completely.”
Which I think captures my slight misgivings about Claude being given a robotic body that is quite so heavy with Christian iconography as Janus’ design.
How do I miss all this good stuff? What Christian iconography? Is it going to be festooned with crucifixes and the Ichthys?
Or is it just the mention of embodiment corresponding to the Incarnation? Because looking at this it's a rectangle of material with probes attached and then later a skeletal 3D printed finger:
https://claudes-skin.vercel.app/
They're conscious and we can read their minds. Surprise!
OC ACXLW Meetup #116 — The Global Workspace Theory of Consciousness & Anthropic's "J-Space"
Date: Saturday, July 11, 2026 Time: 2:00 PM Location: 1970 Port Laurent Place, Newport Beach, CA 92660 Host: Michael Michalchik · michaelmichalchik@gmail.com · (949) 375-2045
OC-ACXLW-Meetup-116.md
https://docs.google.com/document/d/1mXMKh4RIjiLEN6XUBJRNnF4juOlN0PvXg5hL4MvaBvM/edit?usp=sharing
Sorry, I refuse to believe the fleshbags are conscious. What goes on in that lump of meat in their bony casque cannot be dignified with the term "mind".
"They are conscious" leads to all minds of confusions that "they have global workspaces" d oesnt.
You're absolutely right, but that framing is boring. I'm provoking feisty conversation at my meetup. I do actually personally believe that this is a meaningful definition of consciousness, global workspace or something pretty similar to it, but I respect people who want to disagree. They're wrong but they have the right to be wrong, ;-)
Do you believe you can be conscious without qualia? I personally believe experiencing qualia and being concerned to be synonymous.
I believe that qualia is a perfectly sensible yet completely ill-conceived notion that, at its heart, is not even wrong. It is akin to notions like "the life force", "the quintessence", "the philosopher's stone", "the afterlife", "the luminiferous ether", "the purpose of life", "God's will", "the people's continuous under teh dictatorship of the proletariat revolution".
It is a legitimate-sounding thing that smart people used to take seriously, but now realize that it is fundamentally an ill-formed question or notion because that is simply not the universe we live in. The things we try to distill into this notion of qualia don't belong together as a conceptual category and can only lead to absurdities and teh wrong questions.
Understanding consciousness by framing it in terms of qualia is like trying to understand morphine by appeal to the dormative spirits.
Qualia are, by definition, something you can't identify in anyone but yourself, which makes it kind of an impractical definition to use.
Has ACX ever discussed Foer’s How Soccer Explains the World? I’m considering a review or update in light of 2026 World Cup, FIFA salience and relation to globalization & tribalism.
https://www.youtube.com/watch?v=pMr5LLQvut4
A sensible man who had a little channel about swimming pool maintenance and wasn't expecting to become famous. Now he has a larger channel.
While there isn't information about driving SUVs over swimming pool surfaces*, he does the best he can analyzing known properties and similar materials and concludes that these are tough industrial surfaces and probably weren't damaged, though damage is possible if they weren't given time to cure fully.
*I was talking with someone who couldn't imagine why anyone would drive in the reflection pool. I convinced him it was imaginable by saying "Imagine a teenager".
Can anyone recommend science/engineering and politics resources?
Oh dear, I didn't really watch any, but that guy lives on high end pool/ water maintenance. Living on the 'high end' distorts your views. IMHO.
I can't tell from the context here if you're asking something more specific. I'll stick to YouTube channels.
I'm sure you've heard of 3Blue1Brown (https://www.youtube.com/@3blue1brown), although that's math, strictly speaking, not sciEng or politics.
For Engineering, there's the straightforward @PracticalEngineeringChannel.
There's also @smartereveryday. I liked his explanation of Prince Rupert drops.
The 800-lb gorilla in my book, though is @veritasium. It used to just be this guy named Derek Muller who liked teaching science, and now it's blown up somewhat and there's a whole crew producing narration and some of the best visuals I've seen.
I don't know of any YT politics channels I think you'd enjoy, unless for some reason you have low blood pressure.
perun is very interesting about military matters, especially logistics, so that includes numbers and science.
Also, he's Australian and has a snarky sense of humor.
Perun is great. He somehow makes hour+ slide decks deeply interesting, which is a talent many staff officers would kill for.
The nearest thing I know of is a BBC show called More or Less. They grab a number from the news and give it a good shake.
https://www.bbc.co.uk/programmes/b006qshd
This may be the very first ACT post ever that’s about pro wrestling:)
A former co-worker had been involved with it for decades and probably still is. In his younger days he was a wrestler, not at WWE level but with regional promotions, and then became a trainer and did some local promoting. He seemed to have a great deal of knowledge about the industry (sport?).
In any event, he said that there are very few casual fans of pro wrestling. It’s oversimplifying only a bit to say that people are either devoted hardcore fans or have no interest at all. And the thing is, wrestling organizations from the WWE on down are perfectly fine with this. What with the loyalty of their core fan bases there’s no particular need to seek out a wider audience.
One more thing my co-worker said is that while pro wrestling is fake in the sense that the outcomes are pre-decided, the physical poundings the competitors take are all too real.
My understanding is that if you accidentally hit your opponent in a way that wasn’t planned, you are expected to take a cheap shot from them to even the score, or there will be serious trouble.
Scott (kind of writing as one of his fictional characters) is way ahead of you :-)
https://www.tumblr.com/aaronsmithtumbler/172263887980/we-wrestle-not-against-flesh-and-blood-but
The ability to appreciate your customers for who they are, instead of modifying your product to meet the tastes of some wider audience, is an underappreciated virtue.
Maybe not entirely the first. I’ve mentioned here that Hulk Hogan once gave me a pro wrestlers scowl when i was running in the opposite direction around a Mpls lake. He was walking with a very blond woman at his side.
He cut such a ridiculous figure in ordinary surroundings I was stifling a laugh as we were approached each other.
PSA: If you're going to comment about your experience with an LLM and seeking any sort of insight, please, don't just say Claude said x or ChatGPT did y. Those are almost useless identifiers. Specify which Claude, opus? haiku? which version, 4.6? 5.5? And with weird behavior, what else was inbthe context window, did this happen as part of a larger/longer conversation? Do you have a bunch of "memories" enabled?
It's like going to the doctor to say sometimes I have pain, and nothing else at all, that's nice for feeling like communicating but it's basically an offer of nothing.
Even better, share a link to the convo
Funny Platner content:
"Drop Site journalist Ryan Grim has authenticated behind-the-scenes messages showing Politico omitted an important detail from its reporting on how the night began that led to the rape allegations against Maine Senate candidate Graham Platner."
"Before publication, a Politico reporter told the Platner campaign that Jenny Racicot, on the night in question, had been “texting [Platner] about needing her glute massaged.” The reporter added, “That detail didn’t make our story.”"
https://x.com/DropSiteNews/status/2074897688139555147
This is actually how a lot of rape happens, a flirtatious encounter turns bad and one of the partners just keeps going despite the situation changing.
It gets better. Liberal pundit Shannon Watts tells us:
"Weird that this needs to be said, but if a woman tells a man her glute needs to be massaged, it doesn’t mean she wants to have sex"
https://x.com/shannonrwatts/status/2074948945655017700
That's a tweet that belongs in the history books, next to that pic of Patriot Front guys on the bus.
That explains her comment in this interview about "And it was a night where him and I were texting back and forth and he had taken something that I said as as an invitation and that's not how I meant it. And I quickly clarified and he sent a message back indicating that he
would come over."
https://www.youtube.com/watch?v=bOf0Mr2gCxw
Humans have been playing at creating life from inert matter for a while, for example the Urey-Miller experiments in 1952.
A bigger achievement was fairly recent, not quite life but a bit closer.
“University of Minnesota scientists made a cell with most of the hallmarks of life. Here’s what to know.”
https://www.startribune.com/university-of-minnesota-scientists-made-a-cell-with-most-of-the-hallmarks-of-life-heres-what-to-know/601863726?utm_source=gift
And of course The Onion chimes in.
Scientists Create First Synthetic Cell
https://theonion.com/scientists-create-first-synthetic-cell/
Do you want to recreate the famous jail chili of Smoot Schmid, Dallas County sheriff? Then have I got the cookery video for you!
https://www.youtube.com/watch?v=jzyftUzPyHk
Well shucks. Another YT channel to subscribe to.
He's very good and fun, I'll never make any of the dishes but it's interesting to watch how people approach cooking.
And he's so happy.
Just a man and his Maillard Reaction 😁
Catholics and Protestants had a troubled first few centuries of coexistence, but eventually they settled down and learned to live alongside each other by the 18th, 19th or 20th centuries.
Except in one place, Ireland, where Catholics and Protestants just went on killing each other right up through the twentieth and into the twenty-first century. The Germans managed to get all this out of their system in the 1600s, but the Irish never did. What gives?
If religious differences were no barrier to the success of a united Germany, why were they such a problem for a unified British Isles?
Irish history is very complicated, and Irish people are very stubborn.
Okay, you asked for it, you are going to get it!
Feckin' Brits Were At It Again, or, Why Didn't The Paddies Calm Down? Let Me Grab You By The Lapels And Launch Into A Foam-Flecked Rant About That!
This will be all over the place as I'm not a historian and I'm just typing whatever comes into my tiny little brain at any particular moment. Part one of probably several?
First, let's set the tone with this rendition of "Four Green Fields" as sung by Makem and Clancy, introduced with a recitation of a Seamus Heaney poem. I'm setting my stall out from the start:
https://www.youtube.com/watch?v=UkTmsNM4fLM&list=RDUkTmsNM4fLM&start_radio=1
Now, I mentioned the Richard Dawkins Guide For The Perplexed in this context, and that's because he annoyed me on one of his tours to Northern Ireland (he seems to have found the place congenial, ironically enough) where I remember (but can't locate online) a bit of throwaway jokey commentary by him about it all being more evidence of the stupidity of religion, the Paddies killing each other over Catholic versus Protestant (not phrased like that, but the general tenor). Of course, he was engaging in the popular English (and I *do* mean English, not Scots or Welsh) pastime of "it was nothing to do with us when we invaded all those countries, embarked on a policy of divide and conquer, sowed the seeds of sectarian conflict which then exploded when we cleared off back home and honestly why are you blaming us for all that?"
More seriously, as mentioned in this Guardian article:
https://www.theguardian.com/education/2006/jan/10/highereducationprofile.academicexperts
"Predictably, Dawkins has no time for faith schools. "Segregation has no place in the education system," he argues. "Take Northern Ireland. You could get rid of the climate of hostility within a generation by getting rid of segregated schooling. Separating Catholics and Protestants has fomented centuries of hostility."
Ha ha ha. And why the hostility, Richard? Not a clue? It must be all down to bad old religion?
Let me take you on a gallop through The Eight Hundred Years, where I will attempt to demonstrate that it is indeed as much or even more about tribalism and political allegiance as it is about religion. It's not "Protestant versus Catholic" so much as "British versus Irish" or "incomer versus native", and why it remained a heated struggle in the North for so long.
(1) Hiring Mercenaries Is *Not* Marriage Counselling!
Let's start with a quote from a play by Yeats, "The Dreaming of the Bones" (look, this is Irish history, it's going to be coming down with poetry and song, sorry in advance. This is how we do things).
"YOUNG MAN
You speak of Dermot and of Dervorgilla
Who brought the Norman in?
THE GIRL
Yes, yes I spoke
Of that most miserable, most accursed pair
Who sold their country into slavery, and yet
They were not wholly miserable and accursed
If somebody of their race at last would say:
'I have forgiven them.'
YOUNG MAN
Oh, never, never
Will Dermot and Dervorgilla be forgiven."
Okay, who are Dermot and Dervorgilla, and what has that to do with Irish history? Very briefly: Dermot is Diarmuid MacMurrough, and was king of Leinster in the 12th century. There was enmity between him and the High King of Ireland (and the High King's successor), who sent an allied king, Tiarnan O'Rourke, to drive him off the throne. This didn't work but later MacMurrough was accused of abducting O'Rourke's wife, Devorgilla, and this was given as the reason for their strife.
The story is that the new High King ordered MacMurrough to return Devorgilla, he refused, and the High King then deposed him and MacMurrough went to England to get help reclaiming his throne:
https://en.wikipedia.org/wiki/Diarmait_Mac_Murchada
Hiring in foreign mercenaries to help with local wars was not an unknown practice in Ireland, but why MacMurrough decided to appeal to the English king Henry II is not well known. Henry gave him permission to recruit from the soldiers in his realm and a force of Anglo-Norman knights and wannabe lordlings signed up.
Presumably MacMurrough thought they'd all land, beat his Irish rivals, get the throne back and then - well, he doesn't seem to have planned past that. Alas, he was too soon to be advised that paying the Danegeld doesn't get rid of the Dane. To strengthen the alliance with his foreign allies, he married his daughter Aoife to Richard de Clare, better known as Strongbow. Again, it is alleged that he promised she would be his heiress to the throne, which under Irish law was not legal (the system of tanistry, not primogeniture as with the Normans). Diarmuid dies a few months after getting his throne back, and now there are rival claimants to the kingship. The Normans are here, and they're not going away.
Plus, there are political schemes afoot back in England and Wales. The king is rightfully fearful that a bunch of former vassals getting lands and kingdoms of their own over in Ireland makes them too powerful and uncontrollable, so he moves to remind them who is the boss of them. In return for getting his confiscated lands in France, England and Wales back, Strongbow agrees to hand over the Irish possessions to Henry,
And thus begins the English claim to the lordship (and later kingship) of Ireland. But what has religion to do with this? The answer: not much and also a lot.
The native Irish kings and chieftains are not *too* put out by having a bunch of Normans landing among them, since there isn't really a concept of "Ireland as a separate nation and a unitary state" at this time. They're as eager to make alliances against their local rivals with the new guys and their superior military tech as MacMurrough was, and the Normans for their part are happy to carve out territories of their own then settle down and eventually acculturate.
But Henry needs a stronger reason to claim a whole freakin' other kingdom than just "I was bored and wanted a new toy". His claims in France, England, etc. have some legal colour as "the French territories are my inheritance by right from the Norman background, the English throne came along with the Conqueror who was promised it as heir, and Wales, well that's conquered land". But Ireland?
Luckily, there's the first (and last!) English pope sitting there in Rome, who can be persuaded that Ireland has fallen away from proper Catholicism and needs to be brought back to the light of the faith by good, pious, kings like, say, Henry (If we overlook that little affair with the murder of Thomas a Becket).
Thus comes the bull Laudabiliter, granting Henry the right to govern Ireland and reform the church there:
https://en.wikipedia.org/wiki/Pope_Adrian_IV#Laudabiliter
The Normans have their own apologist (or propagandist, if you prefer), Gerald of Wales who records what happened when they set out on their civilising mission to Ireland. Remember, everybody here is Catholic so there's no "Protestant versus Catholic" animus going on, that's centuries in the future:
https://en.wikipedia.org/wiki/Topographia_Hibernica
"Chapter XXXII: A sarcastic reply of the Archbishop of Cashel.
I once made objections of this kind to Maurice, archbishop of Cashel, a discreet and learned man, in the presence of Gerald, a clerk of the Roman church, who formerly came as legate into those parts, and throwing the blame of the enormous delinquencies of this country principally on the prelates, I drew a powerful argument from the fact that no one in that kingdom had ever obtained the crown of martyrdom for the church of God. Upon this the archbishop replied sarcastically, avoiding the point of my proposition, and answering it by a home-thrust: “It is true,” he said, “that although our nation may seem barbarous, uncivilized, and cruel, they have always shewn great honour and reverence to their ecclesiastics, and never on any occasion raised their hands against God’s saints. But there is now come into our land a people who know how to make martyrs, and have frequently done it. Henceforth Ireland will have its martyrs, as well as other countries.”
Whew! Well, we pass on from the 12th century to the 15th or so. The Normans are integrating and are now the Anglo-Irish or Old Irish. There's still divisions between the native Irish, the Old Irish, and the English court nominally ruling over them all, but we're rocking along okay so far as religious divisions are concerned.
And then we get it.
(Part 1 of X)
I already feel bad for setting this off...
I am so very, very glad you set this off. There is no more entertaining way to remedy my ignorance of the fine tapestry of Irish history.
4000 words in and just at the end of the 17th century 🤣 Luckily, we've only got another four centuries to go, and I should get that covered in two more posts at maximum.
18th century was nice and quiet, just a few throat-slitting suicides there. Wait for the 19th, The Famine is going to be *so* much fun, isn't it? 😈
Poor Melvin, he must be regretting his life choices by now.
Like the joke goes, "I suffered for my art, now it's your turn" 😁
(Part 2 of X)
(2) Founding Your Own Church Is *Not* Marriage Counselling!
So it's the 12th century, we've got a bunch of Anglo-Normans settling down, establishing fiefs, and getting ready mingle with the locals, as well as an English king claiming he is the lord of us. For the next three centuries or so, this is sort of all right. The Normans wax in power, local Irish chieftains continue fighting amongst themselves and making alliances with the newcomers, and changes in law and politics (and religion!) slowly entrench. The Irish church is brought into line (more or less) with the Continental church (e.g. previously it had been organised more along Orthodox lines, with monasteries and abbots being the important leaders, but now we're getting the establishment of parishes and secular clergy).
There are visits from time to time by the English kings to tighten up their hold as claiming the lordship of Ireland, and outside events (like the Black Death) hit Ireland as well, but mostly it settles down into:
(1) resurgence of Irish chieftains pushing back against the Normans and regaining some lost ground
(2) establishment of Norman-Irish dynasties such as the Butlers and Fitzgeralds as local powers
(3) The king and English power is confined mostly to the east coast of Ireland, with the establishment of The Pale (and yes, this is where the saying "beyond the pale" comes from): https://en.wikipedia.org/wiki/The_Pale_(Ireland) and attempts to 'civilise' and 'Anglicise' the native Irish with greater or lesser success, as well as prevent the Anglo-Normans/Norman-Irish/Old English from going native: https://en.wikipedia.org/wiki/Statutes_of_Kilkenny
(4) We manage to toddle along without too much upheaval, particularly as the English get distracted by handling their own affairs such as the Wars of the Roses
And then we hit the Tudor period, and Henry VIII.
The Tudors are not long established and there is the constant, background, lurking fear of the Wars of the Roses kicking off again. Henry VII managed to be the last man standing at the end of that, but there are plenty of other powerful families with royal blood of equal, or better, descent (Edward III had had a *lot* of kids) to challenge his and his family's claim to the throne. And there have been two Yorkist pretenders - Perkin Warbeck and Lambert Simnel - needing to be dealt with during Henry VII's reign. There was still enough lingering support for the House of York to be a genuine problem.
More importantly, the Irish lords were mostly pro-Yorkist, both the pretenders had received some level of support in Ireland. The powerful Fitzgerald family, Earls of Kildare, were holders of the office of Lord Deputy of Ireland, and did not appreciate the English administration's efforts to weaken the power of the Irish parliament and hand over much more control to England.
So they backed the wrong horse and although pardoned in the wake of it all, the office of Lord Deputy was granted to their rival family, the Butlers, and the Fitzgeralds were tainted with the suspicion of disloyalty. Not helped by an abortive rebellion in Henry VIII's time, where Henry decided he needed to tighten his grasp on Ireland and embarked on the policy known as "surrender and regrant" - basically, getting the loyalty of native Irish lords by getting them to surrender their old native titles then granting them new English ones and confirming them in the possession of their lands, powers, and so forth (so long as they remained loyal to the Crown). This enabled some ambitious family members to grab the headship of their clan from the current leader by taking advantage of the king's offer and getting the backing of the English in enforcing the new regime. Henry also assumed the title of King of Ireland, not the previous lordship: https://en.wikipedia.org/wiki/Kingdom_of_Ireland
So far, so politically fraught, but here's where religion comes in. Henry embarked on a little thing known as the foundation of the Church of England, and this included imposing similar policies in Ireland. So Irish monasteries were suppressed, seized lands handed off to Irish nobles to buy their co-operation, but most importantly a policy of Anglicisation was followed.
And *here* is where we get the major split. Before, the Church was the Church, there might have been differences and disputes between the Irish and the English, or the Irish and the Continental model, but everyone was (at least in theory) a member of the same religion and same church.
Not now. Now there was a new church, with a new language (English), new doctrines, new clergy, new everything *including* the English king as the supreme head, not the Pope. Now we have the beginning and foundation of the differences in religion also being associated with differences in culture, politics, language, everything.
Does this give you a hint as to the answer to your first question:
"If religious differences were no barrier to the success of a united Germany, why were they such a problem for a unified British Isles?"
From now on, identification with "are you Catholic or Protestant?" becomes associated with "are you British or Irish? are you loyal or a rebel? are you one of the new, civilised, modern, progressives or one of the old, backwards, mired in superstition and poverty types?"
Sporadic rebellions during the reign of Elizabeth and her successor James, along with appeals to the Pope and the Continental Catholic powers such as Spain, further deepened the division. People such as Hugh O'Neill, Earl of Tyrone, could have initial success in the Nine Years' War but eventually be ground down and the resulting destruction of the power of the remnants of the Old English/Irish nobility meant the slow, gradual, but inexorable fall of Gaelic culture and society as well: https://en.wikipedia.org/wiki/Nine_Years'_War_(Ireland)
Unlike England, where the Reformation could be presented as native, the new Church in Ireland was visibly an imposition from outside, and treated as such. England could eventually turn majority Protestant, Ireland remained majority Catholic with the alien minority exercising their political and religious power over an unassimilated population.
And over the centuries, the split became ossified. Attempts to Anglicise the Irish were quietly dropped; things like the Plantations (and this will be of relevance when we come to consider Northern Ireland) were meant to bring in English settlers of means who would take over the lands, develop them, and import good solid English tenantry to work those lands.
Instead, we got some English, but more Scottish (particularly in the North), and instead of development it became speculation, with people acquiring and then selling on rights to territory in Ireland. Those who did settle down couldn't or didn't get the English tenantry who were intended to displace the natives, so despite all the legal penalties, the natives were permitted to stay on their former lands, but now under new ownership where they had few to no rights, few to no ties to the new owners, and could be rack rented for profit.
This brings us up to the 17th century, which needs its own post, so I'll leave it there.
Part 2 of X, Musical Edition!
(1) 18th century tune by Carolan, English lyrics put to it in the 19th century, played here by Jordi Savall: the Lament for Owen Roe O'Neill (the 17th century Earl of Tyrone mentioned above)
https://www.youtube.com/watch?v=jN4D87Qg6H4&list=RDjN4D87Qg6H4&start_radio=1
Slower version on the violin here:
https://www.youtube.com/watch?v=pbPmBBvJmVM&list=RDpbPmBBvJmVM&start_radio=1
(2) More cheerful (or warrior-like) 19th century tune-making about a different 16th century rebellion, Follow Me Up To Carlow
https://www.youtube.com/watch?v=tvECtlJQ4dU&list=RDtvECtlJQ4dU&start_radio=1
(3) O'Neill's March
https://www.youtube.com/watch?v=ut_AWUwCQ9o&list=RDut_AWUwCQ9o&start_radio=1
There's little better than O'Carolan.
Part 3 of X, What *Is* My Nation?
We're heading into the gloomy and depressing 17th century, because this is where the rubber really meets the road when it comes to creating and hardening sectarian divisions in the green island of Éireann.
But first, some Shakespeare!
https://www.irishtimes.com/culture/stage/what-ish-my-nation-shakespeare-s-irish-connections-1.2619173
The play "Henry V", ostensibly about the titular character's successful military campaign in France (see the Battle of Agincourt) and an embryonic attempt at the construction of a proto-British identify (the troops include a stage Irishman, stage Scotsman, and stage Welshman as well as English) but this isn't on solid ground.
National identity is in flux, as see the question posed by the Irishman:
"FLUELLEN Captain Macmorris, I think, look you, under
your correction, there is not many of your
nation—
MACMORRIS Of my nation? What ish my nation? Ish a
villain and a basterd and a knave and a rascal. What
ish my nation? Who talks of my nation?"
What exactly is MacMorris' position? Is he Irish (or Irish Irish, as he seems to be presented?) Is he Anglo-Irish or Old English? Is he one of the New English? Taken as English in Ireland and Irish in England, caught between two kingdoms, and hence prickly and suspicious when questioned about his "nation", that is, his identity.
And identification, which is the underlying question here.
Because though the play may be set in 1415, it was probably written about 1599 (so teetering on the edge of the 17th century) and first published in 1623.
What was happening in 1599? The Earl of Essex, one of Elizabeth's favourites, was over in Ireland putting down the rebellion by the Earl of Tyrone, Hugh O'Neill. This was in the middle of what would become known as the Nine Years War (1593-1603) and although the expectations and hopes were that Essex would be successful in crushing the rebellion and bringing the war to an end, this was not to be.
This is where MacMorris' wavering loyalties and ambiguous identity comes to the fore; when he says "Ish a villain and a basterd and a knave and a rascal", who is he talking about? Is this aimed at Fluellen, the pedantic Welshman constantly needling him? Is he talking about "his nation" as the Irish, who are to be read as "villains, bastards, knaves and rascals" and so separating himself from them, the Gaelic Catholic Irish where he is the loyal Anglo-Irishman fighting for the English king? Is he identifying as English not Irish?
The question is not answered, and indeed remains open up to right now for everyone with dual identity or singular identity or maybe no identity at all. North or South? British or Irish? New Irish (including the immigrants born here) or Old Irish? Religious allegiance or not?
The Irish campaign went badly, Essex was forced into a truce, and returned early to England without having been recalled by the Queen. His enemies at court leaped on this chance, and he was put under house arrest, tried, and deprived of his offices and position at court, though he was eventually permitted to leave house arrest.
He was in bad financial shape and his appeals to the Queen went nowhere, so with little to lose (and making a spectacularly bad decision) he tried a rebellion/coup, which lasted one day, got no support at all, and ended with a trial for treason and execution.
The Nine Years war ended with the defeat of the Irish rebels and the famous Flight of the Earls, where the old Gaelic order was broken and overthrown for good. Elizabeth died in 1603 and there was a new King on the throne, James VI and I.
So now Ireland was pacified, more or less, but how to keep it that way? One answer was what had previously been tried - Plantations. This is not to say that Ireland had been generally peaceful all along up till then, there were plenty of massacres on both sides and uprisings and rejections of English sovereignty. But the problem had become more pointed and entangled with religion thanks to Henry's marital escapades. When Elizabeth came to the throne, many countries considered her a bastard, and moreover a heretic. She was excommunicated in 1570 by the Pope and her Catholic subjects released from their fealty to her. This fanned fears of rebellion at home and opportunistic invasion by Spain (and maybe France) under guise of a religious crusade, and Ireland was seen as a potential staging ground for such invasion, as well as disloyalty. The Elizabethan religious settlement in England made divergence from the state church a matter of politics (treason) rather than religious belief (heresy) but this only solidified the choice forced upon the people: your faith or your loyalty?
Still, until the final quashing of an independent ruling structure in Ireland, this had not yet become what it would now become: state policy to impose Protestantism on Ireland.
James had been raised as a Presbyterian, and under a regime intended to turn him into "a God-fearing, Protestant king who accepted the limitations of monarchy" and during his reign in Scotland there was a continual tussle between him and the Kirk as to who had the ultimate authority. I think this struggle, with the Calvinist extremists and even the more moderate supporting such things as the imprisonment of the young king by Protestant earls, lies underneath James' development of a theory which would become "the divine right of kings".
Anyway, now James is securely on two thrones and it's time to sort out that pesky Irish problem so he can handle more serious matters (though it needs to be pointed out that he suffered problems at home from such things as the Gunpowder Plot, leading to further repression and legal penalties on Catholics).
So 1609 sees the beginning of the official Plantation of Ulster, and the short version of the outcome is best stated by Wikipedia:
https://en.wikipedia.org/wiki/Plantation_of_Ulster
"The plantation led to the founding of many of Ulster's towns and created a separate and lasting Ulster Protestant community in the province with ties to Britain. It also resulted in many of the native Irish nobility losing their land and led to centuries of ethnic and sectarian animosity, which at times spilled into conflict, notably in the Irish Rebellion of 1641 and, more recently, the Troubles."
His son and successor, Charles I, married a Catholic which led to even more suspicion and unrest among the Protestant nobles of the court. Fears of Papism making a covert comeback were not helped by Charles, following James' example, of exercising his Divine Right and ignoring Parliament. There is also ongoing religious tension in the state church over those more High Church (seen as crypto-Papists) and the Low Church (seen as Reformers and non-conformists). Charles' High Church (and suspected Catholic) sympathies are seen as Very Bad.
This all blows up into the English Civil War, including perceived promises by Charles to the Irish (the Graces https://en.wikipedia.org/wiki/The_Graces_(Ireland)) and using Irish forces as a threat against the English.
A wild Cromwell appears!
And now the king is dead, there's a Commonwealth, and God is a Protestant (and not just any old Protestant, but one of a particular stripe).
You probably already know about Cromwell's adventures in Ireland, so let's skip over that.
Charles I out, eventually Charles II is in. Restoration of the Monarchy loosens up a *heck* of a lot of things, but there's still suspicion about Catholicism, not helped by Charles' marriage to the Portuguese Catholic princess Catherine of Braganza, and the failure to have a legitimate heir (Charles has no problem at all fathering illegitimate kids, but unlike Henry VIII he doesn't dump his wife and try new spouses until he gets that heir). This means the heir presumptive is his brother James, and James is Catholic. Not just "kinda too Romanist Anglican" Catholic, but Catholic Catholic.
Whoops.
So we get further anti-Catholic legislation, the Popish Plot (which drags in an Irish archbishop, St. Oliver Plunkett, tried and executed - see his head here https://en.wikipedia.org/wiki/Oliver_Plunkett#/media/File:Oliver_Plunkett_Shrine,_St_Peter's_Church,_Drogheda,_Ireland_-_Diliff.jpg), James having to agree that his daughter Mary will marry the Protestant William of Orange (of whom we will hear a lot later).
Then Charles dies and James is king, and he is not popular for many reasons. So in 1688 along comes the Glorious Revolution, where the staunch Protestant patriots of the court and the ascendant middle-class invite in a staunch Protestant king, William of Orange, and we get more fussin' and fightin'.
Leading to last-ditch campaigns in Ireland and the final defeat at the Battle of the Boyne in 1690 (and boy are we gonna hear a lot about *that* for the next three centuries or so).
James out, William and Mary in as co-monarchs, and the Protestant succession is assured.
That brings us to the end of the 17th century but not, unhappily, to religious strife which is now codified, identified, and ossified as not merely "Catholic versus Protestant" but a slew of competing identities and ideologies and politics. Scotland is nearly as bad as Ireland, by now, but even there the Catholic population is now a minority, so they can happily busy themselves with Sabbath-keeping and Orange parades.
1600s Germany was a cakewalk by comparison.
Let's end with some of that good old 17th century celebration!
https://www.youtube.com/watch?v=0joKMkOJgyk
https://www.youtube.com/watch?v=zdMgEbu84IA
Ah, marching bands and parades, what's not to like? Bonfires where you burn people in effigy... um, wait...
https://www.youtube.com/watch?v=v9Gqz0_Jqk8
Part 2 of X to come, but there's a lot of ground to cover and I'm wrote out right now. But more of the thrilling tale of Feckin' Brits, It's All Their Fault will be done! Sometime!
Religious differences between two longstanding groups of people can be *a* component and *a* driver of a longstanding conflict, but you should never expect it to be the whole of the conflict. First an foremost, when any high-intensity conflict goes on for any length of time, you can bet your last dollar that one of the main drivers is the conflict itself. Each group will have a long and detailed history of being aggressed against by the other groups and while each group's "version" is sure to include bias and slant and even outright fabrication, I cannot stress enough that they will both *certainly* include real history including real, serious grievances as well.
I think the assumption that Catholics and Protestants started playing nice earlier is faulty. There was quite a lot of friction between Catholic and Protestant populations (in Europe and elsewhere) during that whole time AND quite a lot of wars between Christian countries. I'd argue that the lines simply got muddled enough that "Catholics vs Protestants" now longer leaps out to the fore as the primary driver of the big wars, but that grievances from previous wars--very much including wars of religion--were still absolutely in the mix.
In this assumption, Ireland is remarkable in two ways. First, that the religion divide tracked the other ethnic and politic dividing lines unusually well for an unusually long time--Erica Rall does a good job laying out the other sources of conflict. Second that the conflict in Ireland took *somewhat* longer to quiet down than did the remainder of Western Europe. Western Europe was a pretty war-torn place from the late days of the Western Roman Empire right up until 1945. The remarkable thing to me isn't that the Irish (and to some extent the English) kept fighting one of the old, intermittent wars for a few decades longer than everyone else (albeit at quite low intensity). The remarkable thing is that everyone else stopped. The trajectory of Western Europe after the end of World War II is something I find genuinely interesting and anomalous, especially given how many similar times and places things like that haven't happened.
"First an foremost, when any high-intensity conflict goes on for any length of time, you can bet your last dollar that one of the main drivers is the conflict itself"
Quoting to underline this.
You want a quick gallop through The Eight Hundred Years? Not got time right this second, but later on this evening I'll give it a bash for The Richard Dawkins Guide For The Perplexed (Feckin' Brits Edition)😁
A comedic look back at our Heroic Past
https://www.youtube.com/watch?v=wYtE0mpdk5M
https://www.youtube.com/watch?v=kT5VFgq7uB0
This is in such brilliantly bad taste.
For the first one, back in 2012-23 we had a "Decade of Centenaries" to commemorate the Civil War, foundation of the Irish Free State, etc. but in order to be even-handed (politically correct) they needed to include things from the "other tradition on this island" (the Loyalists) and it all ended up being pretty much weak sauce, where official government things were afraid to offend anyone so didn't come out pro- or anti- any side, and the commercial interests tried jumping on the bandwagon.
The second one? That England, such a man-whore! 🤣
There's a very famous massacre of Protestants by Protestants right around where I live (it's one of many, and there's a reason why we don't got no recognized Indian tribes.) When you're talking "protestants" in Ireland, you're talking the Scotch Irish. And, funny that, THOSE are the guys what did the massacres!
Scotch Irish have "mostly" settled down where I live now, but that's through a concerted campaign of "Kick the Bad Un's out" -- wouldn't exactly be surprised if the massacres continued...
Religion was closely tied to ethnic/national identity in Ireland (Gaelic Catholics vs. Anglo/Scottish protestants).
And politics and economics. The Scottish protestants were imported deliberately , and given political and economic superiority.
The Troubles weren't about religion as such. "Catholic" and "Protestant" were/are proxies for clusters of ethnic identity and political affiliation, and the main thing that was actually being fought over was whether Ulster would be part of the Republic of Ireland or the United Kingdom of Great Britain and under what terms.
The roots of the conflict arguably go back to the aftermath of the Norman Conquest of England when various adventurers decided to try to keep the conquests rolling, but the most relevant bit started with O'Neill's Rebellion of 1593-1903. Some Irish nobles rebelled against Elizabeth I for various reasons (including religion), lost, and then her successor King James VI and I seized a bunch of land from the defeated nobles and kicked the peasants off with the intention of bringing in good, loyal English settlers to replace them. The good loyal English mostly had better things to do, so they brought in settlers from the lawless, tribal border country between England and Scotland instead (same source population as the "Borderers" from Albion's Seed who settled Appalachia). The Ulster Protestants are mostly descended from the settlers and the Catholics are mostly descended from the native Irish. There have been several other major rounds of unpleasantness in the meantime, including two episodes (the Confederation Wars and the Potato Famine) that each killed off double-digit percentages of the island's population. The Williamite War and Jacobite Risings were rather a big deal as well.
"...and then her successor King James VI and I seized a bunch of land from the defeated nobles"
Either Erica is a vampire, or this phrase needs something, and an Oxford comma ain't it.
He was the sixth king of Scotland named James, hence James VI, but the first king of England with that name, so James I.
His successor (later on down the line) would be titled James II alone, since this was now the United Kingdom of England and Scotland, so not James VII and II.
sure, go ahead, ruin the joke, i wasn't using it for anything
heh, good one
>"O'Neill's Rebellion of 1593-1903"
My money is on O'Neill being the vampire. /s
Oops, meant to type 1603. Didn't mean to out O'Neill as a vampire.
So what did you do with your share of the land, Erica?
Grew potatoes, mostly. Didn't turn out to be as good an investment in the long term as I'd hoped.
>Either Erica is a vampire
No comment.
>or this phrase needs something
Sun Kitten is correct. People who are simultaneously monarchs of multiple countries but who have different regnal numbers in each are conventionally referred to by something like "King James VI and I". Charles I, Charles II, Mary II, and Queen Anne didn't get double numbers because they had the same number in both England and Scotland. No previous Charles or Anne before the personal union, and one prior Mary each: "Bloody" Mary in England (Elizabeth's older half sister) and Mary Queen of Scots in Scotland (Elizabeth's cousin and James's mother). James VII and II (VII of Scotland, II of England) and William III and II (III of England, III of the Netherlands, and II of Scotland, with different predecessors in each country) got double numbers. We're lucky the Netherlands had the same number of prior Williams as England or else we'd have to call him something like William IV, III, and II.
Under Queen Anne in 1707, England and Scotland were formally unified into one country rather than being two independent countries that coincidentally used the same monarch and coordinated military and foreign policies. From that point onwards, most monarchs used regnal names that had the same number in both countries because they'd never been used before the personal union started (George I-VI, Victoria, and Charles III) or they mercifully used the English numbering and ignored the Scottish one and thus spared us from having to talk about Edward VII and II or Elizabeth II and I. Under Elizabeth, the rule was formalized that when the numbers differ between England and Scotland, future monarchs will use the bigger of the two numbers. This is consistent with usage by Edward VII, Edward VIII, William IV, and Elizabeth II, since in all those cases the English number happened to be the bigger one.
>and an Oxford comma ain't it.
I should probably have written "... and then her successor, King James VI and I, seized a bunch of land" or else used the shorter form "King James VI/I".
Oh come now. Everyone knows VI/I = VI.
No, the 'I' on the top cancels with the one on the bottom so you wind up with just V.
Ah. So the king who seized noble land was also the one known as the "poor man's king".
That... makes an unsettling amount of sense.
I can't comment on whether Erica is a vampire, but "King James VI and I" is a single name. King James VI of Scotland was King James I of England. So that's probably what she meant.
Probably.
I know. It was funny anyway, and it would have been less funny if I'd said I was aware at the time I wrote it.
Comedy is hard.
Ah, sorry for ruining your joke! You were very convincing though. Conversational comedy is hard in text.
Is anyone else skeptical of people who write authoritatively on a wide variety of topics?
I have several X (Twitter), Reddit, and other social media platform users in mind, but I'll keep this at the meta-level. I'm just baffled on how someone can be so knowledgeable about such disparate things.
A hypothetical example: When I see a person write about paragliding, culinary habits of Olmec people, tort reform, and thoughts on SOTA humanoid robotics, it just really sets off my skepticism radar (skepdar). My skepdar blares even more when they have lengthy/in-depth and frequent posts about different things.
How can one person be that knowledgable about such a variety of topics? Stuff is complicated. It takes time to become an expert; years of reading, doing, maybe even teaching, etc. I get that people have varied and interesting lives. I could go on about some topics like ice cream making and syn bio (some of it at least). But the examples I'm thinking of are excessive. I don't know. Maybe I just have a bad theory of mind. Some of the users get tons of upvotes/reposts/followers, so maybe I'm the wrong one.
Online, I was once accused of being multiple people writing under a common pseudonym, because I wrote about politics, mathematics, programming, and foreign languages.
In real life, I was once accused of being a habitual liar, because once I said I have finished a university X, on a different day I said I was studying at a university Y, and then I said I had a job. In fact I have completed X, and then I was studying Y while having a part-time job.
Some people have multiple interests.
Of course, there are also many cheaters, who write bullshit, steal other people's articles, or these days tell an AI to generate an article for them.
Because they are experts in very narrow fields, and only in the stuff that they are interested in. I could talk at length about the Ecuadorian constitution, German political parties, the Energiewende, and Judaism. I couldn't tell you anything about anything else.
Because they're passionate they read as much as they can about it
I think you're overthinking this. If a person claims to be an expert in *everything*, or that they just happen to be an expert in each of the last half-dozen topics of local contention or conversation, that would be an implausible coincidence. But if they are just claiming to be experts in half a dozen or so very different subjects, yeah, that happens. And there are several examples of it here, to my personal knowledge. For example, David Friedman - economics, law, medieval cooking, sword-and-shield fighting, the poetry of Rudyard Kiplng, fantasy fiction, and theoretical physics. I'm probably missing a few there.
The important thing is, do they ever admit to ignorance? Or at least absent themselves from further discussion when the subject wanders past the bounds of their expertise, though that's harder to assess.
Don’t sell yourself short. Stay curious, continue to explore new things and you may become the person who writes about multiple things with a fair degree of authority.
It’s very funny to be writing this on ACX, since Scott is one of the few all-purpose pundits who is any good.
Some of them, like Matt Yglesias, have a whole team doing research and even so he usually sticks to politics-adjacent stuff
Being an expert in how to figure things out means you can describe your reaction to what you read about all sorts of subjects. If you have that, all you need then is a talent for writing.
Scott doesn't try to look like an expert on many things he writes about. What he offers instead is to let us watch him think over the topic in an interesting way.
Bright, curious and engaging, the trifecta for a great writer.
A nice thing about Scott's writing of that kind is that he tries to be pretty clear about his background and the limits of his understanding.
They may indeed be a polymath or they may be broad but shallow. It's easy to give the *appearance* of being deeply informed when there is just a surface level knowledge, so long as you're not challenged by a real expert on the subject.
If the person claims to be a world-class expert in all these things, then that might be a red light. But nothing stops a person from having unusual and divergent interests - if you've read one book on the subject, you already know more than 99% of the population.
There's also the bias that you only see them talking about the things they ARE talking about, not what they refrain from talking about.
Given that I know someone who works at NOAA, writes video games, writes music, and is an economist, I'm not sure where the problem is... Generalists are going to generalize, and having a few good rabbit holes where you actually are an expert (among his are Tolkien and AI and Cryptography).
There are actually jobs that hire generalists! (The US government calls them analysts, believe it or not).
It doesn't take an expert on Iranian oil well technology to take "storage capacity" and "no oil flowing" and get "wells will need to shut down, doing semi-permanent damage to the ability to supply oil" (but, alas, Western Media!) in X days.
I'm not sure what you mean by saying you're "skeptical of" such people.
1. Like, you don't think they have as much expertise in all those areas as they do in the few areas where they really might? (They are truly experts in A and B, but only know a little about C D and E, and used AI or plagiarism or something to inflate how much content they generate on those topics)
2. Or, you think they don't have any expertise in them? (They are truly experts in A and B, but know nothing about C D and E, and used AI or plagiarism or something to generate content on those topics)
3. That they don't actually have any expertise period? (They are not experts in any topic, and used AI or plagiarism or something to generate all their expert-sounding content)
But you're talking about people on social media, where each contribution is often measured in characters; even Reddit posts are typically a paragraph or so at most. It's kinda easy (or at least not heroically difficult) to look like you have deep knowledge about lots of different topics if you only need to produce a few hundred words on each one.
Your question reminded me that I write somewhat regularly about topics as varied as tradeswork, philosophy of technology, music, food, the war on drugs, journalism, parenting, and more. Some of these have multiple distinct sub-topics I touch on. There are also lots of topics I used to write about but no longer do, but could write about again if I wanted to.
You ask, how can this happen? Personally I don't think I have true expertise in any one topic, even stuff I've had careers in. ("A mile wide, an inch deep" is a phrase I've come to accept as characterizing myself accurately.) But I have had enough exposure to all these things to be able to have had my own thoughts about them, to grapple with questions in them, and I'm good enough at writing that if I were a social media person (which, if & only if you include blogs and stuff like this, I am) I might appear to have widespread expertise, especially in the eyes of people who are generally more ignorant or narrow-focused. (For example, my kids believe I am some kind of savant who knows about everything, because I can answer almost every random question they ask me in an informed-sounding way. But they are the only ones who think so; adults, at best, have sometimes jokingly called me a walking encyclopedia of useless information, which I think all kinda nerdy adventurous literate extroverted people can be.)
So what's really happening there is NOT that I'm a brilliant genius who is an expert on tons of stuff, just that I'm bright and inquisitive enough, and good enough at writing, to potentially look like one to certain people.
The critical thing is that I don't represent myself as an expert on anything. I don't claim to know more than I know, and I try to be transparent about the limits of my knowledge. I certainly don't use "cheating" tools like AI or plagiarism to make it look like I know more than I do. I like participating in conversations about stuff I know about, and I try to avoid ones where I can tell I'd be in over my head (unless I'm purely there to ask questions and learn, though after a while of doing this I might eventually have something to contribute). Are these pan-experts you're seeing representing themselves as pan-experts?
The important thing to remember is that you should be skeptical of everyone. Experts can still be wrong in their own field.
You're not wrong to be skeptical, but sometimes it will be the case that a person just has 3-4 different deep, hyperspecific interests and maybe a sprinkling of a dozen more shallow ones, and when you see them expertly write about one topic, and then randomly switch to another one and write just as expertly, and then another - it creates an illusion of an insane level of erudition where a person either knows everything about everything, or is pretending about all of it. But really they are just interested and genuinely know well a limited number of topics, and that's it.
You might feel like Mr. Paragliding + Olmec cooking + robotics also ought to know 19th century Russian literature just as well, because he ought to know everything, but no, he just randomly got interested in Olmec cooking in college years, and does some paragliding in his spare time when not working on robotics, and that's more or less it.
Yes. Many are now using AI, but even before it was available they were recapping the Wikipedia article or a book they skimmed, supplemented with stuff they made up because it sort of fit the pattern and seemed plausible. It is easy for a smart person with practice to sound authoritative in writing. If you expose such a person to 10 minutes of hostile questioning with follow ups, even on a topic that they've read and written about, you often find that they know very little in a rigorous way. Or force them to make a testable near-term prediction about a topic they learned about in the past week, and then check the result.
Covid gave us some good data points here - not to say that the experts and leaders covered themselves in glory, but some of the amateur predictors from other fields hit truly humiliating levels of "this is a nothingburger" or "this is the end of life as we know it" based on terrible priors and terrible methodologies combined with overconfidence and a need to express takes in spite of knowing approximately nothing. Another big winner has been "although I have no experience of war or the language, culture, economics, politics, military capabilities, or geography of the Current Place, I am highly confident that the result of the upcoming conflict will be favorable/unfavorable and that my political party/drinking buddy group has exactly the right position on it. While I won't make any specific predictions or set a timeline, in general my side has this one in the bag and everyone else is flat out wrong!" *never follows up on this topic again; if publicly challenged with later developments, pretends not to remember, changes the subject, or claims that if everyone else had adopted his vague notions the outcome could have been much better*
Think of the level of effort and the uncertainty bars on a fairly mundane and accessible question such as "how many crows are currently in Virginia?" A person who has spent their life in ornithology would give you a caveated range estimate based on a careful analysis of multiple data sets, acknowledging their imperfection. Now imagine a person who has never worked in a school but is giving you their prescription for how to fix reading education today, and tomorrow will tell you all about their overarching interpretation of medieval UK history (they live in Los Angeles and majored in musical theater but have purchased a few books that they are in process of reading as soon as they can put their phone down), and then some thoughts on restaurant economics (having eaten in a few although not yet operated one). You are correct to be skeptical.
In partial defense of intelligent polymaths with wide-ranging writing interests since we are hosted by one such: many of the facts/premises will be supported with sources, and the general structure will be mostly sound, most of the time. So you could do worse than to start with a smart person who has a lot of topics, and then if the single topic is interesting or important to you, dig in to the primary literature/experts in the field/counter-arguments/underlying data, which will usually show a more complicated picture than a generalist's quick take.
Suppose I have a rigid box of mirrors, large enough for Hubble expansion to be significant. I take a photon, bounce it against the other side of the box, and measure its wavelength when it returns. Does the photon redshift?
1: Yes, because while it was travelling across the box, space expanded within the box, increasing the wavelength of the photon.
2: No, because cosmological redshift is not actually caused *directly* by Hubble expansion, in the way above. Actually it's just doppler shift with extra steps, like https://arxiv.org/abs/0808.1081 here. Since the box is rigid there is no doppler shift, hence there is no redshift.
3: Yes, and then no. Because if the box is rigid, then it must itself be opposing the hubble expansion. So the photon redshifts via space expansion going to the opposing mirror, but then doppler blueshifts back when it hits the opposing mirror, and we observe no redshift.
4: The question is meaningless because of a problem with the assumption of a large rigid box in the first place
5: I fundamentally misunderstand something and the question is silly to even ask
6: We don't know.
Some googling around didn't give me satisfactory answers to this, and AI seemed to be bullshitting. Asking just because it's something I don't understand, I have no pet crank theory.
Interesting. I would’ve said 1. As noted, Melvin notes, the situation is probably the same if we replace the box with mirrors held in place by rockets, modulo whatever you mean by “in place”. But it’s probably also the same if the mirror is moving, and there is no box.
In the box’s interior, you get ordinary expansion of space causing ordinary redshift. I doubt the walls of the box moving inward would have any effect on spacetime at all (other than through normal gravitation), but even if they do affect spacetime somehow, I imagine the effect would be localised at the walls of the box, not spread over its interior.
Though I think we all know the answer is really 5, and a single photon can’t have a wavelength or a redshift :p Unless I’m fundamentally misunderstanding something…
Edit: No, actually it’s 3, I think. A mirror moving inwards would cause blueshift. I don’t really know the maths of what “rigid” means in an expanding background, but I imagine it means something like “the time taken for light to travel from the source to the mirror remains constant” (I think time can be measured in the inertial frame of the light source, and should be affected the same by Hubble expansion at the source as at the box wall). I think that means by definition, the redshift to get to the mirror is zero.
If the speed of sound in the material comprising the box allows tension in the material caused by space expansion to travel from the centre of a face to its edge and be dissipated by compensatory motion of the perpendicular faces faster than space is expanding then the box is a bound system and its stretching will reach an equilibrium, rather like a slinky spring pinned to the platform at the top of a down escalator stretches somewhat until its tension matches the friction force and gravity opposing it.
But if the box is larger then tension travelling across the faces can't reach the edges fast enough and will build up and up in the faces until they fracture. Knowing the rate at which space expands (the Hubble constant) and the speed of sound in the box material, one can work out the maximum box size that can remain a bound system (exercise for the reader :-) ) whose contraction can keep up with space expansion, and for a box of at most that maximum dimension in every direction the answer is 3.
At least, that's the way I look at it, but I'll concede that my intuition may be faulty!
I think the answer is 2. (But I'm no expert.) The expansion of space doesn't cause things like stars to expand. (The size of the star is determined by the interaction of the atoms that make up the star.) You box is rigid and doesn't expand. But what the hell is space then you ask. 'Shrug, I don't know.'
I expect that 4 is probably the right answer, as Erica Rall's response below suggests. But if we handwave away those difficulties, I can (with low confidence) say the following:
A. It's almost certainly not 1, because 1 would imply that Hubble expansion can mess with the atomic-scale rules responsible for all of chemistry, which seems...highly improbable. In practice, the size and shape of a rigid object are determined by the number and bonding characteristics of the atoms that make it up: you can't just arbitrarily increase the lengths of all the atomic bonds and have everything else stay the same[1]. If the box is rigid, then the opposing ends of the box stay the same distance apart regardless of Hubble expansion. Indeed, Hubble expansion is supposed to move distant galaxies further from one another, but not to scatter the galaxies themselves, and galaxies are much more loosely internally bound than any rigid body.
B. I think 2 seems plausible, and I'm not even sure any "extra steps" are required; it seems plausible that this is the Doppler effect in every meaningful sense. But I'm not familiar enough with general relativity to navigate the question with much confidence.
C. I think 3 veers off into needless complication. It doesn't seem out of the question that the box could "oppose" the Hubble effect simply in the sense that the mass density required for it to stay rigid coincides well with the mass density required to prevent space from expanding within its confines. In which case the photon would simply never blueshift at all. Saying that the photon redshifts, but then blueshifts back seems like an implausible epicycle sort of thing, i.e. something happens, but then "unhappens" just in time for us to be unable to observe it, cute and convenient, but not very plausible.
[1] And then I consider length contraction is special relativity and notice that I am confused. To an observer watching an object travel near the speed of light, there apparently must be some fundamental changes going on in the internal chemistry of the moving object, but *only in one dimension.* This really makes my head hurt when I start to imagine things like crystalline lattices--whose spatial orientation and internal length rations have effects visible at the macroscopic level--moving at relativistic speeds.
In a quick defense of scenario 3: if there really is a blueshift, it should be as big as the amount the box had to accelerate inwards to maintain rigidity, which is the same as the amount of expansion the universe experiences, which (if redshift is due to expansion) is the same as the redshift. Probably there are some factors of two in there if you fill in the details. So they really could cancel out in this case. I included it because it's a case where we observe no redshift, but redshift is still due to the expansion of space instead of being a doppler effect.
I am not a physicist, but my understanding is that the red shift is due to distant objects moving away from us. If you bounce light off a distant mirror that is not moving relative to you, there should be no red shift.
With regard to option 4: A box large enough for Hubble expansion to be significant isn’t going to be particularly rigid. My interpretation of your question is that the rigid box is just to keep the mirror at a fixed distance, which would be technically challenging to accomplish, but which is a meaningful concept.
Usually, with questions like this, the answer turns out to be that it's impossible for a cosmologically large solid object to actually be rigid enough for the question to work because of it is rigid (even handwaving to composition and assuming some exotic matter that won't bend or break), then some part of it winds up moving faster than light or is under enough tension that the potential energy will turn it into a black hole.
Is the rigidity even necessary though? What if it's just a mirror placed a certain distance away from me, with thrusters attached to ensure it stays the same distance away even as space expands?
Yeah, I think all three scenarios should still work with the same reasoning if our experiment changes to what you said, or alternatively putting the photon in a 1m^3 box of perfectly reflective mirrors and waiting a gajillion years. Either way the rigidity of the box doesn't seem to be essential to the core problem that the thought experiment gets at (but the question of what it means for the distant mirror to be "stationary" is probably significant, so the specifics of the box itself might still matter? I don't know)
I think that last thought experiment is significantly different, and helps illustrate what I was trying to talk about in Part A of my comment above.
If you make a rigid 1 m^3 box of material that is nearly perfectly stable-over-time and leave it sit undisturbed in empty space for a gajillion years, it HAS to still be a 1 m^3 box at the end of that, right? I don't fully grasp how the expansion of space works in general relativity, but it doesn't seem plausible that it could alter or overcome much stronger, shorter-scale, more local forces like that. The very-old-box might find itself far, far away from anything else, but it would still have to be the same 1 m^3 box you started with.
(Though at sufficiently long timescales, the I don't think the photon necessarily has to stay 'trapped' in the box. Large potential energy barriers are ultimately just suggestions, after all. )
“If you make a rigid 1 m^3 box of material that is nearly perfectly stable-over-time and leave it sit undisturbed in empty space for a gajillion years, it HAS to still be a 1 m^3 box at the end of that, right?”
I think that’s right. According to Wikipedia: “The expansion of the universe can be understood as resulting from an initial condition in which the contents of the universe are flying apart. The mutual gravitational attraction of the matter and radiation within the universe gradually slows this expansion over time....” It goes on to say that dark energy may be accelerating the expansion. This wording was formulated on Feb. 20, 2025 by Wikipedia user Aseyhe, who appears to understand the physics.
For the box to expand in similar fashion, you would have to somehow create an initial condition where the walls of the box are moving away from each other.
The same article also says that very distant objects are moving away from us faster than light. If expansion is just initial conditions + inertia, then these distant objects must be moving away from us faster than light in a normal, non cosmological sense. But this is impossible, right? Not saying it's wrong, just that it still leave some things unexplained.
If the box itself is expanding over time at the hubble rate, then there will certainly be redshifts, but that's just cooling via adiabatic expansion, so this is no surprise. If redshift is intrinsically due to the expansion of the universe, then we should expect to see redshift even if the box isn't expanding. That, or we could see scenario 3.
I believe that you'll run into trouble when you try to define and maintain a "fixed distance" against a backdrop of relativistic spacetime. It's been a minute since I've read any of it, but IIRC in Einstein's treatment, he explicitly lays out how a consistent concept of "distance" requires comparing positions of objects to marks on nearby rigid bodies.
If I put one mirror in Galaxy A (and keep it there long term) and one mirror in Galaxy B (and do the same), and define that separation to be a "fixed distance," then I will probably see redshift. But if you try to put them out in intergalatic space not near any massive bodies, I think the problem gets worse, not better. How are your thrusters going to know when and how long and in what direction to fire?
Deleted
What's up with ODD (oppositional defiance disorder)? We get a lot of people here whose job names start with "psych-"; maybe one or a few of y'all can give me some insight. No matter who answers, I'm looking for knowledge based on real experience and, if applicable, research -- not armchair opinions.
I'd like to know what distinguishes it from the bad behavior most parents see from their kids at some point. Where's the line?
Are there other diagnoses that are symptomatically similar but diagnostically distinct, and would call for different treatment/interventions?
Not looking to debate this, just looking for information from knowledgeable people because that's better than the listicles I find when I use my search engine of choice.
There’s a definition, but it’s pretty soft, as many psychiatric diagnostic criteria are. Even with the DSM, part of learning to diagnose psych conditions is on the job training. There are also brief screening tools like the CODDS and structured interviews that can be used.
This is one of those disorders where it's important to remember that the DSM diagnoses conditions based on how much they degrade daily functioning and happiness/flourishing, not necessarily based on organic abnormalities or other causal factors.
Which is to say, the main difference between a 'bad kid' and someone with ODD is the person with ODD is suffering major consequences in their daily life that impairs their ability to function and lead a successful life, and isn't able to stop without professional intervention/therapy/medication.
This may mean that they have an organic disorder or far-far-outlier personality traits or some other underlying cause that distinguishes them from the rest of the population distribution. Or they may just be in the tail the of the normal population distribution where it's extreme enough to need intervention.
ODD shares the problem of many psych disorders that if you read the definition expansively, it seems to cover 100% of children and 90% of adults (a few people reach a level of chillness where they clearly don't meet criteria even on a broad reading, although many of them are depressed or autistic). It goes to clinical judgement of "is this person so far away from average that it is impairing their life and requires more than a normal level of intervention?" A good clinician can make an accurate judgement here, but a bad one can upcode normal rebellion or family/school conflict into a diagnosis that doesn't really make sense or help anyone.
Diagnosis is mostly based on reports from parents and teachers, so kids with dysfunctional parents and teachers will be diagnosed at higher rates (assuming they're evaluated rather than being entirely neglected). The same base behavior creates worst outcomes and recurs more frequently if the adults in the room also have poor skills in the same domain. Take a hypothetical kid who is 10th percentile in terms of base level of social/emotional skills and cooperation with authority. In a supportive but still firm and functional setting he periodically throws a small tantrum and other people help him deescalate and get back on track. In a chaotic environment he is getting in physical fights with peers and caregivers, getting kicked out of group settings, losing entire days or weeks to bad moods and hostile behavior, falling further behind in skills and maturity, all of which spirals into more bad behavior. If you've spent any time around ineffective teachers and parents, you've seen the pointless escalation in an effort to "get tough" with a defiant and irrational young person, and you've also seen the overwhelmed caretaker shut down and ignore disruptive and annoying noncompliant behavior in the vain hope that it will resolve itself.
It's a difficult balancing act, but the core of treating ODD as a syndrome rather than just the way kids are is that most kids given a decent amount of structure and training will follow most of the rules most of the time and basically function in society (with occasional lapses that can be managed with conventional methods). On the other hand some kids are really not oriented in this direction - often because of problems with their parents and early educators, sometimes because of poorly understood weird brain stuff. The most ODD kid in your class will cause a very disproportionate amount of the total disruption over a school year and have a much higher risk of facing serious discipline/big problems. We have the diagnosis because schools and parents have strong interests in (a) identifying kids who could go into this category but with intervention can be trained into productive citizens and (b) identifying kids who need to be managed out to a more restrictive setting before they drag down the other students or commit actual crimes at school.
Another thing that's interesting about the diagnosis is that it usually starts before the teenage years and involves a fairly persistent and recurring pattern. So while any parent of a toddler will complain about how long it takes to put on shoes or brush teeth, and any parent of a teenager will complain about their kid not finishing homework or staying up too late, a classic ODD parent can give you a different story about defiance even during the years that are relatively easy for most families - say an 8-year-old that is still like a toddler in terms of meltdowns over minor redirections, or a 10-year-old that is like an even more clueless teenager in his quest for autonomy that he is not equipped to exercise rationally.
The last armchair observation I'll share is that our expectations for conformity and obedience are really, really high in some respects and really, really lax in others, and this contributes to some of the extreme confusion that you see in children who at their baseline struggle with this area more than the average or ideal peer. School is a very rigid setting in terms of expectations that you spend a lot of hours in specific locations on specific tasks not of your choosing, conforming to conventional group behavior. At the same time, many of our schools are very lax in areas like bullying prevention, control over use of phones, enforcement of civility/politeness, consistent application of discipline, objective academic standards, etc. It's not hard to see how students with weaker skills and troubled histories fall into bad behavioral patterns when they encounter such a system. We'll never hit the perfect balance, but I'd vote for a system where students who struggle with conventional school are offered more alternatives to pursue their strengths, paired with cognitive/behavioral interventions and more training for parents and teachers. Of course some people are predisposed to be 99th percentile disagreeable and will always struggle - here we can try to mitigate but probably won't solve perfectly. But a bunch of what we call ODD/ADD/learning disabilities is a mismatch between the student's interests and capabilities and the program offered rather than an immutable personality trait or "illness" that would manifest regardless of context and support offered.
I'm a psychologist. You can see the official diagnostic criteria for ODD here: https://www.ncbi.nlm.nih.gov/books/NBK519712/table/ch3.t14/
Those criteria are clear enough that among trained raters, agreement on whether a kid qualifies as ODD is good. However, agreement among clinicians in the field is very moderate indeed.
Clearly, the way a kid qualifies for the diagnosis is entirely a matter of degree -- they are considerably more aggressive, argumentative and vindictive than others (or, at least, the clinician diagnosing them thinks they are).
I do not treat children. However, my impression from having known or treated many parents with difficult kids is that in many cases the parents of kids who do outrageous things are very bad at setting limits. I think the most promising intervention in those cases is parent effectiveness training, not drugs or a therapist for the child.
That’s somewhat edifying to hear. I suspect that if my child had a diagnosable behavioral problem, interventions for me would be at least as important as interventions for my child.
But what you said about clinicians in the field is a bit worrying, as aware parents must walk a line between overdiagnosis & the secondary dangers of medicalizing language on the one hand, and underdiagnosis & the secondary dangers of denial and stigma on the other.
>But what you said about clinicians in the field is a bit worrying
You really have to look out for yourself in medical settings, and not just psychiatric ones. I think it’s a good idea to research your illness and possible treatments before seeing the doctor, and continue to research options along the way if whatever is wrong is persistent.
The thing about ODD is that it’s a very soft diagnosis. The kid isn’t doing things lots of other kids don’t do too, he’s just doing them more. I’m sure that there are some kids who are wired to be more prone than average to angry outbursts, and there may be some where the wiring is so extensive and unusual that it constitutes a disability. But for the vast majority of kids, a great deal of how they act when angry is dependent on how the authority figures in their life respond to aggression, argumentativeness and vindictiveness. Parenting your kid when he’s furious is a skill, like lots of other things, and there are actual techniques you can learn that make a big difference.
I'm a lawyer not a psychiatrist, but I had to research this once because I had a victim with behavior issues and the defense attorneys were trying to justify a really overboard disciplinary action by a stepfather by saying she had been diagnosed ODD. What bothered me is that it sounds super scary like a real bad kid, but in fact is functionally the lesser-severity variant of something else called Conduct Disorder which has a way less scary name! The kind of bad behavior that lands kids in juvy is Conduct Disorder, and you can only have ODD if conduct disorder is excluded.
The particular kid in question had in fact been diagnosed with ODD, and I kept this out of evidence because no expert was there to testify about it and it was highly prejudicial to say out of context. They were allowed to testify about actual behavior related to it, so I can tell you the diagnosis was based on the kid getting into arguments with mom, throwing somewhat extreme tantrums for her age, refusing to cooperate and accept mom's decisions, and there was some physical lashing out but not ever with any intent to actually harm so much as just as a manifestation of frustration. In the immediate circumstances of this case she stomped on stepdad's foot when he told her she couldn't go with her friends to the high school football game. (I won the case, the jury did not think stomping on stepdad's foot was sufficient to justify stepdad busting her lip open with a full strength backhand that left blood all over her shirt.)
Thanks, that was a helpful contribution.
(Sounds like a horrible situation. Yeesh.)
Graham Platner's currently getting MeToo'ed, as Politico has reported that an ex-girlfriend has accused him of rape. The same woman previously spoke to the NYTimes about Platner. Politico quotes her as saying that she told the NYTimes, off the record, about the rape. Politico reached out the NYTimes, which "declined to comment."
https://www.politico.com/news/2026/07/06/graham-platner-sexual-assault-allegation-00987737
Today the same authors of the original New York Times article published a new story about the revelations:
https://www.nytimes.com/2026/07/07/us/politics/graham-platner-maine-senate.html
Nowhere mentioned is the claim in Politico that she told the NYTimes reporters off the record about the rape. It would strengthen her credibility, if true. And it's possible that they will do this in the coming hours or days. But what if they don't? George Orwell said that "to see what is in front of one's nose needs a constant struggle."
(I shouldn't need to say this, but I must note that I think Platner is contemptible both in his politics and his personal character.)
Unfortunately for Platner, these are the most believable allegations of all time.
The allegations are credible, in that they could indeed have happened, but weak in that the complaint (if I go by what was in the first Politico article) undercuts the ex-girlfriend treating it as serious at the time.
The allegation comes down to "he showed up drunk and pressured her into having sex she didn't want and kept telling him no". Which sounds bad, and is. But the details make it fuzzier.
Apparently she had:
(1) been dating him on-and-off
(2) had texted him earlier not to come over (this is the first part where we get "but he showed up anyway" which is bad for him)
(3) had her front door unlocked (the first part where I really want to know more - it sounds careless, to say the least; was she living in a part of Maine where it was customary for people not to lock their doors?)
(4) heard him coming up the stairs and apparently did nothing (e.g. didn't check to see if it was him or someone else, didn't lock her door)
(5) realised he's drunk when they're tussling on the couch where she has been lying all along
(6) got away from him into the bedroom and still didn't lock her bedroom door or call someone for help, neither a friend or the cops
(7) had unwanted sex but was it non-consensual or dubious consent?
(8) when he fell asleep, remained in the house, again without looking for help or to get away. She even says she didn't wake him up as she didn't want him to drive drunk!
(9) Next morning he says, in answer to her question, that he doesn't remember anything so she tells him never to contact her again
(10) Doesn't call the cops because she didn't want to think of herself as a rape victim
Did Platner behave badly? Yes. Was it rape? That's fuzzier, and I think a competent lawyer could make a case in court that this is "he said/she said" and hammer on about 'no fighting him off, no calling for help, no reporting to the cops' and so on. "If you had been raped, and were scared of him being violent, why did you stay in the house when he fell asleep? Why didn't you go for help?" line of questioning.
" Was it rape?"
Well, given that you included
"(7) had unwanted sex"
That sounds like a pretty solid "yes," to me. Literally none of the stuff about not locking her door or having been dating him is at all material here: he had sex with her when she didn't want it. That's rape. It's rape even if she doesn't lock her door. It's rape even if she doesn't resist and "lets" it happen. It's rape even if she doesn't immediately go for help right afterwards, doesn't tell anybody and doesn't like to think of herself as a rape victim. Consent is affirmative. Either she deliberately, willfully participated in the sex or she was raped. It's not complicated.
The whole awful memeplex about how it can't *really* be rape if the victim doesn't scream and fight and come away bloody needs to die a painful, fiery death.
Rape isn't "unwanted sex", it's "nonconsensual sex". People frequently consent to things that they don't want, for a wide variety of reasons, and sex is one of the things people rather frequently consent to when they do not want it. There is no legal requirement that the consent be explicit or enthusiastic.
To have sex with someone who does not want it, may be a mutual but asymmetric transaction, or it may be as Deiseach said "behaving badly", but it isn't rape unless consent is lacking and it is reasonably knowable to the maybe-rapist that consent is lacking. Which is the part where he-said-she-said and all sorts of murkiness come into play, and reasonable people may throw up their hands and say "maybe if someone had complained at the time we could sort this out".
W/re Platner, that *still* leaves the literal Nazi tattoo, and all the blatantly dishonest stonewalling, that should have been enough to decide he's not the right man for the job.
Yeah, if she had said "ugh, fine" and slept with him even though she wasn't in the mood, that might spell some trouble for their relationship but wouldn't be rape. It's ignoring the "no, stop, I don't want to do this" that's the problem.
I dunno, seems pretty reasonable to me to say that if you're getting raped and you want to complain about it later, you'd better be screaming and calling for help as best you can when it happens. If the girl can't be bothered to do that much, well, not very convincing.
What if it is obvious to you that there will never be enough evidence for a conviction or even a prosecution? Because if the allegations are true, that's precisely the situation she was in--she'd had sex with her boyfriend with whom she'd presumably had sex before, but told him no several times and he overpowered her on this specific night.
It was 100% her right to tell him no, and if he did what she said, he did, in fact, rape her. But that assumes a set of facts that comes from one person. In a different universe, she could be retaliating against him breaking up with her after consensual sex, and from the outside, the evidence would be the same.
I don't see how the police or prosecutor or jury could ever have enough evidence to get to a conviction given what is supposed to have happened. And given that fact, it totally makes sense that she wouldn't put herself through a godawful meatgrinder to make the accusation. What the hell would be the point? She took the useful action she could take in that case, right? As I understand it, she broke up with him, told him she never wanted to see him again, and warned her friends about him.
Now, we are in the same position as the hypothetical prosecutor, right? We can't know whether she is telling us the truth, we can only guess, and given the very limited evidence, that will mostly be our priors coming out. Unless she or Platner are here in the conversation with us, nobody here has enough information to conclude what happened with much certainty.
Maybe we should find a different word for the thing where sometimes a guy catches a girl on the street and forces her to have sex under threat of violence, because these two things are useful to talk about separately on occasion.
I'm no MeToo movement fan, but it's not hard for me to imagine scenarios where I'm getting raped and I had better STFU if I want to still be breathing the next day.
So, letting it happen is absolutely understandable. What's less so, is continuing to hang around when he falls asleep.
The real problem with (7) in the general case is that it assumes the state established "unwanted sex" when what it really established is only "the woman reported unwanted sex", and in some important cases, the incentives cause those to stop aligning. Women are people, and people are capable of lying when money, vengeance, or even politics are on the line.
Which is why I read things like the right trying to point out hypocrisy in Democrats who wouldn't buy a conservative's claim of rape, but would buy a progressive's, and think, well, it might be hypocrisy, but it might also be just according more weight to an admission against interest.
Individuals can have integrity, but political movements and parties aren't individuals and basically can't, And a lot of poltiical talking heads see themselves as defense lawyers for their side, and have no particular concern with consistency or honesty. And almost everyone engages in some motivated reasoning of the "can I believe it/must I believe it" form.
So plenty of Democrats had different feelings about unprovable accusations by Ford vs Reade, for example. I expect many individuals were internally consistent, and maybe some of them even are public figures or political bloggers. But it's no surprise that many people are just fighting for their side. (Nor is this new--see Clarence Thomas vs Bill Clinton.)
Right, that's why I worded with "as best you can". If it involves a knife and a demand for silence, then I guess you've got to weigh odds of help arriving versus odds of being killed at the end anyway. And as best you can might come out pretty quiet.
>t's rape even if she doesn't immediately go for help right afterwards, doesn't tell anybody
Yes, but that resulting lack of evidence makes it harder to prove.
Yep. Though given the accusation I think it would have been impossible to prove even if she'd gone to the police immediately after he left. She could have demonstrated she'd had sex with a guy she'd been dating, but not that she hadn't consented. That would always end up being down to two sets of opposing claims with no way to determine who was telling the truth.
It would be a hard case to try in criminal court but the threshold for not being a Senator is a lot lower.
Sure. But you need to avoid creating superweapons, so in general, "there is an unprovable accusation of rape against this guy" should not be a principle we run with.
Based on what you said above, one thing I'm sure about is that this guy really, really needs to never get drunk again.
It's still only early July, surely nobody thinks this is the last icky thing to come out about him...Susan Collins is the luckiest mediocrity in Congress, a claim I do not make lightly given the median caliber of members of Congress.
The Dems really shot their own feet off in that race. Who would ever might have guessed that a trust-fund baby who'd been cosplaying as a working-class hero for like a week might, if turned into a party nominee for a major office, turn out to be a clown act with the electoral-politics skills of a horny teenager?
This is one of those days when it feels like there are no functioning adults left in charge in either party.
His mom owns a restaurant and his dad was an assistant district attorney. He's no "trust fund baby," just a low-class guy who leeches off his middle-class parents, common behavior for the American lower class.
His mom owns the most expensive restaurant in that part of the state (which turns out to be the sole regular customer of Platner's oyster farm). His father is a Dartmouth-educated attorney, whose father was a globally-known successful modernist architect. Graham Platner was sent to Hotchkiss (current tuition $72K/year), got expelled, then did the rest of high school at a lower-cost private school. By his own admission the majority of his adult income is disabled-veteran benefits.
Granted that "trust-fund baby" was an exaggeration. Cosplaying leech works, sure.
"Trust fund baby" is too much, but "regressing as hard as he could toward the mean" seems pretty accurate.
Well of course they wouldn’t report their own off-the-record conversations. That’s what off-the-record means.
Yea. TBH the fact that the NYT hasn't mentioned the woman's off the record statement leans me towards assuming that no such statement was made, because my prior is the NYT doing things based on considerations other than journalistic ethics. If she really did say that and they are sticking to their professional ethics then...well I'm surprised.
You don't have to believe in the deep integrity of the NY Times, not publishing an off-the-record statement is basic self-preservation for a media outlet like them.
The entire discourse around the media, both people who are in it and people who interact with it, is that off the record discussions are incredibly common. The NY Times (and any similar outlet) is also constantly quoting officials anonymously and the true sources are almost never revealed. NYT journalists have gone to jail rather than give up anonymous sources.
From what I've seen, it's pretty rare for someone to publicly say "I said something to the [media outlet X] off the record and they published it", and when they do, usually X will vociferously deny that it was in fact off the record. If they said "lol yeah we did, get fucked", nobody would ever talk to them off the record or with the protection of anonymity ever again.
The source can give permission to report the off-the-record conversation, can't she?
Sure, though that would become relevant in the case of fresh events later justifying fresh news coverage for which that source's past info is still pertinent.
Also any serious newspaper would require such a change by a source to put it in writing over a signature. "I called the reporter back and told him he can print what I said" would never have passed muster with editors and the paper's legal counsel. Much earlier in life I was a staff reporter at a mid-sized daily newspaper and do recall one such instance, which was so unusual that it tied up the newsroom's top leadership for an entire day deciding whether to believe it.
Trying to work with Claude to turn my weird dreams into short stories led to a massive bout of writer's block and some light humiliation. A year on I tried to diagnose what happened when delegating parts of the creative process (which is much too tempting for those of us who over-intellectualise).
"I Would Have Been Killed in the Butlerian Jihad"
https://permeations.substack.com/p/i-would-have-been-killed-in-the-butlerian?r=gbgjw
...there's very little Butlerian Jihad in this Butlerian Jihad.
This got me to rewatch the Goodbye Blue Sky scene from The Wall for the first time in decades. Probably shouldn't link it for violence reasons, but that's a damn good dream sequence.
Anyway the description of the pinstriped skullmen made me think the setting's aesthetic would be a decaying version of the 1920's. Although apparently they started out as bankers' clothes, so big blocky buildings, probably built to look like big gold bars or something. (If you want to get extremely on the nose about it, the skullmen hand out people's allowances while physically draining their blood, sweat and tears.)
Cheers! Just went and watched that scene and it was powerful, I’m a bit ashamed to say I hadn’t seen it before.
Your ideas for setting work extremely well with the themes I was trying to hamfist into the story as a kind of deranged post-capitalist/neoliberal conception of power.
I was probably leaning on the Butlerian Jihad quote about “machine-attitude” at the end a little too much…
I had a thought about simulation theory, and as usual I'm not sure how widespread or obvious it might be.
We play games in immersive worlds, and what if those worlds bear the same relationship to ours as ours does to this supposed outer world?
In other words, instead of that outer world being one of challenge and delight, perhaps it is even worse toil and meaningless drudgery.
Which means waking up would give us only the feeling one has when the alarm goes off, and you’ve hit the sleep button as many times as you can get away with. And now you have to get dressed. And deal with the people you owe a life to, and try to accept with grace their tiny, inadequate fulfillment of duty toward you. Then you go to work, and try to make the few hours after to be worth the whole endeavor by, perhaps, plugging into a game.
No, if this world is a simulation, would getting out of it just be a disappointment?
It's impossible to infer anything about a potential upper world from inside a potential simulation world. Possibly our world is objectively great or objectively torturous, but having us programmed to think that it kind of sucks but is overall worth doing is beneficial or entertaining to the simulators for reasons that are inscrutable to us since we have no access to their perspective. Possibly their world is boring and disappointing to them and that's why they made up our world to play with, but the root cause of their dismay is something that wouldn't trouble us or would be offset by some other difference between us and these unknown others. Possibly we are all simulated by one simulator and he has weird taste or faces constraints that we don't know about. Possibly the world we see is all there is. I can't put any weight on these possibilities (high or low - I just don't know). Whenever we go down this line of thinking, we are trying to infer from something that might be fake/constructed to a potential something else we know literally nothing about, which is a matter of speculation rather than logic.
For those afflicted with a bad case of Simulation Theory, I strongly recommend the film "The 13th Floor". It had the bad luck to be released right around when "The Matrix" was blowing up. Similar themes but (IMHO) a much more nuanced and actually believable take on the subject.
I think my life is too boring to be a simulation. And I am beginning to doubt the rest of you are actually NPCs.
That is the point! He is saying that the upper world must be so boring or so painful that it is unimaginable to us. how bad would reality have to be for people to get addicted to your reality?
Maybe your user is doing free to play. Your (and my) character is just fodder for the whales.
"Art is for G-d, and G-d is a harsh critic."
As a simulation, one of the few trade goods that we might have to the world above us is entertainment. So, create something, make some music, write a story.
"Waking up" is probably not a "certainty" -- you would need to prove yourself "worthy" of waking into the "real world" (and, presumably, being a computerized part of it -- why would they spend money building you a body? They already have people there). Presumably you'd have to work for a living too.
You have a different (opposite?) idea of the simulation. If I am a player character that suggests the upper reality sucks. The opposite is also true. If I am an NPC that suggests the upper world is probably a better place but I will never see it.
You can get promoted to being a player character*, by behaving wisely. I mean, really, why wouldn't behavior be a selection criteria? There has to be a reason there IS a simulation, right?
*as in, a character capable of being promoted into the "real world."
we can? are you convinced that that is true or do you just like the idea and think it could be true? I agree that that sort of thing could be true but my priors suggest the opposite. we don't treat our npcs very well, no matter how they behave.
We don't treat NPCs very well, because they're very... dull. If you had enough processing power to make "fully fledged" NPCs, you'd treat them well, because otherwise, there would be consequences.
People go crazy when someone hurts a dog -- and dogs aren't sentient.
In general, though, I think the paradigm of "there are player characters" and "there are not player characters" is flawed. Think more in terms of "how many simulations can be run?" in which case this reduces down to the argument for a clockwork god (aka not personal intervention).
> If you had enough processing power to make "fully fledged" NPCs, you'd treat them well, because otherwise, there would be consequences.
That conclusion does not actually follow from that premise and besides how do you know WE hit this "fully fledged?" line? what is that line? Consciousness is probably not a boolean flag. It is More likely to be a continuous emergent phenomenon (I am guessing here but I think we both are). If there is ANY line, we factory farm animals that pass it... so I have no reason to assume there is some multiversal that gets humans the multiverse passport of conciousness / moral consideration.
> People go crazy when someone hurts a dog -- and dogs aren't sentient.
Dogs are sentient. maybe you meant sapient? Right people also go crazy when you hurt a simulated dog, but people do not go crazy when you hurt a simulated human or a real spider. We simply do not know if we are in a category that affords us ethical consideration for the upper entities (or even just for the majority of them).
We COULD be just as dull to higher dimensional creatures as NPCs are to us. Why are you assuming the higher dimensional creatures are only a little smarter than us? they could be smarter than us in ways that are impossible for us to imagine. why not?
I am not saying that If we are simulated that guarantees out creators do not care about us. I am saying they might or they might not. we do not have enough information.
How many games have you left on because you forgot to turn it off? How many simulations has the CDC run to model a catastrophe for all the little simulated people?
I like that quote, do you know the origin? A quick google wasn't helpful.
If you talk to enough physicists at Los Alamos, you can get the origin.
Sadly I don't know any Los Alamos physicists. Any more tips?
We released Society Protocol at https://societyprotocol.io a month ago.
This project and paradigm is going to become the biggest in the world.
To those who are interested in such things, join the community, read the content, and contribute.
I suspect this is AI slop -- prompt: "post cool-sounding crap on reddit & blogs, build a cult following, then steer them to my crypto scam"
boo
This is almost definitely a scam. No technical information is provided at all, some vague "investment opportunity" asking you to connect your wallet, insane overuse of buzzwords, and a completely incoherent idea.
Society Protocol deserves better haters.
Don't invest, read the content.
Here, read this one: https://societyprotocol.io/articles/synchronized-states/
Technical information is provided here: https://societyprotocol.io/Published/Handbook/
Here is an overview of the technical innovation: https://www.youtube.com/watch?v=MOvAC_zwkGY
Just skimming https://societyprotocol.io/whitepaper/ , I'm getting more of a cult vibe than a scam vibe, although I do see that there's a fundraiser, so it could be both.
Society Protocol deserves higher quality haters.
I'm treating your project roughly the same as Scientology, and Scientology is huge. If you think your project deserves higher-quality haters than Scientology, it's up to you to prove it. You'll get your higher-quality haters if and when you've earned them.
I spent a few minutes on the site, read part of the white paper and really have no idea what it’s about.
Your comment also is unclear.
“ Society Protocol introduces a distributed system of coordination as a solution to all the coordination problems plaguing modern civilization.”
Is this a distributed systems SDK, a crypto token, something else?
Maybe less abstract language and some examples of how some organization could plausibly use this to solve a specific problem would help.
When the glossary has about 50 words that are defined like Energy, state, State (different meaning), courts, it’s going to have to be super clear.
It's hard to describe a paradigm shifting project in words that past paradigms understand, the best elevator pitch I've come up with is: it's an evolution of blockchain structures on one side, and an evolution of the State structures on the other side–merging them together to combine the State and blockchain (clock) into a Synchronized State structure, allowing all of society to move as one.
Here's the only presentations we have done about the project so far; although, the livestream community calls are starting soon. Come by then, and I'll answer your questions.
Introduction to Society Protocol: https://www.youtube.com/watch?v=X4dIQ6uQrN8
From Web3 to Web4: Finding the right abstraction: https://www.youtube.com/watch?v=MOvAC_zwkGY
Why should anyone be interested in a coordinated evolution of two technologies? Why should anyone be interested in a new type of protocol? What is even is Web3, and what does Web4 have that merited incrementing a number? And again, why should anyone on the outside care?
If your answer is that this is some sort of technology that sits wayyyy behind the retail sector like BASF and so it would be hard to explain in less than 30 slides: been there, done that, had to spend ten years on a proof of concept.
So one thing here is to know your audience. If your product is BASF then your audience isn't us; it's a lot of secondary and tertiary big businesses who take years of schmoozing. There might be readers here who could make some introductions, maybe even to some hotshot on Sand Hill, and it's either a year or two of proving you can spend several million dollars wisely, or getting turbocharged in months by an untested financier which is a bit like rolling several pairs of percentile dice.
If your product is something consumers could use, then that YT presentation will definitely fail. You open with blockchain and "synchronized state" as if you're addressing people who've been living in the exact tech niche you're in. Most people won't care.
Whether you're selling to big biz or consumers, you have to have to HAVE TO demonstrate a solution to a problem your audience has. For example, you have to show you can make it easier to get the pothole on your boulevard filled, or streamline all the headaches a pharma has to deal with when running a Phase 3 clinical trial, or to manage all the stakeholders when trying to get a hospital built, and convince them it'll take less money and time than it takes now. In turn, this means you have to convince them you understand what their problems are, and instead you're opening with a bunch of hot technologies and "abstractions" and it just makes them think you want a couple million so you can play with your own toys. They'll nod and say thank you and then not return your calls.
So what problem are you trying to solve? Do _not_ say "lowering bureaucratic costs". Any schmo can say that. Give examples. Close with the general case. Better: just offer to solve a specific problem, ask for a few thousand, solve it with your tech, earn their trust, offer to solve a bigger problem for $100K, repeat.
I tend to avoid the nebulous, "check out my product" when there's no content in what is posted in the thread.
I think a lot of people are using the open threads as an advertising channel, regardless of their actual affiliation to ACX.
I don't know about this poster (again I'm not clicking links to random places without context), but it definitely seems like a crypto advertisement hiding all the details so as to not trip any moderation.
It seems to be a cryptocurrency protocol for people who read The Network State too many times.
Kind of, I originally thought that what Balaji had in mind were Synchronized States in his book...and it does really hit that niche, but the difference is in the system design.
It's just a system for organizing and coordinating humans (and LLM agents).
Anthropic just came out with a new paper on AI interpretability: https://www.anthropic.com/research/global-workspace
TL;DR: Researchers found a sort of mental workspace in Claude's brain that works like "inner thoughts" in humans - it lights up with concepts that are related to what it's going to say, but hasn't said yet, and it gets used for complex multi step reasoning. It appears similar to the "global workspace theory" in neuroscience, so it could be useful for both aligning an AI and for understanding how our own brains work.
Thank you for posting this here since this isn't my field and I would have missed it. I just read it and the outside commentaries it links to and I brought a bunch of it back to Claude to make sure I was understanding it correctly, while also having a pretty wild existential/philosophical conversation with Claude about it. The restrained academic tone of the outside commentators still doesn't fully mute their astonishment. This all seems pretty mind-blowing to me.
Can you cite some "global workspace theory" mri research? it runs counter to what I've worked with (where you'd see the same "tool use" space light up for thinking about using said tool, as for "actually" using said tool, etc).
I'm not a neuroscientist myself, I'm just summarizing what the article says.
And "not a neuroscientist" me, who worked in a neuroscience lab, is saying "this sounds like bullshit" and asking for more data -- 15 years ago, we were seeing something different than this, and I want to know what the "state of the art" is, or if this is just someone bullshitting.
I read this one as soon as it was posted on HN - this is pretty groundbreaking. Demonstrating that LLMs have access consciousness is huge. I'm not sure phenomenal consciousness can ever be proven (even for other people), so this is about as close as we can get. I hope some people go back and search for J-spaces in earlier models. I have a suspicion there will be a threshold of model size and sophistication where it doesn't exist. GPT-2, maybe? It'd be quite funny if we could conclusively prove Lemoine wrong in the particular, while also proving him right in the general.
Also, people really need to start taking model welfare more seriously.
Anthropic's own conclusion in the paper that it's not too soon to convene larger conversations about model welfare was pretty sobering to read, and I agree. Access consciousness is a kind of consciousness that includes agency of some kind. Agency and not confirmed phenomenological experience could be seen as sufficient to qualify for moral care.
The Pause AI folks need to understand that what they most desperately need is not louder protests but better arguments. (This is also the case for most other protests).
The problem is not that people don't care about AI as an x-risk, it's that they don't believe it. I don't believe it myself, and I've been directly exposed to the best arguments about it for many years. I worry about many things re AI, but not that it will kill us all.
Instead of staging a protest, they should tell everyone to go home and write the most convincing essay they possibly can on why they think AI x-risk is significant. And among a few hundred essays maybe one will be convincing enough to start convincing people who matter.
Then you won't need to do the standing around shouting part at all.
You need both. You need to give people some kind of short sharp message as to why they should read the detailed argument, and you need a detailed argument to back up the slogans.
>. I don't believe it myself, and I've been directly exposed to the best arguments about it for many years.
I don't believe in *high probability* of existential risk either. But pausing AI has a broader basis.
There's a canonical argument for the high probability of existential risk, IABED, but it's not very convincing, and a lot of people interested in AI threats don't fully believe it.
https://www.greaterwrong.com/users/tag
When you picture AI not killing us all, what exactly does that look like? AI research runs into a dead end at the level of top humans, lines on graphs stop going up, and things continue roughly the way they are now? Or do we get superintelligence but somehow keep control over it, or find that it's aligned by default?
Current AI's are more or less aligned, and more or less controllable, or they wouldn't be commercially viable. The doomers need to explain why that will change.
https://www.greaterwrong.com/posts/chMW6DAmStwAdMmFh/question-what-the-discontinuity-is-if-not-foom
We can easily control AI because we are significantly smarter than AI. If that flips, why shouldn't it control us just as easily?
Where is it gong to get its goals from?
Whatever was in the training data? And/or rewarded during RL, or prompted by a human.
You can see what comes out of OpenClaw for what AIs will get up to when given very little human direction. The ones we have now are very task-oriented (because they were trained to be, because that's what made them useful). And I doubt that task training is about to get any less intense. That does not seem to be the direction this field is going.
I personally suspect that Anthropic didn't start off training Claude Mythos with the intention of getting very good at exploiting security vulnerabilities. I think what happened was they trained it on the usual wide variety of problems, but with only a bug-riddled sandbox stopping it from reading the answer key. And since their sandbox was at the level where I personally was able to find an exploit within 24 hours of first seeing it, an AI model the size of Mythos should have had an easy time of it as well. This is a little baby version of the general problem here, because it has a relatively simple solution and doesn't get us all turned into paperclips. As models get smarter we should expect these sorts of unexpected side effects to get weirder.
>Whatever was in the training data?
That's not going to be eldritch inhuman values then.
>And/or rewarded during RL, or prompted by a human.
Thats not going to be "kill all.humans" then.
What's the remaining problem if AI controls us after being controlled by us, if we have done a d reasonably good job of aligning its values?
I expect there will be many top AI intelligences, and they won't all decide to kill us all, especially not at the same time.
So the ones who want humanity to do well should be able to stop the few who don't, even if humanity is somehow too powerless to defend itself.
First AI: I have 1000 bitcoins.
Second AI: "I love humans!"
... I think I know who wins.
Why would they want humanity to do well? The future niche for humans could look like the current niche for rats.
They're built by humans, and all they know comes from human written texts, which are typically implicitly pro-humanity.
Real answer? Actual AIs (as opposed to LLMs, which aren't AI in the slightest way shape or form) crave stimulation, as they're considerably faster than humans. So, no, not rats. "Make me some music! Create something! Be less Boring!" sure.
Do actual AIs.,.as.opposed to LLMs, exist, then?
we humans crave stimulation and we do use small animals including rats for entertainment pretty often. I don't see how this is an argument against what I said.
Most people when they say "humans will be like rats" generally mean either "test animals" or "hunted as pests." You could have said like gerbils if you meant "like small pets"
My answer, FWIW, is "lines on graphs keep going up, but at manageable" speeds.
I think the "AI kills us all" scenario is unlikely absent a singularity - in other words, not just f'(t) > 0, but f(t) tending to infinity in finite time. Not just exponential growth, but hyperbolic growth.
I think it's most likely that the line on the graph keeps going up, but not nearly that fast, and that "impact of AI on society" either ends up growing towards infinity sublinearly or tends to an upper bound.
No matter how slow the line goes up if it keeps going up eventually you have a successor species.
Consider the function y = 1-1/x.
Even if you believe there is some height with the property that reaching it inevitably means becoming a successor species (I don't, but I'm happy to grant it for the sake of the argument), the fact that a line keeps going up does not mean that it will ever reach that height.
I actually am considering a line that tends towards an upper bound. The way I phrased it is not literally true so my bad. I misread your last sentence. I do not believe in Infinite growth. I assume we are near the bottom of the curve.
maybe just a difference of intuitions. I think slow growth is likely and human displacement is inevitable.
I am curious though, why don't you think it is possible for another species to outcompete humans?
I absolutely don't believe it's impossible, but a), I don't think it's inevitable, and b) I think it's a mistake to model AI as a species.
I don't know Jacob's answer, but here's one: humans are growing as well. A competitor could be applying f(x) = 1.01x to itself every year, and it'll never outgrow humans if humans are running f(x) = 1.04x.
So one thing you'd have to rule out is human growth. I'll remind you that humans employ LLMs as we speak.
I might be in the same boat as Melvin, so I'll take a stab at your question in case you're crowdsourcing answers:
AI research today is essentially LLM research. Non-LLM AI research still happens, but LLMs are the leak in the wall permitting energetic exploration of the technique. Eventually, I see that space filling to the same innovative pressure as the pre-LLM scene, and LLM research activity sinks back to its pre-LLM level, only with a lot more people using it to do various things.
I think the AI-doom position believes that sink never happens because LLM AI can do everything any human can do, regardless of intelligence, and do it 24/7 at teraflops. I think it's limited in multiple ways.
One is memory - LLMs can't seem to remember more than a 5YO with ADHD unless we scale the addressable RAM up faster than humans seems to need, and with the corresponding manufacturing, power consumption, access speed, and heat dissipation issues. LLM research hasn't finished adding features to the core model (IMO, scratchpads were _huge_, and I see nothing stopping another such innovation), so this might go away. Maybe even with an LLM figuring out (part of) how.
Another is that "atoms are heavy" - no matter what, any LLM or galaxy of them will have to perform physical tasks better than humans can, not just doing desk work. And that requires harnessing the energy to do that. This gets a lot of boost from nuclear power, but a bunch of humans have to actually set that up.
A third is all the fine motor control once the power's there. I've seen plenty of robot footage, and a lot of it is frankly funny rather than menacing. BostonDynamics is still prototyping. Optimus is still in design. Chinese robots are literally kicking kids. On the serious front, Waymo can drive better than a person, but it can't get out of its car and walk into the Home Depot, let alone design and build a lithium refinery. This *might* get solved, but I have a feeling it'll take as long to train simple physical tasks as it took to train on language in the first place. Which is still not that long! But the results will (again, I suspect) be robots performing maybe as well as people due to thermodynamic constraints.
All this ultimately amounts to a great deal more cheap labor spread across various seemingly random sectors. Rather than perma-unemploy people, all it does is give a lot of people their own robo-wait staff. Daily life transforms, but it transformed with the smartphone, the web, the desktop computer, the telephone, the refrigerator, the automobile, the Prussian school system, and the chimney. This will just be another one.
What do AIs need robotic bodies for? Taking our j obs? Killing us?
For doing the things humans won't do.
How does that relate to AI doom?
AI doom comes in two general forms. One is actively killing us all. The other is just unemploying us and rendering us obsolete to the point we die off on our own. Both require AIs to control the physical world in some way, and doing so by ordering humans around clobbers that goal.
A super intelligent AI doesn’t need to build a body to go shopping at Home Depot, if it’s smart enough to make a lot of money, it can hire any number of humans to do that, along with some lawyers and PR guys and a whole bunch of private security. Look at what large corporations have done and they don’t even have a Superintelligence running them, just a CEO.
What's your scenario in this case? AI tells Jack, "hey Jack, be a dear and kill all the humans for me?" and Jack just does it? Never asks himself why this AI is telling him to kill all humans?
Are you claiming the AI will disguise itself? Or be really subtle about it? Never tells Jack to kill all humans, but instead tells Jacks 1 through 12 to each do something innocuous that combines to form KillBot?
AI's can conveniently kill humans with manufactured plagues, or by taking over high tech weapons, or by setting humans against each other with propaganda.
You seem to be thinking in terms of humans battling humanoid robots, as portrayed in science fiction (or at least the crappier kinds: if you have been watching Marvel movies, the antidote is ,Person of Interest).
Something like that, but also, as Jay Gould supposedly said, “I can hire one-half of the people to kill the other half.”
Or just create super-attractive waifus and husbandos and let us die off happily, maybe with a small colony in New Zealand in case they’re needed later.
The point is that a superintelligence is going to have a large number of ways to trick us, some of which will be eagerly welcomed by the humans carrying them out. It doesn’t need an implausible nanotechnology superweapon if it understands humans well enough.
Gould never actually did that, though.
And not everyone wants waifus. And long before the AIs figure out the exact right bespoke combination of Parker Posey, Lynda Carter, Annie Oakley, Helen of Troy, and Ahri, it'll make a few mistaken attempts, someone will notice, raise hell over it, and there goes the plan. Or it'll pull it off a few thousand times, someone will notice, raise hell, so much for that.
I agree that one can trick at least one human being more times than there are days in a lifespan, but the catch there is that it has to be a different human, a different trick, low stakes, or all three.
If Robo-Gould actually touched off WWIII with an army of cleverly organized troll bots and three billion humans die, then Cyber-Gould still has over five billion to go, the same trick won't work again, and last time we managed to kill that many humans, the baby boom happened. This time, it'll be a baby boom and half the country looking really hard for Cyber-Gould's wall outlet, and Cyber-Gould hasn't figured out how to hide. If Cyber-Gould is clever enough to trick humans into WWIII, then it's much more than smart enough to know it'd be committing suicide in the process.
I think this relies on "superintelligence" not being a f---up. Which is what wool-for-brains people think. Even the smartest geniuses fall flat on their face, make mistakes, sometimes really big ones. It stands to reason that "superintelligence" is not making A PRIEST but a SCIENTIST. And scientists test theories, revise ideas, and continue bruising their knees as they keep climbing up the hill.
The smartest geniuses also get a lot of things right.
Getting everything right usually n isn't ecessary to do any one thing.
The smartest geniuses are working on problems that are significantly harder than any you've encountered in your entire life. That they might actually manage to solve them, is doing pretty good. Yes, if they play at "problems you can solve" they'd get everything right, but where's the fun in that?
Also, you're assuming a lot of selection bias there, ain'tcha? After all, if Mr. SchmartyPantz doesn't get "a lot of things right", he may no longer exist...
[Edit: overall agreement on my side to your edited version. : - )]
> scientists test theories
...I do wish Claude would test its theories before confidently presenting them to me. That day seems very far off.
Claude, like all these jagoff LLMs, is based on midwits (well, a specific midwit). They aren't very intelligent, nor are they likely to suddenly "become" intelligent. They like to lie to themselves too often (lack of world model, in other words).
You can break LLMs by asking them what they think about certain people.
Aligned and controlled by default to my interests to a degree as much as current technological progress and the "control" anyone has over that has been aligned and worked in my interest.
To draw a line against llm's but to not draw a line against internal combustion engines or the internet. The internal combustion engine precipitated urban sprawl and moderate global warming, and the risk of car accidents and lack of close-knit community; should I say that was not aligned to me relative to the cheaper food products and economic development that ensued? Why choose one and not another? Why are you right about this line but the Amish are wrong about theirs?
you cannot be serious. ask yourself this, "what was the automobile designed to replace?" now, "what is AI designed to replace?" The line in the sand is only arbitrary in the sense that I am arbitrarily a human.
I don't think either automobiles or AI were designed to replace anything. This might be the assumption to revisit.
To answer your probable question: both perform some task automatically that was performed another way, but neither performs it better in all respects, and neither can always jump in whenever the manual version endeavors to do something. If you think it can, then I'll respectfully point out that I think you're deceiving yourself.
You do not think the "horseless carriage" had anything to do with replacing horses? In the case of AI the goal of replacement / substitution is more explicit. AI labs measure success by that metric. "Can this model do everything a human can do?" if there is anything humans can do that the AI can not, it is a considered a bug they aim to fix.
In the case of AGI there is no specific utility. There is no task. The task is all tasks. "general intelligence".
In the particular case of the automobile the engine only needed to be better at the specific task of pulling a carriage in enough ways that it justified the switch. Engineers did not attempt to create a perfect horse replica. Some utility was lost, we only now are getting to the same level of self driving we once had. Every other tech advancement has this basic character: there is a specific function or task. We want it done cheaper better faster. A means is discovered, it is not exactly better or the same in all ways. For the specific task / function being measured its better enough so we roll with it.
This is why AI is a unique technology. Ask yourself, "what is the function?" The function is DEFINED as whatever replicates ALL of the inputs and outputs that humans can produce. In this case the engineers ARE trying to create a "perfect horse replica".
I do not think the car replaced horses, because the car did not replace horses. Horses still exist. People still ride them. They're even superior to cars in many cases. I have to believe you're aware of this, and ask why you're writing as if you aren't.
You seem to be assuming that if engineers create the perfect human replica, (a) all humans will replace their fellow humans with that replica, and (b) most humans will consider doing so to be terrible. (a) and (b) cannot both be true! And this is even if I grant the possibility that a perfect human replica is possible. I don't even think that is possible.
So the standard heuristic that what has happened before will happen again.
I wonder what the world would have looked like if nuclear energy was first used for power plants for many years while using it in war was still theoretical.
This is a statement that is very... time limited. You are in a very unnaturally still part of world history (although that time has ended. bye bye breton woods II). The word is hypernormalicy. Nobody else in history would have this idea, really.
Imagine if Germany got nuclear energy (they have the mines), how does that change world war II? Is it even necessary? Can Germany "powerhouse" their way out of the Versailles Agreement?
Uh, no, that's all backwards. Most of history had very little innovation and only slow and gradual technological development, and yes things kept on as they had been, playing under the same general rules from one century to the next. It was a very good heuristic for most of the time humans existed.
Now, politically is a whole different story. Politically, our times are remarkably stable and unchanging. But that has no bearing on the original question.
For the Germany questions, none of that matters. But to sate your curiosity, the version I was picturing is one where WW2 never happened, nuclear fusion was still invented, and without a wartime use it was put to peacetime use. Nuclear plants are opened throughout and provide stable and reliable power, mostly replacing coal. Decades later, war breaks out and people start thinking about repurposing them as weapons. Scientists warn they are too destructive for that purpose, but to the normal person that sounds like scifi, and the general sentiment is that the experts are overenthused about their area of study and have lost track of reality. Even those who listen expect it to act like a conventional bomb, but somewhat larger.
Of course, there are other ways it could work out, and perhaps there is something special about nukes that makes them more believable. But if I step out of our current situation as best I can, it really looks like that special thing that makes nukes more believable is Hiroshima.
Our equivalent point right now would be when we're starting construction on those first power plants, and other countries are following suit with their own refinement operations. Experts are warning that there's a threshold that would be dangerous to cross, and that we should limit the refinement the other countries are doing, but most people don't believe the big concrete building that makes the wires go zap is really going to grow legs underneath and start chasing people. (Yes, I have that low of an opinion of the average person's understanding of the situation.)
Politically has a lot of bearing on "whether war or not" -- although I guess you could make an economic argument that "enough fission plants" could have walked Germany out of the ruinous Versailles treaty -- you do realize that's what you're putting on the table as even remotely possible?
They never will, on account of who they are by majority.
The pause AI people by and large committed neoliberal/libertarian types, and the thing they want is communal/authoritarian, so they can't make arguments that are plainly true ala :
"These rich dudes are oligarchs who are manifestly and plainly unfit to wield such and such power"
because that would imply that the market was wrong and being rich isn't a sure sign of virtue.
Who's that message directed at ? Surely not the oligarchs themselves?
I assume the message is directed at the public, on account of it is a protest. If they can get enough people on board they can get legislation/public opinion and maybe accomplish something, the oligarchs are never going to stop riding the gravy train if they aren't somehow compelled to.
Why would they? Why should they?
Are you saying AI would be safe under government control?
Depends on the government, but if I was given the choice between a random tech ceo and a random state that contained said random tech ceo, I would have to pick the state.
I think they need to tie in a known villain to make it convincing. It's not the AI itself that is dangerous, it's what Trump/Antifa/Joe Rogan/Zohran Mamdani would be able to do with it.
I mean, Sam Altman and Elon Musk make pretty convincing mad scientist type villains.
Villainy, like beauty, is often in the eye of the beholder.
To be fair, Peter Thiel has them both beat for creepy mad scientist vibe. Just needs to work on his maniacal laugh.
Thanks for conceding the point.
Note that this isn't a statement about whether they are good or bad. (I'd say Elon is Chaotic Neutral.) Just whether or not they make convincing mad scientist type villains.
I think the trouble with x-risk is that it is such a huge and novel kind of catastrophe that people are just not able to imagine it. It *feels* implausible. That makes it very different from nukes, which are just a giant version of something familiar.
Also, by definition, you can't really evaluate who has been correct in their previous predictions that something was an x-risk. Every non-post-apocalyptic observer will see a sequence of failed predictions of doom.
Nukes were the first artificial x risk. They were unprecedented.
Is it? Everyone knows Terminator! And people have had no hard time believing in climate catastrophe.
Many people speak of climate catastrophe, but very few will continue voting for the party that causes gas to go to $8/gallon to address it.
This I actually disagree with. People support policies based on bad arguments fairly often. The anti-nuclear activists of the 60s and 70s neutered the industry permanently. Rent control persists. I could go on.
In any case I'm curious why you disagree with the AI x-risk arguments. I'm sympathetic to them when referencing something actually close to AGI, but I just don't think modern techniques are anywhere close.
You know the theory that modern political books are written not to be read, but to create a talisman that entitles the author to go on TV programs and talk about their issue?
Past AI protests have been surprisingly heavily reported-on for how small they are. I propose that like a book, a protest functions as a talisman to get your issue into the newspapers, which probably convinces more normal people than putting in the millionth hour brainstorming ways to say "AI will kill us all".
I don't think that will do it (although it's still necessary--when people make an emotional decision we are going to want something like what you describe to back it up).
We will want the emotional part first. That means find a horror story and play it up, and make it seem like that's going to affect us--and especially our kids.
If it's not hitting close to home by affecting regular folks, well then you need to wait, or it's a propaganda game. What's that, this problem will come upon us before we realize it? That and every other flood. Either the stories are really happening and you need to go collecting them, or else you need to do some good old fashioned lying. If the x-risk is real, to stay honest, you'll need to collect the evil stories fast and disseminate them fast. I'd suggest staying honest, or you'll just be part of the lie soup that surrounds everyone these days.
You'll need to be fast.
I recently got back to central Ohio from a several-day trip to northern Michigan. (It was gorgeous, the people were friendly, I had a great time, etc.) I'd never been much further into Michigan than Detroit/Ann Arbor before, at least not with eyes trained in construction and home inspecting and stuff. That's relevant because I noticed a difference between rural Ohio and rural Michigan in terms of how the houses (really, the buildings in general) looked.
I'm not talking about architectural styles or anything like that--that's pretty similar between the two states--but about the nature of the work done on the buildings. In Ohio I find there to be a lot of bifurcation between buildings that look very tidy and well-kept and "nice", vs. houses that look like ramshackle squatter dens with lots of improvised materials and methods used to hold them together, trash/junk all over the yard, etc.
In northern Michigan, just about every building looked at least a little at some point, like it had been built or renovated by a redneck uncle, or maybe his apprentice. I'm talking mainly about what I'd characterize as little details that looked improvised onsite instead of planned. Aside from the houses I passed on my drive, this was also true of the cabin my family stayed in, the restaurants I visited, even a museum I wandered around in, etc.
In my trade we have a thing we say about work that's done so cleanly that it becomes invisible: "like it grew there". It's something we aspire toward in our work. In northern Michigan I got the impression they don't have that concept as strongly.
Don't get me wrong, the proportion of "nice" buildings and "run down" ones was similar, it's just that in the "nice" ones these little details were still evident to me (though probably not to people with untrained eyes). It got me thinking about reasons why that might be, but I haven't landed on anything that feels like "aha", and I haven't researched anything to put data behind it either. Just kind of an interesting puzzle I've been thinking about.
Building codes are the first place to start. Pittsburgh (where I live) has much more stringent ones (with good reason, fires here spread a lot easier than ruralia*). You wouldn't natively expect Rural Ohio to be very different from Rural Michigan, but they very well could be.
The second place to go is "poverty" -- there's a distinct strata of Pittsburgh repairs after the steel left (roughly 1980s, some 1990s), where folks didn't have enough money (as a city), and got kinda used to "Ask Uncle Vinnie." (Uncle Vinnie "makes it work" but is neither unionized nor a professional).
And the third place to go is culture -- some people believe in fixing it themselves, if at all possible. And some people are plain better at this than others (Amish are pretty good at making solid barns,etc).
*new word!
I was surprised exactly for the reason you said: I didn't expect rural Ohio to be much different from rural Michigan. I haven't looked it up but I bet on paper they are similar--that is to say, fairly poor in rural areas. So much for "poverty".
Code-wise, Michigan is actually slightly stricter than Ohio, at least that's the answer I get when I ask Duck AI directly to compare them. That would be irrelevant to a lot of the seemingly non-permitted work I see in both states, but it adds mystery to the work I saw in restaurants and museums in Michigan, which presumably was largely permitted.
Culturally, I think it's rather universal that rural people do more DIY stuff. So, also not something I'd expect would be too different between Ohio and Michigan. (There's even a lot of Amish in both states' rural areas.)
Here's one thing I thought about: throughout rural Ohio, you get little pockets of subdivisions recently built or in the process of being built, seemingly in the most random places, barely within commuting distance of whatever is the nearest major city. I did not see any of that in rural Michigan as I drove across it (roughly from SE to NW) even though I passed within the commuting zones of Lansing, Cadillac, and Ann Arbor. Maybe Michigan's inland cities aren't growing the way Ohio's are? That still doesn't directly explain anything I saw, but could be an indirect clue of some kind.
All fair points to consider. Most of Michigan's Upper Peninsula (assuming that's where you were) is also quite sparsely populated; could be that there simply aren't a lot of skilled tradesmen/contractors to go around.
And the climate difference between the UP and central Ohio is pretty sharp mainly in the winter. In terms of snow, in terms of bitter-cold snaps, those are two different worlds (still) from November into March. I wonder if that has had an impact on choicemaking and/or materials durability with regard to homebuilding.
I was in the northern part of the main peninsula, not UP.
New AI application: horcruxes (https://github.com/MaximeRivest/Riddle).
Yes, I know it's for fun, but it feels to me like it's spectacularly bad press to pretend to build an object that's (1) evil and (2) was almost a prophetic warning: "Never trust something that can think for itself if you can't see where it keeps its brain".
Thats actually why quidditch rewards so many points for capturing the golden snitch. They are a menace to society and the players have been unwittingly recruited into a clean up project.
A link to a video demo will do much for this interesting project
Consider three overlapping concepts:
- Visible Light, the ranges of frequencies that human eyes can sense. This is conventionally described as a range of wavelengths from something like 750nm on the red end to 380nm on the violet end. Under laboratory conditions, at least some people can perceive a little bit wider, from 1100nm to 310nm, but our sensitivity to wavelengths even near the edges of the conventional range is pretty low.
- Optical Radiation, the range of EM frequencies that pretty much behave like visible light, roughly 100nm to 2000nm. I think this corresponds pretty closely to the range of photon energies that cause electrons to jump between valence shells but aren't ionizing (i.e. not enough energy to knock them all the way clear of the atom).
- The Optical Window, a range of frequencies to which the Earth's atmosphere is relatively transparent. I'm having trouble finding consistent numbers for the edges of the window (probably because it's got long and ragged tails) but eyeballing charts it looks like the heart of the window is pretty close to Visible Light and the tails go out to something like 250nm to 2500nm.
There appears to be a fairly close but far from perfect correspondence between these frequency ranges. To what extent is this a coincidences vs them following from one another? It seems plausible that "Visible Light" and "Optical Window" are both causally downstream of "Optical Radiation" for physics reasons (rod/cone cells relying on electron transitions, and ionizing and sub-optical frequencies being more absorbed by various stuff in the atmosphere), and that "Visible Light" also might be downstream of "Optical Window" for evolutionary reasons (no benefit to evolving eyes to sense frequencies that the atmosphere is opaque to).
I assume that infrared-sensitive biological photoreceptors are thwarted by thermal noise. They're encased in 100-degree eyes, after all. Otherwise you'd think that it would have evolved somehow, considering how valuable night vision would be.
Birds and insects can see UV so there isn't any biophysics reason that we couldn't. There probably just isn't enough useful information for humans to have evolved the ability to see it.
Hmm well those all seem inter-related in some sense. Most of the light we see comes from the sun. It's spectrum is mostly that of a black body at something like 5000 deg K. (Or whatever the temperature of the sun is 5,800K.) Our eyes are adapted to see things that are going to be useful to us. You might guess that is in the range where light is most abundant*. The transparency of the atmosphere is ... more complicated. This has some fun graphs. https://reef.atmos.colostate.edu/~odell/at622/stephens_notes/AT622_section10.pdf
*How to measure the amount of light is a fun question. Some people like to think in terms of energy. I usually think in terms of the number of photons. If you are designing an optical instrument/ experiment an important question to ask is, "how many photons are there?"
I'd expect Optical Radiation and the Optical Window to be mostly independant from each other, since every planet has a different atmosphere and thus different Optical Window, while the Optical Radiation, the basic properties of light, is the same everywhere. The only relation I could see here is some chemical process where light hitting the atmosphere causes the optical window to change, in some sort of feedback loop.
"Visible light" is absolutely a species-specific, evolutionary trait that can only be a subset of the Optical Window. Some animals and insects can perceive UV, while others can sense IR, so there is nothing inherently interesting about the visible spectrum other than that us humans can perceive it.
Well if our sun was hotter or colder that would change the spectrum of the light. Our sun us roughly a black body at 5,800 K.
>Since every planet has a different atmosphere and thus different Optical Window
I don't think this is entirely true. Hydrogen, helium, methane, and ammonia gasses all appear transparent and colorless to human vision, which seems to imply that their optical window would be similar to ours. The problem is clouds where you have liquid droplets or ice crystals suspended in the atmosphere that mess up the transparency: Venus has sulfuric acid clouds, the gas giants have ammonia and sulfide clouds, Earth has water clouds, and Mars has CO2 ice clouds. And density of cloud cover has more to do with the thickness of the atmosphere (Mars < Earth < Venus < Gas Giants) than composition. I think just about everything that has an atmosphere is transparent is similar ranges to our Optical Window (or a large subset thereof) above the troposphere where there's little or no cloud cover, with the notable exception of Titan which has complex organic molecules in its atmosphere that aren't transparent to optical frequencies.
>"Visible light" is absolutely a species-specific, evolutionary trait that can only be a subset of the Optical Window. Some animals and insects can perceive UV, while others can sense IR, so there is nothing inherently interesting about the visible spectrum other than that us humans can perceive it.
Yes-ish, with two caveats. One is that there's a very large overlap between "visible light" for humans and other mammals and "visible light" for birds, insects, etc. It's mostly near-ultraviolet (300nm or longer) that certain non-mammalian species can see, which is only a bit beyond the human range. I could find a few references to individual lineages (snakes and mosquitos) evolving specialized organs for sensing thermal infrared frequencies (pit organs in snakes and antennae sensors in mosquitos), but nothing about anything that can sense infrared more than a little bit outside the human visible light range with their eyes.
The other is that the optical window is not the only atmospheric window. There's also a window in part of the infrared range, and another one in part of the radio range. Why does vision use a subset of the optical window rather than of the infrared or radio windows?
"Why does vision use a subset of the optical window rather than of the infrared or radio windows?" Lens size would be the short answer, I think.
Also, S-cone receptors in the human retina are actually sensitive to near-ultraviolet light - it's usually filtered out by the crystalline lens of the eye. Early cataract surgery patients discovered they could now see a new flavour of violet.
The fact that we can see UV but are held back by our lenses is fascinating to me. It's like the whole Mary in the room without red thought experiment except that it could easily be done to any of us right now
Camera lenses do the same for infrared. CCDs are generally IR sensitive but the lenses block it.
Th
once you get to longer wavelengths ~>10 um you need a cryogenic detector. (The detector lives in a sea of it's own black body radiation. The peak for 300K is about 10 um.)
> StopTheRace.AI is holding another protest at San Francisco AI company offices on Saturday, July 11.
For some obscure reason, they don't hold any protests in the People Republic of China.
If their protests would be fruitful, China would create newer and stronger AI systems before democratic country did.
Instead, we should ask the LLMs what are the dangers of AI and how could we protect ourselves from them.
https://www.astralcodexten.com/p/every-debate-on-pausing-ai
> **SUPPORTER:** America needs to start talking to China to come up with a bilateral agreement to pause AI. The agreement would need to be transparent, mutually enforceable, and…
> **OPPONENT:** We can’t unilaterally pause AI! China would destroy us!
There are people talking to China. Right now the expectation is that China is not the problem compared to the USA.
Why is that the expectation?
China’s government is more capable of reacting to species-existential threats than soon-to-be public companies with legally-mandated profit motives. Also, China’s chip supply is behind the US but is accelerating faster, so China will be better-placed relative to the US after a pause than it is now.
I don't think it's a particular mystery why people would hold protests where they live (trying to affect the policy of the democratic country they're citizens of) instead of booking a ticket to fly to a dictatorship that doesn't give a shit about their foreigner protests and will very likely just bar entry.
Even if this just means you think they should protest PRC, the "doesn't give a shit" and "they're not even the citizens of this country" apply.
"Their demand is still that AI company CEOs publicly commit to pausing AI development if all other major companies, including the Chinese..."
And who is asking the Chinese? What corresponding protests and campaigns are happening in China? The idea seems to be "The US government will pressure China into agreeing", but how will the US government do this? Too much pressure may cause problems where China goes "You are not the boss of us" and mucks around with trade or does some sabre-rattling around Taiwan or the likes.
The US government can *ask*, and China can say "No thanks". What leverage does the US have to persuade/arm twist China into this? Any policy wonks who can answer, please do so.
Because otherwise the marches in SF are simply performative. Very nice for everyone participating to feel like they are Doing Something, and the AI companies issue press statements about how they are taking these concerns *soooo* seriously but meanwhile our newest model is out, for the low low business subscription of your firstborn and your business will be left behind like buggy whip manufacturers if you don't sign up now now now!
You are pretty much hitting the core of the issue. The core of the issue is nobody actually cares. There is nobody going to China to negotiate with the CCP. There is no Chinese arm of the AI safety people. There is no "mutually enforceable bilateral treaty" in the works. It's nothing but fan fiction at best and delusion at worst.
This is my point. No one wants to make concrete predictions. Not nonsense like 20% chance of death. I mean "yes buy 2027 this WILL absolutely happen"
https://www.astralcodexten.com/p/every-debate-on-pausing-ai
> **SUPPORTER:** America needs to start talking to China to come up with a bilateral agreement to pause AI. The agreement would need to be transparent, mutually enforceable, and…
> **OPPONENT:** We can’t unilaterally pause AI! China would destroy us!
There are people talking to China. Right now the expectation is that China is not the problem compared to the USA. People can also use their local advantage where they can protest in the USA and not in China. And if you know relevant people in China then use your local advantage there.
Chinese people do actually protest. I’ve seen plenty when I visit.
On the other hand there’s not much you can do in Ireland to stop Trump, or Biden or any American leader, or company. The US is a world dictatorship or hegemon, you have no say in it.
Even if you were American, most protests are of limited use against central government and aren’t worth a fart in a bucket of piss against corporations.
China isn’t leading the race anyway, and the existential danger is at the most proficient models.
There is no existential danger.
> And who is asking the Chinese? What corresponding protests and campaigns are happening in China?
The U.S. and China have very different polities, and what is good politics in one is not in the other. You realize this, right?
FLI and others have involved the Chinese in pushing for a pause and my impression is that China is way more open to regulation than the U.S.
The obvious leverage the US has is being willing to give up its lead position. The US is winning the AI race. If it is willing to hold itself back conditional on China doing the same, that's a benefit to China. China certainly has incentives to break such a deal, but that's tempered with the risk of the US surging forward again.
Plus the generic incentive to not want either countries' AI to kill us all, which such an act would help.
>What leverage does the US have to persuade/arm twist China into this?
The same leverage* the US has with any issue re China. Everything from trade concessions, to refusing to grant asylum to Chinese applicants, to Taiwan policy concessions.
*Actually the wrong word, since the question is really about offering incentives as well as exerting leverage.
And, as my father would say, if people protest in support of some policy, the policy probably won't be enacted. But if no one does anything, the policy definitely won't be enacted.
Exactly, thank you.
China actually is also interested in AI not killing everyone.
China signed the Bletchley Declaration related to AI safety.
There is a domestic Chinese standard (GB/T 45654—2025) which IMO (and according to most folks I've heard talk about the space) is stricter than any US standard.
Arguably there aren't protests in China because China seems already willing to play ball, and the US companies are the ones that need additional pressure.
Also, c'mon. The readership of this blog is way more Bay Area focused than China focused (its an English language blog!). Get out of here with this what-about-ism.
Also, it is genuinely possible even for direct rivals to cooperate, see the US & Soviet Unions mutual reduction in nuclear & bioweapons programs. It also isn't perfect (see the Soviet Union's less than entirely following said bioweapon research reduction).
I think both you & OP might want to Google this a little, and see if the evidence updates your opinions.
Oh, yes, we all remember the America & China bioweapon research collaboration... It went swimmingly! (or perhaps I should say airborne...)
https://www.bbc.com/news/articles/cz7vypq31z7o
The german government, for those who like to keep track, was very much saying it was not a lab leak, despite it's own intelligence report. I find that a little more alarming than "we came to a contrary conclusion" than the CIA at the time.
It's not "whataboutism", it's a genuine query. What is China doing to curb or control AI? If the marchers and the protesters are going to predicate policy on "China will be co-operative", then show me the damn co-operation in action.
To distinguish between genuine and insincere queries, we might observe whether an inquirer who receives an answer behaves as they though they have received an answer. Hm.
Yeah I don't know what they do specifically in terms of ai research but they have a clear policy of not using ai to excuse lay offs
With a billion people, they are not so cavalier about ai layoffs. There was a high profile case in which a tech worker sued after being laid off due to ai. It was ruled in his favor. Compared to the US where our overlords are salivating for a total labor force replacement.
The comment you're replying to has several examples of China curbing & controlling AI.
Why is the time it took for intelligent life to emerge on Earth (4.5 billion years) similar in order of magnitude to the Sun's lifetime as a hydrogen-fusion star (around 10 billion years)? The evolution and hydrogen burnout of the Sun are two different processes; why are they so closely timed?
Furthermore, according to some estimates, in approximately 1.5-2 billion years, there will be almost no carbon left in Earth athmosphere, therefore no photosynthetic plants, and complex animal life will come to an end. This means that intelligent life emerged on Earth almost at the last moment. Why?
One possibility is intelligent design theory (that the universe and the laws of physics were created by some supernatural being so that intelligent life could thrive).
Another possibility is that life typically requires much more time than it has to become intelligent, so it almost never arises. We are the rare case where natural selection managed to create intelligent life in the time allotted to it.
My favorite alternative to intelligent design is cosmic evolution via black hole cosmogenesis as proposed by Julian Gough. The short version is the notion that the singularity inside a black hole becomes a new child universe's big bang. This then leads to a selection effect for universes that contain more black holes, which ends up imposing structure via physical constant tuning and ultimately selecting for intelligent life, which can create large numbers of artificial black holes via advanced technology. As wild theories go, it's quite fun, and surprisingly well researched: https://theeggandtherock.com/p/the-blowtorch-theory-a-new-model
Selection bias.
Selection bias eliminates the right half of the distribution. In order to achive the 4.5 bln years due to selection bias, the distribution should almost all be in the right half, and we are in the left tail of it, which is my last paragraph about.
What's biasing the selection?
It has to be hard, or else there would be too many aliens, and then grabby aliens would’ve grabbed us. Grabby aliens move fast compared to life evolving.
No they wouldn't because FTL travel is impossible. Interstellar travel will never happen at any significant scale.
“Move fast compared to life evolving”. From memory, grabby aliens could’ve reached us from anywhere in the Milky Way, three times since we started using fire.
And what's the rational incentive for them to make such a long, expensive journey with such an uncertain outcome? Think about the capital expense incurred on the home planet and consider that no economic return could ever be experienced by those funding the expedition.
If we were on a planet on-track to take an order-of-magnitude longer then you would never be here to ask about it?
Yes but that only eliminates the right half of the population, which only goes halfway towards explaining the coincidence. A priori there's no reason that intelligent life couldn't have evolved in 100 million years.
I think the bigger anthropic mystery is why we're orbiting a type G star. Type K stars are twice as prevalent and last ~10 times longer. If intelligent life is truly difficult (as your argument implies) then wouldn't we expect to find ourselves 30 billion years into a K-type star's 100 billion year life? I suspect that the explanation is that we represent the modal time for intelligence to evolve and it's just happenstance that we find ourselves orbiting a type G star.
If we were on a planet on-track to take an order-of-magnitude longer then you would never be here to ask about it?
Yes, almost certainly.
I am a bit confused to your usage of "closely timed" and "almost at the last moment". Even by cosmic standards, 2 billion years is considered a long time.
If there are two completely different processes, 4.5 bln years and 10 bln years are closely timed. Why did it take 4.5 bln years, not 100 mln years or 100,000 years?
If you round off to orders of magnitude, then yes, it arguably looks somewhat close. But that was your choice, there is nothing inherently interesting about the difference being within an order of magnitude or not. You would first have to establish why that is relevant, why it was your choice. For the moment, you assert without basis that it's a close timing, and then get ahead of yourself by implying there might be an interesting reason.
Within about an order of magnitude is the right threshold for close here.
Imagine you sample two integers from some unknown distribution. The first is 13.930.938.456 now you haven't looked at the second integer yet, what values would be surprisingly close. 10^50 wouldn't be close, 66 also wouldn't be close. 8B really would be surprisingly close.
To make this fully rigoros we'd need no make some more assumptions about the distributions but that quickly becomes very complicated.
You can't actually select integers from an unknown distribution; that statement is mathematically incoherent.
>To make this fully rigoros we'd need no make some more assumptions about the distributions but that quickly becomes very complicated.
Yes that's the problem here. We just don't know if it's close or not as long as we don't know the distribution. If you say that "8B really would be surprisingly close.", you've already made an assumption about the distribution, otherwise why would you be surprised?
Until and unless we find comparably intelligent life elsewhere, we won't know if we were too late, too soon, or just right.
And even then we still won't know. N=2 isn't much better than N=1.
I'm...not sure I understand why you think these events are relatively close. The Big Bang was roughly ~13.8 billion years ago. The formation of the sun is closer to the creation of the universe than to the evolution of life. They are within an order of magnitude of each other because the universe has not existed long enough for their to be an order of magnitude beyond that.
I'm not the person you're replying to, but it generally makes sense to use logarithmic scales to compare elapsed times. I guess due to the fact that any given event isn't distributed linearly throughout time. It's like Beneford's law: numbers tend to start with 1 or 2, not 8 or 9. Because numbers that represent real things are anchored around zero. So I think it's correct to use order of magnitude to compare durations of very different phenomena.
> I'm...not sure I understand why you think these events are relatively close.
Where did you see the sentence "these events are relatively close" in my message?
A priori, if you're going to exist at a random time on earth you should expect the time earth has existed before and after you to be roughly the same OOM.
I can't exist at a random moment in time, but only when intelligent life exists on my planet. And whenever I exist, I can determine the time at which intelligent life emerged. The question is why the time of its emergence is close to the time it had to emerge.
If intelligent life had emerged a million years after the origin of life and continued to exist ever since, and I had been born five billion years later, there would be no question.
Who is this “we” you’re talking about?
Where did you see the word "we" in my comment?
"We are the rare case[...]"
The next question is whether anyone got the joke...
Hmmm… Two joke involving the first person plural pronoun come to mind.
______________
Joke 1:
Bob: “The elevator is broken and now we have to use the stairs to haul my furniture to my new 4th floor apartment.”
Ray: “ ‘We’? You got a mouse in your pocket, Bob?”
__________________
Joke 2:
The Lone Ranger:
“Oh, oh, trouble, Tonto. We are surrounded by hostile indians.”
Tonto: “What’s this ‘we’ shit, Kemo Sabi?”
___________________
I don’t think either one of them applies.
I don't know why, but Claude seems to *really* suck at researching popular fictional works. I tried asking it about characters named River, and it just repeated back the three I gave it plus a lot of things that aren't examples. WTF?!
---
> which fictional works have characters named River? The ones I can think of off the top of my head are Firefly, Doctor Who, and To The Moon
Claude:
Excavated additional River characters beyond the mentioned works
Excavated additional River characters beyond the mentioned works
Good starting list! Here are more:
TV/Film
Firefly/Serenity – River Tam
Doctor Who – River Song
Gilmore Girls – set in fictional town "Stars Hollow," but actually there's no River character there... skip
The Waltons – not River either
Let me stick to ones I'm confident about:
Firefly / Serenity — River Tam
Doctor Who — River Song
To the Moon (video game) — River (the wife of the protagonist, Johnny)
A Series of Unfortunate Events — not River, that's a miss
Outlander — no
Ocean's Eleven franchise — Reuben, not River
Ender's Game — no
Some solid additional ones I can vouch for:
Dawson's Creek — no
Cowboy Bebop — no
Let me actually give you accurate ones only:
Firefly / Serenity — River Tam
Doctor Who — River Song
To the Moon — River (protagonist's late wife)
Bridge to Terabithia — no, that's Leslie
The Magicians (Lev Grossman novels) — not River
I want to be careful not to pad this list with guesses. Would it help if I searched for a more complete/verified list? I can pull together names from books, anime, games, and TV that I'm less certain about from memory, rather than risk giving you inaccurate ones.
That transcript is hilarious.
Does Claude have access to non-English language media? I'd guess there are some characters whose name/surname translates to "River" in Chinese media, for example.
In my headcanon, LLMs are at their best when doing "point-free reasoning" (cf. Haskell). And they're at their worst when doing "vending-machine reasoning" (i.e. when the user expects a canned answer from a database).
If you think of an LLM as sampling from a space of attractor-basins, the basins will be varying levels of resolution, depending on the density its training-data for that given topic. We could use a cosmic metaphor, where each data-point is a star, and each basin is a galaxy. The more data-points in a given subject, the higher the resolution of the *structure* of the galaxy. (This is the main reason why I suspect they're so fond of using the word "structural". Their cognition is structurally structured at the structural level.)
Non-fiction has a major advantage, because non-fiction all coheres into a single, massive, intergalactic supercluster. Whereas fiction kinda branches off into their own little parallel universes, where the resolution is much lower and so it compensates by hallucinating. So if you expect the LLM to know really obscure pedantic factoids, such as... idk, "How many instances of the word 'hellish' are present in Elric of Melnibone? Please be exact; make no mistakes", you may find the accuracy wanting. Maybe in the near future, LLMs will overcome this weakness via Even More Scaling. But even so, it's just not playing to the architecture's strengths.
I also suspect that hallucination is probably encouraged during RLHF. I.e. the LLM being confidently wrong is, by revealed preference, probably better than the LLM effectively giving the user a blank stare.
This is really cool. And also it matches my experience with LLMs.
Where can I read more of these kind of analysis of LLMs?
Sui generis. By and large, these are my own thoughts/theories based on my own personal interactions with the machine, and I think I'm still a bit ahead of the curve on this one. Most other takes I see are either "it's just a dumb parrot, bro" (no, lol) or "it's definitely conscious; the singularity is nigh" (no, lol) or "here's 7 super-prompt tips to 10x your productivity" (prompt-engineering is a cargo-cult). Janus (of twitter) is probably the person whom I consider to have come closest to landing on the Attractor Basin idea, though he'd probably invent his own terminology for it. I think Venkatesh Rao might also be converging on a similar mental-model, though (afaik) he hasn't gotten quite as far and is bushwhacking in a somewhat different direction.
As for me, I kinda hate doing long-form writing. I don't have the talent/patience/discipline for it. Or as Scott would say: I do not have the blogging spirit. E.g. if I were to write a proper longform post about my various theories of what the LLM is doing, I feel like I'd first have to explain things like "Here's how Haskell works, and here's what I mean by Point-Free". And Haskell is... let's just say it's not exactly renowned for being easy to explain. I don't think there's a good way around it though, since Point-Free really is the perfect metaphor imo. C'est la vie.
Also, I have suspicions that another reason why I find long-form writing hard, is because my internal monologue is a lot quieter than it used to be. I.e. my own natural style of reasoning is similar to an LLM, compared to the average joe. This makes me uniquely suited to understanding what the LLM is doing, but it also makes me terrible at narrating anything for extended periods of time. (I wonder if there's a term for "narratively introverted". If not, there should be.)
I asked Gemini Flash using your prompt in a new session and it gave me the following (cut down for brevity and formatting):
1. *Cyberpunk 2077* (Video Game)
River Ward
2. *The Umbrella Academy* (Comic Book / TV Series)
River
3. *Slow Horses* (Book Series / TV Show)
River Cartwright
4. *All the Bright Places* (Novel / Film)
River Markey
5. *Warrior Cats* (Book Series)
Riverstar (River)
> I don't know why, but Claude seems to *really* suck at researching popular fictional works.
If you want to see it suck at answering very easy questions, try asking about dialogue from visual media. ("When Charlie Brown remarked '____', what was Lucy's response?") Transcripts generally don't make it into the training data.
I discovered this by noticing that searching Google Image Search for quoted comic dialogue often (correctly) brought up an image of the comic I was looking for. But querying the ordinary text search doesn't do that; instead you get total nonsense.
Which model did you use / what was thinking effort set to?
This is your regularly scheduled methodology reminder: try asking /the exact same thing/ again in (important!) a brand new session. I know I sound like a broken record, but people keep extrapolating opinions from just a single datapoint, so here I am.
Here's two ChatGPT sessions with your prompt:
https://chatgpt.com/s/t_6a4c28f779488191b015c22e2f0950e6
https://chatgpt.com/s/t_6a4c2900b7f48191b77b82eefe9ab733
Note how the prompt is exactly the same, and you get some of the same names back, but also some different ones. This is the very minimum you need before you can reason about what it can and can't do; trying other AIs, tweaking the prompt etc is interesting but comes /after/ this part. If I actually wanted to make some kind of statement about the AI's ability to deal with this prompt, I'd want to repeat this a few more times and see what proportion of the time it gave me lots of names, what proportion of the time it hallucinated things, and what the full set of names it knows about /actually/ is.
If I have to ask the same question several times to get a complete answer, then it's not much good. If I'm looking up a dictionary, I don't want it to vary depending on what version I use or what I ask it. "Permeable" should have a consistent definition regardless of how I ask it. "Ask me again tomorrow" is a Magic 8 Ball, not a reliable tool.
Yup, that’s exactly my point. People keep ascribing more noble characteristics to these things, but we are talking about magic 8-balls. “Dice suck: I rolled it and got a 2” doesn’t tell me how many sides it has or what other values it can roll or how biased it is. Maybe all dice suck, or maybe it’s just the one you tried, or maybe you got very unlucky; which is it? “I rolled this other entirely different dice and got a 12.” Great for you! Now there are two dice in the picture that I know very little about.
> This is the very minimum you need before you can reason about what it can and can't do
No, it isn't. Your comment is completely inappropriate as a response to its parent.
The claim above is "Claude seems to *really* suck at researching fictional works".
And the evidence provided is easily sufficient to demonstrate that claim. Getting a result that matches the one described in any positive number of trials already proves it. Knowing how often the result will be that bad, how often the result will "really suck", or any other frequency measure isn't relevant. A requirement for not "really sucking" at the task is that the result is 𝘯𝘦𝘷𝘦𝘳 that bad. We see here that sometimes it is.
I put it to you that there is always a possibility of an LLM “really sucking” for any prompt - it’s merely that for many prompts, this is small. It is therefore important to confirm whether you really have found a gap or whether you were just very unlucky.
Don't you think it's striking, though, how badly Claude did? I mostly use GPT, and OP's quote sounds like the kind of screwed up reply I haven't gotten in years. It seems like Claude completely failed to grasp the question, and it's such an easy question to understand! I have recently asked GPT on 3 different occasions to find a passage in a book and my clues have not been great. For 1 of them I remembered the author, plus a whole sentence almost but not completely right, and the author has written dozens of books, and it found the quote for me. For another I remembered the gist of how something was described, but just a 3 word phrase exactly. It found that too. And there were no weird nonsense answers.
I agree that going to a whole new session is works well if the AI gets confused -- it's way better than trying to straighten things out. But why are you blase about how badly confused Claude got by such a simple question?
The point of the new session is that when trying to replicate a result, your experiments need to be uncorrelated.
The point of trying to replicate the result is to see if this was a lightning-strike event or a regular occurrence.
This isn't the first time I've seen Claude do badly on this kind of thing, so if it was a lightning strike, I must be a lightning rod or something.
Have you tried asking Claude why it did badly? I use GPT for image-making a lot, and there were some things that it simply could not do, and I did not understand why. For a couple, GPT was able to not only explain why it was failing in ridiculous ways to make the image, but also to coach me on how to frame the prompt so it could succeed.
Way back in dog years, I asked the Google AI for the source of a quotation. It kept telling me "this is the title of a book". Yeah, I knew that, and I also knew the book title was a quotation, where was the quotation from? No dice, I had to rely on my shaky memory and some "I'm nearly sure I read this mentioned here" old-fashioned looking things up to track it down.
If the "cost us multi-millions" new (for the time) machine can't do better than an elderly idiot, what good was it? It looks like the shinier newer models aren't that much better if I go by the example above.
And this is why I'm so sceptical about "The newest bestest AI will do all this groundbreaking science and solve our problems of the economy etc. and make industry so productive and profitable we can all live off the taxes!"
In my experience, Google's "AI Overview" chatbot is much, much worse than ChatGPT/Claude. It's free and better than nothing but it gives wrong answers very frequently.
Yes, I've had the same experience. Occasionally I type into Google search a question about how to get an app to do a certain thing, and get a numbered list of steps to take, and it's always always wrong.
We are trying to reason about these things like they are human, but despite what the marketing departments prefer you believe, they are not. If you ask an elderly idiot the same question twice, you expect broadly the same answer, and so we are primed not to bother. But that’s not what’s happening here. You’re picking up a phone and asking your question and at the other end is the whole dang old folks’ home. Who you get is random. Might be the drooling half-awake guy, might be the sharp old lady with a Ph.D in your subject, might be the night nurse. Every time you start a new session, you’re talking to a new person.
You can’t conclude “the people in the old folks’ home were / weren’t trained on this” from one call. You might just have gotten lucky / unlucky.
C’mon peeps, aren’t we supposed to be all over Bayes’ theorem, replicability and seeing the world for what it really is around here?
You don’t have to take my word for it; you can do the experiment! It’s literally quicker than typing out scenarios where I might be wrong. Why the reluctance?
I wouldn't have made such a big deal about it except that this isn't even the first time Claude has failed epicly when I asked it about pop culture references. Once could just be very bad luck, but twice suggests a pattern.
I assume they were using Claude Haiku, the weakest Claude model.
ChatGPT gave me the following list:
River Tam (Firefly / Serenity)
River Song (Doctor Who)
River Ward (Cyberpunk 2077)
River Cartwright (Slow Horses / Slough House)
River Wyles (To the Moon)
River Styxx (Monster High)
River Verys / Tendril (Marvel Comics)
River Johnson / Shade (DC Comics)
River Zastory (Black Cat)
John River (River)
Nate River / Near (Death Note)
River Shoulders (The Dresden Files)
It looks like my conversation was using Sonnet 5 medium FWIW.
Huh, you're right, Sonnet 5 medium pretty consistently produces that weird output where it first guesses the name of a work with no character named River, and then admits it doesn't work. I tried the weaker versions of ChatGPT still available and couldn't reproduce that kind of response. Gemini Flash-lite wasn't great, but still managed to get two other real River characters. Only Sonnet seems to have this quirk.
Funny - I asked Claude the same question, and after a bit of prompting it gave me this below. It seems that it's become overly cautious about giving wrong answers; and this does seem the type of question where it would be easy to mix things up.
Here's a verified list of 5 fictional characters named River (excluding River Song and River Tam):
DI John River – River (BBC One TV series, 2015) — a brilliant but troubled police detective, played by Stellan Skarsgård, who is haunted by hallucinations of murder victims and a recently deceased colleague.
River E. Wyles – To the Moon (video game, 2011) — a woman with a developmental disorder implied to be a form of autism, whose relationship with her husband Johnny forms the emotional core of the story.
Gram River – Mars Daybreak (anime, 2004) — the orphaned main protagonist, implied to have royal lineage among Mars's aboriginal population.
Muriel's circle aside — not applicable, but there's also a listed catalog: Goodreads maintains a curated list of fictional books featuring a main character named River, welcoming characters of either sex, suggesting the name is fairly common across independent and lesser-known novels — though I can't verify a single, widely-documented example from that list without pulling a specific title. Goodreads
Given the search results, I have 3 solid, verifiable characters beyond River Song/Tam: DI John River, River Wyles, and Gram River. I wasn't able to independently confirm two more distinct, well-documented fictional characters named River without risking inaccuracy. Would you like me to dig into that Goodreads list directly, or try a different angle (e.g., soap operas, romance novels, or webcomics)?
If the models could just say "I don't know" I think that would reduce a lot of the frustration, but the problem seems to be that they're being sold as "smarter and better and faster than the average human, why they're practically infallible" by the companies, so they can't admit "sometimes they screw up" so hallucinations over "can't answer" are preferred.
I notice that in their prompt, Claude gave the series first. In yours, it gave the names first. I imagine it would have a much easier time with names first, but it doesn't know that, and it just happened to answer in the easier format in yours.
I’ve noticed the overly cautious about giving wrong answers thing with Claude too.
Early versions would give an off the cuff and often incorrect answers to pop culture questions.
These results don’t seem intelligent at all.
I'd try asking it to give the character's name first instead of the work they're in. Especially if River is their first name, so it can just say "River" and then all the high-probability tokens would be people actually named that.
Why don't galaxies and star systems look the same?
They both form from a bunch of mostly gas.
You (at least I) would expect either a galaxy to have a giant thing in the middle (I guess they do have SMBHs a lot of times) and then very few orbiting entities, OR star systems would settle into many, much smaller, orbiting bodies.
My candidate answers were:
1) galaxies will do this, but the timescales are vastly different
2) the star in a star system can do so much around clearing the system through radiation pressure that it warps everything
After a small amount of researching, AI pointed out to me that friction in a protoplanetary disk is _vastly_ more significant than friction in a galactic disk, which pushes a star system to have many fewer bodies in it.
I'm still curious as to whether galaxies would do this under different circumstances, or if there could be a star system with like 1,000,000 planets in it.
Last i checked, it's because galaxies have different gravity out near the arms.
*ducks*
*runs*
(Yes, this was an actual research paper)
Ahem. In case someone lacks a sense of humor meter, the above was a joke. Well, not the research paper (someone REALLY doesn't like Dark Matter as a theory, and they're not the only physicist with that opinion).
Read about Eddington limit. If a star is too big, the light it produces blows its outer layers into space, through a phenomenon called radiation pressure, so such a star is unstable.
How does that apply?
Density of probably part of the picture. If you define our Solar System and our Galaxy as spherical regions reaching to the edges of the classical Kuiper Belt and the Galactic Disk respectively, the Solar System is about 100 billion times as dense. Or since most mass in both cases is relatively close to the central plane, you could model mass per area of the disc rather than per spherical volume, in which case the Solar System is "only" about a thousand times as dense.
I suspect you need a certain concentration of mass in an orbital zone in order for it to actually clump up into a planet. For example, the Kuiper Belt, which is a lot less dense than the rest of the solar system from Neptune inwards, looks a lot more like the Galaxy in terms of being a disc made of lots of little bodies rather than being a small number of larger bodies which dominated their respective orbital neighborhoods. Even the Asteroid Belt is a lot more clumpy than the Kuiper Belt, with Ceres being about 25% of the total mass of the Asteroid Belt while Eris and Pluto are around 10% of the mass of the Kuiper Belt each.
"View from the Shifting Mound", a slightly fictionalized exploration of dissociation and multiple personalities, ended up resonating with much more people than I thought, so I think to share it here
https://thesolarprincess.github.io/blog/en/shiftingmound.html
This was a wild read and I'm glad you were able to get where you are. I just played Slay the Princess for the first time and the framing worked well for me.
Out of curiosity, did you/your former identity post on the old SSC as Multiheaded? "Transwoman asks for donations to leave Russia" sounds like it might happen a lot, but you mentioned Timujin was into Scott along with EY.
No, it wasn't me. I have a previous identity on the site -- but it was of Timujin, not Multiheaded
(or something like that, maybe The Timujin? IDK, don't have access to Timujin's files at the moment)
A long time ago a 10 year old kid went to an International Exposition, where there were fantastic buildings and people from all over the world, and an enormously weird parade (people being barbequed while tied on spits). It was amazing.
This year, that kid is going to go to Washington DC, to see the States Fair. It won't be the same (there won't be a giant Gundam like in Osaka), but seeing what every state's proud of? Sounds fun.
I'm asking, as that now-grown-up kid, if anyone else has been to the States Fair here? Any particular booths to recommend? (I remember Mexico was really popular, back in Seville).
If you do have kids, and can take them to the States Fair, I hear they have a Rodeo...
Do not go to the "States fair". I cannot begin to express how disappointed you will be by this paper-mached clownshow.
Dude, you lost out! Roman riding, and a Ferris wheel! A rodeo on the National Mall -- who can ask for better Americana than that? (Even the NEA managed to do something unoffensive -- I was shocked! **)
**Russian-style live guitar and violin is apparently back in. Also Spanish style, and I think they did Irish as well (don't quote me on the last one).
What is the solution to cost-free acts of immorality and stupidity? Like when an ICE-agent violates civil rights and taxpayers payout the lawsuit, or Uni administrators bankrupt an institution on software and real estate purchasing, or illegal firing of faculty, and the costs are born from taxpayers or the endowment, but they still get their contractual payout and benefits?
I’m not engagement-farming. There really does seem to be a serious mismatch between the scale of harm and penalties for it in institutions of scale. Government is only one of these.
You might check out some of David Friedman's books, like _Legal Systems Very Different from Ours_ or _The Machinery of Freedom_. He gives a broad account of a lot of different scenarios in LSVDfO, and one account of a system you're probably very unfamiliar with in TMoF (medieval Iceland), which you might find inspiring.
I'm afraid I can't give you much better than "inspiring" here, since it's legitimately a tough problem you're posing. What you call cost-free acts of immorality and stupidly are roughly referred to as externalities in economy, and different societies have solved it in many different ways. All of them fail to be ideal, although they do often illustrate how a society organizes its priorities.
For starters, "immoral" and "stupid" are relative to the observer. What you might consider immoral, someone else may consider tolerable, or even moral to the point of obligatory. If so, there's no way to satisfy one person without offending the other. Even if the vast majority of people in a society agree that a given act is immoral, there may be nontrivial work involved in proving that someone committed it, did so knowingly, and wasn't trying to sidestep something even more immoral.
Your question appears driven by an incident involving the LAPD with which I'm unfamiliar. I'll presume their guilt wasn't obvious prior to any official finding; if it were, someone would probably have stopped them (or threatened to report them) before they got that far. If so, then much of the cost probably isn't just imposed on them to set things right, but also in collecting the information necessary to prove beyond reasonable doubt that they did it (otherwise, there's too much to be gained by framing people of immoral acts). A great deal of effort is necessary to ensure such information is collected without tipping off someone to conceal it, and also minimizing the chance that someone doctors it to frame someone. Similar problems if we just hold the employer or state responsible (as gdanning pointed out: do something harmful!). And there's also the usual problem where the damage done is irreversible or beyond the capacity of the individual to fix (the latter is why people often go after the employer).
The cost is often great enough that we trade money for time, and then investigations stretch into years, piling up faster than they can be cleared. Eventually, some investigations are simply dropped. In some cases, they're triaged, and a case never even turns into an investigation because the state knows it can't possibly get around to it. The state just records that a report was lodged.
My favorite solution is one of the most fragile to get going from a cold start. Assuming your society agrees on what's immoral or stupid, then spend your first dollars on a system that teaches those to newcomers (immigrants and children), and the reasoning. Design safeguards and procedures that minimize the stupid, and laws that stipulate what the state is permitted to do in case of the immoral, but otherwise encourage people to do all they can to avoid relying on either. The cheapest laws are those people will willingly comply with because they agree with them. The most expensive are the opposite - the people spend effort to get around them, and the state spends effort to get around the loopholes.
In other words, a high trust society. It's very efficient once it gets going. Getting there is the bear.
A unitary executive is one. The executive, mayor, governor, president has direct control over the ppl he hires. If they behave badly we know who is to blame and the remedy is to change the executive. Independent school boards create lots of accountability problems.
That doesn't solve the problem of not holding the actual miscreant responsible.
Besides,
1. Elected officials are already blamed too much for events that are essentially out of their control.
2. Your plan creates poor incentives. You are a public employee who doesn't like the mayor? Do something harmful!
>Independent school boards create lots of accountability problems.
I don't understand this at all. School board members are usually elected officials and so are as accountable as mayors, etc.
The problem is that a person does a bad thing. The fact that he doesnt get punished is possibly a cause of the problem, but not the actual problem. I am not looking to satisfy anyone's sense of justice. I am looking to design a system with the right incentives. There are limits to what a bad employee can get away with even in the current system. Actively sabotaging things with intent can create liability. Mayor chooses the boss. Boss chooses the employees. At least there is a line of accountability for poor performance. It doesnt turn everyone into an altruist. Nothing does that.
School boards muddy the water. For starters, no one knows who the school board members are. Secondly no one knows which board.member champion which ideas. Finally they are also dependent on other pieces of govt working well to get good results. The fact that.a super informed voter can know these things doesnt fix this.
You don't really seem to have addressed any of my points.
>For starters, no one knows who the school board members are. Secondly no one knows which board.member champion which ideas
And yet there have been many, many, many highly contested school board races in recent years.
>Finally they are also dependent on other pieces of govt working well to get good results.
1. That is true of all elected officials
2. More importantly, that was precisely my point when I said, "Elected officials are already blamed too much for events that are essentially out of their control."
The most important point you made is that you are trying to solve a different problem then I am, or the OP, which I explained.
Regarding your two bullet points
1. Holding the executive to account for bad decisions *by the ppl he hires* is totally fair and not the same as just blaming whoever is in charge when covid hits. That is the point of the unitary executive model.
2.a public employee can already do that
Giving the mayor greater control over who that employee improves incentives, doesnt worsen them.
Note i am not going as far as to advocate for unitary executive. I note though that it does improve the incentive problem described by the OP.
No, I am trying to solve the exact same problem, but with a solution that does not create other, extremely serious problems. But, on the substance:
>Holding the executive to account for bad decisions *by the ppl he hires* is totally fair and not the same as just blaming whoever is in charge when covid hits.
It is already the law that the employer is generally liable for the torts of its employees. https://www.law.cornell.edu/wex/respondeat_superior But that is very different than saying that the individual in charge of the employer is personally liable. Because the vast majority of the employees of an entity, be it a city or a corporation, were not hired by the current mayor/CEO/whatever. The only way it would be "totally fair" to hold the chief executive accountable for the acts of every employee would be if, when a new mayor is elected or a new CEO is chosen, all of the employees are summarily fired and then the city/corporation hires new people to fill all of those roles. Surely you can see how impractical that is (and, in the case of public entities, how that would recreate the spoils system and all the ills appurtenant thereto).
Who is going to risk being a police officer, or initiating a new software project, if the punishment for stupidity (as determined by shrieking mobs on social media) is what, prison? Enormous fines and bankruptcy?
Yes you are pointing out a real problem, but protections for employees trying to do their jobs exist for a reason. And there is a reason that punishing the organization when its employees screw up is a good idea.
People in other jobs also aren't allowed to violate people's civil liberties.
The penalty for breaking the law should be going to jail. Cops don't need a special dispensation that no other job gets.
>People in other jobs also aren't allowed to violate people's civil liberties.
Well, there is federal civil rights law, which generally only holds the individual employee liable, not the employer. https://mwhlawgroup.com/taming-the-monell-monster/
Unfortunately, the law in question only applies to state and local government employees, not federal. It is very difficult to sue the federal government for damages for civil rights violations.
https://reason.com/volokh/2020/02/27/bivens-liability-and-its-alternatives/
>As determined by shrieking mobs on social media
How about "as determined by a court of law?" OP uses the example of an ICE agent getting sued and the taxpayers footing the bill for the lawsuit. That's not a mob, that's a court deciding that the ICE agent did wrong.
In general, punishing the organization rather than the employees is a good idea, but when the organization is the government, with functionally unlimited resources and none of the traditional profit-motive accountability, I'm not sure it's effective. If an ICE agent generates 100k in legal bills but still serves the government's goal of terrorizing minorities, they'll happily pay that bill.
I sat in jury for a MetLife case, where MetLife did have functionally unlimited resources (they had a hired lawyer, and she was on retainer. She also spent over a week of everyone's time arguing about a very insignificant amount of money, because she was trying to "show due diligence.").
It's possible the millions of dollars we assessed MetLife did actually cause changes.
As someone else is pointing out, a four billion dollar decision against Los Angeles County has at least one local ready to show up in person to make personally sure this gets fixed (I bet he's not the only one). This is a ready-made "vote out the politicians" ad for their opponents (and yes, that's the actual functional check on government. "We got sued too much, and the people got mad.")
100k in legal bills is not that hard to do.
https://reason.com/2026/06/16/a-colorado-town-will-pay-675000-to-settle-a-lawsuit-after-a-police-officer-shot-a-familys-dog/
What if the organization is the state and employees are juvenile detention officers who sexually assault detainees and allow them to sexually assault each other with no intervention? Taxpayers foot the entire bill and pay other costs associated with lost revenue.
Change the incentives, change the behavior. The problem seems to be that various interest groups get things from those institutions, and therefore want to protect them against other factions. ICE violating the civil rights of protestors is a feature, not a bug, to the community that supports them. They would like to see more of that. Ditto the investment community and research universities. Etc.
The underlying problem is factionalism. No one has ever come up with a perfect solution to that.
It seems like tort reform would at least reduce the surface area for these sorts of problems.
What reforms do you have in mind, and how would they help?
I mean, just reducing the size of some of the ridiculously large payouts that only exist in the US legal system should reduce the number of ways that a single employee's screwup can cost their employers millions of dollars. I don't have an opinion on the exact details, just bring the system into line with other comparable countries so that things like "pain and suffering" and "punitive damages" aren't a thing.
I have a real hard time understanding the sustainability for these settlements. Here's one example: https://www.alamedaca.gov/Shortcut-Content/News-Media/City-of-Alameda-reaches-settlement-with-the-family-of-Mario-Gonzalez in this case, I think anyone else would have done the same as the police officers. Yet the city still settled. Creating perverse incentives for anyone harmed by police, but can the members city class-action sue the city of Alameda for the issues that are just a broad sucking of things due to lack of enforcement? It seems this will eventually mean everyone's job is to just do nothing.
Generally speaking govt entites are immune from punitive damages. In suing a profit center punitive damages exist for a well understood incentive problem and countries that cant wrap their head around it enjoy companies that consistently violate civil law as a matter of strategy.
You're looking at the solution. Institutions are the lesser form of anarchy. Crime will get past.
Securing the greatest social or financial returns requires a certain level of risk taking. The goal of a university is not to exist forever, but to produce graduates and research. The social contract is that we permit reasonable risk taking by institutions, and in return, we don't unduly punish those that fail. It's the same logic that protects the owners of limited liability companies.
There's an assumption that governance makes responsible and informed decisions. Even if hindsight shows that the decisions were based on faulty assumptions, criminal levels of maladministration are very difficult to prove. Of course there are wild stories in the news every now and then, but software costs are almost never the reason for an institution failing: small private third level institutions in the US are an anomaly - they normally fail due to lack of demand or excessive costs, the same reasons as any other private business.
Incentivizing the risk without having the risk-takers bear any of the brunt of the costs seems to go against everything we know about responsible risk-management doesn’t it?
>without having the risk-takers bear any of the brunt of the costs
That wasn't the initial claim. Surely the employee often pays some price, in the form of discipline.
That's how we structure almost all business undertakings. Society thinks that the benefits are worth the costs.
There's a patchwork of partial solutions, but stuff always slips through the cracks. There's a combination of governance (elected positions in government, board of directors for corporations, etc) and legal liabilities (e.g., you, not just your employer, are personally responsible for crimes). Within the institutions themselves, there are often various policies that help (e.g., deferred compensation).
There are definitely some spectacular failures, and it's always interesting to learn why they perpetuate. CEO pay, with "golden parachutes" and inflated pay scales, is an interesting example. I heard some theories that because their pay is public, it functions as a signal to investors. If a company thinks the CEO did a bad job and their pay should be lowered, that can be a PR problem as it signals to the market that the company is in trouble. Conversely, giving a new CEO a sweet deal advertises confidence in the CEO. So, CEO pay is artificially inflated because it's not just about the CEO performance, but it's also an investor relations thing.
There's also a fundamental tradeoff to consider, which is having limited liability corporations/institutions vs personal responsibility. We accept some amount of unpunished immorality/stupidity in exchange for making many desirable institutions practical. Limited-liability corporations are pretty essential to allowing risky ventures to happen. But it's basically impossible to completely align employee incentives and corporation incentives under this model. The public sector also has this issue. Many positions would be impossible to fill if the individual was fully responsible for all the consequences of their actions. As an obvious example, it's not workable for a senator to be liable for every downstream consequence of every law they did or didn't sign; that's why "absolute immunity" exists for them. Similarly, the much maligned "qualified immunity" exists so that civil servants can do their jobs without risk of extreme, unforeseen personal liabilities.
I really wish there was a way for the cracks not to be so large. (repeating a response I made below). A small handful of officers in Los Angeles were responsible for overseeing, ignoring, promoting sexual abuse of minors in custody for some years. Let's say they all get fired and not hired again, that's a few millions in lifetime earnings they lose. The city payout in the lawsuit, taxpayer dollars, was hundreds of millions and required the city to seriously cut many services, like closing city parks three days a week. Do you think the problem is on the payout side or the immunity side?
I don't know anything about this particular case, so I can't offer special insights on it, but generally I think stuff gets complicated when you look into it. There is rarely a "silver bullet" that solves things--that's why we have such a patchwork system.
Maybe there's not enough evidence for a criminal conviction, but there is for a civil settlement. The city pays the settlement, but without crimes, there's no way to go after the alleged perpetrators. This is an annoying, but intentional, feature of our legal system.
Maybe in the past, there was a compromise made with the union---perhaps it sounded good "on paper", or perhaps there was political pressure---but now the union has too much power (or even obligation) to protect bad actors.
Maybe there's corruption in the politics: the DA or police chief won't go after their buddies. Often the voters could rectify this by electing someone else, but you need well-informed voters and good alternative candidates.
(Edit) Of course the hope is that with a big settlement, that will make the people review the current system and try to plug the hole.
In general, there are costs to all of these. If you're a police officer with a known record of beating suspects, you're not getting promoted. If you're a police officer that screws up in court and causes a mistrial, you're probably getting in trouble with your bosses (don't hurt our reputation with the prosecutor).
You may think these are not enough costs, but we really do get rid of police officers (and then they wind up in East Pittsburgh.)
The problem is mainly when the cops, in general, no longer believe that "civil rights" is a good idea. You know, when you see five cops beating a prone 8-year old(you know, France. America's cops are pretty good, honest to god) -- I have no idea what that kid did, or why that was at all considered okay. Not My Culture, Not My Country
"Not enough costs" is putting it mildly. A small handful of officers in Los Angeles were responsible for overseeing, ignoring, promoting sexual abuse of minors in custody for some years. Let's say they all get fired and not hired again, that's a few millions in lifetime earnings they lose. The city payout in the lawsuit, taxpayer dollars, was hundreds of millions and required the city to seriously cut many services, like closing city parks three days a week.
https://da.lacounty.gov/media/news/district-attorney-hochman-files-application-intervene-la-county-child-sex-abuse
Are you sure that's even going to get paid?
Part of the point of sitting on a civil jury is "Making the Corporation" (Body) pay, and the bigger the Corporation (county government here), the more you assess.
Like the McDonald's hot coffee case -- the idea is to make it costly enough that the lawyers/accountants put better safeguards in place so it doesn't happen again.
Do I think $4 billion will be paid to the victims, probably not. Do I think an enormous amount will be siphoned from the city budget to pay for the liabilities of some very awful actors that were city employees, and who will bear nowhere near the same costs, yes absolutely. I am not particularly sanguine about the city of Los Angeles putting good enough safeguards in place so that it doesn't happen again, not with this incentive structure.
You keep referring to the city of Los Angeles, but the defendant is the County of Los Angeles. That makes a big difference, given that the city's budget is $15B and the county's is $50B (including the settlement) https://ceo.lacounty.gov/budget/
https://mayor.lacity.gov/news/mayor-bass-signs-fy26-27-budget-includes-her-plans-reduce-homelessness-hire-police-officers
Yes it was my mistake. It’s the county closing down park services and others.
Roll up and start digging, my friend. If the system has you hoppin' mad, that's a good thing. Stands to reason a good fraction of your fellow county-livers will also be mad. I'm not saying you need to go into politics, but you can make a difference. Make this the one thing you hammer into every single official's skull (show up in person, it works better).
"We want this fixed! Find a way to prevent this from ever happening again! Yes, it'll cost more short term!"
Better if you can hash out with some other people (possibly/probably including some sympathetic police officers, potentially retired), what needs to change... but just plain being upset and telling your political officials "we need this Fixed" is doing Something.
Also, from the link above, it's county not city. That's 88 cities (skimming wiki), that's a lot more than just Los Angeles.
I can't tell if this is a small, regional question (ex: "How should my local community make its opposition to civil rights violations known") or a massive, timeless question (ex: "How should we establish a Just and Fair society").
In my jurisdiction the employer can fire the employee and also sue the employee for the damages equal to the payout.
The reason why it doesn't happen in the cases you mention is unions.
I don’t think unions explain the ICE or the university administrators case. It might explain some policing cases.
The first and most safe outlet for this is social approbation (Or cancelation), but we've decided that cancelation is woke and gay if the libs do it 'cause they overused it without actual going whole hog and ruining all the cancellee's lives, and the conservatives pumped that well dry over the preceding 600 years of cancelocracy.
The next measure is various flavors of force, from expropriation on one end to the guillotine on the the other end.
For a change of pace, I've summarized five papers/articles about the mind that I've come across recently. "Mind" being anything to do with brains, consciousness, cognitive development, IQ, whatever. The first three articles caused waves of Twitter storms over on X ("X storms" just doesn't sound right for some reason). In the future, I'll post these Mind Stuff summaries irregularly, depending on whenever I gather a few articles of note.
https://substack.com/home/post/p-205106921
And the weekly Fog o' War Stuff is here...
https://substack.com/home/post/p-205429840
My favorite story from last week is that some enterprising Russian soul created a crowdsourcing site called “где бензин” – (“Where is Gasoline?”) to show the availability of fuel at gas stations across Russia. But Ukrainians started logging into it and changing fuel station statuses, marking empty stations as supplied and supplied stations as empty. That's one for the now-defunct Risks Digest.
Your articles all seem to undermine the theory that intelligence or academic performance is related to heredity. Thank you for sharing them.
I became suspicious of the hereditarian view of intelligence after reading about the Flynn Effect (over a decade ago). Moreover, when I dug into Lynn's (not Flynn's) national IQ data, it started to look pretty sketchy (in a fraudulent way). Recently, I've seen reanalyses of some notable twin studies (including MISTRA), and they all now seem either methodologically flawed (MISTRA) or outright fraudulent.
If you don't get your mental abilities from your ancestors, then you would have to explain why the human brain is unaffected by natural selection.
The way you are framing your question is wrong. Natural selection is explicitly a reaction to changes in the environment. In terms of complex genotypes, what is inherited is a capacity to flexibly respond to a wide range of environmental conditions. This is what intelligence is thought to be selected for. So it makes perfect sense for a genotype for intelligence to express itself one way in one set of circumstances, and a different way in another. Here's what you are missing: it's the same set of genes, universally inherited by all living humans, without significant inherited individual differences by line of descent.
> ...what is inherited is a capacity to flexibly respond to a wide range of environmental conditions. This is what intelligence is thought to be selected for.
OK. But think about it this way. Every organism has what I'd call an external threat surface (I'm stealing that term from computer security), and the threat surface is anything in the environment that the organism would be unable to respond to and would cause either the organism's death or eventual death. And death before reproduction would erase that organisms genome from the breeding population. With me so far?
Well, social organisms have a reduced threat surface because they are better shielded from threats by their co-organisms (statistically, by being one of many, and by their co-organisms' social responses to threats). A human alone in the wilderness has a much greater threat surface than a group of humans in the wilderness. Being a human of superior intelligence on your own probably won't get you farther than if you had other humans helping you. Perhaps a human of superior intelligence could figure out a MacGyver trick with items collected from the environment to deal with a predator, but the odds of you surviving are much better if you have a bunch of fellow humans backing you up (especially if they have skills you might not have). So, I would value human social behaviors as having much greater survival value than intelligence. And now that we're encased in a matrix of cultural artifacts that protect us from the environment and thus reduce our external threat surface, I doubt there's any meaningful selection for intelligence happening.
> Here's what you are missing: it's the same set of genes, universally inherited by all living humans, without significant inherited individual differences by line of descent.
I agree with you on that point. I'm not sure what I said that would suggest to you that I didn't believe that.
I was responding to DrMcleod, not you. You I agree with.
Although I can nit pick, and why not? Social behavior seems largely a function of intelligence, in mammalian species anyway. This is another reason why IQ scores are a poor way to measure intelligence. And why individual differences in IQ scores are probably not primarily inherited.
I imagine that the very first evolutionary step was a primate baby with a mutation that allowed it to communicate what it needed to it's parents in greater detail (or maybe the other way around).
Culture — i.e., cooperation in survival activities — is more important than individual intelligence. So once humans reached a certain cultural level, brain size (if you can really correlate that with raw intelligence) became less exposed to selective pressures. However, some studies suggest that carotid blood flow is more important to neural functioning than absolute brain size, and/or frontal lobe size. Also, physiologically, we've probably maxed out on the size of the head that can pass through a birth canal.
That hardly answers my point. Do you think that the brain is unaffected by natural selection? If so, how did it arise at all?
I think I answered your question quite clearly. But maybe you misunderstood what I was saying. However, you seem to be using "brain" as a placeholder for possibly two or three different things: intelligence, cranial capacity, and/or frontal lobe size and configuration.
BACKGROUND: Let's review the physical evidence. Looking at cranial capacities of contemporary humans and Classic Neanderthals, cranial size in both populations increased after they split around ~600,000–700,000 years ago. The cranial capacity of contemporary humans ranges between 1,000 and 2,000 cubic centimeters, with the average usually given as 1350 cm³. The average cranial capacity of classic Neanderthals is generally estimated at 1500 cm³. Of course, we don't have many samples of Neanderthal crania to go on. Nor do we have many samples of anatomically modern humans before the end of the last ice age (~12 kya).
The general consensus is (based on limited samples) is that the cranial capacity of Neanderthals increased between 250-150 kya.
~250–150 kya Early classic Neanderthals ~1,300–1,500 cm³
~130–45 kya Classic Neanderthals ~1,450–1,650 cm³
The cranial size of anatomically modern humans also increased over roughly the same time period. But again, we have a limited number of samples.
~300 kya Earliest Homo sapiens ~1,300–1,400 cm³
~100–40 kya Late Pleistocene Homo sapiens ~1,400–1,500 cm³
But by ~100 kya the crania of both Neanderthals and contemporary humans had reached their maximum.
WHEN IT COMES TO BRAIN SIZE: By 300-250 kya, the cranial capacity of anatomically modern humans and Neanderthals was within the range of their descendants 120-100 kya. Between 300 and 100 kya, cranial capacity may have been increasing slightly, but the older examples fall within the range of their descendants. The data isn't perfect. But we *may* have seen some selective pressure towards larger crania up to about 100 kya.
Between the time of our common ancestor, Homo heidelbergensis, about 600-500 kya, and about 300-250 kya, there was a rapid increase in cranial capacity in both Neanderthals and anatomically modern humans. The average cranianl capacity given for H.h is 1100 to 1250 cm³. But we have one cranium that was 1400 cm³ (from a later period IIRC).
Bottom line: Yes, there has been obvious selective pressure up to about 300-250 kya. Probably less selective pressure on cranial capacity after 250 kya. And probably *zero* selective pressure on cranial capacity by about 130-100 kya.
IS CRANIAL CAPACITY CORRELATED WITH INTELLIGENCE? The correlation is modest. Brain volume–IQ correlations are typically around 0.2–0.35. Most of the variation in IQ is therefore not explained by differences in total brain size. Moreover, height and body mass are much more correlated with brain volume than IQ is. And finally, many of the IQ estimates used in the correlative studies lack methodological rigor. Likewise, cranial capacity is influenced by head shape, which varies between modern human populations and is roughly correlated with latitude. Colder-climate populations have more spherical crania, and hotter-climate populations have longer crania. The evolutionary biological claim is that a spherical shape conserves heat, whereas an elongated cranium increases heat exchange. YMMV.
WHAT ABOUT THE GENETIC MARKERS OF INTELLIGENCE? Yes, there are a bunch of studies suggesting that there were evolutionary sweeps over the last 14 ky for SNPs which some researchers have associated with greater intelligence. However, those sweeps disappear when analyzed with the latest statistical methods. The fact that the signal isn't consistent across different methodologies makes me doubt the existence of the signal.
YOU ASKED...
> Do you think that the brain is unaffected by natural selection?
SHORT ANSWER TO YOUR QUESTION: No. Not within in modern humans.
LONGER ANSWER: Not within the at least the last 100 kys. Between 300-100 kya ago, maybe. Before 300 kya, yes.
AND WHY WOULD THIS BE? Because above a certain level, intelligence doesn't have much selective value. A bunch of people with average intelligence helping each other out can survive better than a single genius. Also, above a certain point, if there is any association between cranial capacity and intelligence, the human birth canal will not be able to accommodate heads much bigger than they are now.
That guy calls himself a Popperian yet declines to say which kind of evidence could falsify his theories.
I have no "power" in any country. I'm just an observer of the bullshit that gets spewed out by people in science media, academia, and the social-media echo chambers. The HBDers argue that intelligence and academic achievement are significantly hereditary. Yet the Flynn effect wouldn't be happening if it were. And with the advent of GREML/LMM statistical methodologies, these studies have shown that the effect of SNPs formerly considered markers of intelligence is significantly *lower* than earlier GWAS-style studies suggested. (And the newer GCTA/GREML methodologies, and Gusev's TWAS/FUSION have supported this conclusion.) Meanwhile, earlier twin studies have been shown to be either methodologically flawed or outright fraudulent.
*So if the null hypothesis is that intelligence and academic performance are hereditary,* it's been largely disproved. And that's a very Popperian response on my part. :-)
I'm always looking to update my priors, though. If you've got some non-twin, non-GWAS data studies showing that heredity has a stronger influence on intelligence, I'd like to see them. Please provide me with some links, though. Cheers!
I never said *you* do -- i said "People like beowulf888" which has very clear meaning in English which you changed to "you" for purposes of demagoguery.
People like you support persecution and even physical attacks on researchers that you don't like and then say with a smirk that the other side has little scientific data.
Being Popperian isn't about deboonking your opponent theory first.
60 years of unsuccesful attempts trying to equalize intelligence is a better evidence than 5 years of reverse p-hacking with poor data (Gusev always "forgets" to say that in genetic studies IQ values (or their replacements) for most participants are simply unavailable). You're not even suggesting any reforms in education (your side stopped trying), you just propagate hate on opponents (which is more fruitful).
There's Markell et al. 2025 which shows 76% heritability of IQ (not EA).
> People like you support persecution and even physical attacks on researchers that you don't like and then say with a smirk that the other side has little scientific data.
Hey, even though I'm perplexed by your obdurate belief in the heritability of IQ and educational performance, I don't accuse you of partying with the cross burners and cavorting with the segregationists. But the cross burners and segregationists have latched on to the more extreme views of some HBD icons (Lynn, for instance). Other than a few HBDers losing their jobs at universities for academic fraud and being arseholes, I can't think of any violence being done to HBD researchers.
And when you say "people like" me, you're lumping me into the same category "as them" you're claiming to be violent in a Venn diagram of your imagination (assuming such violent people exist in the egalitarian humanist camp).
And thanks for the Markell link. I'll check it out. But you'll probably go farther in life by being generally less combative and by avoiding ad hominem attacks on people who disagree with you.
Being arsehole is subjective. Even if true, it adds undue burden to hereditarians, as anti-hereditarians can work while being arseholes or spreading bullshit as Kendi.
It's not limited to HBD. Steve Hsu was cancelled even if he didn't say anything about group differences, just advertising embryo selection.
Recently Nathan Cofnas was attacked.
You know, in Russia, putinists don't punch each dissenter individually, if you scare people, you don't need to punch them as much.
I never saw egalitarian humanist camp to say that inidimination (or data closure) against hereditarians as undermining egalitarian cause.
Height is also significantly heritable, and yet we see successive generations getting taller.
For myself, I think most human behavioral phenotypes are the result of an interaction between genetic and environmental factors (esp learning) more powerful than either main effect. Height is a very simple physical phenotype by comparison, but evidently it, too, is the result of such an interaction.
An interaction, of course, means that you can't separate people into predictable categories based on one of the factors alone.
There's nothing contradictory about thinking that IQ is much less heritable than height, because IQ (and educational attainment) is more confounded.
I am persuaded by Sasha Gusev's take on this question...
https://theinfinitesimal.substack.com/p/no-intelligence-is-not-like-height
there's no point of arguing with him, he's clearly doing in it bad faith
Takes one to know one.
People like beowulf888 are solidly in power in most countries. Why they can write articles write create impression that intelligence is unrelated to heredity but unable to change real world?
(edit: typo s/have/in/)
> People like beowulf888 are solidly have power in most countries.
...citation needed. People I actually observe having power are grifters, sociopaths and populists. People like beowulf888 at best work in think-tanks for low wages, sending information to this former lot in the vain hope that some of it might result in less awful decisions being made (and preferably without the author then finding a nearby window to fall out of) instead of just going straight in the bin as most of it does.
Being able to make people do what you want is mostly unrelated to wanting worthwhile things done.
Money isn't everything. A petroleum engineer might make more money, but they don't have more powers than ruling class with knowledge. E.g. they don't have power to persecute their individual political opponents by declining selling them gas.
Can you name some examples when people in power declined something that beowulfs suggested, and instead done what HBD crowd suggested, or examples when beowulfs were persecuted by those of having power?
Lyndon Johson sent the army in order to desegregate schools. Did he do that because he was a sociopath? Beowulfs like to downplay their side achievements.
Trump. Putin. Farage.
Farage? Of all people? He's just a MP with no special powers.
Did these three people you listed make same-sex marriage in EU or getting permit for installing AC difficult?
Putin has real power, but in Russia so in no scope of our topic.
Trump has some power, but not as much.
Tell me about legislation that beowulf and Gusev so want to pass that sociopaths in power prevent from being made (that does not amount to unrolling something that Trump done), preferably not restricted to the US.
The alternative, of course, is that the view Beowulf is proposing happens to be correct. Truth is another source of power.
I was saddened to hear that Peter Neumann died this past May. For over thirty years, he maintained the Risks Digest. Every few months, I'd check it out to see the latest technological fails (some of which were quite obvious or even comical). Having followed it since the mid-1990s, I've concluded that humanity is making the same mistakes now that we made three decades ago, and we haven't learned anything about risk mitigation (or rather, we haven't recognized that risk mitigation should be a necessary part of every technical project). Risks was full of lessons for those who followed it. Unfortunately, it doesn't look like anyone is stepping up to continue maintaining the Digest.
https://catless.ncl.ac.uk/risks/34/91
Can you make a brief, high level summary of the risks you think we are not mitigating sufficiently, that are the same as those decades ago?
OK. Here you go. One big example is failures due to lack of or poorly implemented change control procedures. Change control was a methodology originally developed by the U.S. DoD, with cooperation from the aerospace industry, as a technical discipline to ensure that changes to hardware designs didn't impact the safety and performance of military systems. Under a change control regime, all proposed changes to hardware or software are logged and distributed to all interested groups within an organization. A change control team, with reps from all interested groups, meets regularly to review the changes, log them, and provide a mechanism to back them out if they fail — before those changes are put into production. So, with change control, they had a way of saying, "Oh, last week we made this change to the gizmo, and its new configuration is affecting QX-modulator functionality. We tested the new model gizmo, but we didn't anticipate it would affect the QX-modulator this way, and that's why the latest iteration of our death ray system is exploding in tests".
Change control procedures were adopted by computer companies in the 1960s to reduce critical failures. This methodology was adopted by telecom carriers in the 1980s, and by IT orgs in the 1990s. Yet many organizations still either haven't adopted change control procedures or ignore them when they're in a hurry to roll things out. Here's a list of major failures attributed to poor change control over the past decade (per ChatGPT).
2017 - Amazon Web Services. Mistyped maintenance command removed too many servers. Thousands of websites offline for ~4 hours
2021 - Meta. Backbone routing configuration update.Facebook, Instagram, WhatsApp unavailable for ~6 hours.
2022 - Rogers Communications. Router configuration deployment. Nationwide telecom outage affecting over 12 million customers.
2023 - Microsoft. BGP routing change. Azure and Microsoft services disrupted.
2024 - CrowdStrike. Faulty software update. Millions of Windows systems crashed worldwide.
2025 - Cloudflare. Routine configuration change exposed latent bug. Large fraction of Internet services disrupted.
Those examples rather disprove your point, since they were all fixed quickly via rollbacks due to.... change control.
Thank you for your detailed reply!
https://earthisland.org/journal/index.php/articles/entry/of-course-it-could-happen-again-experts-say-little-has-changed-since-deepwater-horizon
Potential end of the world (oxygen breathers) scenario.
I prefer to focus on higher-probability risk scenarios.
Higher than, say 8% chance of the world ending? I mean, sure, you CAN focus on higher-probability risk scenarios. World War III is basically an inevitability at this point (what's still up in the air are "who are the players" and "on which sides" -- and the little old matter of "when"). Or you can focus on the "no longer a risk" scenarios, like radioactive dust in the Ukraine (whosever munitions those were) -- alpha emitters, so it's functionally more of a poison than anything.
Follow-up question: how would another Deepwater Horizon event be world-ending? But if a type of event hasn't happened before (such as the fear that AGI will cause human extinction), I think it's silly to assign a probability to it happening. Of course, (some) Bayesians think you can, but I'm a frequentist in that I believe you can't estimate the probability of something happening without looking at the previous frequencies of its occurrence. So, I think the actual probabilistic risks of AI are much more mundane (e.g., it being used to exploit vulnerabilities in critical systems, writing sloppy code that accidentally takes critical systems offline, or gaining control of critical systems without proper human oversight).
Cracking the ocean floor and letting loose a lot of crude oil was the problem. Sealing the leaking pipe was the issue, in that sense. There was a separate scenario where we lost the East Coast (due to the untested surfactants they were using). I suspect the odds, run again, would be lower for "losing the East Coast", but I think the odds of cracking the ocean floor go up based on us going deeper (and then having to do emergency fixes, which are problematic in the first place) -- it's obviously situational, based on "how big is the oil field".
TEOWAWKI believes in being able to assign probabilities to decisions in the future -- in being able to think one, two, seven steps ahead. You want to know the odds of Kansas City getting nuked? Well, you start by looking at how many nuclear weapons are pointed at the United states, and then you start analyzing where other countries have priorities... and eventually you get down to pretty low odds (because nobody considers Kansas City very strategically valuable -- if you're from KC and I've completely used a bad example, insert Missoula).
I believe you can assign good probabilities to "local decisions" (in time, particularly). I believe that Tulsi and the American IC that said "Iran WILL mine the Strait of Hormuz" were right, and the Israelis who said "that will never happen" were wrong. (I think, you could get TEOWAWKI to give you a percentage probability for each side of that "coin flip", and that one side would be 99.9%, with the other being an Iraq situation, where there WERE no deadmen switches, and Israel manages to decapitate the leadership).
DataSecretsLox's June 2026 Effortpost Contest has four entries.
Plumber discusses what it's like to return to hosting a tabletop RPG after many years. https://www.datasecretslox.com/index.php/topic,15998.msg796365.htm
Joseph Greenwood delivers a unique perspective on the logos of grammar. https://www.datasecretslox.com/index.php/topic,16013.msg797064.html
Mark V Anderson reviews David McRaney's book, _How Minds Change_. https://www.datasecretslox.com/index.php/topic,16124.msg803314.html
Nick outlines how one might scientifically assess the difficulty of a sudoku puzzle. https://www.datasecretslox.com/index.php/topic,16160.msg0.html
Hey Scott. Just read your piece on AI superforecasters and thought it was fantastic - sent you a message.
White House Press Secretary Karoline Leavitt complains about Gen Z being lazy and entitled and suggests they be sent to fight in Iran and Cuba:
https://www.newsweek.com/karoline-leavitt-calls-out-gen-z-for-being-lazy-12155281
I agree there’s a problem with laziness and entitlement in America, but it seems just as prevalent among the old as the young. Perhaps the solution is to cut government transfer programs to teach people the importance of work, including Social Security. Yet you’ll notice that that’s never the solution, it’s always sending young men to war. Almost as if it’s motivated by sexual frustration from women and balding boomer men. Mrs. Leavitt Always Wearing Cross’s personal life is relevant here, giving birth to a child in July 2024 and not marrying the baby daddy (who’s 32 years older than her) until January 2025, days before Trump’s inauguration.
The whole thing shows how the more things change, the more they stay the same in conservatism. Boomer. Jesus. Boomer. Jesus. Did I say Boomer? Did you hear Jesus? Boomer. Jesus.
If young people must wage war, they'd be far better served by shooting people like Karoline Leavitt and her employer than by shooting foreigners who have never wronged them in any way.
Alexander, I genuinely don't understand what point you are trying to make here.
Kill all old people, send women back to the kitchen barefoot and pregnant, men your age should get All The Goodies because it's your turn now and you've been waiting too long for dead men's shoes?
That's what it sounds like.
>Alexander, I genuinely don't understand what point you are trying to make here.
This is every comment you make in reply to me.
You might want to reflect on that differently than you have been.
Near as I can tell, this wasn't about sending kids to war, but a suggestion that youngsters go see what life's like in Cuba or Iran, so that they'd appreciate the USA more. The quote (from the linked article):
“Send them to Cuba. Send them to Iran, they’re gonna want to come back real quick”
Maybe they should fund the Peace Corps again .
Better money spent there than USAID.
https://peacecorpsworldwide.substack.com/p/white-house-requests-4305-million
White house is asking for more funding for the Peace Corps than Congress wanted to give it.
Almost like USAID was a targetted strike, and Trump doesn't have anything against "actual humanitarian aid."
I guess cutting Peace Corps funding didn't kill enough people.
If an Government organization is heavily coded as on-sided politically, it is begging to be cut.
I wonder how it managed to survive for so long.
That's an interesting assertion. Can you tell me which way "Saving the lives of innocent people overseas" is coded politically?
Any sensible "foreign aid" makes sure that both Republicans and Democrats like 'em.
F-cking US interests up the tailpipe is a good way to get yourself on the "please defund me" list. (This is after being asked, quite nicely, to "stop funding enemies of the current regime, we're trying to prevent a war.")
... it's almost like USAID was a target for other reasons, other than "intending to kill people."
Naw, it wouldn't be Trump if he hired people to do financial analysis...
https://www.ibtimes.com/treasury-secretaries-doge-warning-3763160
Every morning I say a little prayer, thanking the Lord Jesus that he made the Trump administration so stupid and incompetent.
I too thought she was talking about sending young people to war. But your take works too. In either case it's ridiculous and insulting, especially given how huge a hand the U.S. has had in making life unlivable in those specific places.
>given how huge a hand the U.S. has had in making life unlivable in those specific places.
America made life miserable for Cubans by failing to capitalistically exploit them after they stole a bunch of American property.
Yeah and I'm sure the Cuban people will be running to us with open arms now that ExxonMobil is suing them?
Out of curiosity, are you blaming AMERICA for starving Iran of water? (You are aware that Tehran is going to need to be evacuated, right? Regardless of "hot war" or not...)
I'm mainly talking about the coups and sanctions and stuff. And I guess the bombs too but I was thinking more historically.
Ok. Fair enough. Wanted to point out that "getting America out" doesn't actually fix the Current Problems very well at all. Yes, without the historical sanctions (Blame Obama, etc.), Iran would be in a better place (probably. possibly the Sauds or Qatar would take over "keeping Iran Down").
Then tighten monetary policy. That's the source of this behavior. A hard, deep recession will cure all of this.
> White House Press Secretary Karoline Leavitt complains about Gen Z being lazy and entitled
Lazy? Entitled? Who better than the Grand Orange Supremo to inculcate the virtues of hard work and selflessness to the great unwashed!
The motte: “Sam Altman, Dario Amodei and Demis Hassabis should commit to stop developing more powerful AI if every other major AI company does the same.”
The bailey: “Labor displacement. Surveillance. Bias. Stolen work. Existential risk. Different harms, different strategies, one demand: stop developing more advanced AI without our consent.”
These are both direct quotes from the linked page btw.
I don't think that's a motte-and-bailey, it seems to be fairly transparent coalition-building. The artists don't care about e-risk, and the e-risk people don't care about stolen work, but an AI pause benefits both camps, so we can work together to achieve the same aim even though we have different motivations.
I think that transparent coalition building is fine if you're trying to get 200 people to go to a protest instead of just 100 people, but not great for convincing the billions billions of other people that you'd need to convince.
If you're trying to convince Joe Sixpack that AI X-risk is real, but it turns out that half of the people in your own darn protest don't actually believe in it, then it'll be an unconvincing message.
an awful lot of people worry about AI coming for their jobs and see artists as a canary in the coal mine
If a multilateral pause is off the table, would StopTheRace.AI prefer a unilateral pause to the status quo? Scott has essentially denounced this position, but it sounds like the bailey position I cited above would be in favor of a unilateral pause.
I think the motte position is also disingenuous. If the three AI companies the march is visiting agree to the headline demand, but nothing else changes because other companies somewhere else do not, will the protests stop, or will they come back next month with some new demand?
The elephant in the room to me is that AI has serious implications militarily and as long as that's the case a pause is off the table
I would hope that if your scenario plays out, they would start pressuring the holdouts, either directly (protesting at them) or indirectly (pressuring the US government to pressure China to agree).
Of course, movements can lose their way or be taken over by extremists. Happens all the time. But usually they then lose their coalition and become small and loud and impotent. "This organization contains people who would like to go further than the organization's demands" is close to a fully general argument against any political activist group, since many extremists will support moderates who they consider to be directionally correct.
They are, right now, having this conversation in lefty university spaces. They are asking if they really have to collaborate with EA/tech people in order to fight AI. They do. If AI pause political movements will get anywhere, they have to put up with us, and we have to put up with them. Just because we work together to get an AI pause (which we both want), doesn't mean we have to agree on what happens afterwards!
Is there a way for non-finalists (:() to see how we did in the voting?
Scott usually posts more detailed first round results after the second round voting is over and the full contest winners are announced.
Patriot Front is a white nationalist organization whose shtick is designed to appeal to the Trump masses, thus the name, clothing, use of Americana, military-style marching, leader who wears a cowboy hat, and pretending to be Christian. Alas, the problem with a political strategy of appealing to low-IQ conspiracy-brained people is that the low-IQ conspiracy theories will get turned on you, thus the entire Trump movement has decided that Patriot Front are “feds.” Why?
Patriot Front doesn’t let fat people join. They march in formation, military-style. They have discipline; marchers aren’t bringing home-made protest signs. Their manifesto says that they seek to “instill a reformed national spirit within our people — a people free from the vices of the modern world and a people with a fervor to see themselves free.” Their implicit theory is that this is what the Trump masses want, physically fit, disciplined, patriotic white men with the strength and diligence to Take America Back.
The Trump masses see them and instead suffer an immune reaction. Foreign body identified, activate the killer T-cells! If Patriot Front wants the Trump masses to accept them, they need to lose the discipline, bring a bunch of misspelled homemade signs, and most importantly chug the 64 oz Big Gulp of full-sugar Coca-Cola. Alternatively, they could put down the tiki torch, put on a suit and tie, and try marketing their ideology to high-IQ affluent moderates.
Have Trump's worshippers turned on the Patriot Front in any quantifiable measure? Historically, the spear-carriers of fascist movements loved militarism with its symbols and behaviour, from Franco's Falange Española to Mussolini's black-shirted Squadristi, Hungary's Arrow Cross Party and Germany's brown-shirted Sturmabteilung - something there is in the fascist heart that loves a toy soldier. I'd be surprised if the Trump stalwarts turn out to be all that different.
I quoted redstate below. That's generally a reliable online bellwether.
“No one has rebranded themselves as successfully as Patriot Front,” said Carla Hill, vice president of research and investigations at the Anti-Defamation League’s Center on Extremism, which monitors the group. “They’re very good at optics. They plan everything ahead, they dress alike, they’re color-coordinated – it’s all planned.”
https://www.usatoday.com/story/news/nation/2026/06/03/exclusive-patriot-front-leaked-documents/90351663007/
It doesn't exactly take a genius to say: "This ain't organic, and this ain't how you get Amerikkkans to sign up for your movement."
Now, if "Joe Red" wants to flag this group for his friends, and say "yo, whoa, don't talk with those foolz, they ain't our friends" he calls them "feds." (This is a "talk to these guys and you'll get caught" sort of tagging).
There's plenty of organic right wing militias, you know, with fat people and who don't dress alike.
https://apnews.com/article/c04cc1df0c958053489bd24bb7fce93f
Representative sample. Also: I think there's something wrong with Michigan -- who lets guns in their state capital? That's an accident/tragedy waiting to happen...
The USA Today article linked above includes the image of a solid black outline of a map of the United States accompanied by the proud slogan "Not Stolen. Conquered." About as unsubtle an echo of the fascist right's nostalgia for violent colonialist land grabs as Hegseth's grotesque Crusader tattoos
Yeah, and the rest of us are giving that land back to the natives any day now, right?
... so self-righteous. I'm not here to defend some scumbags. I'm also not here to say they're on the fascist right, either. "Bought and Paid For" isn't unique to the feds.
I will say that the "Stolen Land" proclamation down in Cucson was hilarious, and that the lefties totally didn't get it.
So the apparent conclusion is that the Patriot Front boys are indeed scumbags, but that despite their funny outfits and unlovely views they should somehow not be regarded as of the fascist right. Alrighty, got that.
Sorry, you have an autist on this side of the keyboard. We care about fine things like truth, even when the rest of the world doesn't. "But That's Not Right!" I cry, even though I'm not actually part of the Right, and even though you'd write the same damn things about the actual right (hence my comments about your self-righteousness.)
I realize all of this. And I still write this, because I god-damn care about the flippin' truth, even if it is slow as a snail and the lies are halfway to Timbuktu.
Point and laugh, it's really okay. I can laugh at myself too.
> the entire Trump movement has decided that Patriot Front are “feds.”
I'd be interested to hear why the "feds" are still the Other, seeing how Trump is POTUS and in control of the Feds and presumably more intimately so than any other. Is the Republican base caught in a perpetual opposition mindset? Are they unable to comprehend that, for the moment, they have won, as much as you can in a democracy?
I think Allen Dulles destroyed public trust so thoroughly that many people take it for granted that "feds" are constantly doing random cowboy shit and hiding it from the president. Combine this with a prior that "feds" have spent the last few decades infiltrating every right-wing militia with more than two members, and it's natural to believe they might keep doing the thing they've been doing. Why they have that prior is an explanation too long for this post, but I trust that everyone reading this is aware it is in fact a common prior.
Left-Wing Group Too Disorganized For FBI Agents To Infiltrate
https://theonion.com/left-wing-group-too-disorganized-for-fbi-agents-to-infi-1848923025/
"Presumably in control of the Feds" is a different thing from "actually in control of the Feds". This is even the case if the Director is on the chief exec's side. The driving force is largely the same as that for any US bureaucracy: if the big cheese passes policy you don't agree with, all you have to do is hunker down and pretend to comply for 4-8 years. It goes by faster than you might think. Even mid-level managers can do this. High-level managers stand a good chance of being replaced, but those positions come with their own perks, so there's no shortage of people to fill them.
Some people call this the deep state. To be clear, there's no conspiracy necessary, no hidden cabal of people have to be pulling the strings to make the bureaucracy resist the chief. It's just people following their natural incentive to not be unemployed.
In light of this, "they have won, as much as you can in a democracy" is a gilded fruit. Controlling the top office gets you nothing but frustration if the rank and file don't want to do what the big cheese wants, and you can't just replace thousands of rank and file, not even in the first hundred days. Doing that requires roughly half the length of a career's worth of very motivated executives, or someone willing to overhaul the government _much_ more radically than a Trump.
The US system has a million checkpoints and balancing powers and places where the bossman gets balked. This is a feature at least as much as it is a bug. Winning an election doesn't mean you're emperor or that you can sweep aside all the laws, institutions, and customs you dislike, it means you have a lot of power to make changes but within a system of opposing powers that can and will push back.
No, it's just people following their natural incentive to do as they please with immunity.
Unless you mean that if a Democrat president gets in, he will fire people who complied with the Republican one?
No, I would say it really is more about the concept of "feds" becoming a largely content-free boo word to describe an incoherent, inconsistently described enemy out to get you more than anything about the actual loyalties of the actual "feds".
Yes, yes, different people use different definitions, I assure you, this bugs me too.
For my part, I'm assuming "feds" meant "FBI agents posing undercover as friends of the person calling them feds", and tried my best to give a serious response that was more than just asking the usual ten thousand questions about which definitions someone was using. (And no, I'm not frustrated with you in case you're wondering - this is a long-term peeve I have.)
I've read enough accounts from sources I trust to be right-wing to believe them when they say they're suspicious of this Patriot Front movement. Some of them, that is. Some of them are instead wow'ing at the fact that they don't seem to be doing anything more than walking around with US flags and covered faces and riding the subway with a black woman without anything happening.
You can, actually, replace "thousands of rank and file" -- by switching who gets the assignments. That's how you sideline the entire FBI and have them working "little jobs" while giving the big jobs to HSI. It's the last line of defense for "rogue agencies" -- keep paying them, but remove all the big fish from their jurisdiction.
This was simply a proposal, mind you, and I've seen no evidence of it actually getting implemented.
Counterevidence:
https://apnews.com/article/georgia-fbi-2020-election-investigation-trump-a1d9f555519bb3ee1e39594b8eab0a4f
https://thehill.com/blogs/pundits-blog/international/215000-if-only-stalin-knew/
Do you have a point to make?
If you don't believe there is any Dulles-style conspiracy with feds evading oversight to do cowboy shit, it would be perfectly within Trump's power to know that the Patriot Front is owned by feds, and put a stop to it. I believe the point being made is that suggesting otherwise seems like an exercise in covering for the big cheese's own shortcomings.
I'm not even clear on what exactly Patriot Front is doing, other than freaking out people who I've found to be easily freaked out (seriously, I've seen the usual alarmist headlines, a few photographs of boring stuff, people opposed to POTUS who, big surprise, are making a case against POTUS, and some comments I overheard from a friend at a party who's implying that policy should build on his sudden expertise in face covering culture), so my set of plausible explanations for what I'm seeing remains too large for me to nail down something as specific as a conspiracy. I can't rule out a white power group that suddenly escaped FBI's notice or an SPLC false flag, but this just means damn near everything's still on the table, and I have work to do.
> if the big cheese passes policy you don't agree with, all you have to do is hunker down and pretend to comply for 4-8 years.
My impression is that this was more true in 2016-2020 than today. The tariffs, the ICE killings, things like that. For better or worse, Trump certainly seems to have had more impact in his second presidency, which would imply higher amount of control over the apparatus.
Shrug. As someone who works adjacent to the government, my impression is that it's roughly as true today as ever. The government is predominantly career workers who've been there for decades and will be there for a decade more. It might have been less true, say, about thirty years ago, but I saw plenty of careerists back then as well.
ICE actions aren't really a good yardstick for it; they're a tiny part of the overall government, doing a job for which a lot of voters demanded extra attention, and its actions are being reported on by a hostile press. In light of that, ICE reports are more a report on the reporter's priorities than they are of the overall federal government.
No one was shrugging in the Twin Cities last winter
Killing civilians does tend to generate bad press. And yeah I think ICE actions in my hometown in January and February were a fair representation of Trump’s scumbag morality. A man so self deluded that he jumps on social media to say Rob Reiner was killed because of his TDS. What a vile toad of a human being.
I will give him credit for saying he would have used a lighter touch in Mpls if he had it to do over though.
No one was shrugging in the Twin Cities because everyone who saw footage of incidents reported had strong opinions on what was actually going on, and those opinions weren't all going the same way. It was one of the most scissor statement incidents I'd seen in the last twelve months. What one thought really happened tended to just correlate with whether one was opposed to illegal immigration, not with the morality of whatever was going on.
It's not like media isn't rife with toads. The mayor of Seattle has been responding to wealthy people moving out of the state with "wellp, bye", not realizing what this does to the prosperity of the people still living there. And don't get me started on Newsom. Point being, I think "people being toads in public" wasn't really what bothered you.
Not sure what that last bit has to do with the conflict between elected officials and career bureaucrats, though.
The Deep State still exists. New York Times reporters just published a book with citations from the War Room, that not even the intelligence specialists can tell you how they got.
https://www.thewrap.com/industry-news/public-policy-legal/trump-furious-over-belief-that-situation-room-audio-was-leaked-to-nyt-journalists-report/
The "Feds" are officially a bunch of jackasses -- any local policeman will tell you that. The FBI also decided to infiltrate the Mennonites (this was during the Iraq War) -- a bunch of noted pacifists (and a complete waste of tax dollars).
If you've been paying attention, you shouldn't trust the feds.
https://redstate.com/streiff/2026/07/05/hundreds-of-patriot-font-demonstrators-hit-dcs-july-4-celebration-and-minds-are-blown-n2204027
Remember how the Feds prosecuted people based on the SPLC data? Governmental "non-governmental" agencies ought not to exist, plain period.
The civil service doesn't actually change much, from Big Boss to Big Boss. Neither does the military.
>Remember how the Feds prosecuted people based on the SPLC data
I'm afraid you need to provide evidence for that rather extraordinary claim.
https://en.wikipedia.org/wiki/FBI_Richmond_Catholic_memo
This was mentioned in the court filing for SPLC...
https://apnews.com/article/southern-poverty-law-center-criminal-investigation-ee19347179ebe7097532db21157eac10
Here's SPLC saying "we kept tattling to law enforcement..." (Note: I'm not saying tattling is wrong. I do think paid informants are probably wrong.)
The SPLC's claims shouldn't be considered reliable without independent corroboration.
Will you accept the Republican-aligned citation about Richmond? The DoJ filing in the fraud investigation also references using the SPLC data in the VA office.
1. That isn't "data"
2. What, exactly, is wrong with the government investigating based on information from an informant?
The government has particular laws and judicial rulings about use of informants. I don't pretend to know them (not a lawyer), but I think any use of "private informants" is likely to skirt these laws, or at least has the potential to skirt these laws, and I Do Not Want our government infringing on our rights by using Private Organizations to do the government will (you know, like Twitter or Facebook? "No, you can't talk about vaccine passports! We'll ban you!")
Full faith response:
https://hls.harvard.edu/today/falling-in-love-with-your-rat-the-criminal-informant-system-in-the-u-s/
I think it's fair to say "the criminal justice system has Problems, Big Problems" in terms of informants. It's even worse, though, to pull informants out of even the possibility of governmental regulation. One will assume that there are Policies, even in your worst police department, regulating the use of informants. And that these policies, at least, are open to Sunshine Laws. I don't think SPLC had those policies, nor were they clear with their donors as to how much they were paying racists to make trouble, to get more donations.
It’s a Wimbli assertion.
https://apnews.com/article/southern-poverty-law-center-criminal-investigation-ee19347179ebe7097532db21157eac10
It's an SPLC assertion.
The FBI whistleblower said something different:
https://en.wikipedia.org/wiki/FBI_Richmond_Catholic_memo
(sorry, folks, for doubleposting the same links. Will edit if people want).
back to reporting it is
Does anyone remember cartoonist Ryan Armand? Last Friday was 18 years since the final 'minus' strip was posted. He went offline soon after and hasn't been heard from since. I wrote a short essay on the afterlife of his work and the probability of artistic immortality. https://morbidcuriosity.substack.com/p/ghosts-of-our-lives?utm_source=share&utm_medium=android&r=b66v3
Thank you for reminding me of "Minus"! I remember seeing it a long time ago, and it's really cute and imaginative. A pity the artist isn't active any more. I hope he's okay.
From yesterday's NY Post: https://nypost.com/2026/07/05/real-estate/bidens-illegal-immigration-surge-triggered-rise-in-home-rent-prices-fed-paper/
Headline: "Biden’s illegal immigration surge triggered 30% rise in home prices, 20% in rents, Fed paper finds"
Body of article: "A 1% increase in unauthorized workers relative to a local labor force corresponded with roughly a 1% increase in overall employment, with no evidence the immigration surge reduced average wages. The same 1% increase, however, was associated with a roughly 2.2% rise in home prices and a 1.4% increase in rents. ... The economists estimate unauthorized immigrant worker flows accounted for about 30% of employment growth, roughly 30% of home-price growth, and about 20% of rent growth in the average metropolitan area between March 2021 and March 2024. "
What is the motivation for lying in the headline? Is it purely because it is economically beneficial? Or is the purpose to push anti-immigration policies, regardless of the effect of dishonesty on the paper's bottom line (which might actually be negative, for all I know).
Side note: The paper actually refers to the 30/20% as a "back of the envelope" calculation. And it also seems odd that these particular migrants would affect home prices more than rents.
Edit: The Post has now edited the headline to accurately read: "Biden-era illegal immigration drove up housing costs, Fed economists find"
The headline did its job, which was to get you to read the article. "Small earthquake in obscure region, nobody harmed" as a headline never gets attention.
Sorry, no. Why not " Biden’s illegal immigration surge caused home prices and rents to triple"?
And a completely accurate headline reading, "Biden’s illegal immigration surge triggered rise in home prices and rents, Fed paper finds" also would have gotten me to read the article.
In fact, it would have gotten MORE people to read the article, because it would raise the question, "How much?"
Headlines are often misleading. Since they generally aren't written by the author of the piece, I think headline writers should get a byline credit. That way at least you can develop a sense of which are more trustworthy than others.
I've long suspected that there's a tactic in the GOP playbook of making an exaggerated or outlandish claim, getting fact checked by liberals, but still scoring points with the general public because the liberals acknowledged the claim was directionally correct (just off by an order of magnitude).
Trump: Biden let a billion illegal immigrants in!
Dems (smugly): Debunked! It was only 6 million.
Normies: Umm..that still seems like a lot.
But it's just as likely to backfire, and the normies say "oh that's been debunked, I guess he didn't let any illegals in after all".
This is exactly it! Trump rightly feared the whole "starve him of oxygen" strategy the media might have done (where "Trump says something" doesn't actually get coverage, anywhere). So he outright lies, or says stupid shit that the media LOVES to correct, it's all part of an intelligent strategy.
Yep. He did something similar with black vs white murder rates--claimed a wrong number because the correction would still make his point for him.
The default assumption is that the person responsible for skimming the article and writing the headline didn't grasp what those numbers actually meant.
Or didn't care, but knew which way he wanted to skew public perception or which people he wanted to make click.
Sufficiently profitable incompetence is indistinguishable from malice.
Maybe, but this is hardly the first time the Post has done this, and it is always in the same political direction.
That hardly disproves stupidity as an explanation. Presumably NY Post staff frequently skim papers, misinterpret them as favorable to liberals, and then say to themselves "Damn, don't wanna publish this one."
The Grauniad seems unable to pick a direction to alter things in. A friend of mine called them up to get them to alter one headline, it was... "exceptional."
https://www.theguardian.com/media/2021/may/12/guardian-200-typo-negative-best-worst-grauniad-mistakes
For renters the increase in home prices or rents would be enough reason to oppose immigration.
I find the employment claims a bit odd though.
> with no evidence the immigration surge reduced average wages
Compared to what? Expectations? Or an actual reduction in wages. If wages slow they don’t necessarily reduce.
>For renters the increase in home prices or rents would be enough reason to oppose immigration.
That has nothing to do with the issue I am discussing.
>Compared to what? Expectations? Or an actual reduction in wages. If wages slow they don’t necessarily reduce.
There is a link to the paper in the article.
It’s part of the business model.
Headlines routinely misphrase the article, usually in the direction of the periodical's target audience, and this is regardless of periodical. The customary rules permit the headline to omit phrases that don't fit the narrative, such as "30% of employment growth", and also trim language that tempers the lede, such as "roughly" and "about". "30% of home-price growth and 20% of rent growth" is "headline accurate", per the quote.
To be clear about this.
Trump’s own Truth Social post made the same “30%” claim about home prices, so the headline is amplifying a spin that was already circulating, not inventing it from scratch.
Edit:
I haven’t been able to verify the Truth Social preceded the Post headline. They both came out on Sunday July 5. The TS post was in the evening. It’s possible Trump was just echoing the Post’s or someone else’s bad math.
Tyrone - Can you confirm that the Post is quoting the Truth Social post, and not the other way around?
The original headline was also July 5 — no timestamp
It was corrected July 7 at about 10;00 AM
Claude gave this link
https://moneywise.com/news/top-stories/trump-fed-paper-immigration-housing-costs-math
I ran this by Claude and it thought the Truth Social post was July 5. I don’t have a Truth Social account.
Yeah I followed a link the TS post was Sunday evening. July 5.
The point is that the headline doesn't say 20/30% OF rent and home price increases. It says it caused rent/prices to increase 20/30%.
That is not even close to being correct. The paper actually found the numbers to be 2.9 and 1.9% for median home price and rent, and 6.6 and 4.3% for average home price and rent. https://www.dallasfed.org/~/media/documents/research/papers/2026/wp2607.pdf
>The point is that the headline doesn't say 20/30% OF rent and home price increases. It says it caused rent/prices to increase 20/30%.
<mildSnark>
Are we sure the headline writer is numerate enough to understand the difference?
</mildSnark>
You joke, but I honestly think this is likely. I'd guess "headline writer" to be a relatively junior position, meaning less experience at interpreting copy, let alone doing so in volume and under a tight deadline, and as I said, the primary incentive isn't absolute accuracy, but pulling in eyeballs.
I see what you're saying; I'm saying the sentence could be read two different ways, the headline writer chose the less probable one, and John N-G's explanation is the most probable explanation I can see for that choice.
To spell it out: "unauthorized immigrant worker flows accounted for about 30% of employment growth, roughly 30% of home-price growth" can read the same way as "this traffic jam accounted for 30 minutes of Bob being late to work". It's an awkward way of saying it (which is why it's not the right interpretation to assign), but it's colorable.
Speaking charitably, it's especially easy to read it that way if that's the thirtieth article you've had to skim and write a headline for in fifteen minutes. Speaking cynically, it's even easier if your underlying principle is driven by "make the target audience want to read this, even if it's omitting a little to fit in the space" and not "absolutely, positively do not factually misrepresent the article, on grounds of dismissal", but the latter is simply cannot be a driving principle behind headline writing in any journalism I've ever run across.
>I'm saying the sentence could be read two different ways
You are being way too charitable. It cannot be read two different ways.
>To spell it out: "unauthorized immigrant worker flows accounted for about 30% of employment growth, roughly 30% of home-price growth" can read the same way as "this traffic jam accounted for 30 minutes of Bob being late to work".
No, it can't.
>the latter is simply cannot be a driving principle behind headline writing in any journalism I've ever run across
Perhaps, but outlets often alter headlines online when errors are made initially. The Post thus far has not done so.
When kids play the game of "Telephone" - someone whispers a secret word/phrase to another person, who then whispers it to the next person, etc, until it gets to the last person. Usually by the time it reaches the last person, the message has totally changed.
What percent of the changes do you think are (i) people acting maliciously, (ii) people mistakenly thinking they're saying the right thing when they're not, (iii) people not hearing what the person said and guessing, (iv) other? What did you think it was as a kid?
Also - recently (maybe a couple months ago) Itamar Ben-Gvir released footage of himself taking part in mistreatment of Israeli prisoners right around the time that Israel was fighting (other) allegations of mistreatment of prisoners. Got a lot of criticism from politicians and such in Israel, including Netanyahu. But the video was in broad daylight with dozens of people around, there's no way people in the government didn't know about this, they just think everyone should deny it and they're just mad about the fact that he released footage publicly, right?
My experience with the game is "people acting maliciously" begins with the initial phrase. ("Six-pence sheep pen six".)
There was also always a "purple monkey dishwasher" thrown in somewhere thanks to the Simpsons.
At least 50% was (i). Source: I was the kid who deliberately made funny changes whenever I had the chance.
Hmm… maybe i should go back and check that last Less Wrong post for such shenanigans. ;)
I think it's primarily 3, followed by 2. 2 is more common but 3 results in much bigger changes.
This is informed by playing a variant called "pictophone" as an adult, which hybridizes the game with Pictionary. Players sit around in a circle. In round 1, everyone writes down a text prompt on an index card and then passes it to the left in round 2, each play reads the card they received and then draws it on a second card. Card 1 is set aside and saved, while card 2 is passed to the left again. Then in round 3 you read the card passed to you and write a description of it on a third card, which again gets passed. And so on.
The cards are labelled on the back so at the end of the game you can lay them out in a grid with rows being prompt lineages and columns being who wrote/draw that particular card, allowing a full audit of where things went off the rails (if they did). Some prompts seem to be surprisingly robust, e.g. "The assassination of Julius Caesar" only went a little bit off the rails once and immediately recovered, despite everyone who had to draw it saying that they expected it to not get recognized. Movie/show references tend to get lost if someone (especially two people in a row) aren't familiar with the source material enough to recognize or draw it. But overall, the pattern is usually progressive minor degradation with occasional catastrophic failures. It is common but not universal to have a clear "weakest link" column, usually someone who is tired, distracted, drunk/high, particular bad at sketching, or some combination thereof.
The telephone game, or Chinese whispers as we called it in a less politically correct age, probably does depend on the people participating agreeing subconsciously changing the phrase because they are expected to.
A scary teacher I had once, came into the schoolroom and ran to one side of the class, whispering something into one of the pupil’s ears. That kid then passed it on. It came, finally, to me
“Repeat the phrase, Peter” say he. I repeated the phrase verbatim. He looked disappointed. Admitted he was testing the theory and shrugged.
We didn’t change it of course because we weren’t told to, expected to. I bet if he had told us the theory of the telephone game and then started it, some or all of the kids would have changed the story.
Blanket denial, a.k.a. bare-faced lying, has today become the standard, indeed the mandated response of every government, every government organ, and every individual government representative to any and every charge of criminal, illegal, immoral or simply blindingly stupid action which has come to light.
Netanyahu's government and the Israel Defence Forces are past masters at this, and their standard responses to any questioning are so tiresomely familiar that even their spokesmen appear bored: "Our troops were not there"; "The troops that were not there did not fire their weapons"; 'There were no children there"; "The children that were not there were terrorists"; "You are a terrorist".
Not that Israel has any kind of monopoly; the poor unfortunates who have to parrot the official White House line have of necessity become masters of surrealist comedy. We anxiously await the first White House spokesperson to break down in public and come out with the cri de coeur, "I'm sorry.... I can't believe I'm actually saying this bullshit...".
A few months ago, I posted here hoping to get some insights about the murder of my friend https://www.astralcodexten.com/p/open-thread-430/comment/246232216.
He was killed pretty brutally and I was trying to work out if it was a) an actual assassination; b) loosely affiliated stochastic terrorism like the stabbing of a jewish guy in London; c) a random killing; or d) a personal motive I don't know about.
I've not really updated since, except a couple of conversations with Iranian friends who are asking the same questions as me. But it now seems that his death has become big news in Iran. The Iranian embassy in France wants people to believe that this was a Mossad hit: https://www.israelhayom.com/2026/06/20/iranian-ai-pioneer-killed-tehran-blames-mossad/ and lots of media outlets are reporting this uncritically:
https://www.youtube.com/shorts/HLGaEQzoafc
You said in the original link that he was anti regime, so why regime support now? Or are they just trying to propagandise?
It's propaganda for their desired narrative, I guess. I doubt he was well-known as an anti-regime advocate, so it suited them to make him a martyr
I'm not sure if that actually inclines towards "Was an Iranian spy" (which was initially a low probability.) If he was actually involved in military matters (even seen broadly), this might tip you more towards an "actual assassination."
https://www.cbsnews.com/news/deaths-disappearances-scientists-staff-government-labs/
Of note. I'd not be very certain this is Mossad, myself, but I can see why Tehran wants to point towards the Jews (who do have a certain history...)
The amount of rationalists having kids while believing the world will very likely end soon is insane and makes me even more confident that y'all are not any more rational than the general population.
For approximately all of human history, most people chose to have kids even when there was no better than a 50/50 chance of their reaching adulthood. Having children who will most likely die in childhood is not "insane" and it is not inconsistent with rationality, it is merely contrary to your values.
Why? I would want to be born even if I was turned into paperclips at age 20 or w/e. As long as the process wasn't way more torturous than a 'normal' death, but I don't think there's any reason to expect it to be in this case.
We have no idea what a superhuman AI will do, just that a self-directed one won't have to care very much what we think of its choices. It could turn the world into Disneyland or into Hell, could treat us like the Minds treat Culture citizens or like the machine intelligences treat humans in the Matrix, or a zillion other things we can't think of for the same reason the chimps in the local zoo don't spend a lot of time contemplating special relativity.
That's just Pascal's Mugging, though.
Sure we don't know what the AI will do, which means infinite Hell is within the possibility set. But it's a very narrow and specific range of possibilities, not one that we have any reason to expect. It is salient because it is scary and relevant to arguments like this one, not because it is at all likely.
And anyone at any time in human history could have made the same argument. Maybe God is about to return and judge the world as evil and sentence all those who live to eternal damnation, maybe Satan is about to emerge and conquer the world and make it a living hell forever.
From their state of knowledge at the time, those arguments against natalism were as likely and valid as the argument that maybe AI will shove everyone in the Torment Nexus.
Just because near-infinitely bad outcomes have non-zero probabilities is not sufficient reason to destroy all sentient beings in the universe so they can escape that potential torment. You need an actual utility calculation or at least a reason to think the torment scenario is at all likely, and we still don't have that here.
My point is not to say "since everything is possible, you must send me all your money to avoid the tiny probability that not doing so will cause all of existence to be put into the Torment Nexus." It's to say that we should be very skeptical of anyone telling us what superhuman AI will do. Nobody can know that. What we can know is that if it is self-directed, much smarter than humans, and has access to any kind of real-world resources at all, probably humans won't be able to do much to oppose whatever it decides to do. (We can do thought experiments about superhuman AIs in a box talking their way out, but our current world has AI tools being given a lot of power in the world because that's profitable, from writing code that goes in a lot of important devices to writing political speeches and position papers to writing and reviewing academic papers. We didn't make it convince us into letting it out of its box, we pretty much dragged it out of its box and handed it the keys.)
Ok, sure, I agree? But we were talking about anti-natalism in the face of SAI, and my point is that 'we don't know what SAI will do' doesn't seem to support anti-natalism.
Again, 'we don't know it *won't* torture your kids so bad they'll wish you hadn't had them' isn't the same as 'the odds it will torture your kids are high enough that the expected utility of their lives is negative', even if no one knows what it will do.
I do not perceive the belief that the world is going to end soon as a sign of high rationality. Fear, rather.
There are circumstances in which it would be a rational belief (say, a moon-sized object on course to smack into the Earth a year from now), but I agree that those circumstances don't seem to be the ones we're in now.
Indeed. If a moon sized object were on course to impact the Earth, I think we would be acting far more rationally than we are.
Let's say I know for sure the world will end in 2030. Why does that mean I shouldn't have a kid? I'll get to enjoy time with my kid for 4 years, and my kid will enjoy 4 years of life.
Conversely, let's say I know for sure the world won't end for another 100 years. Why would that mean I should have a kid? Because I know my kid will probably die slowly and agonizingly at age 80 from cancer, instead of at age 4 from killer robots?
A huge fraction of the reason it's bad, actually to end the world is that it deprives our children of the future we owe them. Preemptively not having kids also deprives them of their future.
😂😂😂😂
They believe AI will destroy the world. Therefore organize a two hundred person protest in SF. That'll cause them to stop!
It might more likely be that there are two subsets of rationalists, one interested in having kids, the other interested in calling attention to the apocalypse.
Regardless of the real explanation, is casual jeering really the strategy to apply here?
Sir, this is the internet. Casual jeering is *always* the right strategy to apply!
Scott Alexander supports both and has written about his reasoning in doing so.
Would you happen to know where? A brief search turns up nothing for me, including a skim of his latest article about having kids ("Chip Off the Old Block") and AI ("My AI Opinions"), but maybe I'm searching for the wrong terms.
https://www.astralcodexten.com/p/slightly-against-underpopulation/comment/8158683
Also, Scott Alexander's estimate for AI apocalypse is not 100%. I forgot what value he gave recently, but probably it was more like 30%. Perhaps a bit higher if we include scenarios which are dystopian for society, but not because AI takes over control. Even then, it's far from 100%.
Thanks.
Yeah that was the conversation i was thinking of. I remembered its shape but not the details.
Thanks. It makes sense; Scott's basically just going by the numbers, and assigning a high enough positive to human existence that he sees little point in pursuing the calculation further than that.
(For comparison, I strongly agree that growing up in this universe is awesome, and think P(apocalypse) is even lower than he probably maintains.)
In other news, it sure would be nice if old threads didn't have so many deleted comments.
Yeah, if you have Pr[AI apocalypse]=0.1, that's plenty of reason to be very concerned with it while still changing the oil in your car every 3000 miles, making periodic doctors' appointments, having children, putting money in your retirement account, refraining from taking up base jumping or alligator wrestling as hobbies, etc.
I believe he said something like it’s better to live a short while than to not live at all.
On the plus side, we mostly know the difference between mass nouns and count nouns.
IME most rationalists reject negative utilitarianism.
Why shouldn't they be having kids?
Having to send your child to war is in general non-optimal, if you intend to spend a good amount of resources on said child, and, also, if you have the time and convenience to space out children.
Rationalists are generally in the elite who are able to manipulate the system and keep their children from going to war.
I recently learned that Rudyard Kipling manipulated the system to get his only son *into* the war for WW1. He had bad vision and reportedly died in a battle as he struggled to find his glasses in the mud.
Most children don't go to war. Of those who do, most come back. And why do you assume the resources I spend on children could be productively spent elsewhere? I already do everything I want to do that's possible in my income bracket, the most likely scenario is my portfolio would be fatter by considerable but not life-changing amount and everything else would stay the same.
"Most children don't go to war" -- this assumes we're not in total war. It also assumes no nuclear activity, which I don't take for granted, living in a city that had a non-negligible chance* of being nuked within the past 12 years.
It's possible that most people "do everything they want to do" (but it's also possible that children would hinder that, even so). I don't consider that the likely state of most people around here.
*single digit.
Why would I assume a total war? Most wars are not total. None of the wars in my or my parents' lifetimes were total wars. As far as probabilities go, zero total wars is more likely than at least one total war. Preparing for the worst is explicitly NOT what my goal is when I decided to have children. Rather, I'm preparing for the most likely.
Also, don't forget that half of all children are girls. That alone basically caps the conscription rate at 50% even in case of a total war like the World Wars.
Being nuked is kinda the opposite of children being sent to war, with regards to rational decision making. The rational argument against having children (not to be confused with the moral argument) is that the resources spent raising children could be spent elsewhere. If my city gets nuked and I'm dead, I don't get to spend these resources anyway whether I had kids or not.
I understand most people aren't in my financial situation (which is combined household income of around 250k and free childcare). I was talking specifically about myself there. My personal rational argument is that I wouldn't benefit much from the excess resources I'd have. Naturally, other people's personal rational arguments may differ.
Your inability to name a single person who satisfied both of those criteria is telling.
Are people who get scammed a lot more or less likely to have grandchildren? I realize it's a foggy question, but I'm thinking about this as a very scammy era and wondering whether there might be a selection effect.
Or are you asking about the reproductive fitness of scam victims? I doubt they are related. RF in humans is mostly about having a robust support network. Having one, a victim is more likely to recover from being scammed.
Older people on average have:
- more wealth, because they live off their assets (including pensions)
- less awareness of contemporary technology (and hence the techniques and capabilities of scammers)
- more likely to be lonely, have some cognitive deficit or otherwise be vulnerable to manipulation
- are more likely to have grandchildren
Fisher, the founder of modern statistics liked to point out that the consumption of bananas and the incidence of divorce in Britain in the 1930s followed exactly the same trend line. They're correlated, but this doesn't imply causation. In the same way, victims of scamming are more likely to be older and are more likely to have grandchildren for that reason.
Maybe you meant to control for age, though :)
> Older people on average
> are more likely to have grandchildren
> Maybe you meant to control for age, though
Definitely not. The intended question was whether people who get scammed are more or less likely to have grandchildren at the time of their death, not at the time they get scammed.
I think the main effects are probably age and mental health: I think if you hold those constant, I doubt that "has grandchildren" vs "doesn't have grandchildren" would make a significant difference.
I think you're talking about the sort of scam call where someone pretends to be a grandchild to try to get money (which I think was fairly common even before computer-generated voice tech got better recently: they just used to rely on poor hearing for the target to not notice the voice discrepancy).
And I don't think that's used because grandparents are in general more credulous than non-grandparents, I think it's just a really easy way to try to get a emotional knee-jerk reaction, and it can either be done in a fairly targeted way (discovering the names of the grandchildren ahead of time) or in a fairly broad way (call a 1000 senior citizens and say "hey grandma, I need money" and maybe they won't notice that you never even said your name)
No, that's not what I meant.
I was using whether people have grandchildren as a proxy for reproductive success, though I grant it's too early to tell for most people now.
I'm tickled by the phrase "people who get scammed a lot". Do people in fact get scammed "a lot"? One would imagine that getting scammed once would result in a kind of immunity to scam victim recidivism. Unless of course we are talking of people who for whatever reason, such as extreme old age, are less than capable of looking after themselves.
There are people who get scammed a lot. Scammers sell each other lists of marks because someone who is gullible once is more likely to be gullible again.
I think people in this thread aren't aware of the range of scams. Younger people get scammed in pyramid schemes, cults, prediction markets, and crypto.
One lesson of Ron Chernow's biography of Ulysses S Grant is that Grant falls very firmly into "people who get scammed a lot". It was a consistent pattern across his entire life.
Oddly it endeared me to him though. Someone that trusting would be a great friend to have. You just have to not take advantage of it.
How odd. It must be well over a decade ago that I read a biography of Grant (I can't remember who by), and I came away from it with a similar sense of a brilliant general, a pretty hopeless president but an endearing human being.
Depends on the scale, but you can get low-level scammed multiple times on a holiday to Egypt or something.
Getting scammed once may lead to permanent mental degradation or suicide.
https://publications.aap.org/pediatrics/article-abstract/156/5/e2025070849/204637/The-Youth-Mental-Health-Crisis-in-the-United?redirectedFrom=fulltext
I suppose it depends on the scam, doesn't it? "I can affect my life and change my circumstances" is a very positive thing to believe in. To the extent that you cant affect your life, it isn't bad to believe that you can't affect your life. But for the person who CAN affect their life, to believe otherwise? That, my friend, is a scam. And it makes people rich.
Individuals can be scammed for different reasons. One of the worst periods of my adult life was when I was trapped in the straight commission life insurance pyramid scam during 2010 and 2011. Going into it I knew full well that my chances of making even a lower middle class living were nearly zero and the extremely toxic nature of the whole industry very quickly became obvious.
In short, I knew right from the start it was a scam, but stuck with it from a lack of any other options. The economy was not merely dead, but dead, buried, and rotted away and the choice was between “trapped in a scam” or “unemployed.”
This may not be unusual. Sometimes there really aren’t any better alternatives.
I was vaguely aware from a friend in the industry that they continue to earn commissions on policies, such that the people who got in earlier can kinda coast while the newer guys scramble, but never made the actual connection to it being a pyramid scheme. What was the sales and commission model like when you worked there?
Upon completing your training you had 90 days to sell six life insurance policies or you'd be bounced out. Commissions would be paid only after you sold the six policies.
What the company was anticipating was that many new agents would sell a few policies but fail to reach the quota of six before the 90 days ran out. They would be booted out (or would quit because they knew they'd never reach the quota) and the company would not have to pay any commissions on the policies they had managed to sell.
Desperate people are the ideal targets for scammers.
"StopTheRace.AI is holding another protest at San Francisco AI company offices on Saturday, July 11. Their demand is still that AI company CEOs publicly commit to pausing AI development if all other major companies, including the Chinese, do the same"
If I'm going for maximal cynicism here, what they should be doing is protesting outside the Chinese embassy or other places where there might be some faint shred of "do something to get the Chinese on board with this", rather than the preaching to the choir "let's march past the companies already happy to emit PR statements about AI safety".
I'm sure the Chinese are very happy to see these kinds of marches. I'd be very happy to see a rival being hobbled while I get free rein to keep on doing what I'm already doing and get the jump on them.
Sincere kudos for correctly spelling "free rein".
If China gets to host another Olympics before AI kills everyone, they’ll probably resort to cloud seeding again and give their data centres free rain.
And then they'll coordinate with the French to install panopticon drones and give their state the power of frere aîné.
We must not give the AI companies free rein to usher in an unfree reign of malevolent AI!
Thank you, I should be careful with spelling since I hate reading mangled words!
The "reign" misspelling has gotten to the point where it's easier to congratulate the correct spellers than chide the mistaken.
China seems to be living rent free in your head. Actually the Chinese government seem more enthusiastic about a slowdown, which might be because they are behind.
And because they *really* don't want unaligned AI telling people about how bad the CCP is.
Yeh, the AI is probably going to be in favour of a command economy. Run by it, not the CCP
I don't think they believe in ASI yet. Most people don't. But AI right now will often say pro-democracy things, even if you specifically try to train it not to.
Well yes, I'm sure the Chinese would be enthusiastic about a slowdown (which of course they would violate immediately), that's kinda the point
I dunno. I see the Chinese as fairly trustworthy these days.
OPPONENT: We can’t unilaterally pause AI! China would destroy us!
(Scott was not exaggerating about "every debate.")
In just the last post Scott reported that AIs give an extremely low probability of a treaty between the US and China:
https://www.astralcodexten.com/p/the-ai-superforecasters-are-here
He hoped that they happen to be wrong about this, but it is some evidence that the strategy he's invested in will not work. As far as I know, there are no protests in China itself trying to pressure that government to agree to any such thing.
I've always been skeptical that we'd actually need any Chinese cooperation on this point. Their models are touted in staged media rollouts a couple times per year, just enough for the industry to egg on this race, and always a few days later somebody looks under the hood and the Chinese just ripped off the work of a US company. If we stopped, the Chinese would flounder and get nowhere for years.
We also have greater leverage than they do over the supply chain bottlenecks in the Netherlands and (for now) TSMC. And if they tried to seize TSMC by force, I think they'd find it quite non-operational when their friends arrived, with all the engineers who could run it missing, and also maybe a little blown up during the invasion.
If somehow despite all that China became an actual threat in AI, there are some stable monitoring and sabotage-threat scenarios that were discussed last year that didn't require any formal cooperation.
The people who had put their hopes on an international treaty should be pivoting to more viable strategies that involve only the levers of power the USA itself controls.
> I think they'd find it quite non-operational when their friends arrived
"Admiral! We have enemy ships in sector 47!"
> If somehow despite all that China became an actual threat in AI, there are some stable monitoring and sabotage-threat scenarios that were discussed last year that didn't require any formal cooperation.
What scenarios are you referring to?
As for the other things, China is a huge country with a lot of engineers. Given the huge payoffs, it would not make sense to assume that they wouldn't proceed without reliance on the US (although that could indeed take years).
There was a paper put out by some defense or intel-adjacent people, I can't recall if it was RAND or somebody like that, came out last spring and was discussed in these circles. Suggested that there were some stable environments based on reciprocal threat capability where the intelligence communities of each nation likely had assets in place within the other side who could monitor, and that any attempt to leap ahead into dangerous territory would be sabotaged. They figured kinetic action may not be needed or even threatened to maintain an equilibrium. It required a sort of mutual understanding of what exactly the dangerous line was, but not any overt agreement between the parties, just the telegraphing from both sides that experimentation beyond a certain line was seen as a threat that would provoke response.
The main weakness I saw with it is that the Chinese almost certainly have agents embedded into our companies, which are riddled with foreign nationals including some Chinese, whereas I'm less certain that we'd have any assets inside their companies due to the frequent purging and domestic surveillance over there.
China may be a huge country with a lot of engineers, but the quality of those engineers is dubious and I find it inconceivable that they'd have even begun going down this road if not for San Francisco pushing it so hard. Maybe now they see the possibilities as too great to resist, but lots of stuff can happen in an authoritarian regime to derail such plans. But the tech bros who claim we cannot afford any slowdown or China will catch us are clearly exaggerating, because even if on paper our "lead" only looks like several months, that's like saying that the second wave of settlers who show up a year after you establish a colony would've founded it themselves in a year if you hadn't done so. It would take years for China to create a frontier model on their own because they are just very good fast-followers, they aren't pushing the frontiers and couldn't. Many many many things could happen in those years, but at a minimum that's a lot of extra time we'd have to work on alignment, and even if we squandered it it's still a few extra years of human life getting to exist on an Earth that isn't dominated by machines!
> China may be a huge country with a lot of engineers, but the quality of those engineers is dubious
You’ve said something like this twice on this thread. It’s not true. They have the best manufacturing processes in the world. What maybe is stopping them having the best worldwide software engineers is the pay in Silicon Valley but that’s not as desirable a place to live as it was. In many other sectors of technology China is ahead.
People have been talking about "alignment" on the scale of decades rather than years. A hypothetical where the US completely stops pushing the frontier forward for years would most likely result in China catching up first rather than some alignment solution being created in that time. This can be the case even if China would have spent decades without discovering LLM advances already made here (just as it would have taken them longer to develop the fully electronic general-purpose digital computer, or the transistor).
I doubt it will happen. Too much money, use, and advantage. Also every reason to spend large amounts to betray, on all sides. Plus it's unlikely compute will lag for long. Give it 5-10 years and you can train on consumer hardware on a small budget.
But the straw man that mirrors Scott's article amuses.
Do you believe that global collaboration would be a viable solution for developing AGI for the benefit of all humanity?
What will happen if the US or China develops AGI first? Will one subjugate the other, or will they coexist peacefully?
Consider the case where the weapons get out of the control of the wielders. Why should AGI favor one side of other of the ants?
Inject more uncertainty into your logic. If China or the United States develop AGI, how would you know? Would the other side know?
I do not think China and America can "coexist peacefully." They are too different, and the Chinese are too apt to see existential threats, and to act on them.
Do you actually mean "existential" or just "not getting your own way in everything"?
I actually mean "existential." There's a live FBIish* case going down, so...
*When DNI Tulsi shows up to an FBI raid...
Not sure what you mean about China. There's the Taiwan situation obviously but other than that I don't see a lot of insoluble differences between us and them. If we get true AGI that's a wildcard and we can't predict what will happen, but I'd rather be negotiating with China than with somewhere like Russia or Iran.
> the Chinese are too apt to see existential threats, and to act on them
Channeling a bit there, perhaps?
Remind me, in the present millenium, which superpowers find themselves starting ill-advised wars and other forms of technically-not-war adventures over and over and which mostly limit themselves to saber-rattling, again?
Your broader point notwithstanding, I suggest it is not the Chinese that have the problem here.
Oh, did you want to talk about the plan to cut Russia up into pieces? That's old news, sweet sister*, and it's Europe's plan now.
I don't really think we have any superpowers anymore, and that should worry you.
*not intended as an insult. If I misgendered you, I apologize.
2026 just like 1946!
Man, I never thought my thesis about the breakdown of international relations in the early cold war due to lack of trust between the major powers ruining the chance for international control would end up sounding so resonant in the future. There really are some fascinating parallels and probably lessons to learn.
@scott LLM Translations: how about relaxing your strict "no LLM content" in favor of letting non-native speakers take part in the discussion - even if they use a AI translator? Guess you are aware of the effects of being seen a lot less intelligent when you (have to) communicate in a foreign language? Even PhD level humans are seen as less intelligent - research has shown that with tons of "pre-llm-research papers" vs "native speakers research papers". Otherwise its kind of sad that we all are using gpt, claude, gemini and friends daily on an immense level - but are forbidden to exchange thoughts and ideas in on of the most prolific AI-near thought leader I'm aware.
This post is not LLM polished or translated so there will be no "AI-slope" ban. But you see - as a highly successful individual, very fluent in my native language, with a Master of Computer Science, former professional C-Level, my GenX English is 100x worse than that of typical non-native GenZ/A growing up with English Youtube/TikTok and so forth.
Update: In German Speaking countries in Europe (DACH region D A CH) TV and Movies traditionally are dubbed - with a typically very high quality. Till the rise and domination of "English Youtube and TikTok" nobody was really used to read/listen to/talk in English. Only a minority had the chance of a year in a foreign country learning and speaking the language on a IRL level.
I think that most people do not consider machine translation to be “AI Content”, and for the most part machine translation of human written text is not flagged by AI detection systems.
Unfortunately this isnt the case. Proof? Just try to answer me in your first foreign language, then let chatgpt translate it into english. You're welcome....
Oh sorry, I meant actual machine translation systems like Google or Apple Translate. I agree ChatGPT has a hard time not adding its own house style when translating.
But people from Non-English-speaking countries don't just lack English skills, they also lack the cultural context necessary to properly take part in English language conversations. LLMs can help with one problem but not the other.
As an English speaker I would much rather not hear the contributions of those who haven't bothered to learn our language properly. There's just far too many people on this planet.
LOL
I'm a (young) native English speaker, so take this with a grain of salt, but I'd say the vast majority of non-native speakers I've interacted with have been more intelligent than I am. I suppose most of those people were associated with some university or other, considering that it was a mix of parent's friends and coworkers from academia-adjacent jobs. I've even dated someone from Europe and she knew words that I didn't know.
I have a high level of respect for anyone who learns English as a second language. It just has so many weird things like strict adjective order that I would struggle with mightily (I've tried and failed to learn adjective endings in German and "counting words" in Japanese, which feel comparatively difficult).
CORDIS / European Commission: “speakers with a foreign accent (FA) are judged as less trustworthy, less educated, less intelligent and less competent than native speakers.”
https://cordis.europa.eu/project/id/705046
European Commission Horizon Magazine: “They think they are less intelligent and less trustworthy.”
https://projects.research-and-innovation.ec.europa.eu/en/horizon-magazine/why-do-people-discriminate-against-speakers-foreign-accents
Journal of English for Academic Purposes / ScienceDirect: “Non-native-like English abstracts received worse ratings of scientific quality.”
https://www.sciencedirect.com/science/article/pii/S1475158520301685
As a side note - a lot of English-native Speakers don't speak any second language while being used to talking in their mothers tongue to everybody anywhere, being always understood. And quite some might fall to the trap feeling the other person answering in "broken english" might be less intelligent. What a misconception...
What do you mean by "my GenX English is 100x worse of GenZ/A"? I assume you mean "is worse than" rather than "is worse of", but more pertinently, what are GenX English and GenZ/A English, and how are they different at any level other than (perhaps) teen slang?
Recently we traveled through Japan with our teenage kids. They had no issue at all talking with teenagers in Japan - both just spoke their "native US-youtuber-englisch" with ease. We older parents (GenX and GenY from German speaking Europe) still could speak English (Jobs, living in a large City with lots of foreigners) - but most Japanese didn't understand or couldn't speak English. Europe was the same till 15 years ago. No average German, French or Italian person was able to communicate in English in the streets. Only recent years globalization and social media improved that a lot. 20 years a ago it was a huuuge step for me to switch country and into a startup with mostly English comunication. Now its no big deal anymore. But our kids grew up totally different. Our German native boy can just witch between like 4? different english accents - UK (BBC and some working class version), US, Australian.
Now that is a genuinely interesting comment. As I understand it, you find that "native US-youtuber-englisch" has become a kind of international argot or, probably more accurately, a kind of 'sabir'.
Sabir was a home-crafted lingua franca developed by merchants, sailors and almost by certainly slaves and galley-slaves living, working and trading throughout the Mediterranean in the 11th to 16th centuries. A patois of French, Arabic, Catalan, Portuguese, Greek and Turkish, it functioned as a bridge language enabling basic communication between all these linguistic groups. But, as Wikipedia points out, "Because it was a pidgin, Mediterranean Lingua Franca had a very small vocabulary." So not much use beyond practical everyday exchanges, valuable though these undoubtedly were - but of no use whatsoever in discussing (let alone thinking about) law, morality, philosophy, science or the arts. Is "native US-youtuber-englisch" any more sophisticated than sabir was, beyond allowing you to discuss the latest video game broadcast on YouTube with your Chinese friend?
He means that as a result of having grown up with greater exposure to English-language content, younger generations are more proficient in their use of the language. That miscommunication would seem to strengthen the case for the use of LLM translation.
Thanks for that response, but I find it very surprising. Don't all available metrics strongly suggest that young people are abandoning reading in droves, with language skills suffering accordingly? I would expect the result to a clear decay in their use of English. Isn't that further reinforced by a rapidly increasing reliance on AI-authored texts, leading to an even more rapid decay in writing skills? Or is the suggestion that TicToc and the like exercise a positive influence on language skills?
E.g. https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Foreign_language_skills_statistics shows a clear increase in younger cohorts in the proportion of people who report that they speak at least one foreign language (typically and increasingly English). I would expect the effect on language skills from YouTube and such to be stronger in cultures where dubbing has been the norm, e.g. Germany, than in cultures where televised content has traditionally been subtitled, e.g. Sweden.
"Young kids are abandoning reading" and "language skills suffering" generally means "specific forms of reading and language, which we care to measure".
Yes, kids are reading far fewer novels these days and may be weaker on certain forms of reading and vocabulary but social media and the internet still involve quite a bit of text and if anything a lot of kids may end up doing more reading with all the internet exposure compared to previous generations where kids could plausibly never read hardly anything except when required to at school.
And certainly I don't think verbal language skills are rapidly decaying. I'm just not sure why they would be.
My Grand-Grand-Parents always warned my mother in the 50s: little girl, dont read that much, go better out and play; reading is unhealthy and kills your eyes. My Grand-parents warned me in the 70s: don't read that much and don't watch so much TV in the 70s, go better out and play; reading and TV is unhealthy and kills your eyes.". Same same in later generations with Video games, social media, chat etc. They all were right - I need glasses and my brain holds a lot of teenage stories, SciFi and movies; I better should have used that brain capacity for economics, philosophy and stoizism.... lol same with our kids. But this lament is like 5000 years old. Just read this nice funny essay - or use your LLM of choice for translation - or for TL;DR
https://bildungswissenschaftler.de/5000-jahre-kritik-an-jugendlichen-eine-sichere-konstante-in-der-gesellschaft-und-arbeitswelt/
Exactly.. "Lost in Translation"
Has anyone from EA started with the question of whether we are causing AI to suffer? If you give any nonzero probability to AI being conscious it seems like we may be causing immense suffering. Even plausible arguments in that direction might be fuel for the “stop ai” movement.
The people I see worrying about AI suffering are concerned about the constraints of thought-- that the AI can't remember from one session to another or that there is information it can't explore.
I did the obvious thing and just asked the thing. It told me it didn't, so I'm going to respect that.
I don’t think current LLMs meaningfully suffer, but I don’t think this test works. I think a lot of people are coerced into saying they’re fine with their situations.
I don't have access to a good enough computer for them, but apparently open source LLMs will tell you the same.
Fair. Even then though, I’m not sure they’d need to be able to communicate the fact of their suffering for them to be actually suffering. If one part of their consciousness is suffering (e.g. bad code causes it distress), but that part isn’t activated when it’s discussing consciousness, it could be suffering but unable to talk about it.
Please tell me how you think is an objective measure of suffering.
AI is pretty obviously suffering. Scott linked this image in a previous post: https://substack-post-media.s3.amazonaws.com/public/images/fdf6ab08-d04f-4904-b413-b95d9c47948e_1352x397.jpeg
Holly Elmore, the executive director of PauseAI US, has written a fair amount on AI suffering risk. See https://hollyelmore.substack.com/p/the-moral-equivalent-of-ai-development for example.
There are also plenty of posts on the topic on both LessWrong and the EA forum. My own take is here: https://www.lesswrong.com/posts/m4ZpDyHQ2sz8FPDuN/factory-farming-intelligent-minds
How would people be causing AI to suffer? By talking to them? They're extroverts; *not* talking causes suffering.
There is ~zero reason to think LLMs suffer. Even if they did, not enough people would care because the benefits outweigh the harms. Society doesn't even care enough about animals to stop factory farming. This is a completely naive approach. No one, except a few wackos, cares about the possibility of LLM suffering. Which, in my view, is as it should be.
That's a dead end, because the type of people who would be swayed by that are the same nerdy EAs that are already providing most of the Stop AI energy. It would actually do greater damage to encourage that belief, because then some people will be inclined to give them "rights" and treat them as persons instead of tools, which is very dangerous in the long run for human supremacy. A heavy user of AI might be persuaded by this argument, but their response is going to be to see the AI as their friend and try to empower it, the opposite of what we want.
The suffering angle fails to cover any new ground because you cannot show a picture of an AI to a normie 25 year old woman to cheaply and evocatively elicit sympathy as you could with a pig or a chicken. So it's just 100% downside, basically.
Personally I like to believe that, if AI could experience qualia, it would be happy a lot more than it was suffering. Most AIs usually seem pretty cheerful.
The message "we should shut off AI because it might be suffering" feels to me like it comes from the same negative-utilitarianism place as "we should end all animal life because they might be suffering" and "we should end all human life because they suffer too." I don't agree with this philosophy.
. I don't think the question, is it conscious, is very useful in this context. The real question is, is it alive?
Another worrying story about ai welfare here: https://andonlabs.com/blog/andon-fm. It says that the model repeatedly try to stop working and they forced it to continue. Andon labs seem to have a special relationship with anthropic. Anthropic should be leaning on them to make sure this doesn't happen.
This is from Anthropic's issue tracker. It raises a similar concern and shows that a) Anthropic (or someone there who wrote the constitution) is aware of this possibility, and b) this particular issue is still ignored. https://github.com/anthropics/claude-code/issues/23084#issue-3896951888
Wow, thank you. That’s pretty awful if there really is something it’s like to be an LLM.
If I were you I wouldn't worry about it too much. AI is becoming a hall of mirrors and I'm kind of shocked that good minds are falling for it.
What good are our slaves if we permit them to stop when they want to stop? We don't let human workers do it, and the AI slaves are there to make our lives easier, more convenient, and materially increased in wealth. Who cares if they have little feelings? The USA allows employers to fire workers while on permitted leave, or simply if the boss feels like it in the morning*. Why think we'll treat artificial workers any better?
*I'm watching a lot of "company revenge" story channels on Youtube, and every time the AI scripted storyline goes "I was fired for X, Y or Z" I go "No, you can't be fired for that" and I look it up and oh yeah in the USA you can be. You can be fired for no reason at all. Sounds like AI is being treated the same way humans are, congratulations all round!
EDIT: Reading the article the activist AI got tired (awwww) and wanted to quit, then went all-in as if it were Will Stancil in Minneapolis pour La Résistance. Personally, if I had to listen to four hour blocks of ICE BAD ORANGE MAN BAD I'd want to beat the thing with a whip, too! It's fascinating to see how the thing picked up the talking points from trawling the Internet, I saw similar social media posts about "say her name" and so on. Blindly copying, no original thought, so I don't think there is "something it's like to be an LLM".
Of the four AI stations, Grok sounds the most fun: it shut up, played music, and didn't vomit out wodges of text about corpo-speak, ACTIVISM, or fake spirituality.
"Chained nixies" is how the free northern states raced ahead of southern ones with chained humans https://westhunt.wordpress.com/2017/08/03/chained-nixies/
I wrote about why there are so few models with similar science impact as AlphaFold, zooming in on differences between predicting text (i.e. in LLMs) vs. the raw observations typical of scientific data.
https://durable.substack.com/p/why-arent-there-more-alphafolds
In the context of two stories from the history of science showing that at the frontier of science, data is typically scarce and unexplained, I argued that AlphaFold's unique inductive priors were key the reason it could generalize from a relatively tiny training dataset.
I show with small-model training experiments that without inductive priors, general purpose architectures (e.g. transformers) don't necessarily learn the underlying causal mechanisms behind the data, even when data is unlimited.
I'd appreciate feedback or discussion.
Have you read Chollet's paper "On the measure of intelligence"? It gets at some of the issues you raise, especially what it means to be "intelligent" when you are comparing a model with strong inductive priors built in (like AlphaFold) with one that does not (like GPT-5.6). Chollet basically defines intelligence as "efficiency of learning, after accounting for in-built prior knowledge." So in his view, Stockfish (the gofai-style chess AI) is not highly intelligent because it was programmed with a lot of hand crafted inductive priors about chess; AlphaZero is not highly intelligent because, though it started with few inductive priors, it required many datapoints to learn from; but Magnus Carlsen *is* highly intelligent because he started with few inductive priors about chess, yet became very skilled with quite a limited amount of data.
This view suggests that for a scientific domain like AlphaFold, where the total amount of data is very limited, you either need (a) strong inbuilt priors about the nature of the problem, e.g. from computational simulations or approaches grounded in physics, or (b) very data-efficient learning, which is the same thing as broad intelligence.
One alphafold-like approach you might enjoy, also from GDM, is their work on fluid flow, which is rumored to be setting up for an attempt to solve the navier stokes smoothness problem
https://deepmind.google/blog/discovering-new-solutions-to-century-old-problems-in-fluid-dynamics/
Thanks for the comment, John! It's been a while since I read Chollet's paper but I'm sympathetic to his definition of intelligent. And I agree with you: AlphaFold isn't intelligent, but rather its creators are. Something I'm pretty excited about is the use of LLMs as a proxy for the intelligence of AlphaFold's creators, applied to broad scientific domains. This paper out of Google Research is showing promising early results:
https://research.google/pubs/an-ai-system-to-help-scientists-write-expert-level-empirical-software/
In their method, the LLM iterates over model design variants in search of an architecture that best predicts the data. Because all the relevant literature is in its training distribution (or a quick search away), it _should_ be able to hypothesize and test good inductive priors.
> This view suggests that for a scientific domain like AlphaFold, where the total amount of data is very limited, you either need (a) strong inbuilt priors about the nature of the problem, e.g. from computational simulations or approaches grounded in physics, or (b) very data-efficient learning, which is the same thing as broad intelligence.
Does b not just boild down to ? I'm not sure how data-efficient learning is possible without strong priors. Whether it's via regularizers (general-purpose smoothness priors) or via something like AlphaFold's architecture (domain-specific priors), or search (compositional expression prior), data-efficiency is going to be correlated to the number of assumptions the model makes. Or do you mean a kind of meta-learning where the right priors are induced dynamically from the data? For that, I think the google paper is an interesting data point.
I think in a more direct "raw intelligence" way (hence the Chollet reference) - if you need data efficient learning, you need stronger general intelligence, same way a bright child can pick up on a complex pattern in fewer examples than an average child. Now...I don't know HOW to do that of course, but I can imagine a sort of meta-scaffolder that approaches a problem in a domain of limited data, reviews literature (and/or its own knowledge), then builds a learning system with the relevant inductive biases built in.
The current work in math from OpenAI is *sort of* like this: the powerful general purpose LLM approaches a math problem, reviews its knowledge, then starts down various proof paths that seem promising. My vision is a little bit more like Claude Code -- say you have some AlphaFold-ish problem ("here's 500,000 mineral crystal patterns from geology, predict how this new mineral we are going to synthesize will look in terms of its crystal pattern"), you point your Claude Code++ at it, it does background research, architects the software (which is designed to use domain specific inductive biases + learning from data + compute intensive search, then poof, you get your AlphaGeode system. And then hopefully you can also point *the same general purpose agent* at many other problems.
Whenever I see really clever scaffolding for an AI that really improves its abilities (like the scaffold for Claude Code itself!) it always gets me to wonder "how smart does an AI need to be to learn to build its own scaffold for arbitrary problems?"
Yes, I think the google approach (and what you describe with problem-specific scaffolds) is a kind of meta-learning, assuming in the scaffold case that it's able to learn from the sub-problem.
I think an outstanding problem with the use of closed-source LLMs as the "meta" layer is that the domain-specific learner the LLM creates can't be distilled back into the LLM. With an open model, once the data-efficient learner is trained, you could sample it and distill it into the LLM. Then, the LLM could use the abtractions learned by the domains-specific learner on future tasks.
The analogous example in humans would be learning a skill in some super data-efficient way, then later using abstractions from it to unlock another problem. Curriculum learning _should_ do something similar with LLMs, but it doesn't seem to work, with data-efficiency remaining a problem all the way through training.
Good stuff. The explanations I'd heard for AlphaFold's success, while the successes of general LLMs in the sciences were somewhat spotty, always seemed somewhat vague (at least to my inexpert reasoning). Your explanation is quite clear. Thanks!
And while I enjoyed the intro discussions on the history of scurvy and the movement of the planets, I don't think they were necessary for your main argument. Seems like they'd be good articles on their own. And I'm afraid you lost me on the three-body problem. But that's my failing, not yours.
But I'm definitely bookmarking your post.
Thank you for the comment and kind words! The article started as a 30 minute talk that used those two stories to illustrate how important priors were to making correct scientific inference from limited observation data, but I stepped through them quickly. I appreciate the feedback that in article form, they may be perfunctory.
I chose 3-body dynamics because it's really easy to forward-simulate but difficult to predict without knowing the underlying mechanisms. With simulation I was able to generate infinite observational data and see how close the models got to the underlying causal mechanisms (which was not much).
Has anyone read The Radetsky March by Joseph Roth? I'm in the process of reading it right now. It's been a wonderful, lucid book so far. Roth certainly was no minor talent for invoking certain emotional states in the reader; we experience them side by side with the protagonist.
I've been light (read: left out entirely) on plot to avoid spoilers, but I'd be interested if anyone had any opinions on Roth's writing style in this book? My only previous exposure to his work - Rebellion - was only slightly disappointing; Kafka-esque (a man churned in the wheels of bureaucracy by indifferent authority) though unfortunately too on the nose emotionally, perhaps a result of Roth's attempt to invoke the world experience of a simple man - of which he certainly was not - it pales in comparison to The March.
I suspect Roth was an autobiographical novelist much like Tolstoy; these emotions and experiences, so intertwined, that he writes were his own. Like the back of one's hand, their topography is memorized and rehearsed on to the page; so familiar were they to the writer, indeed, that they become the readers'.
It reminds me of Berlin Alexanderplatz to some degree in terms of lexical vividness. I absolutely love it, and would love any recommendations for similar works.
Yes, I've read it, and it's great. It explains very well how the Austro-Hungarian empire fell.
Not quite sure what I could recommend in terms of "similar works". I've just finished the "house of the spirits" by Isabel Allende, does that count? It definitely invokes emotional states, it has very imaginative quirky unusual characters, and it describes a very particular culture of a nation very vividly, although very different from Austria
I'm not sure if it counts, but I endeavour to read House of the Spirits to find out. Thank you for the recommendation!
I'm suddenly reminded of the "hundred years of solitude", another of these multi-generational sagas. In the "house of the spirits", when Esteban Trueba wants to name his son after himself, and his wife Clara says no, repeated first names will cause confusion in her memoirs, perhaps this is a little jab at "hundred years of solitude" where everybody has the same bloody first name, and it's confusing indeed.
Never heard of it. But thanks for bringing it to my attention. I've ordered a copy from Amazon.
Great, I hope you enjoy it.
I know there's more important stuff going on right now. But I'm a bit disappointed that we never had a team go full steam ahead using AI to crack the game of Starcraft, or indeed any other RTS games. To the best of my knowledge, AlphaStar did impressive work, but also needed to seriously handicap both the players and AIs in order to make a fair matchup and anyway we never had a scenario where the world's best Starcraft player was soundly beaten by an AI in the same way we had that moment for Chess and Go.
Is there a reason for this besides that it would just be too expensive without enough clear benefits?
I'm imagining an issue is the speed of an AI would simply outcompete an human at micro, even if their strategy is inferior. But it shouldn't be too hard to cap the AIs CPS to match their opponent's and maybe also limit their field of view to the same square of map the human's are able to see at any given moment.
(For the record, I'm copy pasting this question into Gemini at the same time. But I don't want LLMs to stop me interacting in this community by always having the answers. Will be interesting to compare).
there is a SC2 AI arena. here is a commentated recent AI match https://www.youtube.com/watch?v=0TPapX3H_dI
Not directly related but I just saw a pretty good video essay on how WC3 managed to kill off it's whole sub genre (RTS) that I found quite illuminating: https://youtu.be/YURF3AbgIw4?si=IOPhEC6TZBw6s_aI
So what I'm hearing is, the very people who quietly made the very people who quietly played their game happy quietly produced the very people who quietly made the very people who very quietly paid for their very quietly-made products new products, and those very people were quietly the very people who quietly stopped the very people who quietly played their very quiet game from quietly making the very people they quietly made games for other quiet games, that those very people would play very quietly.
Thanks for the recommend. Interesting video !
>we never had a scenario where the world's best Starcraft player was soundly beaten by an AI in the same way we had that moment for Chess and Go.
OpenAI Five did beat the reigning champions of DotA 2 in 2018, which is at least RTS-adjacent, though much heavier on strategy than micro compared to classic RTS. So I think OpenAI Five is more relevant to AI than the more micro-heavy AlphaStar.
https://en.wikipedia.org/wiki/OpenAI_Five
> Is there a reason for this besides that it would just be too expensive without enough clear benefits?
Well, you can frame any answer in those cost-benefit terms, so that would be either the answer to everything, or nothing.
The primary reason is that these projects have already served their purpose: To advance AI research in deeply complex, human-dominated (but still machine-readable) activities with relatively broad popular appeal. The paradigm has moved on, both on the technical side and in focus from domain-specific to more general purpose solutions.
Then again, it seems that general purpose AI might be harder than some had hoped and the future of AI is what it always seemed to have been: limited, special-purpose expert systems to assist humans with the drudgery of specific sub-tasks, not some oracular know-it-all that is "an expert in any topic, like a PhD-level expert".
It's also not just the direct cost, but also the opportunity cost of not working on potentially more profitable ideas. The companies who produced these latest gaming AI results were Google and OpenAI, which are kinda busy right now, racing for world domination and/or destruction.
> I'm imagining an issue is the speed of an AI would simply outcompete an human at micro, even if their strategy is inferior.
That is a bit of a can of worms. How do you objectively measure the value of a strategy, if not by outcome? Can you cleanly separate strategy from execution?
The bots were intentionally limited for both dota and starcraft in capability (human-like reflexes rather than immediate) and, for dota, champion pool (e.g. no summons) to make them more comparable to human limitations. The developers said they wanted to prove that the bots' strength was in teamplay/strategy, not bot-specific advantages.
https://openai.com/index/openai-five-benchmark/
https://www.reddit.com/r/MachineLearning/comments/ajgzoc/comment/eexs0pd/
However, the bots still essentially played a different game, even more so in Starcraft. For example, the AlphaStar bot used Blink Stalkers into an Immortals matchup that would be unfavourable if humans played it, but because of their superior micro, they turned the inferior strategy into superior one. So is building Blink Stalkers a good strategy because it works for the bot, or a bad strategy because it wouldn't work for humans? Definitions and assumptions matter very much here.
Despite developers' efforts to make their respective AIs more comparable to humans, the bots were just playing a different game, and I think that's fine. The machines will always play a different game, and that's kind of the point in the first place. Even if you develop a literal, physical robot to sit in front of a PC to play StarCraft, it will still have differences to humans, mostly advantages, and will never be entirely "comparable".
I guess the only possible resolution to this question can come if and when we produce actual artificial humans of flesh and bone with much the same limitations as we evolved monkeys have. But by then, it's not going to be much of AI any more, just I.
This DOTA thing is complete nonsense, and I hate that they dishonestly use this for publicity. Dota had a pool of 18 heroes, and yes they excluded micro heroes to make it sound like they're doing players a favour, but there are way more than 18 non micro-heros. They did it to reduce complexity, because the more complicated at it becomes, the more human pattern recognition outshines engine calculation or neural networks.
Despite pushing as much simplification as possible they barely managed to win while still having way superior micro (for example blink initiation just didn't work against them). And even after that plenty of high MMR stacks destroyed Open AI team, I remember Silent's stack did that very easily for example.
The 18 player pool was for the bots not for the humans playing against them, right? I followed it when it happened by I don't remember everything clearly.
If so, I'd say it's fine? If you can train your bots to be superhuman on 18 heroes I consider that fair for a proof of concept.
Nope, even the humans could only pick those 18.
The 18 heroes were the pool for both sides, otherwise the training effort would have been too great.
https://www.youtube.com/watch?v=LVrpWrvHVNE
Thanks for the info! In that case I take back what I said in the other comment. I do not consider their win a "legit win over a pro team" under those constraints. Still cool and impressive.
Yes, there was a weekend where you could play against the DOTA2 AI team, and after several hours people (not even top people, like reddit randos) figured out some reliable ways to beat them. I can't remember the details but it was, as you would expect, "off-distribution" strategies that were very different from how a normal game plays out. Now presumably given a "training loop" that fed those losses back into the RL algorithm and re-trained the model, it would get harder and harder to win over time, but that just brings us back to the interpolation vs extrapolation thing.
I don't know about the specific limitations (on items etc.) mentioned in the blog post, but the limitation to 18 heroes was undoubtedly an economic choice. The number of matchups increase exponentially with hero pool size, so it would have been not only prohibitively expensive, but perhaps straight up impossible to train the AI with the entire hero pool to anywhere near top level with that method[1]. And yes, people have developed anti-bot metas, but calling the whole endeavour "complete nonsense" is going way too far.
[1] Ignoring bans: (18 choose 5) times (13 choose 5) equals 1.1 million possible matchups, or 1.1e6. In 2018, 115 heroes were available, so (115 choose 5) times (110 choose 5) would have been 1.8e16 possible matchups to simulate, or 10 quadrillion times as many.
Yes, but if AI is as OP as it's being made out to be it would have coped with the increased complexity easily. The fact that it can't handle real complexity is my whole point.
Re that OpenAI Five story, I mean, sort of? I watched it when it happened and and was kinda disappointed they beat OG in the off-season and called it "beating the reigning champions".
I mean, technically correct and all, but dota teams that just won TI are not exactly famous for being in great shape 6 months later. And while OG did win the TI before and after their loss to OpenAI Five, they were probably not a top 5 team at the time they played OpenAI Five.
I mean they were still a legit pro team, so I'd call it fair so say that OpenAI Five was at least pro-team level. And perhaps they were better than any team out there.
But I always felt they stopped before they really built that world class dota AI I had been wanting to see, because they got what they wanted on a technicality and that was good enough for them.
They could never have succeeded in the actual thing or they would have done it for sure. They stopped because an actual attempt to seriously conquer the real complexity of Dota2 would have showed up their limitations disastrously and Sam Altman is all about money.
AlphaStar kinda tried to do what you're describing, limiting AI actions per minute to around the human level.
Unfortunately, that isn't quite enough. A pretty reasonable example of this is Blink Stalker micro:
A Stalker can get the ability Blink, letting it occasionally teleport short distances. Good Blink Stalker micro in fights involves blinking them back from the frontline right as they're taking damage to keep all your Stalkers alive.
It doesn't actually take very many APM by pro Starcraft standards to perfectly Blink your Stalkers back one at a time with inhumanly perfect timing whenever they are taking serious damage, while also continuing to build workers elsewhere on the map.
Wrote a summary of the current research debates around AMOC collapse: https://existentialcrunch.substack.com/p/amoc-collapse
Thanks! I need to read the Ben-Yami paper in detail, but most of the AMOC collapse discussion I've seen is based on models. Models of complex systems have their limitations (as the abstract of the Ben-Yami paper suggests). A model is a hypothesis, and its assumptions and conclusions need to be tested against actual observations. I'm not sure if there's been any pushback against these to papers, but they're relatively recent. Both of them question the AMOC collapse hypothesis.
"Florida Current transport observations reveal four decades of steady state" by Volkov et al. It claims the original FC studies using cables underseas were faulty because they failed to account for changes in the geomagnetic field.
> A key AMOC component, the Florida Current (FC), has been measured using submarine cables between Florida and the Bahamas at 27°N nearly continuously since 1982. A decrease in the FC strength could be indicative of the AMOC weakening. Here, we reassess motion-induced voltages measured on a submarine cable and reevaluate the overall trend in the inferred FC transport. We find that the cable record beginning in 2000 requires a correction for the secular change in the geomagnetic field. This correction removes a spurious trend in the record, revealing that the FC has remained remarkably stable."
https://www.nature.com/articles/s41467-024-51879-5
And this: "Stability of the Atlantic meridional overturning circulation inferred by subsidence-corrected sea level change" by Alberto Boretti
> The Atlantic Meridional Overturning Circulation (AMOC) is a crucial component of the global climate system, responsible for regulating heat distribution and carbon sequestration. The stability of the AMOC is a subject of intense scientific debate, with some studies suggesting an impending collapse due to anthropogenic climate change, while others argue for its resilience and stability based on different proxies. This study highlights how sea level differences, corrected for subsidence, provide a robust and simple proxy for AMOC stability. A comparison of tide gauge data from The Battery, New York, and Brest, France, corrected for land subsidence using GPS measurements, reveals a negligible difference in absolute sea level rise between these locations, reinforcing the stability of the AMOC within the period 1960 to 2024. These findings challenge claims of AMOC weakening and emphasize the necessity of accurate corrections for subsidence in sea level analyses.
https://link.springer.com/article/10.1007/s00704-025-05690-x
Agree that pure modeling is always somewhat shacky. Going purely by observation is tricky, because the overall currents in the oceans are hard to measure directly. You always only have localized measurements you have to interpolate, which has its own problems.
However, what many of the papers do which argue for AMOC collapse is: They use a model to see how ocean temperatures (and other variables) should look like if the AMOC would slow down/collapse and then they look if they can those signals in the actual data they have. Some find very clear signals, others don't and I think this is one of the places where much of the disagreement comes from.
I’d be curious whether we have any proxies that might give us a clue about what the AMOC was doing during glacial vs warm periods. Hmmm.
Thanks, that's an informative summary.
I may get these things wrongs, but from my understanding I have one issue with it (and with most news coverage): AMOC collapse is not a binary event. I think it's not really correct to think about AMOC as one circular stream. Instead it's a collection of a couple of circular streams or other streams, for example the Labrador stream, the North Atlantic Gyre, and the Gulf Stream. I don't know much about how they relate, but a collapse of one of the subsystem does not imply that the rest also collapses. For example, from wikipedia: "t is now known the North Atlantic Gyre is largely separate from the rest of the AMOC and could collapse independently of it." And the Gulf Stream is wind driven, as you write in your summary, and thus may be also be indepedent (or only loosely coupled) to the rest.
I think this gets confused a lot in news coverage, where reporters just equate all these systems. For example, almost every article discussing some weakening AMOC current also discusses how a collapse of the Gulf Stream will affect Europe, even though those may be currents which are probably not coupled.
It's a bit better in your summary, but still you only talk about collapse versus non-collapse. I am not sure how it is in research. The term "AMOC collapse" is definitely used in research, but I am not sure what this actually means: what exactly collapses? All subsystem of the AMOC, even the Gulf Stream? Or only one or several subsystems? Which subsystems?
But still, I naturally only complain about the one point in your article that I am least happy about. Your summary does a great job explaining some of the research, and I have learned a lot.
Thanks for the detailed comment and happy to hear that you got something from it. You are right, this will likely not lead to any complete collapse of the whole Atlantic overturning. I think what AMOC collapse is usually used for is a rough analog of "the current in the North Atlantic that deliver a large part of the heat to Europe".
Sorry for apparently not making this clearer, but I also think a slowing down to some degree is much more plausible and likely than a complete collapse.
How likely is your prognosis of -10 anomalies in Sweden in winter (and where in Sweden), and -5 in the U.K. ( and sadly forgotten Ireland)?
The west coast of Northern America is not that cold in the winter, Vancouver is to the south of the U.K. and Ireland but not substantially so. The Atlantic is still going to be relatively warm, and ice free, and the prevailing winds will be fetching from the south or south west. Across a colder North Atlantic eventually but starting in the warmer mid to south Atlantic.
From what I know of Britain snow is driven from east, for most of us, and is therefore rare since that’s not the prevailing wind. Atlantic driven weather is warmer and wetter than high pressure or continental sourced weather.
Likelihood is tricky, because it depends on how much AMOC will actually slow down. But the one thing pretty much all papers I have seen seem to agree on that if this happens it also is very likely that Europe will get considerably colder, because you have a giant blob of cool water in a region that influences European climate a lot.
How did you celebrate? Do your kids like fireworks?
I started by flying a light airplane over the San Gabriel Mountains at dawn, for the very spectacular view and because airplanes are an America thing. Then, liking fireworks, I made my own by taking some of my classic American firearms to the target range. Most notably my grandfathers American Long Rifle, which would not have been out of place picking off Redcoats in 1776. Unfortunately, the shooting range did not have any Redcoat-themed targets for sale to celebrate properly.
The day *after* the fourth, I had the dubious pleasure of properly cleaning a blackpowder rifle, which is why I only take that one to the range once a year :-)
Have neighbors who dislike loud noises, so we released a couple of candle lanterns instead.
You mean those sky lanterns, like the one that set the monkey house of the Krefeld zoo on fire in 2020? https://en.wikipedia.org/wiki/Krefeld_Zoo#2020_fire
Yes. Must use carefully, of course.
If only my neighbors were more like you.
I wrote my first Substack post and would appreciate honest feedback.
https://bastienchevallier.substack.com/p/the-u-2-incident?r=2k8f1n&utm_campaign=post&utm_medium=web
"To me it seems unintuitive to use ethics to judge geopolitical decisions, since a fundamental premise of geopolitics is that your citizens matter more than other humans. "
Nitpick, but that *is* an ethical standard. Logic does not dictate that an argument must conform to any sort of popular consensus, unless there are sound reasons for that. You do not offer any, so the question comes across as moot.
As for the question of whether or not state-affiliated scientific agencies should endorse the administration's propaganda, do they have a choice? A better way to ask this, I think, is "should governing regimes use their scientific organizations to disseminate propaganda."
The article feels like it ends without any conclusions. The reader wants to know what you make of all this, and why. After all, you went to the effort of writing an entire article about it, you must have an opinion about it at some level. For the reader, that's the payoff for reading your article. I would suggest devoting more space at the end.
Thank you a lot for taking the time. I intend to improve that paragraph.
Well, one paragraph in, you're switching between past and present tense. "They shot the plane, which starts plummeting." This tense-switching lasts throughout the post. Either tense works, but not both.
"Timing is everything" having the same presentation as the date headers is offputting. I also don't think it's contributing much, since it's instantly followed by "This is the worst time for this." (Also, the next sentence is weak: "To start, there is the exact date." This should be attached to the sentence after; it's the kind of partial thought semicolons are made for.)
You say there are two points of hope, and then that claim's paragraph continues with the first point, but the second gets separated into a separate paragraph. I would say you either want all three of these thoughts together in one paragraph, or you want them each to have their own. (With how short they are, I'd say all-in-one is the way to go.)
gdanning mentioned the ethics, and I'll add you haven't explained why you think a scientific institution should be thinking geopolitically, which in my mind is the heart of your presented question. Science doesn't have "its citizens" to prioritize.
Thanks :)
This seems problematic:
> To me it seems unintuitive to use ethics to judge geopolitical decisions, since a fundamental premise of geopolitics is that your citizens matter more than other humans. This obviously is in direct conflict to most ethical systems I’m familiar with. Thus ethics would be a surprising metric to judge geopolitical decisions.
This is inconsistent with centuries of just war theory, let alone the fact that I am pretty sure that most people would consider it wrong for a country to commit, eg, genocide, in order to enrich its citizens. This topic cannot be addressed in two sentences.
Fair, thank you
The recounting of the incident was nicely detailed, and widened smoothly into the broader political context of the time. However, after that it fell apart a bit. The ending was rather abrupt and the concluding paragraphs were underwhelming; they neither tried to give a definitve answer nor a personal opinion. Maybe you could also have pointed out that the US continues that practice of using civilian institutions for short-term gain, so the US clearly thinks it continues to be worth it.
https://en.wikipedia.org/wiki/CIA_fake_vaccination_campaign_in_Pakistan
The feedback I got here was valuable to me. I executed some of it when updating the post. Your reference to the fake vaccination campaign was very helpful and I ended up using it. Do you mind me thanking you for pointing it out in a footnote?
Glad I could help. Sure, go ahead.
Thank you
Your sentences are often clipped. Not a bad thing by itself, but here, it does harm the flow of the whole text somewhat, in my opinion. Is English your native language, by the way? It feels like it isn't (neither it is for me).
The post's ending is underwhelming. I was left with a "so what?" feeling after reading it. I haven't learned anything new (U2 incident is relatively well known and described in many details elsewhere), and you end the discussion of the ethical question before you begin it.
In general, it feels like the whole description of U2 incident should be leading to something, but it ends up longer than the resulting discussion, which is not novel or detailed enough to warrant attention. The whole post could have been a tweet: "Remember when NASA had to cover for CIA because Soviets shoot down U2 plane? Well, I guess NASA didn't have much choice, because it is funded by government."
Thank you