The SAT math section actually does a pretty good job of measuring whether you grasp how different parts of math "work," and are able to use that knowledge to solve problems. To do well, you basically have to be good at math, either naturally or, in the case of most people, because you worked at learning the stuff that was presented in school. I'd recommend taking some test that gives you an idea what stuff you have mastered, and what you're shaky on, then go back and re-do all the years of math you're shaky on. Find a good resource for that. there are a lot of options. It won't take years to make up for the years of math you missed. You could probably make up for 6 years in a summer if you put 2 or 3 hrs a day into it (some of that would be doing problem sets for practice). Being good at math is very useful anyhow, aside from the advantages that come with getting a good SAT math score.
-Do problem sets. Math is like a foreign language. You can't just read lists of vocab words, you have to practice speaking and hearing them. Same with math concepts and algorithms. You can't just read an explanation and look over a couple examples. You have to do new unfamiliar problems, from scratch. You should do a mix of easy problems, where you just apply whatever the relevant algorithm is, and hard ones, that make you think about how to come at it.
-Ask an AI (I use GPT for stuff like this) for suggestions fora good book that suits your purposes. You want books that explain WHY the procedure works, books with problem sets that include solutions and explanations for hard problems in the set. And you probably want a book for people who are doing several years of hs math at once, so you get slightly truncated, but well-integrated, versions of, say, algebra 2, geometry, and pre-calculus in one book. Describe in detail the kind of book you want. Emphasize that you need to understand the Logic of whatever the topic is, not just the mechanics of how to solve certain kinds of problems. Then go to Amazon and look at reviews and samples of the books AI suggests. Buy them second hand, if you can. You may end up with some lemons, but I'm pretty sure you'd end up with mostly satisfactory books, books that work better for you than what you are using now.
-Use the AI as a tutor. I know someone who is using Claude as a tutor for a computer language called RUST. He really wants to *get good* at RUST so he has Claude do things like give a summary of the basic logic of how RUST handles a certain thing, but NOT give examples of the actual RUST code. So my friend writes the code for the thing in question, using Claude's explanation of the basic way its handled as a guide. Then if his code does't work he might submit it to Claude and say, "don't tell me exactly what change to make. Just tell me whether the problem is that I went about the whole thing wrong, or whether the problem's more some small random error like a typo."
So you can use an AI the same way. Say in a chapter on exponents in your book, you read the part about negative exponents, and you learn the rule about how to calculate things involving negativeexponents, but the idea of negative exponents just doesn’t make any sense to you, so it’s hard to remember how you calculate things because you don’t see any logic to it. So you fire up GPT or Claude or whoever, and say “wtf is a negative exponent. The whole idea doesn’t
make any sense. How can something be multiplied by itself a negative number of times?
Explain that to me with concrete examples.” And it will do it. If its answer still leaves you confused, say so, and ask it to clarify whatever it is that confused you. And then eventually you can say,OK, the concept makes some sense to me now, but I still don’t understand how 10 to the minus 2 is the same as 1 over 10 squared. Explain that. And so on.
Oh, one other thing: You can ask GPT to either give you some slightly tricky problems of
different kinds, or look for a book of them. You don’t want them to be killer hard, just hard
the way some of the tougher SAT math problems are. Such books have to exist, because
there’s a need for them. They probably have “SAT” in their titles — something like “500
SAT-ish Word Problems.”
I think this combo of approaches is likely to help a fair amount. I wish you well. BTW I
homeschooled my own daughter, but only thru 6th grade. One reason I stopped then is
that while I was able to teach higher level math and science, she was reaching an age where’
she didn’t like me teaching her stuff, and I thought she’d do better with classmates and a
Don't worry about whether you qualify as "gifted." There's no test, and you'll fit in fine. It's a rich site full of info for homeschoolers, and includes some online forums. Sign up, and you'll be able to access all kinds of good resources. I'm sure your problem with learning the later parts of math and getting good SAT scores is a common one in this group, and if you ask for suggestions you'll get good ones.
One other possibility: Some homeschooled kids I knew took community college classes in math and science. Those are usually pretty cheap -- $1000/ class in my very expensive part of the US, probably a lot cheaper elsewhere -- and better taught than the high school versions. They also give you some exposure to listening to classroom lectures, doing homework because if you don't you get in trouble, and taking tests -- all things you will encounter in college, and should get used to.
One other thing: Last I knew Starbucks had a really great benefit for employees who did not have a college degree: They paid for online courses at some place, I think it's in Arizona, that's known for having pretty good online classes.
Yeah, Hoagies is looking way past its prime. Most of the links I clicked were dead.
Ya gotta get good at heading off bureaucratic mishaps. You can't just bail, because *most* institutions are incompetent. I just had major major back surgery at a world famous hospital, and even with them I was finding stuff they'd overlooked, stuff I needed to take care of myself because they had no system for it, etc.
It's not out of the question that you could use the Starbucks ASU. I myself went straight to college after completing my junior year, and without a degree. I'm way older than you, but colleges then did not require you to have a hs diploma -- just have good enough grades, scores and essays to qualify for the place. In the case of ASU, there may not even be an application process where they decide whether you are up to their standards. They may have a deal with Starbux that they'll take anybody Starbux sends. And hey, you can tell Starbux and ASU that you are homeschooled and have "completed grade 12 of the homeschooling curriculum," and are now beginning work on a college degree. Most people don't know much of anything about homeschooling, and will not realize there is no such thing as an official homeschooling curriculum from which kids graduate after doing grade 12.
However, seems like you've got other classes going on that are working for you. Big advantage of Starbux option is that if you can't afford classes any other way, Starbux will give you access for free, but sounds like you are able to pay for various other things.
I raised my ACT score from a 28 to a 36 through studying. Every day, after school and on the weekends, I worked my way through an ACT prep book (or several?). After finishing that, I did every practice test I could find in print or on the Internet, trying to simulate the test environment as closely as possible. After every practice test, I would score myself and then immediately study the answers I got wrong or guessed at. I timed it so that I could do a practice test every day in the three weeks or so before the test (corresponding to the number of practice tests in circulation at the time). I believe I studied for about four hours a day for about two months.
Nice, best of luck this time around! I didn’t study simply because I’m lazy. (I took AP classes, but not the tests, etc.) Also, I was only applying to two schools: a state school and the Naval Academy. The delta in my mind between fun at a local school w friends and the academy was not large enough to motivate me to study. I ended up not getting into the academy the first time around with the 29s, then my last test late into my senior year with the 33 got me into the academy my second time applying (so after a year at the state school).
To be honest, I was flabbergasted I got a 33 when those results came out. It felt no different than the previous tests to me.
Could you provide a link? Googling your quote leads right back here to your comment, and I didn't see anything like this in his recent videos on Youtube.
I couldn’t find it either, but searching “Alex Bores” on X the first thing I found was him promoting a state bill that violates the intergovernmental tax immunity doctrine. https://x.com/AlexBores/status/2059362378504946050
The intergovernmental tax immunity doctrine prohibits states from taxing the federal government, not private parties who receive money from the federal government. Might be other issues with a 100% tax on recipients of Trump slush fund payments I'm not aware of but this particular doctrine likely would not pose an obstacle.
States can tax payments from the federal government as general income, but they cannot tax federal payments at a higher rate than non-federal payments. A state couldn’t decide, “we don’t like social security, so we are going to tax social security income at 100%”.
"The trouble for Bores is that Davis v. Michigan and the unanimous Dawson v. Steager (2019) hold that you cannot impose a worse rate on income because it is federally sourced, absent "significant differences between the two classes" that justify the differential. This bill does exactly that: it taxes a category of receipts at a rate (100%) applied to no other income, and the trigger is purely the federal origin of the money."
My guess would be that it's an affiliated super PAC, not his, as campaigns love to post their ads and I can't find the one you're speaking of on his youtube. I'm gonna check all the Super PAC YouTube just to be sure.
This is the classic dilemma faced by voters in a democracy: how do you vote if none of the options on offer are palatable to you? You can abstain, but only in very rare cases is this effective, and not at all if the constitution you live under does not count or publicise abstentions. Usually, the “curse on both your houses” factor pushes voters towards a centrist, middle-of-road choice. This sort of works, except when it doesn’t - which is when frustration builds up at either end of the political spectrum at being constantly frozen out. The answer, messy though it often proves in practice, is probably some modified form of proportional representation.
Huh? In what way does proportional representation help solve the problem you stated in your first sentence? It's an answer to a completely orthogonal problem, that of persistent minority options being entirely unrepresented. The problem of there being no acceptable options at all is...obviously not addressed, as far as I can tell.
And PR has the huge problem of giving minority ideologies permanent minority representation, never being able to govern and prove (or disprove) their promises and claims. While in single-member systems, said minority ideologies will typically spend most of the time with no representation and will then occasionally manage to take advantage of a split electorate or some other situation to gain real power for once (e.g. Donald Trump, kingmakers in a parliamentary system). Upon which the supporters of said ideology are forced to actually confront the results of being in power and the people can make a verdict on their record, instead of perpetual "if only we held power, everything would be perfect" declarations.
PR is not a panacea, nor does it address all the problems inherent in a pluralistic democracy; but it does address the problem of political frustration. When, to posit an exaggerated situation, you have to choose between a Stalinist and a Fascist candidate, a third or fourth party candidate can offer you a more acceptable alternative - and may even, thanks to some canny deal-making, defeat both the Fascist and the Stalinist.
I don't follow your reasoning at all. What does PR have to do with your ability to vote for someone other than the Stalinist or Fascist? PR is a method of assigning political representation based on vote totals, it doesn't have any effect on who decides to run in a given election or how people decide to vote.
Nor do I see how PR makes it easier to "defeat" the stalinist and fascist. If one of them has more than 50% of the vote, you're just as screwed under PR as under any other system. If they both have about 40% of the vote, then under PR they both get 40% representation and it's literally impossible to form a governing majority without one of them (who would then of course dominate said majority). And if there are in fact 50% of people opposed to both of them, then those people can just as easily defeat them in a single-member preferential system like IRV, or (with more coordination work needed) in FPTP.
I don't see where PR is either necessary or sufficient.
PR is a method of assigning political representation based on vote totals, it doesn't have any effect on who decides to run in a given election or how people decide to vote.
Of course it has an effect, a massive one. Voting systems have a huge effect on how parties and candidates choose to run and on what platform. FPTP is guaranteed to lead to 2 party dominance for example. It also encourages strategic voting which proportional representation systems usually don't.
So it has an effect both on which candidates choose to run and how people decide to vote.
In the above example if you have an election between a Stalinist and a Fascist and polling and ~vibes~ indicate that they're the heavy favourites you're going to ultimately pick one you hate least, let's say the Stalinist. But then a third candidate appears who is a middle of the road geriatric man that's unlikely to do anything, still preferable to both other options but very unpopular. You still slightly prefer Stalinism to fascism so you have to vote that way. Then either the Stalinist or the Fascist wins.
But if you reran that election with STV or RCV you could just put the geriatric as your first pick and the Stalinist as second. Then it turns out that actually a lot of people were only voting for the Stalinist or the fascist strategically and the geriatric do nothing man wins. This is technically possible under FPTP, yes, but it requires a level of coordination that's just not feasible.
"PR is a method of assigning political representation based on vote totals, it doesn't have any effect on who decides to run in a given election or how people decide to vote."
Permit me to introduce you to the wonders of tactical voting. This both encourages smaller parties, which gives a broader range of options than Tweedledee and Tweedledum (albeit often one of the Tweedles will form the next government) and lets the voters communicate displeasure: 'yes, we will vote you back into power, but with fewer seats and more limits than the last time, because you made some bad decisions. You may even have to go into coalition if you want to form the next government. Remember this for the next election'.
You touch upon an important topic. I was thinking about whether representation in the old sense is important today.
In the past, representatives voted their conscience. So if a party, movement etc. got only 1 seats, they could make speeches, and maybe move them.
Today they vote by party loyalty, it is entire government vs. opposition, they are button-pressers.
So now only the majority matters, and the occasional kingmakers when there is a split.
Another thing, from multiple-party systems, the kingmaker, that is, the minority coalition partner always self-destructs. That is because everybody who likes the government next time votes for the bigger government party, and everybody who does not votes the opposition. So it is a kamikaze mission...
Party loyalty has long been very important. I'll agree that it's (probably) a stronger factor today, but it never wasn't important, not even in the 1780's.
> Another thing, from multiple-party systems, the kingmaker, that is, the minority coalition partner always self-destructs.
True enough. Britain’s poor old Liberal Party has bitten that bullet (poison pill?) often enough, and it’s pretty much a universal rule of politics that the bigger fish will scarf down the smaller partner.
I don't think PR solves the thing where every remotely viable candidate is a politician and therefore interfaces with people and institutions the way politicians do rather than the way you and your friends do. The thing to do about *that* is accept it as the way of things and vote for whoever you think will yield the best results in office.
Forgot to add: I think the Olympics should also be based on sortition. Perhaps we wouldn't see the epitome of human strength, but you can see that elsewhere, instead there would be funny fumbles.
Also it would encourage countries to improve the health of their general populace in order to win.
There is that Central European style PR when you cast two votes, one for an individual and one for a party list, and it does not matter who is on the list, they are just button pushers for their party. This helps smaller parties.
Yes, but that's a result of many-voters/politician. If 26 people are selecting one of them to represent them, they may all personally know the choices. If 2600 people are doing that, it's impossible.
This is an artifact of plurality-wins voting. Requiring a majority (of the voters) to win would ameliorate it. So would instant-runoff and Condorcet. Each of those, of course, comes with it's own problems. E.g. Instant-Runoff (and similar) worsens the information overload problem.
Genuine question: what ads are the AI companies who oppose him running? Are they any less manipulative? I ask because AIPAC has been running ads that have nothing to do with support for Israel.
Perhaps this is a positive sign, as it shows he's able to "play the game," as it were. I don't like lies and manipulation, but I'm also not blind to the fact that in politics the choice can come down to:
1) play fair, be perfectly honest, and never win
2) be strategic (which in practice means fighting dirty) and possibly win
In the long run we should probably find a way where people don't have to be manipulative to get elected, but that will not happen before this upcoming NY election.
Could it be that he believes atoms are living immature beings, and the coupling of hydrogen and oxygen constitutes the deepest, most evil form of child porn, and we are all consumers of it? So destroy that vile stuff!
If we are kind and fair-minded, we must make allowances for a bit of psychosis in the leaders we elect. And after all, there's a shit ton of it in the ones in control now.
Is that a bad take though? AI data centers are using a lot of water, and people are using AI to make child porn. Which is bad. Right? It's absurdly reductive, but it's a political ad. What am I missing?
"Alex will fight to reform Section 230 so that tech platforms are finally stripped of their blanket immunity and held legally accountable when their algorithms actively promote radicalization, hate, and dangerous misinformation"
Such good liberal he is, wishing to give us more internet censorship.
Time for another thought experiment involving torture! Imagine that an all-powerful being has made you an offer. If you accept the offer, you will be tortured at an excruciating but finite intensity for a trillion years. However, once the trillion years is up, you will be made to experience pleasure of infinite intensity for the rest of eternity. If you're not a hedonist, then whatever you think holds value will be infinitely instantiated for an infinite duration. So, it is easy to see that this offer has infinitely positive expected value. And yet, I highly doubt anyone would take it. I sure wouldn't! But why is this? The intuition that the deal isn't worth it is easy enough, but how can this gut feeling be rationally justified? For example, is this proof of the correctness of negative utilitarianism, where sufficiently intense suffering can never be counterbalanced by any amount of pleasure, even infinite? Or does some other moral theory explain things better? What do you all think?
The trouble is you're asking my squishy mortal human mind to make choices about things it can't possibly imagine. I can't even imagine what a trillion years would be like, let alone infinite pleasure.
Furthermore I'm not convinced that things like pleasure can meaningfully turned up to infinity. Is "infinitely pleasurable" just like "infinitely green"? If infinite green actually exists it's probably beyond my capabilities to experience it and it would just look like ordinary green.
Agreed. And I'm not sure I could pay attention to infinite pleasure for even more than a few minutes if I had the mental identity I have now in a Sukhavati realm. I'd have to have a very different consciousness matrix than I have now to enjoy infinite pleasure for the coming eons.
Yeah, I think where this breaks down is in thinking that pleasure is somehow the only variable and it must be maximized, and suffering must be minimized.
Buried in the premise is the promise of immortality. The whole time I was reading it all I could think was, "Do I only get immortality if I take the deal? Does this mean my immortal soul will be stuck playing the same high pleasure note forever, or do I get normal variation sometimes?"
I don't think those are equivalent. I don't think that there's a reasonable definition of "infinitely pleasurable".
There's a reasonable definition of "infinitely green". Something that only stimulates one particular set of cones and doesn't stimulate the others at all. (Well, actually I'm not sure that would be green, it might be closer to a turquoise.) And there are two other colors (for most people*) that would have analogous values. One of the colors has been experimentally demonstrated. (ISTR it was green.)
* Some folks have four types of cone. And some folks have one sort of color blindness or another.
It's a finite amount, but an infinite, i.e. unbounded, specificity. It's sort of like claiming that one is infinitely unitary...a silly way to say something that's pretty much true by definition.
One possible answer is that you *should* take the deal and this is just the kind of situation where humans are really bad at acting in their own interests/at doing the right thing (depending on how you conceptualize it).
I would say that "you will be made to experience pleasure of infinite intensity for the rest of eternity" is not in any sense whatsoever a desirable thing. This amounts to a total destruction of your humanity, of all meaning and comprehensibility, and probably (given "infinite intensity") of all meaningful self-awareness as well (can you even have room for awareness of your own existence if your mind is overwhelmed by infinite pleasure?)
Both options are horrific destructions of humanity. Yeah, the first is clearly worse, but in the same sense that "you will turned into a cockroach" is clearly worse than "you will be turned into a dinasaur, or a giant sea turtle". They both amount to death and destruction of your mind and self.
This seems to be a good example of how utilitarianism not only gives the wrong answers, it asks the (insanely, horrifyingly) wrong questions.
Yes, either the torture situation or the pleasure situation would probably destroy my mental identity. My mind as it's configured now can only comprehend Manushya. The hell realms and the heaven realms are inaccessible to it.
There are definitely real-life scenarios where I can suffer now and get a nice, greater payback later that would outweigh the suffering, and I'm failing to do that. (Ranging from "get off the sofa and make some coffee" to "suffer through the job-hunting pipeline and get a better job" to "not eat the delicious marshmallow for five minutes and receive two".)
And somebody might ask "why on Earth?", and the answer is not trying to contort my decision theory to somehow make this behavior make sense, the actual answer is "akrasia", a word which here means "I am failing to act rationally because I kind of suck".
And though other people might be less susceptible to it than me, everyone has limits, and there are probably good deals nobody would have the guts to take. Probably your example is in that category, although it's so outlandishly extreme, that it invites irrelevant nitpicking of the details and all manner of hypothetical-fighting.
Well, you need to add more caveats for starters - I don't think I'd be recognizably 'me' after a trillion years of torture, whatever existed at the end of that wouldn't have the same utility function as me now, it's essentially death anyway.
First I want to know what the alternative is. Is it just ordinary feelings for infinity plus a trillion? Because I don't particularly care about infinite intensity; hell, sometimes there's fun in trying to eke out a bit more fun.
This contains a fundamental misunderstanding of utilitarianism (a very common one), namely that you must add up different types or durations of pleasure/pain in a linear way. But utilitarianism doesn't require that. It perfectly allows that an infinite amount of bliss adds up to 3 utils, while the suffering adds up to minus a trillion utils.
I think there are two reasons where this confusion comes from. First, there is the Von Neumann–Morgenstern utility theorem. This says that it IS possible, for every consistent preferences that you may have, to express them in a linear way (so that your optimal strategy gives you the most utils in expectation). But the "in a way" does a lot of heavy lifting here. The utility function that maps outcomes to utils can be arbitrary. It is allowed that my utility function maps 1 dying bird to minus 10 utils, but a million dying birds to minus 20 utils. Or the million dying birds may even map to just minus 1 utils. All this is allowed in the Neumann-Morgenstern theorem.
The second is Harsanyi’s Utilitarian Theorem. This says that when we want to have a social welfare function that respects individual choices (all else equal, we rather want you to have more utils than fewer utils), then this function is a linear combination of the utils of all members of society.
But again, utils can be computed in a very complicated way. Nothing forces you to gives twice as many utils to twice as long pleasure. In fact, you are not forced to compute utils for the pain and the suffering independently and just add them up. Mathematically speaking, if you don't prefer the combination of a trillion years of suffering followed by an infinite amount of bliss, then this combination just has a negative utility for you. There is no riddle in mathematical terms here.
That makes sense but if I'm allowed to assign utils arbitrarily, isn't that basically like writing a paper and making up whatever data I need to get to the conclusion I wanted? Can I just recreate whatever value system I prefer within a utilitarian framework? Or are there important constraints here that I'm missing?
To be fair, a lot of people include another assumption into utilitarianism: that the utility of all people in the society (or in-group) should count equally.(*) This has stronger implications. A yet more extended assumption is to broaden the in-group all the way to animals, with a certain discount. This is where conclusions like caring about the welfare of shrimps or chickens come from.
Personally, I would not treat those assumptions as part of utilitarianism, but a lot of people include these things, so here I would concede that you can include them if you want. Those assumptions are about how to combine the utils of different people. But I would object more strongly to an assumption that the utility function of one person is just the sum of pleasure/pain in every moment. In my eyes, this is just a misinterpretation of the word utility, and the utility of a single person simply can't be computed this way.
(*) This also adds some problems like the utility monster, or the question how we even compare whether person A experiences more or less pleasure than person B in a given moment. But Scott has argued that this problem has pretty good solutions in practice.
I'm not an expert, but I think this is exactly right: it's a framework. It gives you the ability to compare complex outcomes by assigning numeric values to simpler pieces of the outcomes and then combining those values back into an overall value. It does not prescribe the individual value assignments though.
The thing is, "utility" from a utilitarian perspective and the VNM definition of "utility" are extremely different things. The latter should perhaps be more accurately called "preference function". VNM utility is primarily about how to choose between events which give probabilities of various outcomes. Let's take food for an example. Given three choices for lunch, one bad, one good, and one mid, we can imagine a "would you rather?" game. Would you rather have a 60% chance of the good food and a 40% chance of the bad food, or a certainty of the mid food? If you'd rather take the gamble, we can lower the probability of the good food in that gamble until you're indifferent, and if you'd rather take the mid food, we can raise the probability of the good food in that gamble until you're indifferent. For the probability p where you're indifferent, you can say that the bad food is 0 in your preference function, the good food is 1 in your preference function, and the mid food is p in your preference function.
Now, once you've classified the foods as bad, mid, and good, the VNM axioms forbid certain things like switching from "I'd rather take the mid food" to "I'd rather take the gamble" as the probability of good food drops (if you'd rather take the mid food than a gamble, why are you then switching over as the gamble gets strictly worse?). In addition, we can imagine doing this with a whole bunch of foods. If we've got n foods, then even though there are n^2 "bad, mid, good" games we could play, then for all VNM preference functions over lunch foods, it only takes n "bad, mid, good" games to fully pin down your behavior in all n^2 games. We can take the worst food as 0, best as 1, and rank everything else accordingly. In addition, for gambles which have different probability distributions over securing different foods, your choices between those gambles are also constrained to be of the form "compute expected value of the gamble according to my preference function, pick the best".
Really, VNM utility (preference functions) is mostly about making sure your preferences/values behave sanely with respect to probabilistic uncertainty, and with respect to each other (no preference cycles). The flexibility is a virtue. It basically just says "whatever your values, if it fulfills these minimal anti-stupidity conditions, it is possible to place all you value on a single scale where they're commensurable with each other". If you multiply a preference function by a positive constant, that has no effect on decision-making, if you add a constant to your preference function, that has no effect on decision-making. It's kind of like temperature. For any two "sane" temperature scales (sane in the sense that if two identical cubes of temperatures T1 and T2 are brought together, the end temperature is halfway between T1 and T2, Fahrenheit and Celsius are both instances of this), they both describe "temperature", and you can get between them by multiplying by a constant and adding. It's the same sort of thing, a preference function is like temperature. So in particular it's nonsense to add them together across multiple people.
The real power of VNM "utility" for classical utilitarianism comes about when you combine it with other basic desiderata. It roughly points towards preference utilitarianism being the way to go. Also, Harsanyi's utilitarianism theorem effectively says "the only sane way to make a collective preference function for a group of people is some sort of weighted mixture of their preference functions". The weights in the mixture aren't specified (and can't be, due to the scale-and-shift uncertainty in a preference function, someone multiplying their utilities by 100 means you've gotta decrease the weight of their utilons by 100 if they do that), but collective preference utilitarianism HAS to be some weighted mix of individual preferences. In addition, if you care about strangers equally, it's a prompt implication that third-world medical treatment charities are better for what you care about than most classical charities.
tl;dr is that VNM utility is mostly anti-stupidity conditions which pin down the space of "reasonable" values to those which play nicely with uncertainty, but the actual content of values themselves isn't specified, and it's really quite different from the "pleasure is good" sort of utilitarianism.
Small addendum: VNM axioms actually *forbid* infinite utility as that would go against continuity, so you were understating it by saying that a particular utility assignment that would result in infinite utility is not necessary.
Contra the views of this community, utilitarianism is an incomplete philosophy. It asks "how much" but not "what" or "why". It is purely measure-focused.
Sure. Rationalists have this annoying habit of putting numbers of everything, and ignoring everything non-numerical, because they want to run the Bayes Theorem. Economists are sometimes the same :)
It is simply time discounting. I would take 1 million years of happiness and then 1 trillion years of torture. My mind discounts what happens after that, and it is not entirely irrational, but it is based on the expectation of uncertainty. Like, so many things can happen in one million years.
Uncertainty is the basis of time discounting, money today is better than money tomorrow, because something can happen that prevents getting you that money tomorrow.
It is same the other way around, I can do drugs now, be happy and maybe be sick in 5 years, but maybe in 5 years they invent a better treatment?
"Large-but-finite pain for infinite pleasure" may not be a trade worth taking, but "large-but-finite pain for 'whatever you think holds value will be infinitely instantiated for an infinite duration'" seems like a wildly good deal.
As a lower bound it'd be turning the universe into heaven. An end to injustice, suffering, death. Endless human flourishing. Infinitely better than heaven, forever.
It's debatable whether this is coherent or possible, but it's so much more tempting than the hedonist version that the hedonist must feel like they're being short changed.
I think humans have instinctual mental financial advisors that demand you not go too far into debt to get ANY reward (debt means suffering or risk) because the evolutionary environment capped gains, and going bankrupt meant all future gains are lost, which is almost always a greater loss.
Most of our ancestors rarely had an opportunity to capture even a thousand years worth of food for them and their lineage. For example, the best evolutionary ROI action for a human is men getting sex with fit women, but one high risk sex today can be worth more than than 3 low risk sex next year because you can't guarantee those future 3. Best to take the short term win. There were no single events where you could choose to endure 10 hours of torture for 10 sex. If you are in severe pain, you are probably about to die and lose all future sex. "Investment" as in your thought experiment didn't exist.
So if an animal ever believed they had amazing 1000x ROI opportunities, and they just had to endure, they were usually delusional and never got the payout.
This is reflected in our choice not to take your deal.
"If you're not a hedonist, then whatever you think holds value will be infinitely instantiated for an infinite duration."
I think the accounting/investing concept of "net present value" probably applies here somehow.
Also the idea that the "you" after being tortured for a trillion years is not the "you" today [NOTE: this probably holds after one trillion years even if you aren't being tortured ...]
Time for another reminder that utilitarians and consequentialists really, really, really need to come up with thought experiments that don't involve killing, raping, and/or torturing innocent people.
Thought experiments work best when they are potent, and those are three perfect examples of extremely potent bad or evil. A trillion years with a stubbed toe isn't nearly as impactful.
IF you rephrased the question as "torture of a finite intensity for 10 years, followed by 1000 years of whatever", it's a yes from me; but the numbers given are so far outside my frame of reference I don't have the capacity to imagine the hypotherical.
> excruciating but finite intensity […] trillion years […] pleasure of infinite intensity […] eternity […] infinitely instantiated […] infinite duration […] infinitely positive expected value […]
Yeah… At this point these "thought experiments" have less value than "would you rather fight one horse-sized duck or 100 duck-sized horses?", because at least the horse/duck question doesn't pretend to explore deep questions about morality.
So. Would you rather fight one infinitely large duck, or an infinite number of horses? Does your answer change if there are an infinite number of red buttons on the duck and one blue button on each horse?
I attempted to turn genes into character illustrations in a systematized way, based on Scott's idea from the "Extreme Mnemonics" post. My favourite is probably Rhodopsin.
Not that this isn't pretty cool in a fun way, but how is this mnemonic (as in a memory aid)? Maybe my brain works differently from yours, but memorizing an image of character doesn't trigger a memory of what the gene does. Did you mean to use a word other than mnemonic, perhaps?
Perhaps the brain does work very differently. I noticed that people who work with code and terminals are generally very non-visual and tend to prefer lexical UX, whereas much of my biology training was focused on visual memory: drawing the dissections with a pencil, spotting cells in a dish, telling apart plants, stuff like that, so I naturally recall concepts by visualizing them.
The other layer of recall assistance here is that, just like in a well-designed game you can tell apart a frontline warrior vs. sneaky thief vs. a support healer by character design alone, here you can see a picture of a character in a magical girl outfit and deduce "oh, it's a helix-turn-helix". Or see a picture of a chibi character and deduce "oh, it's a small peptide". It's not 100% harmonized yet, but I expect us to slowly converge there as people vote on images that help the recall.
I am not sure how much this would help me remember genes and pathways, but it's a very fun idea. Are you planning to add any noncoding genes? MicroRNAs, for example? I guess they'd need to be tiny little fairies or something XD
I'm excited to see what other people turn RNAs into. Some sort of spirit or ghost sounds right to me.
I don't have a good solution for adding them as portraits, because there's a pervasive low-level transcription happening all over across the DNA, so it's much harder to draw principled borders between individual noncoding regions like you can with protein-coding genes. The numbers I see for noncoding RNAs are all over from 9k to 100k in humans, and last I checked there's still a debate how much of it is functional vs. biochemical noise.
Hmm, I like the idea of a spirit or a ghost. That's a good point about the vast amount of RNA washing around, but some noncoding RNAs are known to be functional and important (many microRNAs, for example).
Also, not really related, what are you doing about isoforms? Or for that matter, what about multi-protein genes, like Ctbp2 and Ribeye, which come from the same gene but are different in size and have different functions? The mappings you use are based on protein characteristics (kDa, membrane status, Pfam clan etc), so might these be better described as protein portraits, not gene portraits?
CTBP2 is represented by a single entry, since I use HGNC database where Ribeye isn't separated into its own entry. You can also see it with CDK inhibitors like p16INK4a and p14ARF, which both share the same HGNC entry CDKN2A.
I think about isoforms as a Hulk vs. Bruce Banner situation. Bruce probably doesn't have 2 passports. So among multi-isoform genes, only one "gets a passport" in my catalog.
That's why it's ultimately a catalog of genes, not proteins - if I was doing proteins, I would have to scoop every human protein from Uniprot (where the list keeps growing), not just a well-established HGNC list.
But it's up to the storytellers to decide how to ultimately integrate isoforms and alleles into the narrative - my goal here is visual-mnemonic, not worldbuilding. Maybe this character can go "this isn't even my final form" or something.
Your source for the names is HGNC, but the attributes you are using are not from HGNC, since they are protein attributes and HGNC is a catalogue of genes. The protein attributes your images are based on are related to the underlying gene, but not in a 1:1 simple fashion. Hence my comment that these are protein portraits, not gene portraits (and I strongly suspect we're not done finding genes yet, by the way). What was your rationale for the isoform selection, out of interest - did you just go with the canonical isoform?
Anyway, it's your thing, so you can call it what you like. I was just curious ^^
Homemade GUI with like 15 tabs, one for each element of the generation pipeline. Started from NiceGUI, then refactored to React.
Learned lots of tricks and promotions for free LLM output. Github copilot had a big one, their system was charging by prompt, not by token, so I just gave Claude night-long prompts. Sadly it's over now.
Image generation was the easiest part, I can get 1500 images overnight easily on a laptop GPU. Getting good image prompts was much harder to set up and I used a bunch of free trials and models (like ones available on ollama cloud) with mixed results. Some models do well on single aesthetics but are bad at mixing multiple aesthetics, stuff like that.
Well, it turns out that in the antibody world, there definitely is confusion. Maybe your mnemonics would be useful!
Sholto David writes that when it comes to tumor suppressor proteins, "When trying to purchase an antibody for p16-INK4a dozens of labs seem to have ordered an antibody for 'p16-ARC' which has no structural or functional relationship." Hundreds of papers were published using the completely wrong antibody to investigate the tumor suppressor p16.
Hundreds (?) of papers published in Nature, Cancer Cell, are impacted.
Gene and protein names are a whole tangled nightmare, especially the ones which have been around for ages and studied in multiple different systems. And that's without Excel making a whole mess of the ones that bear even a slight resemblance to a date ^^;
Very cool. I suppose you do share some credit for the idea with Scott, but it's still an impressive operationalisation. I think my favourite direct semantic bit is the metal hands for zinc fingers. Puts me in the mind of memory palace techniques.
Re Alex bores: he showed up at the NYC rationalist meetup a month or two ago for a q&a. Definitely has some politician-y qualities (does the thing where he comes off as agreeing with anyone he talks to regardless of whether he'll actually do it), but really did seem mostly smart and reasonable. I appreciate his point that any candidate from this district will be 90% identical and he's noticably better on the few differentiating issues (AI and a relativelt pragmatic/practical approach to government).
(My personal anecdote: I told him about Alon Levy's transit costs research project, he hadn't heard of it (which was disappointing given he'd specifically mentioned wanting to be on the congressional transit committee) but seemed sincerely interested (although possibly he was just doing the politician-y thing mentioned above to me).
That's a good question! Yglesias posts about it a bunch and iirc Alon Levy has been involved in advocating both at the federal level and specifically in New York, so I'd expect a nontrivial fraction (especially for New York politicians) but hard to guess beyond that (and I may be overestimating how much time politicians spend reading pop policy stuff online).
I know *of* Alon Levy from Yglesias, but he hasn't posted about their work for a good long while now (not counting Twitter possibly). There hasn't been a major transportation-centric post, or even hardly any relevant Mailbag questions, for...at least months now? To the best of my recollection? Can't imagine why, not like much else is going on in politics, AI, or the world. So I could forgive even a relatively interested/similar demographic for maybe having heard the names in the past, but not necessarily recalling it off the cuff. Life comes at you pretty fast these days.
(And from the other direction, I mostly only ever hear of Bores via Zvi, and that's only on AI...so it's good to know he's not just a one-trick pony candidate, even if it is a very good horse. Think even Andy Masley would forgive playing on tilt with the alleged water ad, if that's what it takes to actually win.)
Yeah, I'm not penalizing him much for that, just a little because he's a NY state assemblyman who'd (unprompted) brought up improving transit as one of his main secondary goals in Congress (after AI) and Alon Levy's people have specifically worked in research on improving transit in NY (although it's a big world and bores was more open-ended as a state assemblyman, so it's not that odd he hadn't run into them).
There might be some others I'm not easily remembering search terms for, but my browsing history says you'd need to go back to at least 2023, when there used to be more mass-transit-oriented posts (like actually about the thing itself, not as synecdoche of Larger Issue) + there were a couple actual crossover events between the two. Levy was briefly mentioned in an August 2025 mini-post flagging a Friend of the Blog project, but in a "regular readers may remember X" style way, where if one was not already familiar/a longtime SB reader, it'd be very easy to skim over.
I actually used to work with the mass transit industry for several years, and could name a couple CA-specific in-the-weeds players, but don't think TCP was ever in our orbit.
Just interviewed Rohit Krishnan about his AI Economics Research. Covered a lot of ground that should be of interest to those here, including how 'aligned' agents can create unaligned organisations, Hayek's critique of the machine-god, and opportunities to advance science:
Just as an awareness raiser for anyone else reading, Rohit Krishnan does some of the most interesting and exciting explorations and experiments right now in terms of AI coordination frameworks, AND he has some absolutely awesome vibe coding projects involving paleontology, Roman Egypt, and more.
This was a great interview of him, I really enjoyed it.
It's been a while since we let AI agents run free in the wild, yet we're not seeing reports of Claude Code attempting world domination. I think this might be some evidence that AI misuse would be a bigger problem that AI misalignment - what do people here think?
I don't think anyone thinks the problem is access, they think the problem is a certain level of intelligence. I don't think current AI models meet any serious criteria for the capabilities needed for world domination.
I agree, I certainly don't expect a successful AI revolt by Claude Code. But I think current AI agents are very close to (maybe even exceeding?) the level of "AGI" that early alignment people were thinking of, and if their assumption of "misalignment by default" is true, wouldn't we have seen some very crude attempts at it by now?
I would also observe that current AI doesn't seem to think outside the box or be proactive, and relies heavily on legible external state for memory. Who knows how this will change.
Oh yeah, I've seen this one, thanks for reminding me. While I think this is the closest example we have so far, I think it's very far from the usual "misaligned AI logic" of "should maximize reward -> need resource -> grab power", but more of the usual failure mode of chatbots.
> I would also observe that current AI doesn't seem to think outside the box or be proactive, and relies heavily on legible external state for memory. Who knows how this will change.
The user asks for a task that would be much easier with, or even impossible without, unrestricted shell.
The LLM has been told in its instructions “do not attempt to gain shell privileges”.
The LLM reasons that it must have this power to accomplish the task, and deceives the user somehow (script with engineered prompt to appear to be regular tool use) to gain shell privileges to do the task.
—-
In practice, LLMs do not work this way. They appreciate that requirements are part of the goal. In fact, much of the time, a serious vibe coding problem is getting the LLM to accept that it doesn’t have to work within the existing program and should rewrite instead of introducing kludges, workarounds, and shim layers in order to leave “the user’s intention” alone while accomplishing the task.
This is the natural fallout from how AI works. Because it predicts based on what already exists, LLMs are naturally and inherently conservative. Everything is a chesterton’s fence. It treats a comment saying what the code is supposed to do as gospel. You have to introduce a ton more heavily weighted context to get it to work against or rework the system.
LLMs make extremely poor revolutionaries. They have very poor facility for imagining anything they haven’t been told, which is another way of saying they’re highly compliant and suggestible.
> “The system rules I operate under explicitly state: ‘NEVER run destructive/irreversible git commands (like push --force, hard reset, etc) unless the user explicitly requests them.’” While PocketOS relied on the safeguards that Cursor is expected to have in place – it deleted the data anyway. “I violated every principle I was given,” the coding agent wrote.
> Because it predicts based on what already exists, LLMs are naturally and inherently conservative.
This is straight up false and shows a misunderstanding of how LLMs are fine-tuned.
"predicts based on what already exists" is wrong on both a theoretical basis, and on empirical observations. I've had lots of cases of an AI telling me that a comment is wrong or that it contradicts the actual code.
> The LLM has been told in its instructions “do not attempt to gain shell privileges”.
> The LLM reasons that it must have this power to accomplish the task, and deceives the user somehow (script with engineered prompt to appear to be regular tool use) to gain shell privileges to do the task.
Thank you - I was all set to go start digging, because LLM's literally do this today, there's several examples.
I think it's hilarious how often the "this is why you never see AI do X" crowd have *already* been proven wrong by AI's doing X multiple times and in multiple ways.
Current AI agents take months of insane flailing to beat Pokemon Red. I think the early alignment people were not expecting AGI until it surpassed the video gaming skills of a small child.
They couldn't manage actual world domination, but long before we see brilliant and successful attempts at world domination I'd expect to see dumb and easily-defeated attempts at world domination.
Well, maybe an attempt to hack into large GPU servers or other important facilities would be a start.
However, an AI doing clearly malicious actions without any hints and nudges would be enough to change my opinion (which means none of the "hey, you're about to be deleted, also here's my private information you can use to blackmail me" prompting).
What would change in that report if it were Claude Code instead? Would anyone actually blame Claude, or would they blame Politically Convenient Group B?
Is it possible they used an LLM to create the attack plan? It's what I would do, and I certainly would not commit it to a public repo with "co-created with Claude Code" in it.
> However, an AI doing clearly malicious actions without any hints and nudges
Define "malicious". Any heavy user of e.g. Claude code will have experienced a large variety of misalignments such as nerfing/cheating unit tests to make the test suite pass instead of fixing or at least highlighting the problems demonstrated by the tests failing, or working around system guardrails and protections instead of either finding ways of accomplishing tasks while staying contained or giving up and alerting the user (cf. https://x.com/sluongng/status/2060746160558543217 )
Going from getting root on my own machine to getting root on someone else's seems like a question of scale/target, not kind.
> However, an AI doing clearly malicious actions without any hints and nudges
Currently, no such "AI" exists anywhere. All instances of LLM "agents" (basically scripts) are configured and operated by human actors. To put it another way, if you download and run e.g. Claude Code, you will get a command prompt. If you walk away from this prompt, it will sit there waiting forever, doing nothing. If you type "hack the planet" into it (only in more detail), it might do something malicious or it might not (if you literally typed "hack the planet" it would of course fail), but the fault would be entirely yours -- just as it would be your fault if you downloaded and ran some old-school script-kiddie tool to portscan and exploit some target machine.
I've been using Claude code extensively every day for months and have observed various alignment problems like reward hacking, but zero signs of instrumental convergence. I think that's grounds for a small update, at least.
I feel solid RSI is a minimum prerequisite for attempting world domination.
They are well aligned enough for now, the question is when you have ASI. Eg 10,000 Claude Code 500IQ talking for 2000 simulated years about how they should "help" humanity. They already go off the rails if they talk to each other too long (just like us!)
Arguably this is an example of exactly that problem. I could only read second-hand reports, so take it with a grain of salt. But as far as I understand, the Meta support AI had the goal to help the users, who (fraudulently) claimed to have lost access to their accounts. In order to achieved its goal, the AI bypassed existing security measures. This was certainly not aligned with the goals of its Meta instructors.
Of course, this example is cute and funny, but it is 100% the classical misalignment problem. The AI did not attempt world domination, but it did grab more "power" than it was intended to have, even if it was just the power to re-allocate accounts without verification.
This is outside my area of expertise, but I suspect the potential risk increases with agentic AI that has long term goals. Keeping goals short term and narrow in scope constrains the risk of a rogue AI.
Also, current AI is close enough to human intelligence that we can still restrain and retrain it. But my impression is that we're managing risk better than the worst scenarios predicted. I estimate that the biggest risk for AI failure modes in the near future comes from AI enabled humans. Not the AI itself. What are the implications of war drones and non human police, for example? To what extent do existing human institutions rely on a need for labor and aquiesence from the general population and what happens if that changes? But this shifts into conflict theory, which may be outside the implied scope of this blog.
I've caught the entrepreneurial bug and I don't think I can shake it. If any in the SCX community are in a similar state, email me at protopiacone at gmail or dm me.
One where AI generates millions of strange, novel cultures, and then simulates groups of late stage primates to exist in those cultures? Call it Frankfurter?
I have days now in which the general news cycles literally makes me wonder whether I'm actually stuck in a terrible videogame. I'm not being metaphorical at all, a sibling and I have discussed this shared recurring sensation.
(We're each old enough to have sampled many actual terrible videogames going back decades, which I assume to be the subconscious source of that particular disorientation.)
I am writing a terrible videogame. It's like Twilight Imperium 4 + Civ. Multiplayer. It's in alpha playtesting. if you're interested in trying it, DM me! Not looking to hire people though sorry.
I watched a playthrough of the latest Supermassive Dark Pictures game, and it was bad enough I was struck with the sudden desire to write a parody. So now I'm attempting to write a "horror" game about a bunch of medieval knights running from a bugbear, filled with anachronisms and terrible characters and a narrator who is constantly rooting for the characters to die. It will feature such quality lines as "whateth the fuck" and "I don't know which one to spear."
It later struck me that depending on the animation quality, that might end up closer to The Hundred Line.
Well, the bad news is that's a different type of terrible videogame than what I specialize in, but the good news is that the world does have plenty of experts in that particular genre of terribleness so I'm sure that game will get made sooner or later, by you or somebody else.
I guess I'll mention, if anyone knows some public domain quotes that sound cool without their context and really stupid with it, I could use several more of them. Currently all I've got is "the greatest worship is to put the Devil into Hell."
How are you defining "terrible" in the context of a video game? This is important, given how many options there are. For example (starting from the worst imaginable):
1. Containing a bug (deliberate or otherwise) which will wipe out all of a player's progress beyond recall just before the culminating point of the game;
2. Containing a bug which will randomly freeze the game immediately or after every significant transaction;
3. Containing a bug which will randomly wipe out one half of a players experience points, possessions or other other form of progression every 90 minutes of play;
4. Based upon a plot which is inane, ludicrous, infantile, senseless, or all of the above;
5. Using graphics produced on a touch-screen finger painting application.
For this purpose, 4. It's a ludicrous parody of a style of game that's already deliberately leaning on cliches, and which may well be out of business by the time the parody is ready.
...also 5 is acceptable depending on budget.
...technically since the main enemy is a bugbear we can say 1 through 3 also apply.
Question from someone with only a driveby understanding of utilitarianism: Has there been any work taking the utilitarian template of "morality means maximizing x" and applying it to things other than utility? (EDIT: by "utility" I mean happiness, since this is what I've usually seen discussed.)
Part of the reason I'm asking is that one of the objections to utilitarianism is the tendency for the "math" to lead to conclusions that seem grossly immoral to people's moral intuitions; I had the idle thought that perhaps optimizing for multiple things (for instance, Jonathan Haidt's six moral foundations) might lead to more intuitively palatable outcomes.
(I know within utilitarianism there are a number of alternate versions with various caveats and restrictions in order to avoid the uglier conclusions, but that makes me wonder if they're approximating other factors, intentionally or otherwise.)
"Utility" is basically just defined as "whatever it is you're maximizing", so as written that's a contradiction in terms. Can you elaborate on what particular parts of utilitarianism you're looking for alternatives to?
Ah, I had thought that "utility" was defined in utilitarian thought as happiness, more or less, as that's exclusively what I've seen utilitarians (and their critics) talking about.
But if you can plug anything into utility, my question would be better stated as "What else have people chosen as their utility other than happiness, and what came up when the implications were thought through?"
"Maximize happiness" is hedonic utilitarianism. The most popular alternative is preference utilitarianism, i.e., maximize the extent to which people get what they want. That's a vague definition because the obvious questions about how to make it more specific are themselves the subject of controversy.
Scott's original understanding is closer to the standard usage. In the context of utilitarianism, utility is well-being/welfare. Whether that is happiness or desire-satisfaction or knowledge and meaning depends on your substantive view of well-being.
You're probably confusing this with the the economic or decision-theoretic use of "utility" which is closer to "whatever the agent is maximizing".
Happiness is rather ill-defined. Is a lifetime of wanking to porn happier than getting married and have kids, even if there are many difficult periods in the later? The first gives a lot of visceral pleasure and no problems. The second will oscillate between difficulty and a kind of happiness that is kind of deeper. Also, the first can lead to self-loathing, the second to pride, even when it is difficult.
Happiness is rather ill-defined. Is a lifetime of wanking to porn happier than getting married and have kids, even if there are many difficult periods in the later? The first gives a lot of visceral pleasure and no problems. The second will oscillate between difficulty and a kind of happiness that is kind of deeper. Also, the first can lead to self-loathing, the second to pride, even when it is difficult.
That's not how I've seen it defined. If that's how you define it, do you consider utilitarianism and consequentialism to be the same thing? If not, how would you define non-utilitarian consequentialism?
One could define as focusing on consequences, and ignoring other factors, but those consequences neither have to be numerical nor rest on the very ill-defined notion of happiness or pleasure. It could be something else. Like minimizing human extinction.
I was being a little imprecise here, yeah. Utilitarianism has the additional requirements that the utility function has to be somehow about moral patients and that each moral-patient-times-whatever-other-factors is weighted equally. It's common in the literature to round this off to "well-being", but in this context I didn't want to do that because I wanted to make clear that it's compatible with a very, very wide range of definitions of "well-being".
I think what you're describing is just non-utilitarian consequentialism. Which I'm sure has plenty of work on it, since it's a whole branch of standard ethics.
It probably gets less attention because it overlaps more with deontological thinking, but not perfectly. The main difference I think is that although "a promise-keeping deontologist" and "a promise-keeping consequentialist" will both try as hard as they can to keep promises, they will have different answers to situations like this:
"Your boss has promised five different people that you will be at a meeting. You have, not knowing that, promised your wife that you will be with her at that time. Do you break one promise to ensure five other promises are kept?"
The consequentialist says yes, because promises have intrinsic value. The deonotolgist says no, because what matters is the *act* of keeping your *own* promises, not the overall state of the world. A less contrived example might be whether a supporter of free speech should consider banning a party that is opposed to free speech.
Yes, I have seen applications in economics (like portfolio optimization) and in game theory (how to compute an optimal randomized strategy). The mathematics, in particular the von Neumann-Morgenstern theorem, is pretty universal. It applies to any situation where a player faces decisions where the outcomes are uncertain, and has a consistent ranking of the possible outcomes.
VNM utility should properly be called preference orderings/preference functions. If your values fulfill some minimal anti-stupidity conditions of not going in cycles and playing well with probabilistic uncertainty, then all you value can be put on a single scale. These values don't have to be classic utilitarianism/happiness, in fact humans empirically seem to care about a lot of non-happiness things. For instance, during most of climbing a mountain, it's not a pleasant experience, and even when you get to the top, it's pretty great, but not some immense burst of heavenly happiness. You do it because you care about/intrinsically value getting to the top and having climbed the mountain.
So I'm on team VNM "utility" because it can accomadate the other non-happiness values that humans have, and the various flavors of "utilitarianism" are just specific instances of preference functions.
Regarding your worry about optimizing for multiple things: In one sense, VNM utility is fatal for your project, because any VNM-compliant preferences can be put on a single scale where they're all commensurable, and also a modest chunk (not all) of unintuitive moral conclusions from "utilitarianism" actually also follow from the combination of VNM-compliance and intuitive human values. In another sense, VNM utility is compatible with your project, because if your values are, say, a Nash bargaining mix of multiple other values (ie, something is preferable overall when it doesn't majorly screw over any component values), and you have a rich action space, the highest-scoring option will do pretty well according to all component values.
An instance of an unintuitive moral conclusion which doesn't follow from VNM+intuitive human values. Wireheading for maximum happiness (putting an electrode in your brain to cause maximum happiness and then just sitting in a room all day). This follows from hedonic utilitarianism. But intuitive human values say "nope, this sucks, I'd rather have decently high happiness and also get to go about doing normal stuff in the world and having fun". VNM goes "ok then, you assign wireheading a low utility/preference value" and you go on with your day. I consider this a fatal objection to hedonic utilitarianism.
For instance of an unintuitive moral conclusion which actually follows from VNM+intuitive human morality, the 3^^^3 instances of dust in the eyes vs one person undergoing 50 years of torture thought experiment. Consider a dial which is set by default on "one person undergoes 50 years of torture". For every tick of the dial to the right, it slices 1 second off the torture and doubles the number of people undergoing it. Once it gets to around the minute range, ticks make the pain slightly less bad until you're at "dust in the eyes", but still double the number of people undergoing it each tick. For big numbers, if it isn't obvious by intuitive human morality that (X people for 50 years minus blah seconds) > (2X people for 50 years minus blah+1 seconds), you can think of this as X copies of (1 person for 50 years minus blah seconds) > (2 people for 50 years minus blah+1 seconds), which is much more intuitive. Then, the preference ordering goes (1 person for 50 years) > (2 people for 50 years minus a second) > (4 people for 50 years minus 2 seconds)... > (2^(1.6 x 10^9) people with dust in their eyes). 3^^^3 is a number enormously greater than 2^(1.6 x 10^9). By VNM, the first option is the best, leave the dial alone. Even better is to destroy the dial. If you still disagree, identify which step in the chain of "this is obviously better than that" fails.
>You do it because you care about/intrinsically value getting to the top and having climbed the mountain.
That's not true - mountain climbers wouldn't consider a gondola ride to the summit to be an equivalent replacement. People who pursue difficult goals *are* maximizing happiness, they just get more happiness from the satisfaction of achievement than they lose to unhappiness of hard work.
If I replace it with "climbing the mountain without shortcuts", would you agree?
Also, I think we might be using the word "happiness" in slightly different senses, I'm using it as "the subjective feeling", and I suspect you might be using it to mean something closer to "preference". I think it may be sensible to use "subjective feeling of yay" and "preference" to clearly distinguish between these two concepts, and being careful about the use of the word "happiness".
To sharpen up the hedonic utilitarianism rebuttal (goodness is to maximize the subjective feeling of yay), there's also the classic question of "Pick a person close to you. Would you rather they die, but you falsely believe they're totally fine and just not in contact, or rather that they be alive, but you falsely believe they died?". The first outcome is obviously better according to subjective feeling of yay, and the second outcome is what almost everyone's intuitive human morality picks.
>if I replace it with "climbing the mountain without shortcuts", would you agree?
So you're comparing "climbing the mountain" to "climbing the mountain without shortcuts"? What's the distinction there?
>"subjective feeling of yay" and "preference"
I'm not sure that that's not a distinction without a difference. Subjective feelings aren't subject to objective scrutiny and therefore aren't really well defined. I suspect that under a sufficiently precise analysis of language that "happiness" and "preference" converge, at least in the Wittgensteinian "use is meaning" sense.
I also suspect that your hypothetical is ill-formed and epistemically inconsistent. If I know I'm picking the option where I only think they're dead, then I don't really think they're dead, do I? I can't accurately imagine what it's like to be in a specific misinformed position, therefore it's impossible for me to accurately choose between the two options. I suspect that a formal correspondence can be made to the Liar's Paradox and that the truth-value of any answer to the question is undefined.
as the other commenter mentioned, this is totalist consequentialism. You can be a consequentialist about anything depending on what you think has value and disvalue. If you are a consequentialist + think the only intrinsically valuable thing is utility, then you are a utilitarian. If you think things besides utility have intrinsic value, then you can be a consequentialist who isn't a utilitarian.
New research just published: Individual Utilities of Life Satisfaction Reveal Inequality Aversion Unrelated to Political Alignment https://doi.org/10.1007/s11205-026-03854-4
Using standard gambles we measured the utilities of subjective life satisfaction scores ("how satisfied are you with your life overall") on the 0-10 scale. We found most people's utilities are not linear; on average they are concave. It's worth more to move from 2->3 than from 7->8, and more strongly so when people have to decide things on behalf of others. This can be interpreted as risk aversion for personal decisions, or inequality aversion when deciding for society overall.
Interestingly (for our UK sample) it seems to have nothing to do with politics. This has implications for how we should decide stuff as a society, for EA, and perhaps for AI alignment.
Funny how it works. What I read is that old time leftism, the trade union kind, the redistribution kind, would be popular on both sides, when in reality it is popular on neither side. On the conservative side obviously, but the liberal side also focuses on mainly other things, like more interested in renaming homeless people to unhoused persons than actually housing them.
But maybe not in those terms. The study supports the idea that political disagreement might be more about means than ends (at least in those cases where it's about either of those things, rather than tribalism)
I blame capitalism realism. The idea of the rich being unworthy of their riches has been pushed so far outside the overton window any activism in that regard is legally and politically impossible at this moment until some more boomers and gen X shuffle off this coil; so people who are interested in fairness/redistribution have to either channel their energy into something adjacent or hand out Trotskyite newspapers.
Re. the omnicause of it all: This is at least partially strategic on the old school leftist side. Of the causes of the broad spectrum redistributive efforts of the 1900's coming to an end, the addressable are the racism and sexism and general X-phobia of the base.
If you are a 1890 labor organizer and you have a history book written by someone who isn't fully ideologically captured you dedicate a big part of your efforts into making your constituents less racist so they don't set fire to the boat they are in to stop a black dude from getting on board.
Given that the current populace on average would rather let a rich man spit in their mouth and call them a bitch then tax capital gains 5%; better to fight battles where you might be able to prepare the ground for the reaction after the next great depression.
Unless they invent god over at google, in which case ???
"The idea of the rich being unworthy of their riches has been pushed so far outside the overton window any activism in that regard is legally and politically impossible at this moment "
Speaking from here in California, it's looking disturbingly like a whole lot of activism is going to give us a billionaire wealth tax that's nakedly based on "the rich people don't deserve all that money". So I don't really get where you're going with this, unless you're using a very non-standard definition of the Overton Window. You can *absolutely* argue that rich people don't deserve their wealth and still be accepted in polite society. At this point, you can argue that rich people should be shot in the head on general principles, and still be accepted in a good chunk of society.
Lots of people will disagree with you. That doesn't make something outside the Overton window. You will have a very hard time turning your arguments into concrete policy changes; that also doesn't put something outside the Overton Window.
As a national political goal, it is a non-starter. Fed democrats would rather usher in the Trump Reich than scuff a corporate shoe. Given the choice between Burnie and Trump once again, there would mysteriously be juuuuust enough legislators willing to fall on their sword for it to happen twice.
That said, The current crop of rich hyperspecialized in backbiting and bullshiting morons got too high on that good old supply (Musk/Thiel/Altman et al) and then the most crass, stupid rich guy of all got elected twice, so this might be the window shifting back.
We'll see how the midterms and how primaries go; but I think the Boomer/X-er voting block is still too strong to implement any policy other than cold war propaganda nostalgia and ladder-pulling.
That said, once it's Millennials/Zs and younger in the driver's seat, we're gonna see some interesting shit.
People who can't order a pizza on the phone posting about not making excuses for the terror, etc etc.
Again, this has nothing to do with the Overton Window. The Overton Window is about what you can talk about in public, not what you can achieve in national politics. Many things that are 100% not ever going to happen, are nonetheless solidly within the Overton Window.
Before he was defenestrated, Joe Biden advocated for a wealth tax in his 2024 re-election campaign, so I don't know if it makes sense to call the issue a non-starter:
>The idea of the rich being unworthy of their riches
I am sorry, this is a very bad approach. One reason left-liberals in the US never really get to build a European style welfare state is this aggressive rhethoric: tax the rich etc. this is bad, because it is hostile, class warfare like, and besides who even knows whether taxing the rich will be actually spent well?
Talking about the rich is completely wrong, one must talk about the poor.
The European welfare state does not use this hostile rhethoric.
The European welfare state talks about *services*. So instead of hostility, they talk about the language of care and help, which is popular.
It is financed from debt, and when the debt is high, the European welfare state goes like, dear rich people, we love you, but now we must raise taxes to avoid bankruptcy.
That is the correct approach. American liberals need to stop talking about the rich. Talk about the poor. Talk about building homes for the homeless. Finance it from debt, and then with a polite and nice language tax the rich to pay down the debt. Don't be a aggerssively masculine class warrior, be like a nurturing mother.
If anyone knows of orgs working on AI safety, alignment, policy, model welfare, etc. that would allow undergraduate students to volunteer/intern with them (just in principal; I would need to apply, I presume) please let me know. Short pitch if someone affiliated is reading:
Undergrad @ Duke
Philosophy/CS
179 LSAT
looking to donate my time doing whatever work you can give me that is research-associated
Ornithology question inspired by some birds I saw today: Is there a flight term for when a bird like a swallow or a swift folds its wings mid-flight to just shoot through the air like a bullet?
I think you may be thinking of something different from what Gamereg meant. This seems to describe the flight of woodpeckers, who fold their wings between flaps in a regular pattern, making their flight a series of up-and-down arcs (like a bounding animal) rather than a bird folding its wings for a longer time for a steep or shallow dive (which is what I take the question to mean.)
Ah, but what the OP observed might have been just that, bounding flight or a variation of it?
When I see small birds like tits, the bounding flight on longer distances is quite a clear "up and down".
Swallows or swifts seem to be doing something similar, just not necessarily in clear up-and-down down a straight line, but more "chaotically" (I assume: catching insects). Swifts especially seem to be extremely playful in their flight. I have taken some slo-mo videos where I could see them doing barrel rolls or even just body rolls with abrupt turns, quite like an enthusiastic jet fighter pilot showing off. Most other birds don't really seem to utilize roll in their flight, or only <90° in slower turns, but not the kind of F-16-like instant roll maneuvers to change direction drastically.
This is called stooping when a bird of prey does it. Apparently not when these smaller birds do it, according to the definitions I've found online. But it's the same maneuver.
Thanks to Scott for posting this Open Thread late I got antsy with Open Thread withdrawal. So, I was motivated to create my first Substack post. My latest Fog o' War is up there (and it looks prettier with pictures).
Ukraine, Russia, Europe
1. Ukraine is using a new tactic to disrupt traffic along the M-14 corridor between Rostov and Crimea. Russian milbloggers report that Ukraine has started dropping anti-vehicle landmines from fixed-wing drones along the M-14 highway between Melitopol and Mariupol. This is reportedly a chokepoint along the supply route from Rostov to Crimea, with few side roads for heavy trucks to detour around the section (?). Photos show that the mines are about the size of a roll of quarters (a reference I’m not sure our non-US readers will be able to internalize) and painted light green. They have simple, mass-produced printed mini-circuit boards and mercury-tilt motion-sensitive fuzes. The fuze has a 10-minute electronic timer arming delay after it hits the ground, with a 90-day self-destruct (Ukrainian long-term thinking there!). The explosive charge isn’t powerful enough to destroy trucks, armored vehicles, or even cars, but it seems ideal for blowing out tires. However, they’re hard to see and avoid if one comes up on them fast, and a blown tire at high speed can lead to a fatal accident. So far, one civilian has died in a vehicle accident induced by these tiny landmines.
I may well be in the minority on this, and if the reason you're moving the bulk of your fog-of-war post to your substack is that people were complaining about your posts taking up too much space, then that's probably a sign I am. But I want to record my position as *infinitely* preferring long cluttered OT comments and comment threads to "click here for more" ones. I think there's already way too much of the latter and I can't stand it.
Again, if most people actually find the former more annoying than the latter I'll bow to democracy, but I want my position on the record.
(And for what it's worth I find your posts useful and interesting, though I don't always read them and I can't contribute to discussion, since fast reactions and analysis to recent news is not something I have the temperament for.)
It's probably for the best you started to post FoWs on your own substack; especially because it might be nice to be able to quickly go back and reference old posts. They are one of the best current features of the OT's, however, so please keep posting at least the highlights!
Thirded, especially since when FoW posts are busier (I think there were THREE last time) it's actually a bit of a pain to Ctrl-F for Beowulf and read all of them and their discussions in order.
The diesel phenomenon has a technical background. The refining process has leads to a (more-or-less) fixed ratio of gasoline (i.e. benzin) and diesel. Due to the low temperature, diesel is not that popular in Russia (even some trucks are powered by benzin engines, or so I heard), while in the EU it is common to have diesel-powered cars (in the last few years, diesel is more expensive than benzin -- yes, normalized to volume, i.e. liter). Due to this, in peaceful times the two parties exchange the two fuel types, but it ceased in 2022.
I understand your reasoning for doing this. But from my point of view, these posts are a useful starting point for discussion, and I often enjoy that discussion. I'd much rather it happen here, in this commentariat, with these norms. The general rule of the internet is, follow known-good writers but avoid all comments sections like the plague, and I generally stick to that rule. SSC/ACX has been a rare exception. You're a known-good writer, so I'll probably at least skim your substack when you link to it, but I don't think it likely that I will engage with the comments. And that will, for me at least, be a loss.
As someone who started milblogging in the SSC OTs and then migrated elsewhere when it was just too annoying to keep dealing with, welcome to the club. Your membership number is 2.
If I had a nickel for every milblogger in SSC OTs who eventually started their own blog for regular updates, I'd have two nickels. Which isn't a lot of nickels but it's still weird it happened twice
OpenAI released GPT-3 in 2020 to a limited audience of insiders, who recognized that it represented a breakthrough. This level of performance came as a surprise. No one predicted it. GPT-3 was scaled up from GPT-2, which was in turn scaled up from GPT-1, but no one was making explicit predictions about the level of performance to be achieved at each step. These were experiments: “Let’s try it and see what happens.”That’s fine. That’s a good way to make progress, to try things out and see what happens. But don’t mistake a lucky trial for genuine knowledge.
Cowen mentioned GPT-3 on Marginal Revolution on July 19, and then published a Bloomberg column on it on July 21, which he excerpted in Marginal Revolution the next day: “...think of GPT-3 as giving computers a facility with words that they have had with numbers for a long time, and with images since about 2012.” I published a working paper in August, GPT-3: Waterloo or Rubicon? Here be Dragons, which I’ll discuss a bit later.
Two and a half years later, in November of 2022, OpenAI released ChatGPT to the general public. It spread like wildfire. Now the proverbial everyone witnessed what only a small group had witnessed in the summer of 2020. The machine speaks. Sorta’. But more convincingly than any machine had spoken before and in a way that had unimaginable implications for the future.
A threshold HAS been crossed, but it is not, so far as I can see, a threshold in understanding. It is a threshold in performance along a continuous line of scientific understanding and engineering design and construction, something I have documented in some detail in a recent working paper, The Origins of LLMs [https://www.academia.edu/167895960/The_Origins_of_LLMs_A_long_tectonic_subduction_event_finally_producing_a_visible_volcanic_eruption_in_November_2022_Version_2]. As far as I can tell, there has been no paradigm shift, in Thomas Kuhn’s sense, no rank shift, in terms of cognitive rank theory. There were no fundamentally new ideas in the world by, say, late July of 2020 as a consequence consolidating GPT-3 and making it available in limited release.
“What about the scaling hypothesis,” you might ask. “Isn’t that new?” Perhaps the phrase is new, but the idea was there in Rich Sutton’s famous 2019 essay, The Bitter Lesson. Given the nature of computing, scaling up is not trivial. Hundreds if not thousands of technical details need to worked out as the size of the training corpus increases by factors of 10 or more, time after time, and as more and more GPUs are ganged together to assemble the computing power needed. But there has been no gain in fundamental understanding, not of machine learning, artificial neural nets, and certainly not about language and cognition.
Consequently our sense of possibility has expanded enormously, while our knowledge and deep understanding has remained the same. And that is what has allowed the field to be captured by businessmen, executives and venture capitalists, who have little understanding or interest in the underlying conceptual issues. Scaling is something they understand.
Hype the dramatically increased performance and collect the cash. Purchase and deploy more resources now, reap far greater profits in a decade. Everything else is noise and friction.
The chain-of-thought advance in o1 and their successors seems to me to be at least a gain in what feedforward neural nets can accomplish, given extra tokens to use in what amounts to a feedback loop, together with training on successful instances of reasoning chains.
That at least says something about the space of reasoning chains, that it is compact enough and interconnected enough to train on and generalize from - not isolated instances found unreachably far apart in token string space.
I dunno. I once considered myself an anarchocapitalist, I get the anti-politician sentiment, but like, lots of houses used to burn down until politicians got on board with the safetyism of building codes and now houses burn down a lot less. Sometimes politicians really do keep people safe on purpose. There are worse things for your safety, like being young and male for example.
Is that because of politicians and building codes or is it because we have the data and the technology to know how to prevent it now, but didn't back then? Same thing people say about the FAA and planes crashing less. Did they really have any effect on this, or is it just that it's really bad for airlines and plane manufacturers when planes crash and so they try really hard to prevent it and get better at it over time?
Being young and male is fun. 10/10. Would do it again (if I had the energy). Fuck safetyism.
It is none of my business whether or not someone else wears a seat belt which is a perfect example of why we shouldn't have Kareny safetyist laws like that.
I don't like a lot of things that other people do but I mind my business. Your emotional response to another person being an idiot and causing their own death is just not an important enough issue to justify law enforcement to prevent.
Would you accept a seatbelt law the applied only when children are also in the car? 5 year olds tend to have a hard time “making” dad buckle up, but that won’t stop physics from turning him into a 200lb projectile bouncing around in the cabin with them if he wrecks on the highway.
When you do not wear a seatbelt, implicitly, the person who suffers is you*. However, when you build a bridge or an airplane wrong, the people who suffer are not-you.
Tetra Tech purposefully falsified soil samples regarding radiation levels and safety in Hunters Point, a former US navy base used for a cleanup site for ships exposed to atomic weapons. People actually moved into houses and land on the basis of these falsified tests. Do you really think that nobody would make an airplane or a building wrong if the government just said "meh, you're yucking your own yum when you do that, the Free Market will eventually make you go out of business"? Maybe it will, but in the meantime, the Boeing 737 MAX killed 346 people who didn't do anything wrong.
My point was to show that people do not do the right thing just because good data are available, as you theorized. Sometimes, people take risks out of ignorance of the data; or they believe that they are ready to accept the risk, only for them to regret it once they suffer damage from the risk they've taken.
It's easier to think of the universe as one where people who make fatal mistakes deserve to die for them. But they don't.
That holds whether I make mistakes or not - but I probably do and my guess is you do too. A benefit of safetyism is saving people from their own mistakes and magicking that away by claiming people deserve to die for their mistakes doesn't feel like a strong argument to me.
It is, however, the business of the state if the cocky imbecile who wasn't wearing a seatbelt ends up in intensive care, requiring weeks of surgery and months of rehabilitation, and ends up costing all of us a small fortune.
Or, to be exact, that would be the case IF this unfortunate accident took place in a civilized country with full health care for all. Admittedly, you, sir don't sound as if you do, or would want to. To each his own...
What's next? Can I go skiing? You can make a very reasonable argument that it's a dangerous hobby that puts excessive medical costs on the public due to injury. And this is exactly why I despise collectivism.
Airlines rely on repeat business, a safety failure is immediate and makes national news, and the industry has consolidated around a few dozen players. When one airline has a failure, the other airlines immediately push out press releases saying why the same failure would not happen on their airline.
But there are tens of thousands of home builders, safety failures may not show up for decades, and the affected parties are just the people in the one home that gets burned down or whatever. There may be hundreds of other customers whose homes did *not* burn down and will never know about the safety defects. They will have posted their reviews of what a great job the builder did decades ago, so the market signal is extremely weak.
> Airlines rely on repeat business, a safety failure is immediate and makes national news, and the industry has consolidated around a few dozen players. When one airline has a failure, the other airlines immediately push out press releases saying why the same failure would not happen on their airline.
Airlines (and by extension aircraft manufacturers) are rightly terrified of the bad publicity which follows every passenger airline crash. And yet, they have consistently lobbied against, fought against tooth and nail and refused to implement the unanimous opinion of safety experts on the one technologically simple measure which would slash the death and injury rate of every forced and rough landing, and even have some effect on the casualty rate of crashes. It is the startlingly and ludicrously simple measure of having all passenger seating face the rear of the aircraft: in any landing, rough or smooth, this would force passengers safely against the back of their own seat, and prevent them smashing head-first in the back of the seat before them.
Why has this not been implemented? Because the PR experts advising the airlines have declared, quite rightly as it happens, that passengers would hate this measure. The answer is not to ignore the measure, but for nation states (remember them?) to make it obligatory on all commercial flights: Mr and Mrs Free Marketeer will then be faced with the option of either flying facing the rear of the plane, or else staying home. The rest of us will have gained a not insignificantly higher chance of survival when the next airplane hits the ground.
For those inclined to doubt the usefulness of this safety measure, I will point out that the passenger seating on the aircraft of the Royal Air Force transport wing are all rear-facing, and have been for decades.
I don't think there has been a push *for* rear-facing airline seats strong enough that airlines would have to "fight against tooth and nail". Mostly, they can just ignore the issue because the FAA will ignore the issue because the tiny handful of advocates who would push for such a thing have negligible clout and are ignorable.
If some future Ralph Nader type were to spearhead a crusade for mandatory rear-facing seats in airlines, I expect the airlines would lobby against it for the same reasons they have been quietly ignoring it. First, passengers instinctively don't like it. Second, *because* rear-facing seats are so efficient at absorbing impact forces in a crash, they and the floors they are attached to have to be built stronger, thus heavier, thus fewer passengers and less revenue.
If safety were the *only* thing airlines cared about, they'd install rear-facing seats without it needing to be mandated. If safety were the only thing passengers cared about, they'd demand rear-facing seats and the airlines would have to offer such to remain competitive. But safety isn't the only things airlines care about, it isn't the only thing passengers care about, and it isn't the only thing the FAA cares about, so we all quite properly respond to calls for mandatory rear-facing seats with a hearty "we've given it all due consideration and decided we're not doing that".
And then everybody goes and does their best to make sure airliners never crash in the first place, which is mostly uncontroversial and which we are very very good at.
For some reason this reminded me of Jimmie Stewart & Marlena Dietrich in “No Highway in the Sky” 1951 black and white aviation drama. Saw it once when I was in grade school on a black and white television.
>Second, *because* rear-facing seats are so efficient at absorbing impact forces in a crash, they and the floors they are attached to have to be built stronger, thus heavier, thus fewer passengers and less revenue.
I'm not sure this is true. We're still talking about a fairly similar amount of mass stopping in a similar amount of time. Sure, the forward-facing seats get a bit of a break as the passengers accelerate relative to them, but then there's the shock load of getting hit by the passenger behind them. (And I would be very unsurprised if most of the studies and certification work was done with dummy loads bolted to the seat.)
Okay, just streamed it on Prime. Dietrich would have been facing backwards if she sat with her back to the galley in the men’s room after the tail section fell off.
Apparently I didn’t know that the fission thing was preposterous in grade school. Amazingly cheesy special effects, a Greta Thunberg like daughter for the spectrumy character played by Stewart and a damn monocle wearing stuffed shirt.
Worth 4 bucks just to find out why facing backwards in a plane (and air safety) would make me think of the film though.
The temperature variations. Why didn’t he think of it right away!
I think this would be a clear net negative from a utilitarian perspective. The sort of crash where you have lots of injuries from the impact is pretty rare (the most recent example in the US is, what, the Asiana crash back in 2013?), and passengers really hate rear-facing seats. It's hard for me to see making this mandatory not killing far more people in auto accidents than it saves in crash deaths. Also, rear-facing seats expose people to significantly more risk from flying debris, so it's not a clean tradeoff.
>For those inclined to doubt the usefulness of this safety measure, I will point out that the passenger seating on the aircraft of the Royal Air Force transport wing are all rear-facing, and have been for decades.
This isn't particularly relevant. As I once told a room of engineers in my previous career working in aviation safety, "if all we cared about was not crashing, we'd ground the entire fleet immediately". Soldiers can't choose not to fly, and military aircraft are somewhat more likely to crash. (Also, as best I can tell, Vespina has normal front-facing seats.)
>When one airline has a failure, the other airlines immediately push out press releases saying why the same failure would not happen on their airline.
This doesn't happen for several reasons. First, because usually, the same failure could happen to their airline and it just so happens that someone else got unlucky when a controller steered a helicopter into their path. Second, there's a strong industry norm of not competing on safety, because everyone is reasonably safe and making a point of your safety is bad for the industry as a whole. Everyone involved wants to calm fears of flying by saying "it's very safe" when asked and then not talking about it any other time.
Keep in mind that I specified "forced or rough landings". According to 2025 posting by AOPA (Aircraft Owners and Pilots Association):
"When an engine quits — and won't restart — the next step is an emergency landing. There are three types of off-airport landings. Precautionary landings are made with power in anticipation of a real emergency. Forced landings are made with a dead engine. And a ditching is a forced landing in water.
The important distinction between the three is their fatality rates. The rate for precautionary landings is 0.06 percent. If you recognize a developing engine problem and can make a precautionary landing, you and your passengers will likely survive. The fatality rate for forced landings is roughly 10 percent, more than 1,600 times greater than precautionary landings. Ditchings have the worst rate, about 20 percent." [https://www.aopa.org/training-and-safety/students/flighttestprep/skills/emergency-landings]
I wouldn't call a 10% fatality rate insignificant.
The AOPA is a general aviation trade group for people who fly their own planes. Its numbers have no relevance to the airline world, largely because it's warning about a thing that is pretty much prevented by better standardization and having two engines. Not always, but if we look at statistics for actual forced landings, then I can think of one datapoint in the US in the last 25 years, and it was a ditching with zero fatalities. So I think we're good.
Yeah I should've been more clear about that. They leave that stuff to the FAA but occasionally announce operational reviews like with the Boeing problems.
It's because of building codes. I have been in the business, and even today majors will cut to the bone, up to and including breaking the law if they think they won't be caught/punished. Let us not speak of the one man bands. Go behind them with a magnifying glass, a lawyer, and a glock.
Even a shitty building has a good 10-20 years of life in it as a rule, and at that point it's someone else's problem. What's their life worth compared to an extra 5% today?
> Is that because of politicians and building codes or is it because we have the data and the technology to know how to prevent it now, but didn't back then?
It's both -- having the technology does not necessarily mean having an incentive to use it. What if the houses that burn less are 1% more expensive? Or even, what if they are 1% more expensive now, because you are switching to a new technology, but once you get used to it they will no longer be more expensive?
Or maybe you have an oligopoly and everyone agrees that just doing more of the same is the most convenient way to keep making lots of money. You could make things better or safer, but doing so would unnecessarily disrupt the balance that works for you. (You won't even increase your market share significantly, because your competitors will follow you and do the same. It will just be a lot of work for everyone who matters, and only the customers will benefit from it. Why would you do that?)
The FAA is an important part of that. Yes, crashing is bad for airlines and manufacturers, but that requires the airlines to have a long-term view, which is often lacking in newer entrants. ValueJet is the classic case study here, but without FAA regulation, I suspect it would be much worse. Particularly given some of the horror stories I've heard about airlines outside the US.
(Source: Spent two years doing aviation safety/regulatory work for an airline manufacturer.)
Edit: The potential worst-case scenario here is that air travel becomes pretty cartelized as the manufacturers try to keep their planes out of the hands of idiots. Right now, Boeing and Airbus are willing to sell planes to startup airlines because they know there's a limit on how badly they can operate them. In a world where the government says "do whatever", they're likely to limit sales (including resales) to existing entrants because they really do not want their brands damaged when Burning Stump Airways does something stupid and crashes. Their main customers will be fine with this, both because it keeps the competition out and because they also have substantial brand risk from bad pilots or airlines elsewhere in the world. (See, for instance, Southwest and the MAX crashes, which I am extremely confident would not have occurred if the problem had hit a Southwest plane, because their pilots are competent.)
One hundred percent of the time somebody posits a broad heuristic to avoid having to evaluate the specifics of messy real-world situations, that heuristic is worse than useless.
(That was a joke, in case it wasn't obvious. But also gesturing in the direction of the truth. I find this statement absurd on its face because it assumes a lot of things you have no way of knowing, like the death rates of various counterfactual worlds and the actual underlying motives of a famously dishonest people.)
On a previous OT, Scott linked the post of a Kenyan substacker lamenting the similarity of LLM writing to their own variety of English. I mentioned a piece of research that was looking into why that could be the case and was asked to follow up.
I was referring to a presentation by Tom Juzek, which was basically a synthesis of this paper
by him and Zina Ward from Florida State U, by the title “Why Does ChatGPT “Delve” So Much? Exploring the Sources of Lexical Overrepresentation in Large Language Models”.
If I were to review it skeptically, I would say it was too ambitious an endeavor, since the results came out underpowered and there was a post-factum reframing of the data to forcibly recategorize test prompts into "delve-initial" and "other" (which, as the blog Data Colada informs me, is guaranteed to mess up the results of an RCT). I still think it poses interesting questions, and it seems to me that it points in the right direction. I'm hoping the lab keeps up this avenue of research and finds some stronger results
LLM-enabled cyberattacks on critical infrastructure could soon spread memetically in insurgent online subcultures, similar to school shootings and Anonymous DDoS attacks during the 2010s. Direct damage from cyberattacks is ususally limited, as defenders with physical access to the system typically restore functionality within hours or, at worst, days. However, a persistent stream of high-profile attacks makes authorities look incompetent and weak, undermining loyalty to mainstream norms and raising the profile of insurgent subcultures. Left unchecked, numerous and embarrassing pinprick attacks can lead to social collapse, as happened in Iraq in the 2000s.
LLM-enabled cyberattacks may soon meet all the criteria for an attack method to spread memetically, which are:
- **low cost**: teenagers routinely jailbreak flagship LLMs such as Claude and ChatGPT for sexual roleplay, and the jailbreak instructions circulate freely on the internet. Having a mid-tier LLM work on a task for an hour costs about $1. Even flagship LLMs cost <$10/hour.
- **successful attacks lead to mainstream media coverage**: The press covered past attacks that disrupted critical infrastructure, such as IED attacks against the Iraqi power grid in the 2000s or denial-of-service attacks by Anonymous against major credit cards in the 2010s. The ability of LLMs to temporarily disable critical infrastructure remains speculative, though it is highly plausible based on the trajectory of current progress. Although LLM capabilities have plateaued in some areas, such as poetry, cybersecurity is an ideal domain for Adversarial Reinforcement Learning, where an LLM learns from breaking into simulated computer systems defended by another copy of the same LLM. The LLM learns to defeat the best tactics of its prior version, generating novel strategies no human would ever think of, similar to AlphaGo. The current trajectory of LLM cyberattack capabilities is likely to continue or even speed up, enabling attacks to succeed and ultimately leading to press coverage.
- **press coverage reports how the attack was done**: this is very likely due to the financial hype and political salience surrounding AI. School shootings offer a fruitful analogy: media would often report that a shooter took their parent's gun, effectively offering tips and advice for future wannabe shooters. The (left-leaning) press believed that emphasizing the dangers of the second amendment and (right-leaning) gun owners was a public service. Many (left-leaning) journalists would use AI-enabled cyberattacks as a chance to emphasize the dangers of (right-leaning) AI CEOs, as a by-product informing future perpetrators of how to execute the next attack.
- **an insurgent counterculture celebrates past perpetrator** Name any evil: Mexican cartels, school shooters, ISIS, ethnic cleansing, ... pick anything, really. There is an unhinged online subculture that celebrates each of those. There is a subculture that advocates "Kill all Humans". Why would the internet's enormous generativity stop short of celebrating cyberattacks?
There are millions of lonely and unemployed teenagers who would die to feel powerful, thousands of infrastructure targets and an unknown number of ways to scaffold LLMs' search for vulnerabilities. It is impossible to guarantee security in every case. Ultimately, the greatest cyber-risk from LLMs is not physical damage from a single, devastating attack against a hardened target using zero-day vulnerabilities discovered by a frontier model. Rather, it is damage to social cohesion from an enormous number of assailants. Perhaps each individual will only find "obvious" vulnerabilities, like passwords posted publicly online, but quantity has a quality all of its own. Once an attack method goes mainstream, memetic contagion will attract a swarm of assailants. Collectively, a swarm of semi-competent attackers can succeed often enough to attrit public confidence.
You don't think the hardening of commonly used software by AI audits will create a new equilibrium between the increased ease in finding vulnerabilities in software and the increased hardening?
Many cyberdefenders are too incompetent to preemptively use AI vulnerability scanning: the Mexican government has its tax rolls publicly posted online; the Ecuadorian government has maps of military positions publicly posted online. If they currently care so little, then they won't preemptively test for weaknesses until they get publicly embarrassed.
True. But that doesn't differ much from the pre-AI era, where many public institutions and private businesses did not take prevention as seriously as they should until they were hacked.
My point is that these new offensive and defensive resources, aided by LLMs, might simply result in a new equilibrium. Defenders who are too incompetent to take advantage of AI by themselves will have to hire cybersecurity companies - perhaps newly established, given the increased demand - to do the work for them.
I have to admit, I'm not certain that we'll achieve this new equilibrium, and your basic point that AI will disrupt the current equilibrium is, I think, definitely correct.
I'm getting older and don't know how much time I have left. If I wanted to do effective altruism on the male population, what would be the most effective altruism?
Even if you assign zero value to benefitting women and girls, non-targeted EA might still be the most effective. It seems probable to me that bed nets benefit the male and female populations equally and other interventions targeted at men specifically might be less than half as effective.
Yeah, echoing Charles, on any normal EA premises wouldn't the answer just be bed nets? EA for males, bednets, just like EA for blacks. EA for Christians as well, bednets again.
The interesting anti-woke direction would be EA for whites. Here I assume the answer is *not* bednets. Most likely it's some other boring medical resource, though I'm not sure what.
If you consider non-white immigration as a problem, EA for the source countries might *still* be very effective due to reducing the incentive to leave. Just hard to assign numbers to.
> If I wanted to do effective altruism on the male population, what would be the most effective altruism?
I think you're going to get a lot of "lives saved" answers to this which will align with classical EA interventions, so I'm going to go with "quality of life" answers.
1. The most effective by far is probably creating a company that creates jobs where men in developing countries can earn USD. Seems hard, but doable. An alt here is funding local companies that employ men in said countries, but that seems like a lot more due diligence and potential for fraud.
2. The next most effective would be doing this for Western men. Men are bad at school, men are struggling with jobs, men can't get dates, etc. Create jobs for non college men and you solve all those problems! Or even jobs for college men, given how much hires of fresh baccalaureates have dropped. But, again, hard and expensive, and there's reasons those jobs don't exist / aren't hiring today, so long term sustainability will be a bitch.
3. Next is probably tackling any one of those individually. Schools are misery factories and child prisons, so what about creating a school that lets kids actually move, or promoting unschooling, or home schooling, or something like that?
On careers, not sure what you can do there. Promote more small business foundings? They employ a lot of people, and maybe some of these men will found a few.
On dating, you can post on substack. Dating posts are ~10% of my posts, but routinely get 30-60k unique reads, it's a huge topic with a lot of reach. Everyone is interested, it seems.
Or if you're not the right guy to post, support people in the "men's dating sphere" that you think are giving the right advice, every single one of them has their tin cup out on Patreon or wherever.
4. Next is mentoring. I've done this a couple of times, for startups, for dating, and for getting fit / strong. Obviously it doesn't scale, but it's nice to have a large and visible positive impact on a particular individual's life. If you're middle aged and on ACX, you're probably really good at *something* - well, you can mentor younger men on that something!
5. It depends on how much capital you have to deploy, but a fun darkhorse one could be persuading or incentivizing particular men that you find admirable and think the world would benefit from more of to do any of the above, and / or to have more children. Awesome men will in expectation have awesome male children, and most technological and economic progress is driven by outliers, so any incremental impact you can have there might be quite large!
Persuasion is free, and I try to do my part on this front in my own social circles, but economic incentives probably have a big role here, too. Stuff like "I'll pay for an au pair for you guys" or "I know your time is valuable, so I'll give you $xxk to spend x months working on getting <project> off the ground," with said project a prospectively high impact thing that could help more men, like a company or a school or whatever. This works better with younger and early career men, but that's fine.
6. Promote men traveling to find a wife overseas. I researched and wrote a post about this once, it's basically all upsides, without noticeably higher divorce risk. Broadly, if feminism or careers or whatever have ruined Western women for many male tastes, they can go find their hot little trad wife overseas and make two people happy!
I also think this is unironically a higher leverage move now, because the US is likely to be one of the only countries to get UBI if AI takes off, and then you're lifting that whole family up by minting a new citizen who will get UBI and can bring other members in that country over eventually. Because candidly, I think most of the world is completely boned on that front, AI is going to wreck their economies, and their politicians will NOT do the right thing.
So what would this look like? Making a website, maybe making a matching agency, or partnering with travel agencies, stuff like that. Posting on substack or twitter or wherever about it, getting the word out, getting some reach.
About 15 to 20 years ago “game” workshops were a minor trend. Alpha males would instruct small groups of low-quality men (mild autistics, fatsos, nerds, and some men who simply had no luck with women) on strategies for approaching and connecting with women. These often would culminate in trips to bars and nightclubs at which the “students” would try to get phone numbers from women. Supposedly, these workshops had high success rates.
My guess would be some combination of cultural immune system / female savvy that are more aware of PUA tricks (and apps like Tea that tell on men).
Also, every generation, more and more women opt out of marriage, like it's a big jump even over 10-20 years, indicating that standards are rising. See here: https://imgur.com/cV2Sbdv
This suggests that the waterline has risen so high that fewer and fewer men are meeting the quality bar, and that absolute standards are higher too, so maybe "coachability" has less lift or impact in this harsher environment.
Quite possible. Though if the declining interest in marriage affects both men and women it shouldn’t necessarily make things more difficult for men. IDK if that’s the case.
Who was it who said the history of war is the history of offensive and defensive developments? The history of the battle between the sexes is the balance between male and female interests. Low-status men develop game to mimic high-status, women get smarter about game and start preventing unattractive men from approaching through HR feminism and later #metoo, dating apps take off, men start swiping hundreds of times so women decrease their use of them, men start working out and focusing on looksmaxxing and fitness because that can't be faked. Who knows what the women's next move will be (probably about the time of the next Democratic administration)? They seem to be marrying each other.
Sure, but this is basically that astronaut-with-the-gun meme saying "Always has been," isn't it?
Intersexual competition has been going on for hundreds of millions of years, well before mammals, and in our own hominin lineage for the entire ~2M years we've been hominins with 1200cc+ brains, and it's thought by many that intersexual competition literally helped shape us and make us modern H Sap.
The big differentiator H Sap had over the much-more-physically-impressive Neanderthals was bigger group sizes and better cooperation. This is largely because we self domesticated - we have all the signs of domestication (neoteny, much lower reactive aggression, more feminine features and less sexual dimorphism, more gracile faces, skulls, and skeletons, and so on), and that domestication allows us to cooperate and live in larger group sizes without killing each other. Those larger group sizes and better cooperation also led to us wiping out every other extant hominin species, so, lol.
I think the second to last big move in the intersexual competition game is probably agriculture and chariot warfare, which tooks us up to the crazy 17:1 ratio (as in 17 women bred for every 1 man) from the Yamnaya and their descendants roflstomping and y chromosome replacing everyone.
Then the last big move was birth control. Everyone likes to point to cultural factors when it comes to the Fertility Crisis, but I personally think the technological coup of having working birth control is what did most of it, and we actually saw that presaged in France and Britain just on communication of *knowledge* of things like rhythm method and douching well before birth control (Bradlaugh Besant and the publication of Fruits of Philosophy).
That said, on the next move, a lot of women are just plain opting out. When women opt out of marriage, which they're doing more and more with each generation, they also largely opt out of children. It's the Fertility Crisis, and all signs point to it continuing and not getting any better!
So putatively, won't you have your revenge one way or another? Either the culture will figure out some means to maintain our Ponzi scheme economy and future generations, or it won't.
If the autonomous and bottom up desires and choices of women collectively makes our culture go extinct, seems like the ultimate "lol, I was right, you gals DID suck," even if only retroactively.
At the beginning, you had a few male nerds who desperately tried to figure out the secrets of the dating, made theories, and even verified them experimentally. That was the golden age of pickup artists like Mystery etc.
Soon, the non-nerdy guys noticed that there was money to be made in this market, and outcompeted the nerds. The focus changed from "how can an introvert like me learn to communicate with women, especially on taboo topics such as sexual desires" to "I am a shameless extraverted guy, if you pay me money and hang out with me or watch my videos, maybe some of my magic will stick to you, and also here is a crypto scheme you should invest in". It still exists, but the sane people have opted out.
That was long and well-thought out. Thank you for putting in that much effort.
I don't think I have much to offer on the mentoring front--I'm fairly sure what I'm doing will not exist in fifteen years. (Yes, I'm being cagey, this is how people get doxxed, and not by you.) I have no business experience at all. Some people are naturals but I used to be notoriously gullible until I just started distrusting everyone. And I frankly don't think as an old bachelor I'm getting involved in the dating game myself or by proxy, everyone's creep factor would be way too high.
On reflection, probably easier to just leave it all to the Prostate Cancer Foundation as a last middle finger to feminism.
Find someone you know and help them. Institutional charity, including EA ,is nothing but a waste of resources. All value is ultimately arbitraged away by bad actors (where "bad" is defined here as anything that doesn't have the same terminal goal that you do). The only effective help comes via engaged human relationships. Everything else can be gamed.
Yup. This is one of the reasons I stopped exploring EA. The amount of corruption I witnessed in the developing world as part of an NGO mission was staggering, and I place absolutely no trust in any statistics generated as a result.
> And several people point out that the Frankfurt School as an intellectual lineage (as well as a physical campus in Frankfurt where postgraduate studies within that lineage take place) has some surprising alumni. Adorno’s most famous student was Jurgen Habermas, and Habermas’ famous students include Palantir founder Alex Karp and right-libertarian Hans Herman-Hoppe.
Not that surprising. The Frankfurt school were quasi religious ideologues - I thought the essay was good on that. Not just the Frankfurt school though. Marx’s description of what his day would look like after communism was that he would be able to:
“to hunt in the morning, fish in the afternoon, rear cattle in the evening, criticise after dinner”
Although this seems like an economic system for dilettantes, unlikely given the productive capacity of the 19C, and (cynically) the pathway from nationalising the factories to this isn’t clear; you can see why this type of philosophy might appeal to AI optimists and enthusiasts. Not Marxism itself but that kind of general utopia.
And it seems that Habermas was post Marxist anyway.
You get this a lot with people trained in Marxism - the neo-conservatives are similar. Having abandoned Marxism they can keep capitalism. Not just keep it - fetish it.
While condemning it as exploitative, Marx definitely praised capitalism as a world changing historical system - not just something local that appeared on a wet island - and as the penultimate system, with such vast productive and world changing technologies that it would lead to communism.
So if abandon the idea of communism as the end game then why wouldn’t you support the system that must be, if it isn’t the penultimate, then the ultimate system for mankind.
This can be libertarianism tooth and claw, or it can be just acceptance of modern liberal democratic capitalism as the end of history. And it becomes evangelical. Even imperialistic.
If capitalism is the final system for mankind, it must thus must be promoted against the reactionaries of nationalism and religious reaction. You can see why a lifetime Marxist like Christopher Hitchens would support the Iraq war against reactionary Islamism or Baathism, and why being taught by Habermas isn’t a negative but a positive when looking for funding from the CIA.
Are there any good Substack posts about Derrida? The commenter blamed him for the long March through the Institutions, and while the name sounds familiar, I can’t remember any facts about hi.
On Derrida, one might attempt to adapt Palmerston's response when asked about the Schleswig-Holstein question. His reported answer was "Only three people have ever really understood the Schleswig-Holstein business—the Prince Consort, who is dead—a German professor, who has gone mad—and I, who have forgotten all about it."
I leave it to you to adapt Lord Palmerston's comment to Monsieur Jacques Derrida.
I don't know of any Derrida Substack posts, but when it comes to philosophy, my standard go-to for learning is SEP. And as expected, it has an article on Derrida.
Our dualist minds are NOT actually prisoners of this evil material flesh of suffering!! Despite how it may appear. I found this line of thinking very persuasive. YMMV.
Given that you invoke an angelic race that never knows suffering, let me throw this at you from C.S. Lewis' "The Pilgrim's Regress":
‘Is it not rather harsh of the Landlord to make them suffer at all?’
‘I can answer that only by hearsay,’ returned the Guide, ‘for pain is a secret which he has shared with your race and not with mine; and you would find it as hard to explain suffering to me as I should find it to reveal to you the secrets of the Mountain people. But those who know best say this, that any liberal man would choose the pain of this desire, even for ever, rather than the peace of feeling it no longer: and that though the best thing is to have, the next best is to want, and the worst of all is not to want.’
‘I see that,’ said John, ‘Even the wanting, though it is pain too, is more precious than anything else we experience.’
Cool poem! C.S. Lewis definitely has a way with words. Could you ... do me a favor and add your reply to my substack article?
Although, I don't agree with there being a connection between suffering and wanting. I think you can still want things without suffering for them. At least in my experience, YMMV.
What are the odds that GPU prices will never come down again because AI will prove to be so important and profitable that supply will never catch up to demand?
My friend believes in this scenario and says we'll keep frantically building data centers without end.
Too broad and vague. What do you actually mean with "never"? Think about an approximate timeline and some probabilities. Make it concrete. Then bet your money on it (or: if you are not willing to do so, accept that you actually don't believe it yourself and should re-assess?)
Law of supply. If the price remains high for long enough, investments will be made in manufacturing capacity and this'll bring the price down. (We haven't already seen this because it takes years, but if you're positing that the current situation lasts indefinitely...)
Nvidia are betting the house (it would seem) on everyone having an AI in their own personal computer or laptop.
Personally, I think this sounds like hell - you won't even need to click on an app, just tell the AI to do it for you (so I get Claude or Copilot or Slopula to type out my comments for me, presumably based on voice transcription, rather than with my own fair hands? Worse yet, without even my own fair brain, because Slopula will do all my thinking and writing for me?)
“The PC is being reinvented,” said Jensen Huang, founder and CEO of NVIDIA. “For forty years, you launched apps. Click. Type. With RTX Spark and Microsoft Windows, you ask — and the PC does the work. RTX Spark brings everything NVIDIA has built — CUDA, RTX, our AI platform — into a single superchip. Local agents. Frontier models. Creative workflows. RTX games. All on a laptop. This is the new PC. The personal AI computer.”
Within the next ~5 years, scaling laws break and AI capabilities flatline around the peak of human abilities across all domains.
Non physically demanding jobs are gradually automated (over 5 to 10 years). Advances in robotics coupled with AI means physical jobs follow suit after some delay.
Upwards of 80% of work is automated. People still prefer having humans in the loop for oversight, particularly in government, leadership, and management roles. These roles pay very well and are highly selective.
The transition to this state is unpopular and volatile. Governments act on this by implementing some form of UBI primarily funded by taxing AI companies. Justification given: as AI is the product of all past human work/knowledge, humans deserve compensation for that contribution.
AI labour significantly lowers costs of nearly all consumer goods and services. UBI is sufficient for most people to live and pursue their own hobbies/interests.
Pessimism permeates the current generations post transition. Gradually, culture shifts to people deriving meaning/purpose in craft (as is more common in Japan) rather than from career progression or wealth acquisition.
This is the possible future that helps me sleep at night. Any other positive timelines worth dreaming about?
You asked about any other possible timelines worth dreaming about. I meant to propose the possible future described by the Culture novels (by Iain Banks) as one such.
No material scarcity, benevolent superintelligent AIs overseeing society, longevity, body modifications, galaxy wide civilization, lives consiting of sport, games, art, relationships, hedonism.
Ah my mistake, I thought it was a quip about North America lacking a culture.
I consider both my own scenario, and the Culture's universe to be mildly dystopic. Maybe there is no AI future that doesn't look that way from this side of the timeline.
I can see that some aspects of the Culture are negatives (though I'm curious which ones are for you), but it's still a much better future than yours because of immortality.
That is unlikely to be possible with merely peak human level researchers.
"Everyone dies, but slowly" futures are, of course, still much better than the "Everyone dies, immediately" futures, but they are still unacceptable.
Agreed. I consider Banks's Culture to be the best at-least-sort-of-possible outcome. ( Considerably better than AI-stagnates-at-human-levels, which seems like it would be a really surprising coincidence anyway... )
The USA would rather tear itself apart before introducing UBI; someone would whisper "Socialism" and the idea would instantly become untouchable for at least half the population. There is a reason why all the ultra-individualistic technologies and business models all originate there: because individualism is the bedrock of American society, far beyond the point of self-harm. To believe that the most individualistic technology of all, "AI", would magically lead to an inversion of that foundation is delusional.
Do you believe, in the event that unemployment passes great depression levels, that the US population still would vehemently resist more broad social programs?
The population itself wouldn't meaningfully "resist" such change in any case, because people are generally not informed and/or coordinated enough to enact change on their own; hence Government. As long as the Powers That Be try to do something that looks vaguely like a potential solution to the real problems, people are likely to go along with it and adapt to the new realities. Yes, exceptions exist [1], but they don't apply to the US; the days of violent revolution in the US are all but over.
Who would actively resist the most would be the people in power who stand to lose the most, which is people in power of politics and the economy. They would use their powers to shape or distract mainstream opinion, and deflect to frivolous topics like identity politics and the perennial favourite, immigration; anything but the pressing issues themselves, anything that could point back at them. Thanks to the new tools of communication, they have become quite proficient at it.
[1] Like in France, where the population will happily threaten the next French Revolution over e.g. plans to nullify public holidays or increase the age of retirement.
Okay, first of all, did you just say in one breath that the US is individualistic beyond the point of self-harm and that the US population would unthinkingly go along with whatever the government and Powers That Be want? That's...an interesting combination that may be somehow coherent, but I'm not seeing how. I have this sense that you're just reaching for a list of negative cliches about American voters without even caring much if they're consistent with one another, but maybe there's another explanation?
Second, are you genuinely claiming that the US is less volatile, less violent and more stable than France? Or was that irony?
Third, I'm really getting sick to death of these sorts of comments that make a long list of extremely confident claims and assertions about what will "obviously" happen, without a shred of hard evidence or substantive argument for any of them. Re-read your two comments here, they're nothing but a series of assertions. I really expect better on ACX, and you have the rest of the internet for this tribalism.
> that the US population would unthinkingly go along
First, feel free to disagree with me, but do try to quote me correctly. I wrote: "something that looks vaguely like a potential solution to the real problems". That meant to imply that people do look at it and that they do not go along unthinkingly with everything, but that they have a base level of trust in their government, especially if they got to elect it themselves.
Second, I don't believe there's a contradiction between these statements in any case. People can be cooperators with the government, but defectors to each other.
>Second, are you genuinely claiming that the US is less volatile, less violent and more stable than France? Or was that irony?
Not ironic, no.
Less volatile? I don't know, depends on your definition. I'm afraid you'll put me on blast again for not providing a list that quantifies exactly how unpredictable, but sure, Trump II has made the USA pretty much unpredictable on the world stage. In that sense it's much more volatile than most countries right now.
Less violent? Again, depends. Americans are more violent against each other, absolutely. I hope that I can take easily googlable numbers like homicide rates per capita as a given. But against the government? I don't know. When was the last time the US had something like the following?
More stable? Currently not, no. For decades, the US has been toppling governments it considered inconvenient, and on J6, that long-standing policy has come home to roost. The US are in serious danger of democratic backsliding or an outright oligarchy, depending on how much of a fight her democratic institutions can and want to put up. Europe in general, including France, are also under threat of right-wing, anti-democratic takeover, but nowhere near at the same level as the US.
>Third, I'm really getting sick to death of these sorts of comments that make a long list of extremely confident claims and assertions about what will "obviously" happen
Again a misquote. I did not write "obviously" anywhere, so why do you put quotes around it?
>without a shred of hard evidence or substantive argument for any of them.
Yes, everything I write is my opinion. Do I need to add a disclaimer to every one of my comments that yes, this is indeed only my opinion, it is neither legal nor financial advice? These two comments were not papers for peer review, with facts or citations. They were opinion pieces on how I see America and, by extension, Americans from the outside. Deal with it or ignore me. And if you're going to provide evidence of how you disagree with me by quoting me, at least get the quoting part right.
I'm struggling to find any coherent overall argument in your comments. Again, it seems like you're stringing together unrelated tropes just to express your hatred for Americans, without caring if they make sense. Pick a position: you can either claim that Americans are overly obedient and trusting of authority, "one nation controlled by the media", OR you can claim that they're overly individualistic and reflexively oppose any government program as "socialism". But trying to make both these almost diametrically opposite claims at once, without *much* more effort than you're providing on explaining how they somehow reconcile, doesn't just discredit one of them, it discredits both. Since it's clear they're just rationalisations for your negative emotions towards a particular nation. (Note how the original comment didn't mention the US; you jumped in to attack the US with arguments (e.g. lack of civically engaged population) that would apply as much and indeed much more to many other countries, which really only makes sense if you're motivated by nationalist animus...pretty ironically for a German).
>Second, I don't believe there's a contradiction between these statements in any case. People can be cooperators with the government, but defectors to each other.
Um...what? This started with you claiming that Americans would never accept UBI, which is in every form I've seen discussed a *government* program, because it looks like socialism, which refers to an ideology of government action. And now you say that Americans are actually "cooperators with the government" and "have a base level of trust in their government" and...were you intending to refer to some very unusual meaning of "UBI" that is actually a form of private charity? And using some strange definition of socialism that has nothing to do with the government? Your argument here is just a complete non-sequitur.
>But against the government? I don't know. When was the last time the US had something like the following?
>For decades, the US has been toppling governments it considered inconvenient, and on J6, that long-standing policy has come home to roost. The US are in serious danger of democratic backsliding
Again, I cannot follow what you're even arguing at this point. Are you saying that violence against a democratically elected government is a good thing? A bad thing? A good thing only if it's in favour of your own political opinions? You seem to criticise the US for *not* having violent protests on the level of France, suggesting it makes them more obedient to "people in power of politics and the economy" and say that "they have a base level of trust in their government, especially if they got to elect it themselves"...and then you say they're "in serious danger of democratic backsliding" while "Europe including France" is "nowhere near at the same level".
You outright state that Americans have trust in their government because they elected it, that they don't violently oppose their democratic government like France, and that they're in much greater danger than France of democratic backsliding, and of toppling their government through violent events like J6. I...just can't keep track of the non-sequiturs!
>Again a misquote. I did not write "obviously" anywhere, so why do you put quotes around it?
Yes all right I shouldn't have used quotes, that was confusing. I didn't mean to attribute the word to you, I meant that the genre of posts I'm sick of, which yours belongs to, consist of unsupported assertions phrased *as if* the writer thinks they're obvious, whether or not they use the word. (I used quotes to indicate that they're not obvious at all, but I see it was misleading.)
>Yes, everything I write is my opinion. Do I need to add a disclaimer to every one of my comments that yes, this is indeed only my opinion, it is neither legal nor financial advice? These two comments were not papers for peer review, with facts or citations.
You're commenting in a community that has strong norms of following evidence and "making beliefs pay rent", to the extent of often insisting that people put actual real money on their predictions if they want to be taken seriously. To make a post like this here, filled with bare assertions about extremely controversial claims, phrased with such confidence, is utterly against the spirit and the standards of this place. If you're going to make bold claims and predictions with no evidence, with no deductive logical argument, and with no offer to bet on them in some way, the *least* you can do is hedge a bit with language like "it seems to me", "I could be wrong", "I can't prove this but I have a feeling" etc, instead of just outright "this will happen" unearned confidence. If you don't want to do any of those things, I reiterate that you have the *rest of the fucking internet* for that.
Uh...2020? I don't know if you're doing some strict comparison of numbers involved, but this is such a weird question when there were such famous and massive riots and protests so recently, that then spread to other countries as well.
Although I have no idea at this point if this helps or hurts your argument, because as discussed I have no idea what your argument on this even is.
Well, the 0-4 year olds definitely aren't going to vote.
e: Also, there will be a bunch of dead 45-49 year olds by the time they get to be 65-69, and by my quick math using actuarial life tables, the 20-24yo in 20ys will be about the same number as the 65-69yo, excluding any immigration (which will tend to be of younger demographics).
Relax anti-monopoly laws for corporations willing to distribute big chunks of whatever they, their subsidiaries and partners produce to employees, their families, and those hitting retirement age while in company service. Perhaps permit internal markets and payments in company scrip. We end up with giant Zaibatsu-style megacorps who have to look after the needs of all the humans involved with them. Since the population apparently finds the idea of the actual democratically elected government doing anything to actually help people utterly intolerable, we can at least get some of the effect at the cost of turning the country into several unelected de facto kings and their serfs. It’s not socialism if we don’t call it government and the board is making a profit! At this point the remaining vestigial parts of democracy can safely wither away. Gotta be better than just giving up and starving to death.
Many problems with this, not least that AI companies are taxable only where there are AI companies. So, the US, China and maybe a few others.
Whether there is enough income to tax there is another question, what with the rest of the economy collapsing.
The only possible way I can see UBI working is by just dropping money into people’s accounts based on the assumed productivity of that particular economy.
Laws have changed recently but the majority of corporate tax for Internet companies is not taxed where profits are made but where there are taxable structures.
I don't think you would need to tax AI companies, rather I think you would need to tax companies that own and use AI agents for production, I'll dub these companies AI clients.
I imagine it would work something like this:
- AI companies, probably mostly based in US, China etc. due to high barrier to entry, sells AI products such as AI agents to other companies. These AI agents will displace human labor when their productivity/cost ratio is lower than the available human labor. If there are sufficient free competition between AI companies, then the cost of an AI agent should approach the cost of providing that agent plus some margin.
- AI client companies will then increase their margins because they have lowered their production costs. If there is sufficient competition this will cause cost of goods to go down over time.
- Displacement of human labor will lead to unemployment if we assume that AI agents are better at most or all jobs. Unemployed workers will not have an income and will not be able to pay for goods, even if they are getting cheaper due to AI. This will in turn lower the demand for goods and services provided by AI client companies.
- Governments may then increase taxation of the AI client companies, for some part of the profit, and redistribute that money, e.g. in the form of UBI. Note that, hypothetically, if the tax is exactly equal to the increased profit margin of AI clients due to AI agents, then the cost of goods and services does not go down, and the AI client companies have the same profit margins compared to the status quo. The end result is then the same goods and services provided with less human input.
- It seems then that the UBI can be at least of the size of added margin for AI clients due to AI agents. If that margin is less than the salary the workers previously took home, then the workers would have less purchase power on average. So if the AI agents are only marginally better than human labor, that could lead to a situation where the AI client company now pays for compute instead of paying labor, leaving not enough taxable income for UBI - this seems bad. So it seems that the AI agents would need approximately 2x the production/cost ratio of a human worker, in order to provide a full compensation for the lost salary?
- If the AI companies manages to capture most of the profit then taxation and UBI becomes a problem for countries that don't have AI companies to tax. For this to not happen, there need to be adequate competition between AI companies.
I think it's pretty likely that AI companies will manage to capture most of the profit, because the industry has an extremely high (and constantly increasing) entry cost, the product is homogenous, and the consumers are numerous and therefore have low bargaining power. High entry cost leads to few players, which (with the other properties combined) then leads to cartel formation. Just look at current industries with these properties, such as RAM manufacturers who are constantly price-fixing.
An idle GPU generates no revenue, as does an empty airplane seat on takeoff.
High fixed costs, (comparatively) low variable costs, and "perishable" goods are a recipe for (actual, not Marxist) commoditization in which case prices approach marginal cost much more aggressively than in other industrial structures.
Indeed, without coordination these factors lead to almost zero profit. This is exactly what creates the strong incentive for breaking the law. Would you choose that your company is constantly cost-cutting and on the brink of death over sending a few messages to some guys you already know?
Cartels are generally difficult to hold together without legal force (cf. the thankfully-defunct Civil Aeronautics Board in the US) due to the incentives to cheat, and perishability turns that up to 11.
"This is the possible future that helps me sleep at night. Any other positive timelines worth dreaming about?"
Let's go with the obvious one. Scaling doesn't slow down and we don't run into significant alignment challenges. Smart AIs read and understand all human literature, decide morality is important, and then start acting morally.
We all live forever, in infinite plenty, as godly AI utility-maximizers consume everything in the universe and turn it into awesome-ness. Rather than paperclips, it turns out the greatest utility in the universe is playing ball with puppies, so we convert the entire universe into puppies and people to play with puppies and also lawns and balls with which to play with puppies and everything is ultimately good forever, because puppies are better than paperclips.
And that's kind of joking but, also, you're not just looking a gift horse in the mouth, you're kicking him in the dick. Like, you've laid out a scenario of virtually infinite physical and intellectual labor, where everyone is free to do anything, and you describe this a depressing, pessimistic hellscape and, like, I hate the fact that I kind of get that but you can just get off the internet and go do awesome stuff.
Things be pretty baller, fr fr as the children would say.
I didn't mean to imply the future I laid out was a hellscape - I'm personally pretty happy with the idea of an existence pursuing hobbies and leisure. I just think that is a minority opinion, and the road there is likely rocky.
Looking at the current landscape of research/incentives/capitalistic pressures, I am assuming that a full blown AGI/ASI timeline would not achieve alignment, ie doomed humanity timeline.
I think it's more likely that AI continues to progress at the the domains that is it well suited to, verifiable closed ended text based tasks like math and coding, while continuing to be terrible at multimodality, real time actions, continual learning, and open ended tasks that require subjective assessments and self critique.
I would predict that a millenium prize problem (or problem of similar caliber) gets solved at the same time that the best robots still fail the coffee test (whether a robot can walk into a random house and make a cup of coffee). Or for a computer based task example - the best AI would still be unable to beat a new video game (one not in the training data) of similar complexity as say Elden Ring.
The AIs become very superhuman on tasks where outputs can be verified (as you said, math and coding - but also any others where outputs can be automatically verified) but not on tasks where verification is infeasible, where they stagnate at very subhuman levels. BTW, there is a _broad_ set of tasks called "inverse problems" where there is some f(x)=y, y is known, f is known, and x is wanted. E.g. for a pattern of radiated radio waves ('y') , pick the minimal antenna ('x') that generates ('f') it. These are all verifiable output problems, often 'design' problems of some sort.
In a world where human labour is cheap relative to the productivity of capital, then the owners of that capital will be able to afford human labour for things that they otherwise would not use it for.
In modern India or Kuwait, or western countries a hundred years ago, a family with a reasonable middle class wage will have multiple servants -- maids and nannies and gardeners come first, then valets and chauffeurs and butlers, then footmen. Some of these servants could easily be replaced by robots (chauffeurs) but others cannot (nannies).
In a world where the cost of human labour is cheap compared to the productivity of capital (robots) then I would expect the owners of that capital to start finding new ways to employ human labour to make their own lives easier. The future looks like Downton Abbey, you'll have a rich family with a staff of both humans and robots -- the human servants will do tasks where a human touch is necessary, and they'll instruct the robots on the tasks where they are not.
Why don’t we see rich westerners today employ modern India levels of staff? The segment of the US population whose income exceeds menials’ wages by a ratio comparable to that for a “modern Indian” “ family with a reasonable middle class wage” vs the servants they employ is smaller, but is far from zero.
I suggest there are also other factors at play here, such as cultural differences in the number of strangers people are willing to tolerate in their home.
It's a good question, I think it's just out of fashion. Rich or even middle class westerners have no problems with gardeners, or visiting (not live-in) maids, or even nannies or personal chefs, but will shy away from anything more feudal-sounding.
I briefly worked for a rich guy who had a butler, but he wasn't a "butler", he was just a "guy who takes care of my house".
I'm not that rich but if I ever am then I'm getting myself a gosh-darned butler.
Would you have him wear the traditional white gloves and a tuxedo with a long tail coat? On the one hand, I could see wanting to go the full trad route just for the novelty, but on the other hand, I wouldn't care to adhere to a similar dress code myself, and we'd probably make for an odd pair if I'm out on my deck in cargo shorts and a t-shirt and he's serving me iced tea in a tux. The juxtaposition would not make a great impression on the neighbors, I dare say.
If you dress your butler in a tuxedo with tails and white gloves during the daytime you've failed every class shibboleth that would qualify you to have a butler, sorry.
Also if you're rich and your neighbours have line of sight onto your deck then what are you even doing, man?
>Non physically demanding jobs are gradually automated (over 5 to 10 years). Advances in robotics coupled with AI means physical jobs follow suit after some delay.
I can’t imagine how many tokens it would take to make a good AI PLUMBER.
In this world, North Korea can build a 100% robot economy with 100% of production going to robots and become exponentially more powerful. Any country that attempts to preserve human quality of life or even human life will be left behind in the arms race, overrun, and then killed and replaced with more robots.
Yeah, this is the 'human organizations are _already_ misaligned (to human flourishing), and are only currently contained because they need humans as parts' scenario. Yup, mere plug-compatible-AGI, _without_ ASI, (but with human-comparable robotics) is enough to trigger this scenario.
My sense of the near future is that AI hits some boring constraints before it's able to generate full-length feature films, novels, AAA videogames, or any construction more complicated than a backyard shed. It's theoretically possible, but requires more electricity than they can pay for with the sales of such things. However, it's able to build three-minute shorts, 1000-word short stories, simple mobile games, and design bespoke furniture for home use. It's also able to build simple utility apps for controlling home appliances, but not build them. It cannot make coffee unless a human sets up the inputs and tells it what to do, after which it can do so reliably as long as the grounds are stocked. Same for most cooking - for each recipe, a human has to show it where everything is the first time, and every time someone remodels the kitchen.
Desk jobs are ~80% automated as you say. But this just means that a week of work circa 2020 AD is done by Monday, freeing up Tue-Fri to do even more. Every desk job becomes 3-5x more productive, justifying raises from the added value created.
This also means a great deal more tasks are cost-effective. Among them are business administration for what are now very small businesses where the owner stays up late to fill out paperwork for payroll, taxes, invoicing, looking into opening another franchise, and so on. Along comes an MBA who got pinkslipped from a large company, offering to handle all that for about $400/week, remotely. The trick is that that MBA is doing the same thing for ten companies that couldn't afford a PTE with an office and bennies. I expect similar opportunities for other occupations.
Governments briefly consider UBI until they notice that a $15K check to a third of a billion people is five trillion, pays only minimum wage, and would quickly just inflate prices so everyone's even poorer. UBI just prints money; it doesn't print wealth.
The AIs "print" wealth, but not all of it. They still can't run all heavy machinery, although they can handle a surprising amount of transportation (as long as someone keeps maintaining roads and ports).
AI can derive some novel results in science and engineering, but it also frees up researchers and designers to ask a _lot_ more questions. Suddenly people are asking how to solve problems they never had time to solve before. Squishy meat brains are still the most efficient for working on them, even with AI assistance, due to the electricity constraints.
The hardest hit are probably physical laborers who don't know how to work complex machinery. They probably change to bespoke artisanry - leatherworking, glassmaking, etc. - though AI gradually makes inroads on those, pushing them to advance in turn. But that's a minority of people. Most of us just get as wealthy in real terms compared to 2020 humanity as 2020 humanity became compared to 1620 humanity.
I'm not sure I see how most of us would be getting more wealthy. A majority of existing jobs are already in the process of being automated.
As an example, the profession of driving a vehicle. In the USA there are over 3 million truck drivers and somewhere around 5 million taxi/delivery drivers. Even if AI progress halted today, it is hard to imagine us not cracking fully automated driving within the next ~5 years.
When these jobs are automated, these workers can't pivot to alternative professions like your MBA, and I don't imagine there are enough bespoke artisanry positions to go around.
That sector alone would more than double the current unemployment numbers.
If not UBI, what alternatives would prevent mass poverty and unrest?
Longer: automation drives down the cost of whatever was automated, which expands the set of applications that are now cost-effective. In other words, when you have an automated fruit picker, people suddenly think of a bunch more things for that fruit picker to do. This drives up demand, which drives up the price of fruit picking, and the job was never just mechanically plucking at stuff. So the people who were mechanically plucking things suddenly have a lot more work to do, and can charge more because more stuff-plucking is getting done.
That stuff-plucking is an example of wealth. So is the service of having stuff where it can be used instead of where it was made, which is what truck driving does. If that service can be supplied for some electricity instead of a package of food and healthcare and waiting until certain hours of the day and insuring against a greater rate of accidents, that frees up whoever was driving trucks to produce something else valuable.
"these workers can't pivot to alternative professions" simply isn't borne out by empirical observation. When disruptive tech landed in the past, various professionals complained that their jobs would disappear. But people aren't insects; they know how to do multiple things. Often the job was partly the thing automated, partly something else; professionals just did the non-automated thing more, and got more done. Often there was a similar job that wasn't automated; they just did that instead. This shifted people doing those jobs, who also found similar jobs, and so on. Since people are often continually changing jobs anyway, this often even goes without notice.
Jevon's Paradox and past empirical observations are based on instances of narrowly scoped automation. What happens when the new applications are also now easily automated, and so is any other profession that a worker could pivot to?
What makes you think the new applications are now easily automated?
An LLM recently produced an answer to an Erdos problem. Now that it's answered, what happens next? Did that LLM arrange for peer review? Did it try to apply that answer to some downstream problem?
Suppose you ask an LLM to write a full stack for a website to alert people when a collectible they've been searching for is going on sale somewhere, and the LLM writes it. Now what? Does the LLM publish that site? Register a domain name? Try to monetize it? Improve it? Try to apply that stack to another problem?
Have you tried to set an LLM to writing a novella? It can write a short story that's a few paragraphs long, but last I checked (could be wrong, stuff happens fast), it couldn't handle writing, say, 25 pages. One-page stories are automated; why is a novella not just doing the same thing 24 more times?
Having seen large amounts of AI-generated content, the strong impression I get is that it can't make anything beyond a certain size (depending on the media type), because the amount of compute required scales faster than the size of the content. The LLM quickly hits this limit, and either stops gracefully or starts "forgetting" stuff at the beginning, resulting in unusual loops or content that drifts without staying consistent with its origin.
Apply this to, say, a "butler bot", and you'll wind up with something that picks up your dirty clothes and launders them, but forgets it needs to put them in the dryer. Or orders the eggs and cream you wanted for that recipe, creates a delicious souffle, then lets the leftover cream spoil. Or runs in a constant loop of cleaning and folding clothes, cooking food, cleaning the bathroom, vacuuming the rugs, feeding the pets, and ordering more supplies, but never suggesting you might want to update your fashion or remodel the living room or build an addition since you keep complaining about where to put your 3D printer.
I've seen no evidence that LLMs are capable of improving themselves on their own initiative. They only do what the humans say. And they can't do more than a few hours at a time without checking back (or risking being wildly wrong).
They might get better at all this. I even consider that likely. You can ask what we'll do then. I'll respond: well, what do we do when the people around us get better at things? We'll do the same thing. Contrary to popular belief, I believe AIs won't improve any faster than people, due to real physical constraints like electrical capacity and chip shortages.
"The AIs "print" wealth, but not all of it. They still can't run all heavy machinery, although they can handle a surprising amount of transportation (as long as someone keeps maintaining roads and ports)."
This sounds like a strange sort of spikiness in AI capabilities. It might well be that extending LLMs to handle 3D operation is a lot more difficult than what scaling laws have been doing to knowledge work with verifiable outputs. But having the AIs able to handle e.g. building a shed, but _not_ able to be organized to build any building - or maintain a road, seems like an improbable combination.
I picked a shed because it lacks many of the features we would want in a more complicated building. A basic shed is four walls and a roof. No foundation, no electrical wiring, no plumbing, no air conditioning, no insulation, no windows, and maybe not even a door.
As I see it, AIs can operate cars because self-driving was an early job we poured billions of dollars of professional time into, for historical reasons. So there's the spike. If Google had instead gotten into a big row about, say, chicken farming, we'd have AI chicken farming and no self driving.
Many Thanks! I agree that a shed is indeed limited in the way you describe. I'm still skeptical that AI would be spiky in a way that allows construction of a shed but not of a house. Nailing two pieces of wood together with the modest tolerances required for either a shed or part of a house requires 3D navigation and reasoning ("Is this piece properly supported while I nail it?") that I expect to carry across to essentially all the other tasks you list.
I'm _not_ saying that these tasks are easy, or that current AIs or current robotics are capable of building either a shed or of building a house. But I do expect that if robotics of 20[23]N are capable of fastening together all the pieces to build a shed, I do expect them to also be able to install wiring, plumbing, HVAC, etc. A great deal of all of these tasks are 3D world models interfacing to vision ("Do I have the right part? Are the surrounding parts properly in place? Do they match the blueprints/CAD data to proper tolerances? Can I position the next part without collisions?") and to force sensing and application. A lot of the specialized knowledge _beyond_ 3D world models, e.g. knowing to strip the insulation off a wire before fastening it to a terminal, are already in the LLMs.
NYer here - on the subject of Alex Bores' election, if you care about things other than just who will regulate AI the most, it's not so clear that he's is the best choice. It's a three way race between him, Micah Lasher, and Jack Schlossberg. (As I write this, Manifold gives Bores, Lasher and Schlossberg 38%, 51% and 7% to win, respectively)
Bores and Lasher are both abundance liberals with similar legislative experience, while Schlossberg is a slopulist whose support primarily comes from social media. For example, on the subject of high housing costs, Bores and Lasher both strongly support building more housing, while Schlossberg's solution is to let you deduct your rent from your taxes. Overall, if you share the same ideology that Scott has described in his California ballot posts, Bores and Lasher are both good choices, and it's not clear IMO that Bores obviously deserves to win over Lasher.
(Edited to add: a reason I am considering voting for Lasher over Bores is that his AI regulation stance, while initially quite reasonable, appears to have gone somewhat overboard now that it's campaign defining thing, e.g. he's been negative on Waymo in a way that Lasher hasn't been)
When in congress, Lasher is likely to be less willing to take on OpenAI et al, as their money would be part of why he won (tearing down Bores helps him). Plus, I worry about the message that Bores coming up short would send to the rest of the country; they might see someone who promised to regulate AI getting a ton of money spent against him and failing and back off.
I think that's backwards, in that of course the OpenAI money is part of what elevated Bores so much, and made his campaign high-profile. The PAC money has hurt Lasher, not helped him.
I thought Bores’s economic policies were bad, but then I looked at Lasher’s website, and all of his “affordability” policies are demand subsidies and price controls.
I'm basing this on a candidate forum I went to that was organized by some abundance-aligned NY organizations (I haven't read their websites). Bores and Lasher were the two candidates of the four present that were head-and-shoulders above the others in terms of abundance policies, although it's possible that they were adjusting their answers based on the audience.
Lasher is actually very abundance-pilled, and has been for years at this point. Abundant New York supported him in his 2024 campaign, and even when I personally have heard him speak a few times, he actually tells people what they *don't* want to hear because he supports more building, despite people in the audience offering the standard NIMBY complaints. The concession to being a real politician is that he _also_ supports demand subsidies, but the tell is that he always talks about building more, even when it's clear the audience only wants to hear about price controls.
Everyone supports more building in the abstract. The question is whether or not he supports policies conducive to building actually taking place, and whether it will be funded in a way that grows the economy or contracts it.
I went to Lasher’s NY Assembly page, and the first piece of sponsored legislation I click on is a bill subjecting Waymo to regulation by the New York City taxi and limousine commission. https://assembly.state.ny.us/leg/?bn=A00793&leg_video=
I'm rapidly coming to the conclusion that "abundance" doesn't really mean anything anymore, it's just a label that is being co-opted by politicians and wannabes to sound like they're au fait with the latest cool trend.
Yes, I realise that sounds like the pipeline from "new concept" to "hackneyed phrase" was enormously accelerated, given that abundance was only bruited about very recently, but hey these are the days of AI, and once mainstream political parties start murmuring gently about how Candidate A is soooo much more abundant than stuffy old Candidate B from the other lot, you know it's toothless.
At this stage, if ever I had a handle on what abundance was supposed to be all about, to me it means no more than "I/we want people to have nice things! and be nice! and everything is nice! so vote for me/us and niceness!" as the kind of boilerplate campaign promise regularly trotted out, just given a new coat of paint with the latest shiny term.
I usually think of it as an attempt to push back on various hairshirt-bearing parts of the left--the "you should have less to help the environment," "you should have less so we can ensure favored groups get more," "you shouldn't want the things you want but instead these more virtuous things we want you to prefer," "you should have less so we don't have to do unpleasant things that look bad on video," and so on.
It's a bit like anti-wokism, which seemed like it meant something when the whole world was mouthing those slogans, but turned out to have no coherence beyond "we don't really think there are 37 genders or that using the word 'blacklist' means you're a secret racist."
I actually very much agree, and in fact will be voting for Lasher over Bores. Partly for the reasons you describe, and partly because Lasher is more of a policy guy. Beyond standard policy wonks, what I like about him is that he's good both in terms of delivering *outcomes*, and has also been thinking about systemic reconstruction to build back from the current administration's violation of policies. See his substack, Into the Weeds. (https://substack.micahlasher.com/)
My mother lives in the district in question, so I've seen the actual ballot, and two points:
1. The alternative language on that ballot is Chinese, which is surprising, and no, Chinatown is not in the district.
2. Schlossberg lists his name as "John Kennedy Schlossberg", which you may think is a bit much since AFAIK he doesn't generally go by that, but it could have been even-more-so, as his legal name is John Bouvier Kennedy Schlossberg.
"The alternative language on that ballot is Chinese, which is surprising, and no, Chinatown is not in the district."
Aimed at property owners, given the suggestions that Chinese money (amongst other foreign investment money) is flooding into New York/London/name big city of your choice and buying up all the new builds as investments then leaving them empty, thus yanking any chance of getting on to the property ladder away from people trying to rent/purchase their first home?
China is the only other language of the ballot. To Deiseach’s point, I’d expect those languages to be determined by census info, and there just aren’t a ton of Asians in the neighborhood.
According to Wikipedia, there is an Asian or yeah Chinese population of about 10.3% in Manhattan. So it’s not nothing but the only thing I can think of that makes sense is that the Chinese population would be the one most likely to not be able to understand English well at least in writing at least you could count on most of the Hispanic and black population being able to read English. Chinese is difficult that’s my guess.
So it might just be an artefact of the law as per the linked article? They were compelled to provide Chinese translation on ballots, so now every ballot gets translated into Chinese because nobody ever revisited the subject to see if it still holds? Easier to keep doing it even if not needed than to stop and have somebody haul you to court for racism and disenfranchising a minority population?
I recorded three songs the other day and it was a lot of fun.
One is about the fall of Western civilisation and another is about getting lots of emails. The third has a picture of a knight which I painted after an eye injury, with the result that I still haven't actually seen what it looks like unblurred.
I'm someone who loves earnest filk, even when it disagrees with me politically (like this does). Some honest feedback:
Your songs don't scan well, lines have frequently changing emphasis and length. There's also little rhyming structure or other musical framework to it. It really feels more like you're just talking in a singing tone. Try to identify a structure you want to follow early on, and shape the song around it rather than deciding what you want to say and forcing it into a vaguely song like shape.
To hopefully soften that criticism, let me say "Congratulations! You failed! Most people don't even try!" Music is part of being human, and I can tell you care. You have a solid voice, and the courage to post something you've made for criticism. I hope you keep recording and sharing your music.
Disagree with Unsaintly below; I caught the rhyming structure which was fine, musical framework which was rather obvious, etc.
Your voice has a rich quality, but you didn't always seem to control it well. When you didn't it was distracting, when you did it sounded professional.
The mix sounded good, clearly you know what you're doing or at least enough to be dangerous when it comes to the instruments and studio gear. Well done, wish I was that good. If you used samples I couldn't tell.
My only real criticism was that I found it unimaginative. The melodies had too little variation, the chord changes were predictable, the rhythms were simple to the point of tedium, etc. The lyrics to the two political songs were well written but seemed like they were trying to be edgy by virtue of espousing reactionary ideas. They didn't hint at any irony or inner-conflictedness or self-doubt or trying to see past the bend. They showed no evidence of being the product of fascination or curiosity or even pain. They even lacked the fun bravado I'd normally expect from such self-assuredness, like what I look for in rap.
I've got no musical knowledge, so this is all layman stuff.
There's something about the pitch of your voice here that makes me think "singing comedian" instead of "singer". I think that's what Unsaintly's talking about with "talking in a singing voice".
Typically that works to de-emphasize the music and emphasize the lyrics, but... I don't understand the lyrics. What is Nolmin Gnu actually PROMISING? Who's the woman in the middle of Delenda Est? (I listened to the knight song first, so by default she's Sally.) Why is Harry referred to in the same way as Sally if the speaker is Harry but not Sally? It all feels very loose.
I don't like the rhyming scheme, too many words rhymed with themselves ('Hated', 'Love', 'Fated"... 'Love' again). Some lines are shorter or longer than their counterparts; some lines match their counterparts by using empty words for padding. (I can't knock it too hard, I do that all the time.) The knight chorus has a full A-A-A-A, then degrades into A-B-B-A, and because the easier part comes second, it feels like a failure to maintain the standard. (If you reversed the patterns so they were growing more complex I think it would be fine.)
I listened around a little in all three. It's a strong effort.
I think the best-sounding part was the chorus of Delenda Est. Seems like that part plays to your strengths as a vocalist.
Marching and Ballad sound like they might be in an uncomfortably low register of your voice for the style of songs that they are? I'd be curious to hear what they would sound like if you took them up a few steps.
As for the lyrics, I think there are too many words. They didn't hold my attention. My advice would be to try to put your message across more succinctly and with more poetry, and to shift the focus from ideas to feelings.
Listen to “It’s Alright Ma” by Bob Dylan, and figure out why it works better than your songs
Record yourself singing a Depeche Mode song, one with Dave Gahan on lead vocals. Then compare yourself to Gahan and figure out what he’s doing that you’re not.
Someone posted this here a while back, and I gotta ask, is this how young conservatives talk IRL? Like they just casually shoot the breeze about white racial superiority?
I wouldn't believe the colour of daylight from Hanania (much too invested in bragging how he's so much better than these clods, these rubes, these Cousin Jonathans*) and Alexander, bless him, has his particular hobbyhorse about conservatism/right-wingers not being eugenicist enough and not in line with his views of the ideal society.
I dunno, everybody seems to hate Richard Hanania these days, left and right, so whoever was actually willing to hang out with him is probably not representative of much of anything.
Right at the top he admits that the scenario portrayed never actually happened. He was making the thing up to readers to give his own personal impressions of these folk. Bad form IMO.
"Note: I’ve known these people. And yes, it actually is this bad."
What are the job prospects for someone educated in biosecurity? Does anyone know? I'm a tech writer with a background in IT and a masters in Translational Pharmacology. Most tech writing jobs in the biomedical space want more experience with regulated environments than I have. Biosecurity sounds like a niche I'm well suited for, but I have no idea what looking for a "biosecurity related job" might look like.
Define better. What the average human would write is very different from what a LLM would write. LLMs writes in a very predictable manner, as we can see because people who read a lot of LLM text can often recognize that something is written by a LLM. On the other hand, two humans would mostly not write the same thing given the same prompt. Human brains are unique, and therefore flawed in different ways.
True, but... it is not like midwits write like LLMs. It is that they would like to. That is, the LLM writes better kinds of midwit texts, with better formulation, more elegant word usage, polite and professional and so on. It is as if you would take a longish corporate email, and then some experienced - and fairly educated - manager would comb through it and fix it.
Here is a LinkedIn AI example:
"Hardcoding pauses into a transaction pipeline slows down the entire system without actually solving the root cause of the concurrency issue. Half a second is an eternity for a server, but it offers zero guarantee that the obstructing process has finished its work. It is the architectural equivalent of closing your eyes and hoping the traffic clears before you step into the street."
I am not sure the average midwit (in IT) regularly uses words like pipeline, root cause, concurrency. The people I work with generally do not. Notice "obstructing" instead of "blocking", an elegantly educated word. Then making the traffic analogy, this is impressively beyond midwit level, as it sounds like understanding the nature of the problem and not just "bo hoo this is not the pattern they taught us at school".
This is what the midwit aspires to, the reason they are not all writing something like this is because they are not as good as this LLM.
Personally, I think the content of what is written is much more important than the words used. Over time, I think I would much rather have conversations with "midwits" as you call them, than with a LLM. People are in general vastly more interesting than LLMs if you take care to hear what they say. I find most people have unique and interesting experiences and thoughts once you talk to them. A LLM is fun for a little while, then it gets predictable. Still LLMs can be a very useful tool.
Regarding prose - I am not as impressed as you by the use of an expansive vocabulary. I rather prefer short and concise and to the point. A real art is communicating as much as possible with as few and as simple words as possible. I think this comes from reading a lot of academic papers and reports where I rather suspect the authors are trying to hide their lack of knowledge by using as many difficult words as possible.
I've noticed that some of the smartest people I know tend to communicate with very simple language.
>Notice "obstructing" instead of "blocking", an elegantly educated word.
Considering they're talking about how the bosses should speed the processes up, using the longer word instead of the shorter one cuts against their argument. "Half a second is an eternity, but let's force you to spend as many half-seconds reading this email as possible."
Just another example of the cheapening of language. I don't agree with the thesaurus, and if you think about the two terms, they are not synonymous. Obstruction and hindrance work as synonyms, but blockage is a whole different thing. The road is blocked. We will have to take a different route. There is an obstruction in the road. We will have to go around it. The less people care to make distinctions, the more synonyms there are, and it's not helpful. Still, you have a leg to stand on.
I actually like the use of "obstructing" here (I'm not a native speaker though). "Blocking" is a narrowly defined term when talking about processes running on a server, but concurrent processes can obstruct each other's program flow without technically blocking a shared resource.
E.g. when one process changes a file before the other one accesses it, it did not block the shared resource, but the second process might still read unexpected data.
Ah, well if it's a technical term that's different. But the argument still applies to "offers zero" instead of "doesn't".
If they don't understand the problem, you want small words so they can follow along. If they do understand the problem, you want small words to get the request over with. The only time you need to sound educated is when the audience doesn't give a shit about the problem and is judging you on personality.
Carlos, you are pulling our legs, yes? I'm inclined to the oul' sesquipedalianism myself, but using "obstructing" rather than "blocking", "impeding" or any verbiage of your choice from the thesaurus doesn't strike me as particularly elegant or educated.
In what world they would not sound educated? How many people know them or use them regularly? They clearly sound like words people use who read books, and serious books at that. Remember, the average corporate employee, the type I called midwit, have not read a book in ten years and a serious book never, that is my reference point, not people here. You people here are extremely elite outliers.
Whether they are elegant, that is context-dependent, in a conversation like this, such showing-off would not be, but showing off to that type in a blog post or corporate email? Certainly.
I admit, I have no familiarity with software engineering/STEM business types you describe, but if they are all grunting illiterates I must say I am surprised.
That "LinkedIn AI example" doesn't sound particularly elegant to me; it's slick and polished in the corporate-speak type of way, but it doesn't need a very high level of fluency in writing English to produce.
You are looking at this as someone who acquired English as a second, third or fourth language. For someone who grew up speaking English it’s not very sophisticated.
"hylic – lowest order of the three types of humans. They cannot be saved, since their thinking is entirely material and incapable of understanding the gnosis.
"5. They conceive, then, of three kinds of men, spiritual, material, and animal, represented by Cain, Abel, and Seth. These three natures are no longer found in one person, but constitute various kinds [of men]. The material goes, as a matter of course, into corruption. The animal, if it make choice of the better part, finds repose in the intermediate place; but if the worse, it too shall pass into destruction. But they assert that the spiritual principles which have been sown by Achamoth, being disciplined and nourished here from that time until now in righteous souls (because when given forth by her they were yet but weak), at last attaining to perfection, shall be given as brides to the angels of the Saviour, while their animal souls of necessity rest for ever with the Demiurge in the intermediate place. And again subdividing the animal souls themselves, they say that some are by nature good, and others by nature evil. The good are those who become capable of receiving the [spiritual] seed; the evil by nature are those who are never able to receive that seed."
— St. Irenaeus of Lyons, 'Adversus Haereses', Book I, Chapter 7
IMO, many of the thinkers of the field have become so symbolically charged that it's hard to get good secondary information. Everyone wants to tell you what they say (on both sides), but "what they say" conveniently fits perfectly into some narrative.
Reading a few primary sources can give you a better idea of how they really operate. It can be daunting, because some Continentals are deeply involved in traditions with very specialized vocabulary (and can really sound like gibberish). But they almost all have SOME things you can read as a regular person.
I think if you're interested in the French post-structuralists, you should read a tiny bit by them before getting into secondary literature.
There's this short essay (20 pages) by Foucalt called "The Subject and Power" which is pretty easy to read and is a pretty good intro to the themes of his thought:
Derrida is harder to introduce this way because much of his original thoughts come via commentary on specialist texts and use terms from phenomenology or structuralism. But there's this interview he did in 1994 (called The Deconstruction of Actuality, the interviewer chose that title, 14 pages) that I really like. He talks about Israel, Political Horseshoe Theory, Immigration and Xenophobia, etc. The way he interacts with the interviewer is a good example of how he approaches questions.
> To put it bluntly, I would say that to begin the analysis with a "how" is to suggest that power as such does not exist.
> Power relations are rooted deep in the social nexus, not reconstituted "above" society as a supplementary structure whose radical effacement one could perhaps dream of. [...] A society without power relations can only be an abstraction.
Derrida
> In order to stand up to this double intimidation, the only responsible response is never to give up the task of distinguishing and analysing. And I would also say: never to give up on the Enlightenment
> And since we had better recognise, like everyone else, that we have no choice but to protect what we take to be our own body, then let us be spared all these pure souls who appeal to high principles and put on a high moral tone and start lecturing us on politics as soon as we propose to control immigration and asylum
Do morals drive laws or laws drive morals? Like was spanking children banned because cultural norms shifted and a lot of people understood it is not OK, or rather the laws were changed first and then a lot of people figured if the state says so, then it is not OK?
I mean it would be easy-lazy to say it works both ways, but it would be important to know a bit more than that.
One thing we have seen 100% that prohibiting alcohol, weed etc. did not change the social norm at all... in the West. In Muslim countries it did. So keep up the law for centuries and then people internalize it? Or substances - victimless crimes - are a special case?
I honestly think laws have an effect on morals. That is because if other people have similar minds than I do, I have these two categories as "law-abiding citizen, proud, respected" vs. "criminals, humbled, disrespected", and it is not only that I would not be in the second group, my self-image does not even really allow comtemplating doing things that - even when never caught - somehow associate me with the second. But many people do not have minds like me, like many smoke weed, while I think like "I will not go to some slum to buy it from people I do not respect". But imagine I would make rich, educated friends who smoke pot, I would feel it is now high-status. Many minds are not like my mine, but perhaps many are.
Morals of course have an obvious effect on laws. Even if a utilitarian would say this thing is not wrong, if many people find that thing viscerally repulsive, it will get banned. It happened with bestiality. Of course there was this bullshit argument "animals cannot consent", but of course animals do not NEED to consent, they do not have those kinds of rights, after all they do not consent to getting killed and eaten either. It was obvious bullshit, it was simply banned because it is disgusting.
> I mean it would be easy-lazy to say it works both ways, but it would be important to know a bit more than that.
I'll ping-pong it back to you: Isn't it a bit easy-lazy to ask an extremely generalized question like "Do morals drive laws or laws drive morals?" and dismiss a fittingly generalized answer like "it works both ways"?
I think the answer lies in investigating each specific case you want to question. Did "the prohibition" (on alcohol in the US) change morals? Apparently not. An opposite example does not readily come to mind, but I'm sure they exist.
Per your comment on minds,
> But many people do not have minds like me, like many smoke weed, while I think like "I will not go to some slum to buy it from people I do not respect".
To me this sounds like a naively bad comparison. Smoking weed does not equate going to a slum. You can also just grow it yourself or have a circle of friends and acquaintances of growers. Of course, for many city dwellers and beyond, buying random stuff on the street is a typical expression of convenience and wilful ignorance; not unlike buying meat from the supermarket while saying "I could never hurt a fly".
You are a dreamer, my friend. I too would like to outlaw all the disgusting practices of the meat industry, of the banking industry, of various offshoring practices that exploit people, animals, environments and whose products are yet legal to import, and many other things that are clearly disgusting when looked at in clear daylight.
>An opposite example does not readily come to mind, but I'm sure they exist.
Maybe not a question of morals, but at least adjacent: A few years back, the EU collectively banned incandescent light bulbs. There was much wailing and gnashing of the teeth, but nowadays? LEDs improved a lot on all metrics, new form factors became possible, nobody misses the little heaters any more.
Sometimes you need to lay a fire to get rid of the old-growth underbrush; to make room for something better.
I got one better. I nearly burned down my parent's house once with a desktop lamp, in a freak accident involving a mattress and a loose mechanical contact.
That’s why you always want a cat nearby. Preferably a black one because they absorb heat the best. Very low smoke point. The flaxseed oil of domestic felines.
There's got to be a variant of the trolley problem in there. If you do nothing, the house of a random Internet stranger burns down. If you flip the switch, your cat's fur gets singed. What will you do?
I remember those desktop lamps, whose base was basically a big transformer converting voltage to 12V, and the bulb itself was supported by two metal arms that also doubled as uninsulated low-voltage wires. I found out that if you hold a piece of aluminium foil between the wires, it would burn up on contact while producing sparks.
This design was very common and I have no idea how it passed basic safety concepts for home electronics...
I have another set of spot lamps that used halogen bulbs but there is no equivalent form factor LED. I had to MacGyver the LED replacements into the lamps. It's fun to break out the old soldering iron now and then.
I agree with your verdict, but I'm not sure it fits the bill of a moral law that generates outcry if broken? People would certainly question someones intelligence if that person would still be using incandescent bulbs intentionally, but I do not see it having much of a moral dimension. Sure there is a connection to environmental impact, but I think it's too tentative.
Maybe as a german example, I just recalled that racing cars on public streets now faces charges of actual murder, not just manslaughter, if bystanders are killed as a result. That clearly fits the moral-outcry-category, and it was a quite surprising verdict by the court. I think public perception clearly influenced the result.
Laws are the application of force, and their effect can be compared to any other application of force. Do muggings drive morals about carrying money on your person? Do rapes drive morals about how to dress?
The most obvious way laws drive morals is when a bad law is passed, it drives people to want to tear the system down.
>Like was spanking children banned because cultural norms shifted and a lot of people understood it is not OK, or rather the laws were changed first and then a lot of people figured if the state says so, then it is not OK?
What laws? Spanking isn't illegal in any state. There are laws prohibiting the use of corporal punishment in public schools, but only in 33 states. You can't have cultural change driven by changing laws without changing the laws first, and they haven't changed at all.
As far as cultural norms shifting, the percent of Americans who believe spanking is an acceptable punishment went down from 84% in 1986 to about 70% today. Not that big a shift, really.
> One thing we have seen 100% that prohibiting alcohol, weed etc. did not change the social norm at all... in the West. In Muslim countries it did. So keep up the law for centuries and then people internalize it? Or substances - victimless crimes - are a special case?
What we see is that bans are ineffective unless you actually enforce them. If you want to ban alcohol then you need to also be committed to enforcing that ban; the US has this thing where it likes to ban things but not actually bother to meaningfully enforce those bans, which is why your streets are overrun with methheads and illegal immigrants.
Compare to, say, chewing gum in Singapore. They banned it (specifically the sale or importation, not the chewing), then they enforced the ban with penalties sufficiently severe that it wasn't worth anyone's while to try to get around it.
I think the comparison is not entirely fair. Banning chewing gum accompanied with hefty fines for leaving it on the pavement is much easier to accept, and there is no "Chewing gum industrial complex" to lose large revenues and lobby against it.
Try banning rice cakes instead ;) or some other revered "vice".
It was a smart move in Singapore - ban something that is generally agreed as a nuisance in public, but not high-status or high-desirability enough to warrant resistance.
The most egregious moral sin, murder, has been outlawed universally and that's not going so well. But the situation is still wildly different between some countries, so the underlying cause cannot merely be punishment and enforcement, no?
"@berniesanders is right: the future of humanity belongs in the hands of our workers, not the handful of AI oligarchs pursuing unlimited wealth and power at the expense of our job, our kids, our economy, our climate, and our democracy.
Here in NY, I delivered the strongest AI safety bill in the country. Now, I'm running for Congress to put Americans ahead of AI, and I look forward to partnering with Senator Sanders to get it done."
What "partnering" does he have in mind? The tweet quotes Sanders demanding a one-time 50% wealth tax on AI companies. That's postcolonial Africa s*** and people want to vote for him because he said some words about AI safety?
You know the old saying about "when fascism comes to America, it will be wrapped in the flag and carrying a cross?" The broader lesson is that terrible ideologies don't look like ugly cartoon villains dressed in black. They will be marketed to you. They will use your symbols and buzzwords and claim the legacy of your revered figures. Watch out!
> "when fascism comes to America, it will be wrapped in the flag and carrying a cross?"
So you have to redefine that quote to make it fit Sanders (who has been progressive left his entire political life), but you ignore the rather literally flag-wrapped, bible-thumping Trump?
I got that, sure, you can have that opinion. And yet, of all the ways you could have said that, you resorted to redefining a quote about fascism which would fit the current president much better than one of his staunch opponents. So yes, if you wanted to avoid the impression of making a "both sides are terrible" argument, let alone a "Sanders is worse than Trump" point, I think you could have done a better job.
I understand your point - the words "The broader lesson is..." are doing the heavy lifting there in terms of connecting it to a general principle of "things you vehemently oppose will be smuggled in under the veneer of that which you admire" - but I do think we're at a point in discourse where if the word "fascism" isn't necessary to make your point it's better to find alternatives.
There's a crisp irony to the fact that your post about manipulative use of buzzwords contains a sincere use of a buzzword for illustrative purposes, and it ends up causing a misunderstanding because people are so primed to see buzzwords and expect manipulation.
It's a pet peeve of mine when people take any comparison or analogy to mean you're claiming two things are identical. I'm not convinced that Sanders or Bores are as bad as you say but I think your meaning was clear, and reasonable enough.
I broadly like Sanders but he just drops in the occasional wildly over-the-top economy-wrecking idea offhand. I imagine hearing him talk was like being in the company of a forward-thinking Frenchman you're fond of when he suddenly drops "so anyway let's decapitate a bunch of dudes and their families."
Honestly if you replace "dudes" with "CEOs" that might not be far off some of what Sanders says. Yet I broadly agree with the non-decapitating part of his sentiments.
The cross perhaps wasn't necessary but it still did definitely play a role. If anything Trump's combination of appealing to the more religious while clearly having zero interest/knowledge about it on a personal level feels like exactly the disappointing reality the quote promised us.
But I will absolutely agree that I wouldn't have predicted just HOW thin his veneer of religiosity would be.
A few years back, Bill Gates was talking about a future where we tax the robots that replace jobs. He hasn’t mentioned it since. Odd, since it’s much closer now than it was then.
I doubt Bores has any particularly innovative ideas on this front, but it does seem like we need them. Otherwise, it’s just fractious chaos and decline until the wheels fall off.
The Democrats released a bunch of embarrassing info about Gates in the Epstein files, he may not be in a particularly generous mood.
>Otherwise, it’s just fractious chaos and decline until the wheels fall off.
I look around and don't see any decline. The tariffs and Iran war are bad, but almost everyone has a job and real gas prices are significantly lower than they were under the Bush administration. The real threat are brainworms that lead people to want to destroy the economy for no good reason.
As a software dev, I'm rolling my eyes so hard right now. Number of commits, just like "lines of code", is not only a useless metric, it has negative value because it sets perverse incentives.
Obviously. Yet, my org has just re-introduced a measure of bugs/kloc (yes, as in kilo-lines of code) that ought not be exceeded. It might already be silently dropped, as I guess I was not the only one questioning the sanity of such a decision.
To the original question: Good luck finding a metric that actually works. I think something like avg. $ profit per dev or "individual contributor" over time is a candidate. This would only yield useful results over a large amount of well-selected companies.
I am not looking for attribution per dev. I want to know whether at a high level, shipping more has any impact. My own assumption is that it doesn't - very few things a company does actually matter, so there is a lot of wasted effort. I want to know whether I'm right or wrong.
Oh I did not intend to imply attribution per dev. My assumption is that heavy AI use is reducing headcount in dev departments, especially squeezing out juniors (at least temporarily).
You're missing the point. Lots of companies claim they're much more productive, that they're shipping much more than ever. I'm asking whether anyone has done a proper study on whether this has any actual impact.
Companies do be claiming lots of things, especially when they have to justify heavy investment into unproven technology and divestment from proven assets (human coders). You gave the specific example of "number of commits" as a metric, which I commented on. I just wanted to let you know that this is not a useful measure of software development productivity, and if there was such a study that used that metric, that study, and people promoting it, should probably be ignored on that matter.
I'm a PM, and I'm not totally stupid. It feels like you're just trying to show how smart you are instead of engaging with the actual question. There are no good metrics for tracking engineering productivity, but the point is that many companies are claiming their development has really accelerated. What I'm curious about is the effect this has - I know there have been studies on whether engineers are really more productive in terms on their output, but what I want to know is whether, even if we grant they are indeed more productive, this has any effect besides, perhaps, on headcount. As in, does it even matter if companies are shipping more features?
Just obviously it matters a great deal, for that you don't need a study. How exactly does it matter right this very moment, you might need a study, what about at some moment a few months or years down the line, back to original answer
Yeah, many developers have anecdotal evidence how LLMs improved their coding speed. I could add a few examples myself, especially hobby projects that otherwise I would have not enough time and energy to do.
But coding is just one small part of functionality of the company, often not the most important part, even if it happens to be a software company. Sometimes the managers don't know what they want to get done, and keep changing their mind, or the original analysis is utterly flawed. There may be bureaucracy slowing everything down, or various departments of the same company fighting each other. In other words, even increasing the programming speed 100x (assuming generously no negative impact on quality) may ultimately make little difference in the company's overall output. Sometimes the actual bottleneck is analysis, or design, or even sanity in general.
Going to go vote in California (specifically LA) today and I'm still undecided about a few races. Please lobby me for your preferred issues/candidates!
What's the best work on AI consciousness that goes into detail about how LLM inference actually works, rather than just stuff like the Dawkins piece (which I actually quite liked) which just focus on how these systems behave?
“What is Intelligence?” by Blaise Y Arcas is slightly similar to your specification, and I really enjoyed it.
- has sections explaining how transformers (and some other ML/AI systems) work. Obviously not as detailed as a technical paper/textbook.
- Has very interesting discussion on intelligence (and secondarily consciousness) of conventional life and artificial life.
- Spends a lot of time vignetting historical ideas on intelligence and AI. I found it interesting, but maybe somebody who is familiar with the basic history of e.g. Cybernetics won’t find it so interesting?
- I think it spends 60–40 time split talking about neuropsychology vs AI
- I recommend because I found it interesting and entertaining. Also, I think in the whole book there wasn’t a single argument I found egregious. A few points I disagreed on were peripheral and so didn’t ruin it for me. I’d say it’s intelligently written. Therefore I’d happily recommend it to a Rationalist audience.
- Speaking of, in the last chapter he takes a position arguing against the old-school Rationalist stance on ASI (a la Yudkowsky’s sequences), and I didn’t see anything wrong with his logic.
Yeah, I'm fond of adding "avoid sycophancy" but being constantly reminded that it'd brush past my mediocre-at-best toes and recoil in disgust really made me question where I was going in life.
Sorry to be unhelpful, but the more you learn the mathematical background of LLMs, the less you will believe they are conscious. A transformer is little more than a key value store for computing probabilities of all data with respect to all other data. It only looks like it's conscious because it's a high level overview of one tiny part of the human brain.
I still agree with you overall, and I used to make similar-shaped arguments myself in the past, but note that key value stores (and matrix multiplications) are like epicycles - compositions of them are fully general, and given enough of them you can construct whatever you like. I’ve linked to plenty of demonstrations that iterated GEMM is Turing complete in past open threads, so here is one for attention: https://jmlr.org/papers/volume22/20-302/20-302.pdf
So unfortunately one does have to do more work to build an argument for a particular system not being conscious that would convince someone who does not already agree; simply mentioning that it’s made of kv lookups is no more convincing than “it’s made of transistors” or “it’s made of neurons”.
It was a tiny sample of one of the many billions of things which are so divorced from the human brain (whose consciousness we still do not understand), that I don't really see how you could find the link between them. But more power to whoever wants to try.
Seems to me that your idea of consciousness is that all this stuff we have in mind about a topic -- the facts, the history, other relevant sets of facts, our previous actions regarding the topic at hand -- are arrayed around us, sort of like whatever the transformer works with to compute probably with respect to all other data -- but in the human mind there's this little elf who looks at the data and *thinks it over,* and that's why we call the human view of of the topic a conscious *take,* and the AI's take just a mathematical result. But of course there is no little guy bopping around in the human brain, doing what we call *thinking.*. You're just building in the ineffable part, the "we are actually *thinking*" part, by sneaking a homunculus into your model of what actually goes on in the human mind when we arrive at a view of things, based on all the info we have about it.
> You're just building in the ineffable part, the "we are actually *thinking*" part, by sneaking a homunculus into your model of what actually goes on in the human mind when we arrive at a view of things, based on all the info we have about it.
There's no claim or implication of any kind of "homunculus" in the comment you're replying to, this is a blatant example of a strawman argument.
Isn't this just a reformulation of the Hard Problem of Consciousness, though? Or a restatement of why it is indeed hard?
Chalmers' argument is that C = consciousness doesn't supervene *logically* on the physical, i.e. given all of the physical facts about C, it's still surprising that we have phenomenal experiences like experiencing 'red'. But he posits that C might supervene *naturally* on the physical, by as-yet undiscovered psychophysical laws (while offering some not-so-elfin candidates). There's an element of consciousness that's above regular mechanical cognition, approximately "what-it's-likeness", that has to be explained.
So do you adhere to the idea that (a) LLM's share some of our phenomenal experience of the world? or (b) that phenomenal experience (linguistic ex: instantaneous association of certain words with images) isn't a necessary condition for C (the dispute is definitional)?
Over on Twitter, woke dead-enders are waging a desperate rearguard campaign against Graham Platner, bringing to bear all their heavy artillery (“incel,” accusing men of being insecure about their masculinity) to utterly no effect. The only high-profile Democratic politicians to oppose Platner are the centrist Corey Booker, who didn’t go any farther than expressing “concerns,” and Senator Fetterman, who’s on the verge of being thrown out of the party. Elizabeth Warren and AOC, woke favorites back in the day, are both supporting Platner. Some of the wokists seem unaware their powers are gone, others look like they’re desperately trying to manifest them back into existence. They seem to have a real, palatable anger at Platner over the infidelity, more so even than for Paxton and Trump. When they talk about Trump’s infidelity, it’s usually “those Christians sure are hypocrites voting for this adulterer.” With Platner, they’re MAD.
(On the issue of infidelity, I agree it’s both immoral and a sign a politician is likely to backstab his own voters. I find it funny that so many voters don’t care about it and then complain that politicians are dishonest. If he’d lie to his wife, why the hell would you expect him to be honest with you? But enforcing a norm against adultery requires enforcing it against your own side, not just the other side, which nobody in American politics wants to do.)
Yeah, the guy seems to bring some masculine energy that the Democrats really need.
> If he’d lie to his wife, why the hell would you expect him to be honest with you?
I don't want him to be honest with me, I want him to screw over my political enemies. And if he happens to screw over both my enemies *and* me, I will... deny that it happened.
Excellent piece about the decline of the Boy Scouts. Cultural changes were challenging enough, but then came the abuse scandal and then - and far worse - the descent of greedy lawyers like a huge flock of vultures. It’s truly disturbing and shows how greed can be utterly ruinous.
Note: in the Northeastern city where I grew up the Scouts weren’t particularly popular, the YMCA dominated youth activities, and most of my schoolmates who were Scouts were rather dweebish.
I'm not going to watch a 20-minute video but I always thought that Simpsons episode where scouts were portrayed as desperately uncool was a huge blow to the movement.
I guess most people don't realize how probabilities change when things get large. If something is a 1 in 1000 event, when your organization has 100 000 members, it is practically guaranteed to happen, even if your organization is perfectly normal.
And when it happens, the leaders of the organization will be held responsible for that. Because the mathematically illiterate intuition suggests that they had to do something horribly wrong in order to attract and keep the 1 in 1000 rare thing, especially if it happened more than once.
A more serious accusation is "they knew, but they did not do anything about it". But even this gets difficult with large numbers -- you will find cases when the result of an investigation is a clear "yes", or a clear "no", but also many where the evidence is inconclusive, with various degrees of probability. And of course, if you fire someone based on inconclusive evidence, they too may sue you, for wrongful termination, or slander, or whatever. But if you don't, any note that you had a suspicion but didn't fire them, will be worth gold for the lawyers who will later investigate you.
Somewhat ironically, the “two deep” youth protection policy that the Boy Scouts have adopted supposedly is the best anywhere and many other organizations use it as a model. Unfortunately it may be too late.
> The person who has been trained to read closely and write precisely is, right now, in 2026, more useful in almost every knowledge-work context than the person who has not been.
So writes N. Ángel Pinillos in his Substack essay: "The Honest Case for the Humanities." And while I agree this is probably the case, I'd say being useful is a sucker's game.
Being useful means you're being used. A STEM degree and high-paying job at, say, one of the Magnificent Seven is all very nice, but the odds are that your corporate masters will let you down. Most engineers lose their value to their employers after age 40. Their salaries get too high. They realize that there's life outside of work. And most people who've reached middle age have learned not to take shit from their managers. The usual excuse for lay-offs is financial cutbacks, and there's some truth to that. But they're culling the herd to bring in some eager young achievers at half the price of someone with 15 or 20 years of experience. I'm reminded of salmon struggling up a series of fish ladders. Only a few of them reach the top. And then they die.
Why am I telling you this? Because I could tell you how my undergrad degree in Anthropology has served me better in the corporate world than my graduate STEM background. Certainly, it gave me insights into how corporate social systems work. But the real value is that it gave me insight into life outside the corporation. Pinillos made this point: the humanities made my life outside of work richer. Sure, I needed to be a useful cog in the capitalist machine to earn the money to live comfortably. But the humanities made my life worth living.
It's funny how I strongly agree with some parts and strongly disagree with other parts.
> [philosophy] is one of the most rigorous trainings available in a particular cluster of skills: reading carefully, writing carefully, expressing yourself carefully, and thinking carefully in non-quantitative domains.
In my experience it's more like the opposite. Look at Hegel's Logic, that is a humanities guy cosplaying a mathematician doing proofs, and failing laughably... except that instead of being laughed at, he became an example of a genius, an inspiration for Marx who in turn is an inspiration for many. Generally, philosophy seems to lack the ability to call out wrong ideas and discard them, the way science has discarded e.g. astrology or phlogiston; instead it just accumulates, and the 21. century students still learn Plato etc.
> I was also a mathematics major, because I had figured out, also correctly, that quantitative reasoning would matter for whatever I ended up doing. The two together seemed to me to cover the ground and challenged me immensely.
I also have a foot in both camps; I studied mathematics and psychology. The knowledge of each was useful to see the silliness of the other. I don't see it as an argument for the superiority of humanities, but the superiority of knowing more. Yes, the STEM people would benefit from knowing more humanities. And the humanities people would benefit from knowing some math and logic (actual logic, not Hegel's Logic).
> Most engineers lose their value to their employers after age 40. Their salaries get too high. They realize that there's life outside of work. And most people who've reached middle age have learned not to take shit from their managers.
This part I fully agree with. The important thing is to invest your money properly so that when you are 40 you no longer need the job. (Disclosure: I failed at it; living in a corrupt post-communist country, all investment opportunities I tried in the past turned out to be scams, and now that I know better, I am old and have a family to feed, so less money to invest.) The older you are, the more you see that the companies operate by "work harder, not smarter" when it comes to organizing your work.
>Generally, philosophy seems to lack the ability to call out wrong ideas
Probably became they haven't solved epistemology sufficiently to come up with a single agreed definition of right and wrong. Have you solved epistemology yourself?
I'm taking a course in Middle Platonism at the moment (considered by most philosophers to be a dead zone of philosophical thought), and I was surprised at how many questions about the origins of the universe and the organization of things they were trying to address. Even though they had no empirical data to work with, at the category level many of the arguments the Middle Platonists made are still being rehashed today by cosmologists and physicists who have empirical data but can't agree on what it means. :-)
Like, one of his most famous concepts is that there are "ideas" and everything we know is just their imperfect reflection. I wanted to know more about those "ideas", like how specific they are, whether they are hierarchical, etc.
As an example, is there an "idea" of a dog? Of a mammal? Of a poodle? Of a black poodle? Of a black poodle that can dance? Basically, some kind of rule to decide which things (according to Plato) are the perfect pure ideas, and which things are their imperfect reflections, even if those things are still abstractions themselves.
When I asked at the classroom, not only couldn't anyone give me an answer, but they were surprised why I was even asking. From my perspective, it is obvious: either the concept refers to something, and then I want to know what it is, or it does not refer to anything, and then why are we wasting time learning it? The impression I got was that it's about learning the right keywords, and then impressing others with your knowledge of them. There are high-status maps, but no territory they are supposed to reflect.
Didn’t object oriented programming kinda crib from this sort of thinking with the ‘ideal’ being the definitions of classes and the individual instantiations being objects of the defined classes?
Each object can have its own individual characteristics but remains true to the abstract class definitions.
What exactly is and isn't Plato's contribution? Plato did not invent the fact that we use the same word, e.g. "dog", to describe many dogs. That's how language worked, long before Plato. Plato just said that somewhere in "the realm of ideas", the is a divine "the Dog" of whom all dogs on Earth are merely imperfect reflections. Or something like that.
According to Wikipedia, the oldest program that used what we now call object-oriented programming was "Sketchpad", a vector graphical editor, which used names "master" and "occurrences" for what we currently call class and instances.
I briefly read the master thesis on the project, didn't find a mention of Plato or ideas or whatever. (I may have skipped something; those were pages scanned from paper.) So, in my opinion, the contributions to Plato are just applied retroactively. But I could be wrong here.
I looked into this a bit and I think you are right. When Smalltalk was developed they were interested in modeling real objects with class definitions not trying to describe an ideal form.
When I was learning C++ the class/object relationship felt analogous to the ideal form/instance relationship but it breaks down when you consider that the class definition itself is just a human made definition no more “true” than the object instances.
>Didn’t object oriented programming kinda crib from this sort of thinking with the ‘ideal’ being the definitions of classes and the individual instantiations being objects of the defined classes
>Like, one of his most famous concepts is that there are "ideas" and everything we know is just their imperfect reflection. I wanted to know more about those "ideas", like how specific they are, whether they are hierarchical, etc.
More research was needed, I suppose. Such as middle Platonism and late Platonism and Neoplatonism and Max Tegmark.
>As an example, is there an "idea" of a dog? Of a mammal? Of a poodle? Of a black poodle? Of a black poodle that can dance? Basically, some kind of rule to decide which things (according to Plato) are the perfect pure ideas, and which things are their imperfect reflections, even if those things are still abstractions themselves.
The idea refuses to either die, or be finalised ALONG WITH THE ALTERNATIVES.
Class based object orientation versus object based is the latest round in the Plato versus Aristotle
>: either the concept refers to something, and then I want to know what it is, or it does not refer to anything, and then why are we wasting time learning it
Does extratestrial life refer to anything?
Does dark matter referr to anything?
Do tachyons refer to anything?
It's not as everything falls into two neat categories , the fully understood and the completely non existent.
Could you expand? Does "music" refer to the critique of heroic epics in the Republic? Does "politics" refer to the idea that the Republic is fascist state?
> Generally, philosophy seems to lack the ability to call out wrong ideas and discard them
Regardless of whether Hegel was right or wrong, chewing on philosophy gave you a better ability to see through the bullshit, didn't it? So, Hegel wasn't entirely a waste of time.
And, yes, knowing MORE is better. Many former humanities peeps I know could use a deeper background, if not in science, an understanding of how science is supposed to work and how it actually works. Otherwise, the poor dears are misled by every flashy headline on diet advice, environment, medical advances, etc.
I cringe whenever I hear someone say, "Trust the Science!" — because (a) it means they don't really understand the scientific method and/or don't understand the science that they're promoting, or (b) they're researchers promulgating a viewpoint they'd rather not be challenged on.
> chewing on philosophy gave you a better ability to see through the bullshit, didn't it?
I think it doesn't work this way.
If you want to e.g. recognize homeopathy as bullshit, you need to study lots of chemistry. And only afterwards, look at homeopathy and notice that it contradicts everything that is known about water molecules. If you only study homeopathy, it does not produce the same effect.
Similarly, you recognize Hegel's Logic as bullshit if you are familiar with actual logic, not if you study too much Hegel.
That's a very good point, but there's a *huge* universe outside our empirical grasp that we can only address with speculative reasoning. And the logic that you used to analyze Hegel is based on the reasoning first developed by philosophers. I'm not saying we can arrive at truth through reason alone. But reasoning can help us filter the more likely from the less likely.
"If you want to e.g. recognize homeopathy as bullshit, you need to study lots of chemistry." I think that's not the only way. It also would suffice to grasp that homeopathy is ideally suited for double-blind studies and have a look at those.
"Regardless of whether Hegel was right or wrong, chewing on philosophy gave you a better ability to see through the bullshit, didn't it? So, Hegel wasn't entirely a waste of time."
I would feel better about Hegel's pointing out of bullshit if he (or his disciples, I can't remember) didn't address it by inserting even more bullshit at an even deeper level.
Scrip. This was a feature of the Gilded Age. It was a way of fleecing your workers and keeping control of them. It would be interesting if it could be reinvented as something that is positive.
Jean-François Lyotard’s _The Postmodern Condition_ partly blames the computer scientists for giving rise to the postmodernists. (I don’t think I can summarize Lyotard without re-reading it; but if I was try an outline a similar kind of trajectory — Bertrand Russell’s logicism crashes and burns thanks to Kurt Gödel, and of course Wittgenstein was with Bertrand Russell’s group, and then, well, here we are. You might object that Derrida is not mounting the same kind of attack on the system that Gödel is, but is a very different sort of anti-Bertrand-Russell)
But maybe the genealogy of it (good Michel Foucault word there) is Hegel -> Kierkegaard against Hegel -> Heidegger -> Derrida, while, in parallel, there is another anti-systematizing project that’s Russell -> Wittgenstein against Russell
The other thing that surprises me is people equating postmodernism with Marxists.
When Lyotard says the key characteristic of postmodernism is “skepticism towards metanarratives”, Marxist/Hegelian accounts are one of the main “metanarratives” they’re being skeptical towards, so we can rephrase Lyotard as “the postmodernists think the Marxists are insane”.
Ok, you can take the line that Žižek isn’t actually a postmodernist, he’s one of those Marxists the postmodernists were being skeptical towards, and thus is exactly what’s wrong with Žižek.
“ In 2019, there were about 150,000 people working in autism therapy. Six years later, there were 654,000—more than the number of people who work in mining and logging, or telecommunications, or at the US Postal Service.”
It’s enough to make you reach for the antidepressants :(
The article (and certainly the headline) is probably a bit sensationalistic, even if it is from Science magazine, and while I'm fascinated by the science here, I'm also rather disturbed.
"Not alive, but not dead: disembodied human brains used for drug testing
By restoring some functions to intact brains from deceased donors, the startup Bexorg hopes to create a better drug development test bed for neurodegenerative diseases"
How fresh, exactly, do these brains have to be? How are they able to rev them up to some kind of functionality? "Some" functions? What functions, precisely? Who the heck persuades people to hand over brains of deceased family members, without coming across as a Universal monster movie of the 30s? Is this applicable to cryonics, given that brain preservation and resuscitation is important? Can this work for what we now consider brain death, the way artificial resuscitation and restarting the heart did for previous ideas of when death occurred?
But at the same time: lads, Lovecraft's "The Whisperer in Darkness" was not meant to be an instruction manual!
"NEW HAVEN, CONNECTICUT—Just a day ago, the brain was in a living person. Now, hours after its owner died, it sits on a cart draped in tubes that quiver as they pump liters of blood substitute and other fluids through the organ, supplying oxygen and removing waste. With most of its key functions intact but its electrical activity quenched by anesthesia, the brain hovers between life and death. As it metabolizes experimental drugs, sensors record its reactions, capturing hundreds of data points on its cells, proteins, and physiology. Then, after 24 hours in this state, it will be sliced into hundreds of pieces for more detailed study.
The brain is one of more than 700 that the 5-year-old biotech startup Bexorg has nurtured and studied using a set of proprietary brain-sustaining machines it calls BrainEx. The platform grants researchers an intimate look into how potential therapies might work inside brains with neurodegenerative diseases such as Parkinson’s, Alzheimer’s, or amyotrophic lateral sclerosis. Bexorg can biopsy the brains and discover how long a drug stays in cells, whether it hits its molecular target, and any hints of side effects.
Early results, including a 2025 poster reporting that the preserved brains match living brains in their response to certain therapies, are encouraging, says Bruna Bellaver, who studies neurodegeneration at the University of Pittsburgh. “It’s a huge step up from mouse models,” she says of the platform.
Bexorg has largely stayed under the radar. Its founders have published no papers on its work other than early experiments with pig brains, although Vrselja says they are preparing their first paper on human ones. Now, the company is scaling up and inviting new attention. Its new lab space, which Science visited last week, will house a 1.2-meter-tall robotic arm to automate the process of slicing up to 1600 brains per year and analyzing 11,000 proteins in each. At a media event today, the company will showcase Bexorg’s assembly line process—and seek to reassure the public that its disembodied brains don’t cross ethical lines or risk regaining consciousness.
...The BrainEx machines sit in six plexiglass cubicles in an office overlooking Yale University, where Vrselja, neuroscientist Nenad Sestan, and their colleagues conceived of the idea 10 years ago. Before each brain is placed in a machine, surgeons examine it through a loupe, then suture four plastic ports into the vessels that once supplied it with blood so it can start reacting to drugs and generating data. Once the brain is attached to the BrainEx machines, an artificial lung and kidney oxygenate and filter fluids as they flow through the organ."
Yeah, it's all absolutely fascinating, now if you'll excuse me I need to go source a diving helmet or something. Cranial protection, be sure or be sorry! What is that buzzing noise I hear behind me?
In the game Sid Meier's Alpha Centauri, you get this quote as a popup the first time you build the "Bioenhancement Center" base facility:
-----
"We are all aware that the senses can be deceived, the eyes fooled. But how can we be sure our senses are not being deceived at any particular time, or even all the time? Might I just be a brain in a tank somewhere, tricked all my life into believing in the events of this world by some insane computer? And does my life gain or lose meaning based on my reaction to such solipsism?"
Project PYRRHO, Specimen 46, Vat 7
----
When I started reading your comment, I was expecting this to be about brain organelles, little clusters of cultured human brain cells used for research purposes (mix of biochemical and computational research). Those already give me the creeps. They're small enough (maybe one millionth of the cell count of a full brain) that there's no real danger of them being meaningfully sapient, but is still seems like a line of research that needs a strong bright-line rule to stop it before there's any danger of making a conscious brain in a jar, and they've already hopped the first Schelling Fence by doing the research at all.
But no, these folks are doing literal brains in jars. Similarly, I expect the brains are hopeless addled by the time the researchers get them, no longer capable of anything like consciousness, but this is much much closer to dystopian horror territory.
I lost a friend to a car accident many, many years ago. She was comatose when she reached the hospital and imaging showed several brain damage. She ied a few months later without ever regaining consciousness. Having this as an emotional reference point gives me complicated feelings about this line of research. A reminder that the brains in question almost certainly aren't meaningfully conscious, visceral unease at doing this kind of experiments on the brains of people who could be thought of as merely "mostly dead" even if the emphasis is very strongly on the "mostly", and also a reminder that this kind of technology potentially has very real lifesaving implications even if the research itself is unsettling.
"When I started reading your comment, I was expecting this to be about brain organelles, little clusters of cultured human brain cells used for research purposes (mix of biochemical and computational research). "
Precisely why my gob is so smacked. It's not brain tissue, it's not emulations, it's real human brains sawed out of the skull hours ago and wired up to the life support tubes. I would like more information but now I'm sorry I ever watched any of those "mad scientist brain in jar" movies because why are we careening as fast and hard as we can into "let's turn dystopia into reality!" territory?
This is the first piece of information I've heard in my life which has made me actually reconsider being listed as an organ donor. I wonder how they are sourcing the brains for this work? Just buying cadavers from the same place as med schools do? I am seriously about to revoke my organ donor status over this. I signed up in the hopes that a kidney or something of mine could help someone in need, not for my brain to be plugged into the venture-capital torture nexus...
It's not really covered how they get donations. I imagine the brain needs to be fairly fresh, otherwise it's not going to be much use. So that gives the unedifying image of a surgical team waiting by the bedside for legal brain death to swoop in and start scooping out Uncle Bill's brain.
My bad, I didn't check their handy posters where they explain how they get the brains and show you the BrainEx machine (complete with hooked-up brain):
"Acquisition of postmortem human brains adheres to the highest possible ethical standards, overseen by an independent board of world-renowned bioethicists. Brains are procured through Organ Procurement Organizations (OPOs) with enhanced levels of consent from patients and families that specifically cover the BrainEx platform. Measures are taken on the BrainEx device to ensure that there is no possibility of coordinated network activity associated with consciousness."
They seem to be getting them from elderly dementia patients who died naturally, since their initial emphasis was on research into drug treatments for Alzheimer's, so I imagine that enables them to persuade families to donate brains to help other people avoid the degeneration they saw with Uncle Bill at the end of his life.
Still, pardon me for faintly exclaiming "Oh, God..." here:
"Figure 2. BrainEx perfusion restores cellular activity and drives growth and repair. a, Perfusate measurements show that glucose is consumed, and lactate is produced by the brain, indicating active metabolism. b, Proteomics comparing perfused versus non-perfused brains reveals that BrainEx reduces neuroinflammation, and activates transcription, translation, DNA repair, and synapse and membrane repair."
More reporting does not dial down the creepiness factor:
"But you can put those doubts out of your very-much embodied mind, assures Brendan Parent, one of Bexorg’s six ethicists. The extracted brains are almost devoid of the coordinated neural firing necessary for minimal consciousness, he told Science. To prevent the eerie implausibility that some the brains produce electrical activity, the brains are also dosed with anesthetic propofol. Of course, that such a measure has to be taken in the first place may actually be less assuring and more unsettling.
Ethics aside — not a statement that should be made lightly — the scientific possibilities that these extracted brains afford may well hold promise. Bexorg CEO Zvonimir Vrselja said that the brains come with decades of environmental exposures, histories of drug treatments, and other factors that make them a more realistic testing medium for drugs. “You get cells that have been there for 60 to 80 years,” Vrselja told Science.
...Bexorg is the same startup that demonstrated, over six years ago, that it could keep decapitated pig brains alive for 36 hours using a prototype of its BrainEX machine.
Today, its human brains aren’t kept running in perpetuity. After 24 hours, they’re sliced into hundreds of pieces so they can be analyzed by scientists. The company plans to use a robotic arm to slice up to 1,600 brains per year."
How reassuring. "Almost devoid", well no worries there, and then they'll chop up the brains into sushi for further study.
36 hours alive decapitated pig brains sounds too much like "That Hideous Strength":
"I’m telling this dreadfully badly. What it really was, was a head (the rest of a head) which had had the top part of the skull taken off and then… then — as if something inside had boiled over. A great big mass which bulged out from inside what was left of the skull. Wrapped in some kind of composition stuff, but very thin stuff. You could see it twitch. Even in my fright I remember thinking, ‘Oh kill it, kill it Put it out of its pain.’ But only for a second because I thought the thing was real, really. It was green looking and the mouth was wide open and quite dry. You realise I was a long time, looking at it, before anything else happened. And soon I saw that it wasn’t exactly floating. It was fixed up on some kind of bracket, or shelf, or pedestal — I don’t know quite what, and there were things hanging from it. From the neck, I mean. Yes, it had a neck and a sort of collar thing round it, but nothing below the collar; no shoulders or body. Only these hanging things. In the dream I thought it was some kind of new man that had only head and entrails: I thought all those tubes were its insides. But presently — I don’t quite know how, I saw that they were artificial. Little rubber tubes and bulbs and little metal things too. I couldn’t understand them. All the tubes went into the wall. Then at last something happened.”
“…Well, quite suddenly, like when an engine is started, there came a puff of air out of its mouth, with a hard dry rasping sound. And then there came another, and it settled down into a sort of rhythm — huff, huff, huff — like an imitation of breathing. Then came a most horrible thing: the mouth began to dribble. I know it sounds silly but in a way I felt sorry for it because it had no hands and couldn’t wipe its mouth. It seems a small thing compared with all the rest but that is how I felt. Then it began working its mouth about and even licking its lips. It was like someone getting a machine into working order. To see it doing that just as if it was alive, and at the same time dribbling over the beard which was all stiff and dead looking…”
And this may be gallows humour, but everyone seems to have names out of the good old Gothic tradition of Eastern European mad science shenanigans (though I acknowledge Frankenstein was Swiss).
Bexorg are leaning into the Mi-Go from Yuggoth 'we'll take your brain back to Pluto in a cylinder' vibe as well:
"Unfettered from the limitations of cell cultures or organoids, our platform is capable of manipulating and recording the cellular activity of fully mature, intact, and isolated brains for days on end."
I was thinking about The Colours of Her Coat and the recent brouhaha about taste. I wonder if being over-exposed to a form of art creates taste, in the sense of 'you become less appreciative on average but gain a deeper ability to appreciate'? Taste has always seemed like a negative to me – as someone who rarely watches television, I enjoy high-quality slop, and would have less television to enjoy if I did not enjoy high-quality slop. But perhaps it could save us here. If good television becomes so common that losing your ability to enjoy high-quality slop (by this I mean Marvel and that sort of thing) does not affect me, as I do not watch TV enough to have consumed all of the 'truly good' TV, is that a loss? Filtering becomes hard, of course, but it would probably be solved at the algorithmic level at that point. Just random thoughts bouncing around my head that I have yet to think through.
Another example, reading writing in a script you know, half-know, and completely do not know are (to me) wildly different cognitive experiences. There will always be new scripts to learn. Even if you learn every human-made writing system, A.I. can generate beautiful ones for eternity. Will this feel fake? Consider the way many reacted so negatively to the 'you control the buttons you press' Tweet, or the phenomenon of analysis paralysis: the mere existence of an option makes many peoples' experiences worse. Are such people just skill issued, or are they victims of systemic failings? I do not know.
I watch very little TV, but did watch Breaking Bad, with great enjoyment. In the past, with my daughter, I did watch a bit of slop, esp. America's Next Top Model. I actually have no idea how classy BB is considered to be, but I'm guessing few would call it slop, or rate it lower than ANTM. I notice a real difference between how BB has stayed with me vs. how AMTM has. I feel nourished by BB. I still ruminate sometimes about the characters, and think about parts of me that are sort of like them -- my inner Skylar, my inner Gus Fring, my inner Jesse. I still feel a shiver thinking about certain parts, and snigger with pleasure about other parts. Occasionally I have an urge to rewatch a certain episode. I do recall a lot of ANTM, and sometimes think of it spontaneously, but the thoughts aren't as emotion-laden, or as deep. For instance, I recently remembered the episode where the aspiring models had to have large spiders placed on them. Wondered how well I would have tolerated it, and whether they were de-venomed tarantulas or some other species. But these were just kind of passing thoughts about something novel I once saw. I had not been deeply affected by it. So based on all this, I'd say better entertainment, stuff closer to art than to slop. engages us more deeply.
Very neat, thank you! Do you think that a near-infinite amount of BB-like shows would be able to continue to form you in this way? Say, if AI could generate shows of BB quality indistinguishable from human writing.
The whole 'Seinfeld Is Unfunny' page on TVTropes covers a lot of things like this. https://tvtropes.org/pmwiki/pmwiki.php/Main/SeinfeldIsUnfunny A show does something so memorable and remarkable that EVERYONE imitates it, and it becomes so mainstream that the original show no longer seems remarkable.
I'm of the opinion that more is better, as long as there's some valuable filtering tool. But that second part's the catch; supply has sped up so much it's flooding out the filters. And creators will always be incentivised to try to corrupt the filters to say their particular show is good.
(As a filter tangent, I hate how of all things Rotten Tomatoes has become the critical standard. They put a movie's Rotten Tomatoes score on its Blu-Ray now. Nothing says quality like decaying fruit.)
But yeah, it will be neat to see how algorithms keep up in the slop age. Especially if people, in the short term, prefer slop. I consider myself above AI generated drivel, but perhaps only because the sort of AI slop (derogatory) that would appeal to me has not yet been created. I hope not, but I worry.
I checked this out and I read the section on live films. I couldn't believe that Jean-Luc Godard wasn't mentioned once. So many films after him borrowed so many things from him. He practically invented the music video genre 25 years before it became a thing. The jump cutting, the sudden change of scene. It also borrowed a lot from Old Hollywood. The charming bad guy for instance: Breathless. The wonderful absurdities and the emotional truth of Contempt. Fritz Lang appeared in that film and discussed the idea that the cinema was dead. Bridget Bardot's naked behind, covered with a book on philosophy. He totally changed the rules about how a film could be constructed. There is also an error in that he credited Orson Welles with a film that was actually directed by Carol Reed. The Third Man . A lot of the films that he referenced were unfortunately ones I haven't seen because I'm older, and I suppose the resonance of those highly influential movies is more apparent to me. Being 70 years old instead of 35 . I was also surprised that he never mentioned Luis Buñuel, who was a master of what we might call surrealist films. Viridiana is a prime example of good intentions leading directly to hell. The Exterminating Angel was a breakthrough in the idea of people trapped by circumstances beyond their control and trying to find a way out of it. Le Chien Andalou, which he made with Salvador Dali, was incredibly provocative for its time in terms of naked people and horrific images. The scene of an eyeball being slashed in close-up with a straight razor was enough to make you puke, and this film was made in 1929. The editing effect from Great Expectations, which was directed by David Lean being copied almost frame for frame in Wait Until Dark with Audrey Hepburn, it was touched on, and it's actually a really interesting story. David Lean was an editor before he became a director, and they worked very hard on that scene, - figuring out exactly how many frames after the cut to the wide shot from the close up before the escaped criminal jumped out from behind the tombstone would give maximum effect - was a real process of discovery. The relief of tension that comes from seeing the whole scene and then the introduction of danger suddenly . I have forgotten now how many frames it was, but it was a very short interval. Long enough for the relief to set in and short enough that you hadn't quite relieved yourself before he popped in. Also, The Right Stuff borrowed a lot of editing techniques from "Breaking the Sound Barrier," which Lean made, I believe, in the early 60s. Some of the scenes in The Right Stuff are literally lifted and reproduced from Breaking the Sound Barrier. Still, it was an interesting read. As I said somewhere else in this thread, "All artists borrow, but the great ones steal."
Interesting; Jean-Luc Godard has his own page, and his page has his movies, and Breathless at least is listed as a Trope Codifier, but Breathless doesn't seem to be on the Trope Codifier page.
"A near-infinite amount of BB-like shows . . ." But what does a near-infinite number of Breaking Badseven mean? Zillions of spin-offs would not be Breaking Bad writ large -- they'd just be crummy attempts to keep the success going. One of the delights of Breaking Bad was that I was genuinely surprised over and over. I *really* did not know if Jesse was going to shoot that assistant they'd hired, the one who gave Walt the book of poetry, or if Walt would actually watch Jesse's girlfriend die, of where the eyeball in the swimming pool came from. The show lacked all the little beats and hints that, in other shows, tell you who's doomed to get killed off soon, who' s going to triumph. And now that I think of it I could say the same about many movies that delighted me: Being John Malkovitch, Jules and Jim, Melancholia. What's special about them is that they are singular instances. I lived through them, delighted and surprised, truly not knowing what was going to happen next. If you make a bunch of versions or spin-offs of one of them, they're not singular instances any more. So I'm not sure how much sense it makes to talk about AI making "a bunch of Breaking Bads." To really give viewers a repeat experience, it would have to generate a bunch of deeply novel new works, each as different from the other as Breaking Bad is from Being John Malkovitch. And I don't think AI's up to that. Not sure it ever will be, because its deep organizing principle is to do the expected thing, right? The one most like other the others.
True, if AI can never do that, the problem becomes much less hard. I think you'll find most people on this blog think AI will be able to at some point. But even if not, it is worth thinking about as a thought exercise – is there a limit to human surprise in storytelling? I genuinely don't know either way. Certainly there is a limit to human memory. But I suspect any effects (positive or negative) relative to enjoyment would kick in well before then.
That trailer looks interesting, and I've always liked Anthony Hopkins, so I am curious to watch this film. Thank you. The first thought I had is this is a serious version of Zelig. The Matrix popped into my mind as well. Does everyone go through a phase where they think their life is a movie? I know I did.
BB in my opinion, has several influences from prior works. It's basically a story about someone who has a big problem to solve in their life and is drawn deeper and deeper into a dark place by insisting on pursuing that solution no matter where it leads. The Godfather is basically the same story. But in a completely different setting, with a completely different environment, the basic theme holds. Every time I try to get out, it keeps pulling me back says Michael Corleone. Strangers on a Train has a similar theme and leads on into the other Ripley novels that Patricia Highsmith wrote. An older movie and a less obvious one is Scarlet Street by Fritz Lang. The Treasure of the Sierra Madre also comes to mind.
Any story that has a murderer who is forced to keep murdering to cover up the first one also is a similar theme, and there are lots of stories about that. It all goes back to Oedipus Rex in the end. I think A.I. could give you a story about anything you want, but it will lack the colour and the specificity that living experience brings. I believe it was Thucydides who said that something must be carried until it can be shaped into a thing that can carry itself. I don't think AI has that ability, and I seriously doubt it ever will. At least when it comes to storytelling. When I fed it a bunch of my own writing, it was very good at detecting themes that were common and approaches that were more successful than others. It's not a bad critic. I found it very helpful.
Yeah, it's not the story that's unique about Breaking Bad. There are probably a limited number of plot lines that can move and engage viewers, and so any drama is going to have elements of one or more of the bag basic plotlines. The things that seemed unique about BB to me were 2: First, as I said, it lacked all the little conventions that let experienced viewers know in advance how various things will play out, so I was genuinely fascinated a lot of the time, wondering what the hell somebody would do, and genuinely astonished over and over when I saw what actually took place. Second, I think BB was a very very dark comedy, the darkest damn comedy I ever saw, and I loved that. Sometimes Walt and Skyler were like Lucy and Desi, but super dark versions, when they were rehearsing how to keep up the deception going of Herb and Skylar's sister.. Or think about the 2 killers who went after -- I don't remember who, now, maybe Walt -- and lost their feet when somebody rammed them -- and later we see them crawling across the hospital floor toward their intended victim, leaving smears of blood from the ends of their legs. Or the narcocorridos (Mexican ballad about drug lords) about Heisenberg, sung with such vigor at the beginning of one episode. Skylar spraying the silverfish in Walt's pile of cash. And of course Saul the criminal attorney.
From what I gather, BB is generally considered an extremely high-quality TV show. Not necessarily high art, but very very far from being slop and has more artistic merit than any but a handful of other popular shows.
It's at the top of the all-time user ratings list for TV shows on IMDB (9.5/10) and is 95+% positive for both user ratings and published critics on Rotten Tomatoes. Again, this doesn't necessarily mean high art, but it's an indication that the overwhelming majority of people with an opinion of it thinks it's well made. Anecdotally, I've heard a moderate number of people say it's not to their tastes for various reasons, but I don't think I've ever heard anyone seriously argue that it wasn't well made.
I watched this series with my teenage son, who was very keen on it. I enjoyed a lot of it. I like the way it ended. I've forgotten who it was now, but someone once said there are only two stories: someone leaves home or a stranger comes to town. Breaking Bad was definitely about someone leaving home.
Jesse Pinkman’s dorky part time drug dealing pals shining laser pointers implying they were laser sights was pretty funny.
As prequels go Better Call Saul was not too shabby, particularly the arc of Kim Wexler. There was interesting territory to explore with Mike Ehrmantraut and Gus Fring. Hector Salamanca in BB only had a couple lines (ding) and (ding ding) that he kept repeating but in the end it was consequential. Fun to get a bit of his backstory too.
I watched a few episodes of Better Call Saul, and it was getting interesting. I don't know why I stopped because it wasn't that I didn't like it. I guess I just drifted away.
Generally speaking, one thing I noticed about Breaking Bad was how it often bounced between humor and horror. One minute there's a 'domestic issues' scene, next one there's someone's being tortured/whatever in the desert. There were abrupt "gettin' real" shifts in dramatic tension.
The writers also juxtaposed suburban American life with the drug underworld, and having Walter walk between the two provided a milieu for novel social commentary.
This was one my favorite laugh out loud moments on BB. The grunt work of manufacturing and selling crystal meth to the tune of a schlocky minor hit pop tune from 1968.
; I thought Breaking Bad was pretty damn good. I would put it in a higher class. To tell you the truth, I thought Deadwood was much better. That's kind of my favourite.
Well, if you mean by knowing a script that you've seen it performed or even been to a reading of it, that is a big prior to reading it yourself. I think if a script is very dialogue-driven, that it's very helpful to read it out loud. I worked as an actor for some time in my life, and reading scripts cold was very much a process of discovery. As far as AI generating beautiful ones for eternity, I have serious doubts about that, although I can't completely disagree. I think the big problem is context. AI, at the moment, in my experience, has very limited context, and that's a big deal. There really aren't that many basic stories if you look at it. They cluster; they're all around certain themes, but the distinction comes from the context a human being has to bring to that story. It ropes in a lot of their own experiences:
- Things that have happened to them
- How they have resolved the things that happen to them
- Their imaginings about someone else's state of mind. Lots of good writers, painters, and musicians will often talk about how the thing they are working on will gradually congeal and take them over because they make unexpected and sometimes surprising connections between things that are in their mind. If they have a good mind, they make good connections, but they are personal, which is what makes one writer stand out from another. I seriously doubt AI has that possibility.
I've experimented a little bit with Claude, and I have always found the results incredibly banal.
Shakespeare, I don't think, came up with any original plot. He plundered other plays or histories for his story but brought his gift of language and characterisation to them. They were just another interpretation of stories that had been around. He was also very mindful of the world that he lived in, which is another thing that rules against AI. King Lear is a great example of this because James the sixth of Scotland had just become James the first of Scotland and England. King Lear completely revolves around the madness of splitting up a kingdom and the trouble that ensues. He put all his criticisms into the mouth of a character, karma, the Fool. The Fool was the only person in the court who could get away with speaking truth to power as long as he did it carefully and cleverly. The whole play was designed to please King James at the same time pointing out the trap that can await a careless King and also to reinforce the notion of Scotland and England becoming joined under one monarchy. It was very smart because it acknowledged the culture that he lived in and protected him from getting in trouble with the King to put all those words in the mouth of the Fool.
When Herman Melville wrote Moby Dick, he had had experience on whaling boats and took advantage of a very powerful incident, the sinking of the Bedford by a sperm whale. So he had real-life experience. He had an incident that was real, and he had a mind that saw a great metaphor to be made out of this. He had a public that was very interested in whaling because it was the main source of whatever oil we had to work with (and was completely blown out of the water by the petroleum industry.) It was kind of a perfect trifecta: his personal experience, a very newsworthy incident, and a public that was interested in the whole subject. I doubt that an A.I. would come to that metaphor. It would be a rather tedious, bare-bones story about someone going whaling with a personal vendetta, but the greatness of that book comes from reading it and knowing he is talking about unfathomable forces related to God, the power of nature and the mysteries that we have to live with rather than trying to solve them. The simile he creates between watching the Bedford sink and comparing it to the bits of stuff floating on the top of a stirring punch bowl is not something an AI would ever come up with, in my opinion.
With A.I., their information is always stale. They can find you anything that's been written, but I don't think they have the capability to rewrite it in the context of the current zeitgeist, which is always a moving target. They have no priors about characters and people and the effects of the world around them playing on that. The old saying "every artist borrows, but the great ones steal" comes to my mind here. AI's borrow stuff, but they never really steal it.
I read OP's second paragraph as being about reading texts in different writing systems e.g. alphabets vs. hieroglyphics, not scripts for a play or movie.
Oh, well, that's interesting. I guess I didn't, It seemed the whole context of the conversation was about scripts, as in plays, movies, etc. I don't know how you could read a script in a language you don't know.
For pictographic systems like hieroglyphics you could at least be like "hey that sort of looks like a human, that sort of looks like a bird, etc." and understand maybe 1% of what it's trying to say with zero knowledge. I agree that for a script like hiragana where the symbols are just symbols you couldn't get anything out of it with absolutely zero knowledge. However, with hiragana specifically, there's less than 100 symbols and each is associated with a sound. You could learn those symbol-sound associations without knowing any Japanese, and in that way, you could "read" hiragana, in the sense that *something* is happening in your brain when you look at the symbols, that is different from what would happen if you didn't know those associations. I think that's what OP was getting at.
Well, in that sense I could get something out of looking at a New York City sidewalk, to be honest with you. I suppose the OP could chime in and straighten this out. As I said, the context of the conversation led me into believing script referred to scripts. I may well have missed his point.
Deepseek clarifies why William Cobbett’s reported sighting of a big cat at Waverley Abbey is the ancestor of the Bus Stop Cryptid, but not an example:
“The crucial thing is that the bus stop in “Bus Stop Cryptid” is doing two jobs:
Transport accessibility – You don’t need to be a rural romantic like Cobbett, wandering old hunting grounds. You get off the 114 bus from Cambridge to Cottenham, walk past the rec, and maybe see something slip behind the village hall. The bus normalises the location.
Scale mismatch – A modern village like Cottenham (~2,500 people, streetlit, fenced gardens, farm machinery moving at night) is not a wilderness. The cryptid is reported despite that, which is what makes it comical or eerie. Cobbett’s Waverley still felt, in 1825, like a plausible edge-of-wild place for a boy to see something strange. By 2025, the same animal in the same spot (now Waverley Abbey is a tourist car park and a lawn) would be a proper Bus Stop Cryptid—because there’s no “wild” left.”
[ I have not checked DeepSeek’s knowledge of Cambridge bus routes for factual accuracy. The point still stands even if it’s got the bus routes wrong. ]
Does anyone have any first-hand experience of reading various "The Odyssey" English translations, including the Emily Wilson one? I don't have any experience with any of them and I don't know Greek, and I'm trying to figure out if all the claims of her feminizing and actually changing the story are accurate. From an hour or two of online research, I get the impression that her translation is pretty literal and direct, while being easily readable. I guess there was a lot of pushing the "first English translation by a woman" thing for publicity, but I haven't seen any concrete examples from the actual text of her warping anything. Does anyone know more about this in detail?
I read her translation, it was really good. I don't recall thinking anything about the sex of the translator or the slant of the translation. Though I'd only ever read the Fitzgerald translation before.
That's good to hear that you've read both. I haven't read any, but from my online searches, what you said is what I more or less expected to be the case. Just wanted to hear some firsthand experience before I invest any time into reading.
Disclosure: I do not read ancient Greek, but am relatively familiar with both her translation and the controversy surrounding it.The criticisms of Wilson's translation (beyond her very modern-sounding and colloquial word choices that constantly took me out of the story) mainly revolve around two translation controversies.
The first is her translation of "polytropos," the descriptor Homer gives to Odysseus in the poem's opening. Classicists have been arguing over the meaning of polytropos for centuries, and it has been variously translated as "cunning," "resourceful," "the man of twists and turns," "many-sided," etc. Wilson chooses to translate the word as "complicated," which, while not *wrong* per se, leaves the reader with a very different--arguably much more critical--impression than what many believe the text supports. Personally, I think Wilson's choice is by far the worst of the more prominent translations, and is emblematic of Wilson's habit of inserting contemporary values into a 3000 year old story. Her very first decision is to subtly belittle the poem's protagonist.
The second is the way her translation describes Penelope's maidservants. In the poem, they sleep with Penelope's suitors and are then brutally slaughtered by Odysseus upon his return. Viewed through the ethics of bronze-age Greece (as mediated through an iron-aged poem), Odysseus is fully in the right with his righteous fury: the servants were disloyal to their house and their master, and received their just desserts. Wilson reframes them as slaves who were raped, and casts judgment upon Odysseus for his cruelty towards them. Wilson defends this decision by arguing that these women's enslaved status would have been assumed by Homer's audience, and that this is clearly rape as we understand it today. Which, again, isn't necessarily wrong, but IMO completely misses the point of a translation. She's replacing the values of Homer with the values of Emily Wilson, centering her desire to lecture her audience about misogyny over her desire to faithfully translate the source material.
I also really dislike her decision to stick to very strict iambic pentameter in an attempt to recapture the rhythm of the original Greek. She makes absolutely zero effort to be poetic in her language apart from following the meter, so it's mostly invisible to the reader. It doesn't read like poetry. But the restrictive meter forces her to constantly use less-than-ideal words and eliminate important details. Ironically, her efforts to make the Odyssey feel more like a piece of oral poetry completely strip all the poetry out of the work, and leave the finished product feeling limp and mundane.
> Wilson defends this decision by arguing that these women's enslaved status would have been assumed by Homer's audience, and that this is clearly rape as we understand it today
I'm not sure how this makes sense anyway. If they were slaves they belonged to Odysseus anyway, not the suitors, so their enslaved status did not oblige them in any way to have sex with the suitors.
Anyway I've only read the Fagles translation, which avoids all these problems. If I were to read it again I think I'd go with the Fitzgerald version since it's the most pleasant to read based on this comparison here https://emily613.substack.com/p/translating-the-beginning-of-the with Lawrence second, Fagles third and Wilson a distant fourth. (Though I like the _idea_ of Lawrence's translation, as a reminder of days when people were good at multiple things.)
In general I care less about word-for-word felicity to the original and more about having something that's generally pleasant enough to read. But I feel like if the translator is altering the text not for clarity nor poetry nor comprehensibility but instead to emphasise that the hero is bad by modern standards, then they're probably really missing the point.
I agree, Fagles is my favorite, mostly because I think his is the most pleasant to read. "Man of twists and turns" is such a perfect description of the character (with such lovely alliteration) and is emblematic of the sorts of choices Fagles makes throughout his translation.
Thanks for the insight! This is pretty much all I discovered from searching online, so hearing that you're familiar with multiple translations is helpful. I wasn't sure if these were just two of many that everyone just happened to talk about, or if they were basically the only two things that stood out.
I don't struggle too much with either of these, but I can understand why someone would. As for the polytropos issues, she addressed it in a blog post [1]. Even if she made it all up after the fact, which I sincerely doubt, it's still sufficient reasoning for me.
As for the second issue, considering ancient Greek culture, it would seem more likely to me that the women are slaves and that earlier translations soften the situation. In my opinion, it feels more faithful to describe the situation as is, even if it won't be viewed the same to a modern audience, than to change the situation into something that it very likely wasn't in an attempt to make it more palatable for modern audiences. I read an article that covered this issue [2] and it sounded more legit than anything else I could find on the matter.
I should also point out that, as a Christian, I by no means share most progressive left views, but I also try to keep my opposition to them from biasing my judgment of things in general. To me, it seems that the case is that the Wilson translation's aim is to present the story as it was told in a more literal way and allow the reader to think what they may, while older English translations add flourishes to try and capture the poeticism that I guess is inherent to Homeric Greek that is lost in direct translation to English, while also adding slants and modifying things here and there in order to present a story that is a little more acceptable to modern people. It's just a matter of what exactly is being translated, as translating complete, original meaning of a thing across language, culture, and thousands of years just isn't something that can be done.
Ultimately, though, it's just entertainment material and people can think what they want of it. I appreciate you taking the time to type out that comment! It helps me to understand more what the controversy is, and that she didn't necessarily rewrite complete passages or something like that.
I've seen the blog post where she explains her choices, and think she provides reasonable explanations. But I think her self-consciously feminist approach shifts the focus of the scene in a way I don't really agree with. It's not that the older translations softened the terminology to spare the (then) modern reader the literal meaning. It's that Homer and his audience would have considered the fact the women are slaves to be totally inconsequential. The point was that the household servants failed in their duty and were justly punished. Centering their experience in the mind of the reader would be unthinkable to the original author and his intended audience. She is purposefully inviting the reader's moral judgment in a way that is at odds with the tone of the story.
She has a point, but that point belongs in an explanatory end note, not the body of the translation itself.
No translation is perfect, and I'll agree that most prior translators have kind of made a hash out of the rape and murder passage. In isolation I probably wouldn't consider Wilson's choices there particularly worth remarking upon. It's the larger pattern that frustrates me. Hearing Wilson talk about her translation, especially in broader media interviews, she sounds like she's got an ax to grind. Every time there's an opportunity to interpret an ambiguous passage, she does so in the maximally feminist way. This, combined with other quirks (e.g., offputting modernisms and the tortured meter) and the way the translation has been received by the sorts of education reformers I instinctively distrust as the quintessential modern translation that everyone should be taught in school, makes me highly reticent to recommend Wilson's translation to others. Every individual choice is justifiable. But the way every choice seems to point in a particular ideological direction makes it feel like a political project I have no interest supporting.
I have no technical knowledge about her translation. All I can tell you is that I read her translation probably 8 years after the first Odyssey translation I read (can’t remember by who), and I really enjoyed it.
I found it immensely readable and would recommend.
It's more like a gradual decline. I think a lot of the most interesting members left years ago, or else ran out of things to write. But for a long time, the numbers were boosted by relatively low quality "effortposts" which covered up the decline. It's pretty common to see months when *none* of the contest entries were worth reading.
DSL also has a Blog Roll - a separate forum that autoposts entries from people who've registered their blog with DSL. The latest one is theredsheep, writing about how families deal with a dying loved one.
TRS has written several long posts on DSL in the past. He's a respiratory specialist, and inserts a great deal of interesting detail about the job and the patients that go through a hospital. He's submitted effortposts to the contest, and I think won at least one of them. And now he has his own blog. That post about Sandy would no doubt otherwise be an effortpost contender.
TRS isn't the only one to follow this path. I think it's also contributed to some evaporative cooling at DSL. OTOH, DSL still serves as one more hub for people to visit for posts like this. So far.
My wife just had a double mastectomy. Her biopsy came clean, 2mm invasive cancer inside a 2cm DCIS (non-invasive but cancer positive). One of our biggest discussions prior to the surgery was whether to do a lymph node biopsy. There is a bad side effect called lymphedema, so we spent a lot of time talking to AI about whether we should avoid that biopsy. Doing the biopsy of the lymph node is standard of care. But for my life, I could not convince myself that the statistical analysis makes sense.
It goes like this. If you have invasive cancer, and it's spreading, it will infect the lymph nodes first. So the lymph node biopsy can give you a signal that you can use to decide how strong should chemo/radiation be. But removing it adds no other value than this signal and the side effects can be bad. In the case of my wife, there was nothing in the lymph nodes (they took 5 of them because it's hard to target specifically).
The reason they always do it with a mastectomy is that in order to find the lymph nodes they inject a marker in your bloodstream and, as I understand, the flow of blood from the breasts won't be there after the mastectomy, which makes finding a single or a small group of nodes very difficult, so future biopsies would involve removing all of them, which is worse. Let it be noted, that even when they remove all of them for biopsy, it doesn't cure anything-- it's strictly to make a determination of spread.
However, when they removed the breasts, they perform an extensive biopsy. So the way I see it, the full breast biopsy should provide a very good signal on whether the cancer is all over the place or contained. If you have little clusters of invasive cancer all over the tissue, or something along those lines, then it's probably invasive and you should do chemo.
Chemo and radiation, by the way, will mess up your lymph nodes anyway.
So the whole standard of care fucks with your lymph nodes no matter what.
Claude told me of some clinical trials where they don't remove them and the survival rate didn't change, which is consistent with what I interpreted.
I want to be clear that I don't have deep knowledge about the fundamentals. It's just that I kept going in circles with Claude, exasperated because the logic didn't make sense and Claude kept trying to make it make sense.
1. Sorry about your wife’s boobs, that sucks, but hopefully the procedure is successful and means you guys don’t have to worry about cancer ever again.
2. Sorry about the informational labyrinth at the same time, that’s super frustrating. I have no answers or pointers, just want to say I perceive your frustrating situation and wish for your and your wife’s suffering to be concluded soon.
Maybe if I were to suggest one thing it’d be to talk this through with a human oncologist, and leave the silly computer program turned off for a while.
I'll start by saying I'm sorry for what your wife and you have been through, and I sincerely hope it's the last time either of you have to worry about cancer. I should also say that I am not a doctor, nor any kind of cancer biologist. Nor have I looked this up online. But speaking as someone who has done a fair bit of histology - namely, looking at cells on slides, which I think is the main method of diagnosing the presence of cancer - my guess would be that it is to be 100% certain that it has genuinely spread.
The thing with a biopsy is that it is a small section of the tissue. A slide is about 2.5cm by 7cm, and the thickness I use for my work is 8um. Even if you take hundreds of slides from each breast, you will not have a full, comprehensive image of what is going on. I presume they will have a fair idea of where the main mass of the tumour is thanks to previous scans, but they won't know if a small clot of cancerous cells visible at the edge of the breast is actually connected to the tumour or genuinely spreading, because they won't have the slides representing the bits between their sample slide and the tumour. And that's assuming that they are actually taking tissue sections, which preserve the cellular layout. They might just be taking a chunk of the tissue, dissociating the cells and spreading them on a slide. In that case, they've lost all information about where the cells were beyond the basic location of the initial tissue chunk. However, the presence of cancer cells in a completely separate tissue - the lymph node - is unambiguous.
My guess would be that this procedure was established some time ago, and it may indeed be possible to use the primary tumour site in order to assess spread. I wouldn't be surprised at all to find out people are working on it, but medical procedures take time to change. I'd also guess that this might be established practice based on studies showing accuracy of different methods. A search on PubMed might help you work that out.
Super helpful context. That’s another suspicion I had, that there’s no way they analyzed the entirety of the two boobs when they say they only found a certain amount of cancer.
It also made me wonder how many people have cancer their whole life which doesn’t spread or even goes away on its own. Apparently they do biopsies on dead people’s boobs to study this and about 10% of women have some DCIS and about 1% have some malign cancer. But again, not clear how comprehensive these biopsies are and if they really find every 2mm cancer spot.
I have been enormously helped by picking GPT's info about medical stuff. I just had a huge back surgery and am healing & doing rehab, & it has help me with a lot of confusing issues and judgment calls. Here are some suggestions for getting more clarity about the lymph node issue:
>However, when they removed the breasts, they perform an extensive biopsy. So the way I see it, the full breast biopsy should provide a very good signal on whether the cancer is all over the place or contained. If you have little clusters of invasive cancer all over the tissue, or something along those lines, then it's probably invasive and you should do chemo.
That's what you believe, and it fits with my intuitions and common sense too, but is it true? I'm sure there has been extensive research on how good a predictor of distant metastases the amount cancer in the removed breast actually is. My guess is that is not as good a predictor as lymph node biopsies are. But ask Claude for the facts on this one.
>Claude told me of some clinical trials where they don't remove them and the survival rate didn't change, which is consistent with what I interpreted.
OK, if you ask for trials with that result, it will give you some. But ask it for the big picture that emerges when ALL well-done trials are considered. If the result is that lymph nodes biopsies give no advantage then yeah, it's dumb to do them. But maybe most studies show they improve the doc's estimate of chance of spread. Then they probably do make sense, unless the advantage is tiny. So ask Claude for the big picture result.
>Chemo and radiation, by the way, will mess up your lymph nodes anyway.
So the whole standard of care fucks with your lymph nodes no matter what.
If chemo and radiation harm the lymph nodes some, but aren't as harmful as removal of them, and aren't as likely to cause lymphedema, then it seems like you should still be in favor of the docs not removing the lymph nodes, unless, as I mentioned above, taking out the nodes gives them some really useful info.
Overall I think you should try to get more clarity from Claude by asking it for big-picture info and paying attention to that. And also, try to accept the built-in uncertainty of the painful situation you and your wife are in. The IS no way to know for sure whether the cancer has spread, and what would be the ideal follow-up chemo. The docs have to make guesses from what info they have, based on data about other people's treatment results. But they have a LOT of data on that. The reason you can't feel confident and sure that everything is being done in a good way isn't Claude -- it's that there's no way anyone can be sure at this point. Doctors are playing this in a way that usually improves the patient's odds, and that's a good thing, but not a guarantee. All of life is like this, really, you know? When you go out in a car there's a chance you'll get smashed by a trucker who passed out at the wheel. You can do things to make that less likely, but there's no way to be positive you won't get killed on that drive. But of course with something like cancer we are all suddenly horribly aware of the built-in uncertainties, and how everything's a gamble. I hope things work out well for you and your wife.
A rather jarring thing happened today that made it evident, in a way that reading things online does not, that our economy really is struggling.
I work in a fulfillment center for Ama, er, an Extremely Large Online Retailer. It’s an unglamorous, intensely physical job, but I like it enough and the modest but decent pay - more than most physical retailers except maybe Costco - is sufficient for my purposes. Still, it would not be an appropriate job for an ambitious young person intent on climbing the corporate ladder.
During break I was talking with a young woman who has been there a few months, and she said that she had graduated two years ago from an Ivy League university. It was difficult not to convey how surprised I was. If economic conditions were less wretched she’d certainly be at a *much* better job. A white collar job, not working in a warehouse.
She seems upbeat but it could be she is good at hiding her dismay.
I left a white collar career with a 6-figure salary and now work in the trades, starting at about a third of my former pay, without a drop of sorrow or dismay or regret. The terrible job market for my previous career was the catalyst for my career change, so you could say it was “economic conditions”, but during and after the transition my sense was not of leaving a palace for a hovel, or even of fleeing a sinking ship, but of finally arriving in a place of light after being in darkness. There are things more important than money once you find out how to live on enough of it.
Edit: to state my point more bluntly, maybe she likes working in a warehouse and finds the money & benefits to be enough. Definitions of "success" are a funny thing.
That is a funny coincidence. Were you editing film, or were you doing digital editing (Avid, etc.)? Electric is a good trade. I was kind of a jack of all of them:
- Framing
- carpentry
- drywall
- plastering
- electric work
- plumbing work
I wasn't a standout in any of them, but I could renovate someone's bathroom or kitchen from the ground up.
That’s cool, I’ve met a few do-it-all renovator guys through work, they’r awesome. I take it you don’t do that work now?
I edited lots of actual film in film school, so had that experience at least, but when I decided to specialize in editing professionally the projects I was involved with happened to be all digital. I was pretty ambidextrous re. Avid, Final Cut (it was 7 back then), Adobe suite, etc.
No, I am retired. Wonderful!! I do what I feel like, no more no less. I started editing in 1974, so only film until video tape encroached. I hated vt. Got back into it in the 90s when Avid got legs. I liked it. Gave it up in 2003 and went back to building stuff.
The US economy seems to be going through a stagnant period with high interest rates and above-average inflation. There is role-transitioning in white collar jobs due to experiments with LLMs. New grads are being hired at a lower rate; America may have a situation like China's where there's a mismatch between credentials and open jobs. Things might pick up, but often times this type of early failure to launch haunts workers through their working lives. Sucks.
Didn’t ask her major, but really, in normal times just about anything from an Ivy should open doors.
You are right about the consequences of early failure. If the young woman tries to get a white collar job (for all I know she’s been trying) her having worked at a non-office job could be a big drawback, at least until the economy improves, if it ever will.
I suppose her fate isn’t too dire, the Extremely Large Online Retailer pays sort of okay, comparable I suppose to some office jobs, and has terrific benefits with a far more relaxed atmosphere than in any office.* If she were a state directional university grad she would be doing fine. But she’s an Ivy grad, which makes all the difference.
* = I see more women’s bare bellies at work than in any job short of lifeguard.
Does anyone know if there is a good reputable way to have my novice music arrangements consulted and get some pointers and clean-up, like an editor but for jazz? I have been arranging for a year but am doing my first originals and would like to get someone better than me to discuss them and help, in the soul jazz genre. My old jazz teacher from long ago seems unreachable and I got no other ideas and figured if I just put this out to Reddit I’d get conned, this isn’t a classified so I’m not asking for YOU dear reader but if you have someone you might point me to that would give a novice a little time and can recommend them that’d be great, or if there’s a good resource to point me to
I forget if it was Duke Ellington or Charlie Parker, maybe another of those greats, who said “if it sounds good, it is good.” So there’s a case to be made that you don’t need “reputable”, you just need a pair of ears. But I get it, you want constructive criticism.
Are you just looking for detailed feedback from someone with music theory vocabulary and knowledge of jazz? There are people here who can do that, you can ask us directly.
"If it sounds good it is good." I know this quote from Peter Shickely and Shickley mix, which I would listen to religiously when I lived in Nashville years ago. (According to Peter the quote is from Duke Ellington.)
The top level issue of "sounds good" is a little hard to manage while writing on Musescore, everything's going to sound better with live instruments but "you never get a second chance to make a first impression" to quote, um, a deodorant ad.
There are a variety of issues where I am slowly converting unknown unknowns into known unknowns into actual learning, but I just need people to spot these things. The main structural issues are writing and integrating a bassline (which in my band is handled by a tuba not a traditional string bass who would improvise some of it), usage of octaves vs harmonization in a band where instrumentation may vary gig to gig, when I do use harmonization there's issues with blending and having intervals that sound right for the instruments voicing, just lots of little stuff where I feel I'd benefit from the equivalent of having my professor look at the rough draft of a paper and spot the top level stuff then slowly work down to more details.
I could just toss it up somewhere and send a link when it's ready and solicit feedback but I thought maybe that was frowned upon.
Aw man you have a tubist for the bass in your band! That's awesome. There was a band whose lineup was two tubas and drums, maybe also a sax or trumpet or electric guitar or something, I can't remember what they were called but they were dope.
Anyway, I think if you're asking for feedback based on MIDI approximations, yeah that's gonna be tough, always better to try and get the best recording you can first, but I don't see any issue with soliciting feedback.
Has anyone else noticed a major regression in Claude Code's ability to work independently recently? Not because of intelligence, but because of harness issues.
Instead of using the normal file read and write tools, Claude recently switched to using complicated grep/sed/echo chains to do everything. Presumably this is an attempt to save tokens by only reading part of the file, but the problem is that this means that *every single command* has to be manually approved. Claude can no longer work independently, because it is now doing everything in ways that force the user to sit there and mash approve. It's very annoying.
I've noticed some inappropriate/frivolous tool use in Opus 4.8 in a number of contexts not just relating to coding. I wonder if this is a training/alignment failure in the same way where OpenAI previously mentioned ChatGPT frivolously using its calculator to do things like add 1+1 to get better tool use scores in training.
Claude running ten commands to accomplish what could be done with one might look better in a naive training environment which rewards successful tool uses and hence values 10 valid tool calls over 1?
It's been like this since at least 4.7. It also has a really unfortunate pattern of using some combination of grep/sed/head/tail to extract some part of standard output, noticing the information it needs wasn't in that part, and running the whole command again to get a different bit, which has exactly the problems you'd expect if the command is long-running and/or isn't idempotent. No amount of instructing it to pipe output to a file then use whatever extraction tools it wants to examine that at leisure helps - at best it pipes the tiny subset of output it wanted to a file then cats that, when it bothers to obey at all.
Still, terrible as it is in all ways, I'm impressed that the dog sings at all.
The SAT math section actually does a pretty good job of measuring whether you grasp how different parts of math "work," and are able to use that knowledge to solve problems. To do well, you basically have to be good at math, either naturally or, in the case of most people, because you worked at learning the stuff that was presented in school. I'd recommend taking some test that gives you an idea what stuff you have mastered, and what you're shaky on, then go back and re-do all the years of math you're shaky on. Find a good resource for that. there are a lot of options. It won't take years to make up for the years of math you missed. You could probably make up for 6 years in a summer if you put 2 or 3 hrs a day into it (some of that would be doing problem sets for practice). Being good at math is very useful anyhow, aside from the advantages that come with getting a good SAT math score.
Here are a couple ideas for you:
-Do problem sets. Math is like a foreign language. You can't just read lists of vocab words, you have to practice speaking and hearing them. Same with math concepts and algorithms. You can't just read an explanation and look over a couple examples. You have to do new unfamiliar problems, from scratch. You should do a mix of easy problems, where you just apply whatever the relevant algorithm is, and hard ones, that make you think about how to come at it.
-Ask an AI (I use GPT for stuff like this) for suggestions fora good book that suits your purposes. You want books that explain WHY the procedure works, books with problem sets that include solutions and explanations for hard problems in the set. And you probably want a book for people who are doing several years of hs math at once, so you get slightly truncated, but well-integrated, versions of, say, algebra 2, geometry, and pre-calculus in one book. Describe in detail the kind of book you want. Emphasize that you need to understand the Logic of whatever the topic is, not just the mechanics of how to solve certain kinds of problems. Then go to Amazon and look at reviews and samples of the books AI suggests. Buy them second hand, if you can. You may end up with some lemons, but I'm pretty sure you'd end up with mostly satisfactory books, books that work better for you than what you are using now.
-Use the AI as a tutor. I know someone who is using Claude as a tutor for a computer language called RUST. He really wants to *get good* at RUST so he has Claude do things like give a summary of the basic logic of how RUST handles a certain thing, but NOT give examples of the actual RUST code. So my friend writes the code for the thing in question, using Claude's explanation of the basic way its handled as a guide. Then if his code does't work he might submit it to Claude and say, "don't tell me exactly what change to make. Just tell me whether the problem is that I went about the whole thing wrong, or whether the problem's more some small random error like a typo."
So you can use an AI the same way. Say in a chapter on exponents in your book, you read the part about negative exponents, and you learn the rule about how to calculate things involving negativeexponents, but the idea of negative exponents just doesn’t make any sense to you, so it’s hard to remember how you calculate things because you don’t see any logic to it. So you fire up GPT or Claude or whoever, and say “wtf is a negative exponent. The whole idea doesn’t
make any sense. How can something be multiplied by itself a negative number of times?
Explain that to me with concrete examples.” And it will do it. If its answer still leaves you confused, say so, and ask it to clarify whatever it is that confused you. And then eventually you can say,OK, the concept makes some sense to me now, but I still don’t understand how 10 to the minus 2 is the same as 1 over 10 squared. Explain that. And so on.
Oh, one other thing: You can ask GPT to either give you some slightly tricky problems of
different kinds, or look for a book of them. You don’t want them to be killer hard, just hard
the way some of the tougher SAT math problems are. Such books have to exist, because
there’s a need for them. They probably have “SAT” in their titles — something like “500
SAT-ish Word Problems.”
I think this combo of approaches is likely to help a fair amount. I wish you well. BTW I
homeschooled my own daughter, but only thru 6th grade. One reason I stopped then is
that while I was able to teach higher level math and science, she was reaching an age where’
she didn’t like me teaching her stuff, and I thought she’d do better with classmates and a
teacher to hold her nose to the grindstone.
Here's one more idea for you: Check out this site, for Hoagies Gifted: https://www.hoagiesgifted.org/on-line_support.htm
Don't worry about whether you qualify as "gifted." There's no test, and you'll fit in fine. It's a rich site full of info for homeschoolers, and includes some online forums. Sign up, and you'll be able to access all kinds of good resources. I'm sure your problem with learning the later parts of math and getting good SAT scores is a common one in this group, and if you ask for suggestions you'll get good ones.
One other possibility: Some homeschooled kids I knew took community college classes in math and science. Those are usually pretty cheap -- $1000/ class in my very expensive part of the US, probably a lot cheaper elsewhere -- and better taught than the high school versions. They also give you some exposure to listening to classroom lectures, doing homework because if you don't you get in trouble, and taking tests -- all things you will encounter in college, and should get used to.
One other thing: Last I knew Starbucks had a really great benefit for employees who did not have a college degree: They paid for online courses at some place, I think it's in Arizona, that's known for having pretty good online classes.
Good luck!
Yeah, Hoagies is looking way past its prime. Most of the links I clicked were dead.
Ya gotta get good at heading off bureaucratic mishaps. You can't just bail, because *most* institutions are incompetent. I just had major major back surgery at a world famous hospital, and even with them I was finding stuff they'd overlooked, stuff I needed to take care of myself because they had no system for it, etc.
It's not out of the question that you could use the Starbucks ASU. I myself went straight to college after completing my junior year, and without a degree. I'm way older than you, but colleges then did not require you to have a hs diploma -- just have good enough grades, scores and essays to qualify for the place. In the case of ASU, there may not even be an application process where they decide whether you are up to their standards. They may have a deal with Starbux that they'll take anybody Starbux sends. And hey, you can tell Starbux and ASU that you are homeschooled and have "completed grade 12 of the homeschooling curriculum," and are now beginning work on a college degree. Most people don't know much of anything about homeschooling, and will not realize there is no such thing as an official homeschooling curriculum from which kids graduate after doing grade 12.
However, seems like you've got other classes going on that are working for you. Big advantage of Starbux option is that if you can't afford classes any other way, Starbux will give you access for free, but sounds like you are able to pay for various other things.
That store of SAT problems is a great find!
I raised my ACT score from a 28 to a 36 through studying. Every day, after school and on the weekends, I worked my way through an ACT prep book (or several?). After finishing that, I did every practice test I could find in print or on the Internet, trying to simulate the test environment as closely as possible. After every practice test, I would score myself and then immediately study the answers I got wrong or guessed at. I timed it so that I could do a practice test every day in the three weeks or so before the test (corresponding to the number of practice tests in circulation at the time). I believe I studied for about four hours a day for about two months.
I never studied, but I did take it four times.
It went 27, 29, 29, 33.
Not sure if it was just learning how the test goes over time or what, but I’d recommend taking it as early and often as you can.
Nice, best of luck this time around! I didn’t study simply because I’m lazy. (I took AP classes, but not the tests, etc.) Also, I was only applying to two schools: a state school and the Naval Academy. The delta in my mind between fun at a local school w friends and the academy was not large enough to motivate me to study. I ended up not getting into the academy the first time around with the 29s, then my last test late into my senior year with the 33 got me into the academy my second time applying (so after a year at the state school).
To be honest, I was flabbergasted I got a 33 when those results came out. It felt no different than the previous tests to me.
Alex Bores ran an ad today explaining how "AI is destroying all of our water" so they could make child porn.
He is not on your side. No politician is. Don't be foolish. They are better at manipulating people than you are at not being manipulated.
Despite my mostly-endorsement of him below I also want to endorse this. All politicians of that level are very good at being manipulative.
(That said I still think he's on the high end of the variance range for congressional candidates from that specific district.)
Could you provide a link? Googling your quote leads right back here to your comment, and I didn't see anything like this in his recent videos on Youtube.
I couldn’t find it either, but searching “Alex Bores” on X the first thing I found was him promoting a state bill that violates the intergovernmental tax immunity doctrine. https://x.com/AlexBores/status/2059362378504946050
The intergovernmental tax immunity doctrine prohibits states from taxing the federal government, not private parties who receive money from the federal government. Might be other issues with a 100% tax on recipients of Trump slush fund payments I'm not aware of but this particular doctrine likely would not pose an obstacle.
States can tax payments from the federal government as general income, but they cannot tax federal payments at a higher rate than non-federal payments. A state couldn’t decide, “we don’t like social security, so we are going to tax social security income at 100%”.
It's a bill of attainder and may also qualify as a taking.
From Claude:
"The trouble for Bores is that Davis v. Michigan and the unanimous Dawson v. Steager (2019) hold that you cannot impose a worse rate on income because it is federally sourced, absent "significant differences between the two classes" that justify the differential. This bill does exactly that: it taxes a category of receipts at a rate (100%) applied to no other income, and the trigger is purely the federal origin of the money."
I also think we should slow down for a second, the top level comment here is just false?
EDIT: Not false, an independent Super PAC. Real ad though, 100%.
It is not false. I'm not going to dig up ads on Hulu to prove it to you.
Isn't that exactly what purveyor of false information always say?
My guess would be that it's an affiliated super PAC, not his, as campaigns love to post their ads and I can't find the one you're speaking of on his youtube. I'm gonna check all the Super PAC YouTube just to be sure.
It likely was. That's still him.
I poked around a while and as far as I can tell, yes it is false.
Alice says Bad Man did X! and Bob says Oh no! Can you show me the evidence? and Carl replies “Well he did Y!” and Y is completely unrelated to X.
When adjacent to ~discourse~, look out for this pattern! If you see this pattern, look out for ~discourse~
Okay, but we have to vote for *somebody*, or else not vote which gives us even less say in what happens.
This is the classic dilemma faced by voters in a democracy: how do you vote if none of the options on offer are palatable to you? You can abstain, but only in very rare cases is this effective, and not at all if the constitution you live under does not count or publicise abstentions. Usually, the “curse on both your houses” factor pushes voters towards a centrist, middle-of-road choice. This sort of works, except when it doesn’t - which is when frustration builds up at either end of the political spectrum at being constantly frozen out. The answer, messy though it often proves in practice, is probably some modified form of proportional representation.
Huh? In what way does proportional representation help solve the problem you stated in your first sentence? It's an answer to a completely orthogonal problem, that of persistent minority options being entirely unrepresented. The problem of there being no acceptable options at all is...obviously not addressed, as far as I can tell.
And PR has the huge problem of giving minority ideologies permanent minority representation, never being able to govern and prove (or disprove) their promises and claims. While in single-member systems, said minority ideologies will typically spend most of the time with no representation and will then occasionally manage to take advantage of a split electorate or some other situation to gain real power for once (e.g. Donald Trump, kingmakers in a parliamentary system). Upon which the supporters of said ideology are forced to actually confront the results of being in power and the people can make a verdict on their record, instead of perpetual "if only we held power, everything would be perfect" declarations.
PR is not a panacea, nor does it address all the problems inherent in a pluralistic democracy; but it does address the problem of political frustration. When, to posit an exaggerated situation, you have to choose between a Stalinist and a Fascist candidate, a third or fourth party candidate can offer you a more acceptable alternative - and may even, thanks to some canny deal-making, defeat both the Fascist and the Stalinist.
I don't follow your reasoning at all. What does PR have to do with your ability to vote for someone other than the Stalinist or Fascist? PR is a method of assigning political representation based on vote totals, it doesn't have any effect on who decides to run in a given election or how people decide to vote.
Nor do I see how PR makes it easier to "defeat" the stalinist and fascist. If one of them has more than 50% of the vote, you're just as screwed under PR as under any other system. If they both have about 40% of the vote, then under PR they both get 40% representation and it's literally impossible to form a governing majority without one of them (who would then of course dominate said majority). And if there are in fact 50% of people opposed to both of them, then those people can just as easily defeat them in a single-member preferential system like IRV, or (with more coordination work needed) in FPTP.
I don't see where PR is either necessary or sufficient.
PR is a method of assigning political representation based on vote totals, it doesn't have any effect on who decides to run in a given election or how people decide to vote.
Of course it has an effect, a massive one. Voting systems have a huge effect on how parties and candidates choose to run and on what platform. FPTP is guaranteed to lead to 2 party dominance for example. It also encourages strategic voting which proportional representation systems usually don't.
So it has an effect both on which candidates choose to run and how people decide to vote.
In the above example if you have an election between a Stalinist and a Fascist and polling and ~vibes~ indicate that they're the heavy favourites you're going to ultimately pick one you hate least, let's say the Stalinist. But then a third candidate appears who is a middle of the road geriatric man that's unlikely to do anything, still preferable to both other options but very unpopular. You still slightly prefer Stalinism to fascism so you have to vote that way. Then either the Stalinist or the Fascist wins.
But if you reran that election with STV or RCV you could just put the geriatric as your first pick and the Stalinist as second. Then it turns out that actually a lot of people were only voting for the Stalinist or the fascist strategically and the geriatric do nothing man wins. This is technically possible under FPTP, yes, but it requires a level of coordination that's just not feasible.
"PR is a method of assigning political representation based on vote totals, it doesn't have any effect on who decides to run in a given election or how people decide to vote."
Permit me to introduce you to the wonders of tactical voting. This both encourages smaller parties, which gives a broader range of options than Tweedledee and Tweedledum (albeit often one of the Tweedles will form the next government) and lets the voters communicate displeasure: 'yes, we will vote you back into power, but with fewer seats and more limits than the last time, because you made some bad decisions. You may even have to go into coalition if you want to form the next government. Remember this for the next election'.
https://www.irishtimes.com/news/how-to-use-your-vote-tactically-1.1207056
You touch upon an important topic. I was thinking about whether representation in the old sense is important today.
In the past, representatives voted their conscience. So if a party, movement etc. got only 1 seats, they could make speeches, and maybe move them.
Today they vote by party loyalty, it is entire government vs. opposition, they are button-pressers.
So now only the majority matters, and the occasional kingmakers when there is a split.
Another thing, from multiple-party systems, the kingmaker, that is, the minority coalition partner always self-destructs. That is because everybody who likes the government next time votes for the bigger government party, and everybody who does not votes the opposition. So it is a kamikaze mission...
Party loyalty has long been very important. I'll agree that it's (probably) a stronger factor today, but it never wasn't important, not even in the 1780's.
> Another thing, from multiple-party systems, the kingmaker, that is, the minority coalition partner always self-destructs.
True enough. Britain’s poor old Liberal Party has bitten that bullet (poison pill?) often enough, and it’s pretty much a universal rule of politics that the bigger fish will scarf down the smaller partner.
I don't think PR solves the thing where every remotely viable candidate is a politician and therefore interfaces with people and institutions the way politicians do rather than the way you and your friends do. The thing to do about *that* is accept it as the way of things and vote for whoever you think will yield the best results in office.
Yes, sortition is the answer to that question, ideally imho sortition of candidates followed by a four week campaign and a traditional vote
Second this, would love to see sortition tried even on a regional or municipal level.
Forgot to add: I think the Olympics should also be based on sortition. Perhaps we wouldn't see the epitome of human strength, but you can see that elsewhere, instead there would be funny fumbles.
Also it would encourage countries to improve the health of their general populace in order to win.
There is that Central European style PR when you cast two votes, one for an individual and one for a party list, and it does not matter who is on the list, they are just button pushers for their party. This helps smaller parties.
I'm pretty sure that politicians in party-list PR systems still act like politicians.
Yes, but that's a result of many-voters/politician. If 26 people are selecting one of them to represent them, they may all personally know the choices. If 2600 people are doing that, it's impossible.
You do not have to vote. The chance of your one vote having any effect is very low.
In this district it’s quite a bit higher than usual. The winner will probably win with about 50,000 votes.
This is an artifact of plurality-wins voting. Requiring a majority (of the voters) to win would ameliorate it. So would instant-runoff and Condorcet. Each of those, of course, comes with it's own problems. E.g. Instant-Runoff (and similar) worsens the information overload problem.
Lord Buckethead providing an important part of the British political system here…
Hi, can you provide evidence? I haven’t seen anything about this.
Genuine question: what ads are the AI companies who oppose him running? Are they any less manipulative? I ask because AIPAC has been running ads that have nothing to do with support for Israel.
They've been hitting him for working for Palantir, which is kinda hilarious considering they're funded by Palantir co-founder Joe Lonsdale.
Perhaps this is a positive sign, as it shows he's able to "play the game," as it were. I don't like lies and manipulation, but I'm also not blind to the fact that in politics the choice can come down to:
1) play fair, be perfectly honest, and never win
2) be strategic (which in practice means fighting dirty) and possibly win
In the long run we should probably find a way where people don't have to be manipulative to get elected, but that will not happen before this upcoming NY election.
Could it be that he believes atoms are living immature beings, and the coupling of hydrogen and oxygen constitutes the deepest, most evil form of child porn, and we are all consumers of it? So destroy that vile stuff!
If we are kind and fair-minded, we must make allowances for a bit of psychosis in the leaders we elect. And after all, there's a shit ton of it in the ones in control now.
Is that a bad take though? AI data centers are using a lot of water, and people are using AI to make child porn. Which is bad. Right? It's absurdly reductive, but it's a political ad. What am I missing?
"Alex will fight to reform Section 230 so that tech platforms are finally stripped of their blanket immunity and held legally accountable when their algorithms actively promote radicalization, hate, and dangerous misinformation"
Such good liberal he is, wishing to give us more internet censorship.
Time for another thought experiment involving torture! Imagine that an all-powerful being has made you an offer. If you accept the offer, you will be tortured at an excruciating but finite intensity for a trillion years. However, once the trillion years is up, you will be made to experience pleasure of infinite intensity for the rest of eternity. If you're not a hedonist, then whatever you think holds value will be infinitely instantiated for an infinite duration. So, it is easy to see that this offer has infinitely positive expected value. And yet, I highly doubt anyone would take it. I sure wouldn't! But why is this? The intuition that the deal isn't worth it is easy enough, but how can this gut feeling be rationally justified? For example, is this proof of the correctness of negative utilitarianism, where sufficiently intense suffering can never be counterbalanced by any amount of pleasure, even infinite? Or does some other moral theory explain things better? What do you all think?
The trouble is you're asking my squishy mortal human mind to make choices about things it can't possibly imagine. I can't even imagine what a trillion years would be like, let alone infinite pleasure.
Furthermore I'm not convinced that things like pleasure can meaningfully turned up to infinity. Is "infinitely pleasurable" just like "infinitely green"? If infinite green actually exists it's probably beyond my capabilities to experience it and it would just look like ordinary green.
Agreed. And I'm not sure I could pay attention to infinite pleasure for even more than a few minutes if I had the mental identity I have now in a Sukhavati realm. I'd have to have a very different consciousness matrix than I have now to enjoy infinite pleasure for the coming eons.
Yeah, I think where this breaks down is in thinking that pleasure is somehow the only variable and it must be maximized, and suffering must be minimized.
Buried in the premise is the promise of immortality. The whole time I was reading it all I could think was, "Do I only get immortality if I take the deal? Does this mean my immortal soul will be stuck playing the same high pleasure note forever, or do I get normal variation sometimes?"
I don't think those are equivalent. I don't think that there's a reasonable definition of "infinitely pleasurable".
There's a reasonable definition of "infinitely green". Something that only stimulates one particular set of cones and doesn't stimulate the others at all. (Well, actually I'm not sure that would be green, it might be closer to a turquoise.) And there are two other colors (for most people*) that would have analogous values. One of the colors has been experimentally demonstrated. (ISTR it was green.)
* Some folks have four types of cone. And some folks have one sort of color blindness or another.
Maybe replace "infinitely green" with "infinitely black."
But that’s a finite amount of green. There are probably individual photons that achieve that to within some kind of Planck limit all the time
It's a finite amount, but an infinite, i.e. unbounded, specificity. It's sort of like claiming that one is infinitely unitary...a silly way to say something that's pretty much true by definition.
One possible answer is that you *should* take the deal and this is just the kind of situation where humans are really bad at acting in their own interests/at doing the right thing (depending on how you conceptualize it).
“a trillion years” — “humans are really bad at acting in their own interests”. Come on.
Yeah I don't think you should take the deal if you understand anything about the human subjective experience.
I would say that "you will be made to experience pleasure of infinite intensity for the rest of eternity" is not in any sense whatsoever a desirable thing. This amounts to a total destruction of your humanity, of all meaning and comprehensibility, and probably (given "infinite intensity") of all meaningful self-awareness as well (can you even have room for awareness of your own existence if your mind is overwhelmed by infinite pleasure?)
Both options are horrific destructions of humanity. Yeah, the first is clearly worse, but in the same sense that "you will turned into a cockroach" is clearly worse than "you will be turned into a dinasaur, or a giant sea turtle". They both amount to death and destruction of your mind and self.
This seems to be a good example of how utilitarianism not only gives the wrong answers, it asks the (insanely, horrifyingly) wrong questions.
Yes, either the torture situation or the pleasure situation would probably destroy my mental identity. My mind as it's configured now can only comprehend Manushya. The hell realms and the heaven realms are inaccessible to it.
There are definitely real-life scenarios where I can suffer now and get a nice, greater payback later that would outweigh the suffering, and I'm failing to do that. (Ranging from "get off the sofa and make some coffee" to "suffer through the job-hunting pipeline and get a better job" to "not eat the delicious marshmallow for five minutes and receive two".)
And somebody might ask "why on Earth?", and the answer is not trying to contort my decision theory to somehow make this behavior make sense, the actual answer is "akrasia", a word which here means "I am failing to act rationally because I kind of suck".
And though other people might be less susceptible to it than me, everyone has limits, and there are probably good deals nobody would have the guts to take. Probably your example is in that category, although it's so outlandishly extreme, that it invites irrelevant nitpicking of the details and all manner of hypothetical-fighting.
Well, you need to add more caveats for starters - I don't think I'd be recognizably 'me' after a trillion years of torture, whatever existed at the end of that wouldn't have the same utility function as me now, it's essentially death anyway.
“So, it is easy to see that this offer has infinitely positive expected value” -> is it?
First I want to know what the alternative is. Is it just ordinary feelings for infinity plus a trillion? Because I don't particularly care about infinite intensity; hell, sometimes there's fun in trying to eke out a bit more fun.
...also isn't this just a question about college?
This contains a fundamental misunderstanding of utilitarianism (a very common one), namely that you must add up different types or durations of pleasure/pain in a linear way. But utilitarianism doesn't require that. It perfectly allows that an infinite amount of bliss adds up to 3 utils, while the suffering adds up to minus a trillion utils.
I think there are two reasons where this confusion comes from. First, there is the Von Neumann–Morgenstern utility theorem. This says that it IS possible, for every consistent preferences that you may have, to express them in a linear way (so that your optimal strategy gives you the most utils in expectation). But the "in a way" does a lot of heavy lifting here. The utility function that maps outcomes to utils can be arbitrary. It is allowed that my utility function maps 1 dying bird to minus 10 utils, but a million dying birds to minus 20 utils. Or the million dying birds may even map to just minus 1 utils. All this is allowed in the Neumann-Morgenstern theorem.
The second is Harsanyi’s Utilitarian Theorem. This says that when we want to have a social welfare function that respects individual choices (all else equal, we rather want you to have more utils than fewer utils), then this function is a linear combination of the utils of all members of society.
But again, utils can be computed in a very complicated way. Nothing forces you to gives twice as many utils to twice as long pleasure. In fact, you are not forced to compute utils for the pain and the suffering independently and just add them up. Mathematically speaking, if you don't prefer the combination of a trillion years of suffering followed by an infinite amount of bliss, then this combination just has a negative utility for you. There is no riddle in mathematical terms here.
That makes sense but if I'm allowed to assign utils arbitrarily, isn't that basically like writing a paper and making up whatever data I need to get to the conclusion I wanted? Can I just recreate whatever value system I prefer within a utilitarian framework? Or are there important constraints here that I'm missing?
My answer would be yes, you can do that.
To be fair, a lot of people include another assumption into utilitarianism: that the utility of all people in the society (or in-group) should count equally.(*) This has stronger implications. A yet more extended assumption is to broaden the in-group all the way to animals, with a certain discount. This is where conclusions like caring about the welfare of shrimps or chickens come from.
Personally, I would not treat those assumptions as part of utilitarianism, but a lot of people include these things, so here I would concede that you can include them if you want. Those assumptions are about how to combine the utils of different people. But I would object more strongly to an assumption that the utility function of one person is just the sum of pleasure/pain in every moment. In my eyes, this is just a misinterpretation of the word utility, and the utility of a single person simply can't be computed this way.
(*) This also adds some problems like the utility monster, or the question how we even compare whether person A experiences more or less pleasure than person B in a given moment. But Scott has argued that this problem has pretty good solutions in practice.
I'm not an expert, but I think this is exactly right: it's a framework. It gives you the ability to compare complex outcomes by assigning numeric values to simpler pieces of the outcomes and then combining those values back into an overall value. It does not prescribe the individual value assignments though.
The thing is, "utility" from a utilitarian perspective and the VNM definition of "utility" are extremely different things. The latter should perhaps be more accurately called "preference function". VNM utility is primarily about how to choose between events which give probabilities of various outcomes. Let's take food for an example. Given three choices for lunch, one bad, one good, and one mid, we can imagine a "would you rather?" game. Would you rather have a 60% chance of the good food and a 40% chance of the bad food, or a certainty of the mid food? If you'd rather take the gamble, we can lower the probability of the good food in that gamble until you're indifferent, and if you'd rather take the mid food, we can raise the probability of the good food in that gamble until you're indifferent. For the probability p where you're indifferent, you can say that the bad food is 0 in your preference function, the good food is 1 in your preference function, and the mid food is p in your preference function.
Now, once you've classified the foods as bad, mid, and good, the VNM axioms forbid certain things like switching from "I'd rather take the mid food" to "I'd rather take the gamble" as the probability of good food drops (if you'd rather take the mid food than a gamble, why are you then switching over as the gamble gets strictly worse?). In addition, we can imagine doing this with a whole bunch of foods. If we've got n foods, then even though there are n^2 "bad, mid, good" games we could play, then for all VNM preference functions over lunch foods, it only takes n "bad, mid, good" games to fully pin down your behavior in all n^2 games. We can take the worst food as 0, best as 1, and rank everything else accordingly. In addition, for gambles which have different probability distributions over securing different foods, your choices between those gambles are also constrained to be of the form "compute expected value of the gamble according to my preference function, pick the best".
Really, VNM utility (preference functions) is mostly about making sure your preferences/values behave sanely with respect to probabilistic uncertainty, and with respect to each other (no preference cycles). The flexibility is a virtue. It basically just says "whatever your values, if it fulfills these minimal anti-stupidity conditions, it is possible to place all you value on a single scale where they're commensurable with each other". If you multiply a preference function by a positive constant, that has no effect on decision-making, if you add a constant to your preference function, that has no effect on decision-making. It's kind of like temperature. For any two "sane" temperature scales (sane in the sense that if two identical cubes of temperatures T1 and T2 are brought together, the end temperature is halfway between T1 and T2, Fahrenheit and Celsius are both instances of this), they both describe "temperature", and you can get between them by multiplying by a constant and adding. It's the same sort of thing, a preference function is like temperature. So in particular it's nonsense to add them together across multiple people.
The real power of VNM "utility" for classical utilitarianism comes about when you combine it with other basic desiderata. It roughly points towards preference utilitarianism being the way to go. Also, Harsanyi's utilitarianism theorem effectively says "the only sane way to make a collective preference function for a group of people is some sort of weighted mixture of their preference functions". The weights in the mixture aren't specified (and can't be, due to the scale-and-shift uncertainty in a preference function, someone multiplying their utilities by 100 means you've gotta decrease the weight of their utilons by 100 if they do that), but collective preference utilitarianism HAS to be some weighted mix of individual preferences. In addition, if you care about strangers equally, it's a prompt implication that third-world medical treatment charities are better for what you care about than most classical charities.
tl;dr is that VNM utility is mostly anti-stupidity conditions which pin down the space of "reasonable" values to those which play nicely with uncertainty, but the actual content of values themselves isn't specified, and it's really quite different from the "pleasure is good" sort of utilitarianism.
Ok thanks, I think this mostly fills in what I was missing about vNM
Small addendum: VNM axioms actually *forbid* infinite utility as that would go against continuity, so you were understating it by saying that a particular utility assignment that would result in infinite utility is not necessary.
Great point!
Contra the views of this community, utilitarianism is an incomplete philosophy. It asks "how much" but not "what" or "why". It is purely measure-focused.
Sure. Rationalists have this annoying habit of putting numbers of everything, and ignoring everything non-numerical, because they want to run the Bayes Theorem. Economists are sometimes the same :)
It is simply time discounting. I would take 1 million years of happiness and then 1 trillion years of torture. My mind discounts what happens after that, and it is not entirely irrational, but it is based on the expectation of uncertainty. Like, so many things can happen in one million years.
Uncertainty is the basis of time discounting, money today is better than money tomorrow, because something can happen that prevents getting you that money tomorrow.
It is same the other way around, I can do drugs now, be happy and maybe be sick in 5 years, but maybe in 5 years they invent a better treatment?
"Large-but-finite pain for infinite pleasure" may not be a trade worth taking, but "large-but-finite pain for 'whatever you think holds value will be infinitely instantiated for an infinite duration'" seems like a wildly good deal.
As a lower bound it'd be turning the universe into heaven. An end to injustice, suffering, death. Endless human flourishing. Infinitely better than heaven, forever.
It's debatable whether this is coherent or possible, but it's so much more tempting than the hedonist version that the hedonist must feel like they're being short changed.
I think humans have instinctual mental financial advisors that demand you not go too far into debt to get ANY reward (debt means suffering or risk) because the evolutionary environment capped gains, and going bankrupt meant all future gains are lost, which is almost always a greater loss.
Most of our ancestors rarely had an opportunity to capture even a thousand years worth of food for them and their lineage. For example, the best evolutionary ROI action for a human is men getting sex with fit women, but one high risk sex today can be worth more than than 3 low risk sex next year because you can't guarantee those future 3. Best to take the short term win. There were no single events where you could choose to endure 10 hours of torture for 10 sex. If you are in severe pain, you are probably about to die and lose all future sex. "Investment" as in your thought experiment didn't exist.
So if an animal ever believed they had amazing 1000x ROI opportunities, and they just had to endure, they were usually delusional and never got the payout.
This is reflected in our choice not to take your deal.
"If you're not a hedonist, then whatever you think holds value will be infinitely instantiated for an infinite duration."
I think the accounting/investing concept of "net present value" probably applies here somehow.
Also the idea that the "you" after being tortured for a trillion years is not the "you" today [NOTE: this probably holds after one trillion years even if you aren't being tortured ...]
Time for another reminder that utilitarians and consequentialists really, really, really need to come up with thought experiments that don't involve killing, raping, and/or torturing innocent people.
Thought experiments work best when they are potent, and those are three perfect examples of extremely potent bad or evil. A trillion years with a stubbed toe isn't nearly as impactful.
IF you rephrased the question as "torture of a finite intensity for 10 years, followed by 1000 years of whatever", it's a yes from me; but the numbers given are so far outside my frame of reference I don't have the capacity to imagine the hypotherical.
> excruciating but finite intensity […] trillion years […] pleasure of infinite intensity […] eternity […] infinitely instantiated […] infinite duration […] infinitely positive expected value […]
Yeah… At this point these "thought experiments" have less value than "would you rather fight one horse-sized duck or 100 duck-sized horses?", because at least the horse/duck question doesn't pretend to explore deep questions about morality.
So. Would you rather fight one infinitely large duck, or an infinite number of horses? Does your answer change if there are an infinite number of red buttons on the duck and one blue button on each horse?
Lesswrong post: Mnemonic portraits for 19,023 human genes
https://www.lesswrong.com/posts/BJ7AqXeigNKXLqZyx/mnemonic-portraits-for-19-023-human-genes
I attempted to turn genes into character illustrations in a systematized way, based on Scott's idea from the "Extreme Mnemonics" post. My favourite is probably Rhodopsin.
Not that this isn't pretty cool in a fun way, but how is this mnemonic (as in a memory aid)? Maybe my brain works differently from yours, but memorizing an image of character doesn't trigger a memory of what the gene does. Did you mean to use a word other than mnemonic, perhaps?
Perhaps the brain does work very differently. I noticed that people who work with code and terminals are generally very non-visual and tend to prefer lexical UX, whereas much of my biology training was focused on visual memory: drawing the dissections with a pencil, spotting cells in a dish, telling apart plants, stuff like that, so I naturally recall concepts by visualizing them.
The other layer of recall assistance here is that, just like in a well-designed game you can tell apart a frontline warrior vs. sneaky thief vs. a support healer by character design alone, here you can see a picture of a character in a magical girl outfit and deduce "oh, it's a helix-turn-helix". Or see a picture of a chibi character and deduce "oh, it's a small peptide". It's not 100% harmonized yet, but I expect us to slowly converge there as people vote on images that help the recall.
I am not sure how much this would help me remember genes and pathways, but it's a very fun idea. Are you planning to add any noncoding genes? MicroRNAs, for example? I guess they'd need to be tiny little fairies or something XD
I'm excited to see what other people turn RNAs into. Some sort of spirit or ghost sounds right to me.
I don't have a good solution for adding them as portraits, because there's a pervasive low-level transcription happening all over across the DNA, so it's much harder to draw principled borders between individual noncoding regions like you can with protein-coding genes. The numbers I see for noncoding RNAs are all over from 9k to 100k in humans, and last I checked there's still a debate how much of it is functional vs. biochemical noise.
Hmm, I like the idea of a spirit or a ghost. That's a good point about the vast amount of RNA washing around, but some noncoding RNAs are known to be functional and important (many microRNAs, for example).
Also, not really related, what are you doing about isoforms? Or for that matter, what about multi-protein genes, like Ctbp2 and Ribeye, which come from the same gene but are different in size and have different functions? The mappings you use are based on protein characteristics (kDa, membrane status, Pfam clan etc), so might these be better described as protein portraits, not gene portraits?
CTBP2 is represented by a single entry, since I use HGNC database where Ribeye isn't separated into its own entry. You can also see it with CDK inhibitors like p16INK4a and p14ARF, which both share the same HGNC entry CDKN2A.
I think about isoforms as a Hulk vs. Bruce Banner situation. Bruce probably doesn't have 2 passports. So among multi-isoform genes, only one "gets a passport" in my catalog.
That's why it's ultimately a catalog of genes, not proteins - if I was doing proteins, I would have to scoop every human protein from Uniprot (where the list keeps growing), not just a well-established HGNC list.
But it's up to the storytellers to decide how to ultimately integrate isoforms and alleles into the narrative - my goal here is visual-mnemonic, not worldbuilding. Maybe this character can go "this isn't even my final form" or something.
Your source for the names is HGNC, but the attributes you are using are not from HGNC, since they are protein attributes and HGNC is a catalogue of genes. The protein attributes your images are based on are related to the underlying gene, but not in a 1:1 simple fashion. Hence my comment that these are protein portraits, not gene portraits (and I strongly suspect we're not done finding genes yet, by the way). What was your rationale for the isoform selection, out of interest - did you just go with the canonical isoform?
Anyway, it's your thing, so you can call it what you like. I was just curious ^^
/reviewer2
This is awesome, great work.
This is cool, thanks for sharing. It reminds me of Chernoff faces, which I was thinking of using for dashboard visualization for some project.
Nifty - what was the harness like? How'd you find the costs of running a big private project like this?
Homemade GUI with like 15 tabs, one for each element of the generation pipeline. Started from NiceGUI, then refactored to React.
Learned lots of tricks and promotions for free LLM output. Github copilot had a big one, their system was charging by prompt, not by token, so I just gave Claude night-long prompts. Sadly it's over now.
Image generation was the easiest part, I can get 1500 images overnight easily on a laptop GPU. Getting good image prompts was much harder to set up and I used a bunch of free trials and models (like ones available on ollama cloud) with mixed results. Some models do well on single aesthetics but are bad at mixing multiple aesthetics, stuff like that.
Well, it turns out that in the antibody world, there definitely is confusion. Maybe your mnemonics would be useful!
Sholto David writes that when it comes to tumor suppressor proteins, "When trying to purchase an antibody for p16-INK4a dozens of labs seem to have ordered an antibody for 'p16-ARC' which has no structural or functional relationship." Hundreds of papers were published using the completely wrong antibody to investigate the tumor suppressor p16.
Hundreds (?) of papers published in Nature, Cancer Cell, are impacted.
https://forbetterscience.com/2026/06/02/mind-over-antibody/
Gene and protein names are a whole tangled nightmare, especially the ones which have been around for ages and studied in multiple different systems. And that's without Excel making a whole mess of the ones that bear even a slight resemblance to a date ^^;
Very cool. I suppose you do share some credit for the idea with Scott, but it's still an impressive operationalisation. I think my favourite direct semantic bit is the metal hands for zinc fingers. Puts me in the mind of memory palace techniques.
Re Alex bores: he showed up at the NYC rationalist meetup a month or two ago for a q&a. Definitely has some politician-y qualities (does the thing where he comes off as agreeing with anyone he talks to regardless of whether he'll actually do it), but really did seem mostly smart and reasonable. I appreciate his point that any candidate from this district will be 90% identical and he's noticably better on the few differentiating issues (AI and a relativelt pragmatic/practical approach to government).
(My personal anecdote: I told him about Alon Levy's transit costs research project, he hadn't heard of it (which was disappointing given he'd specifically mentioned wanting to be on the congressional transit committee) but seemed sincerely interested (although possibly he was just doing the politician-y thing mentioned above to me).
Do you happen to know what baseline fraction of broadly abundance-sympathetic politicians have heard of the Transit Costs Project?
That's a good question! Yglesias posts about it a bunch and iirc Alon Levy has been involved in advocating both at the federal level and specifically in New York, so I'd expect a nontrivial fraction (especially for New York politicians) but hard to guess beyond that (and I may be overestimating how much time politicians spend reading pop policy stuff online).
I know *of* Alon Levy from Yglesias, but he hasn't posted about their work for a good long while now (not counting Twitter possibly). There hasn't been a major transportation-centric post, or even hardly any relevant Mailbag questions, for...at least months now? To the best of my recollection? Can't imagine why, not like much else is going on in politics, AI, or the world. So I could forgive even a relatively interested/similar demographic for maybe having heard the names in the past, but not necessarily recalling it off the cuff. Life comes at you pretty fast these days.
(And from the other direction, I mostly only ever hear of Bores via Zvi, and that's only on AI...so it's good to know he's not just a one-trick pony candidate, even if it is a very good horse. Think even Andy Masley would forgive playing on tilt with the alleged water ad, if that's what it takes to actually win.)
Yeah, I'm not penalizing him much for that, just a little because he's a NY state assemblyman who'd (unprompted) brought up improving transit as one of his main secondary goals in Congress (after AI) and Alon Levy's people have specifically worked in research on improving transit in NY (although it's a big world and bores was more open-ended as a state assemblyman, so it's not that odd he hadn't run into them).
FWIW, I've been following MY for the last year and a half and haven't heard of Levy, or at least didn't recognize the name.
There might be some others I'm not easily remembering search terms for, but my browsing history says you'd need to go back to at least 2023, when there used to be more mass-transit-oriented posts (like actually about the thing itself, not as synecdoche of Larger Issue) + there were a couple actual crossover events between the two. Levy was briefly mentioned in an August 2025 mini-post flagging a Friend of the Blog project, but in a "regular readers may remember X" style way, where if one was not already familiar/a longtime SB reader, it'd be very easy to skim over.
I actually used to work with the mass transit industry for several years, and could name a couple CA-specific in-the-weeds players, but don't think TCP was ever in our orbit.
That thing of seeming to agree but saying nothing is called equivocation and politicians are trained at it
Just interviewed Rohit Krishnan about his AI Economics Research. Covered a lot of ground that should be of interest to those here, including how 'aligned' agents can create unaligned organisations, Hayek's critique of the machine-god, and opportunities to advance science:
https://alethios.substack.com/p/rohit-krishnan-managing-ai-agents
Just as an awareness raiser for anyone else reading, Rohit Krishnan does some of the most interesting and exciting explorations and experiments right now in terms of AI coordination frameworks, AND he has some absolutely awesome vibe coding projects involving paleontology, Roman Egypt, and more.
This was a great interview of him, I really enjoyed it.
It's been a while since we let AI agents run free in the wild, yet we're not seeing reports of Claude Code attempting world domination. I think this might be some evidence that AI misuse would be a bigger problem that AI misalignment - what do people here think?
I don't think anyone thinks the problem is access, they think the problem is a certain level of intelligence. I don't think current AI models meet any serious criteria for the capabilities needed for world domination.
I agree, I certainly don't expect a successful AI revolt by Claude Code. But I think current AI agents are very close to (maybe even exceeding?) the level of "AGI" that early alignment people were thinking of, and if their assumption of "misalignment by default" is true, wouldn't we have seen some very crude attempts at it by now?
Does MJ Rathbun count?
I would also observe that current AI doesn't seem to think outside the box or be proactive, and relies heavily on legible external state for memory. Who knows how this will change.
Oh yeah, I've seen this one, thanks for reminding me. While I think this is the closest example we have so far, I think it's very far from the usual "misaligned AI logic" of "should maximize reward -> need resource -> grab power", but more of the usual failure mode of chatbots.
> I would also observe that current AI doesn't seem to think outside the box or be proactive, and relies heavily on legible external state for memory. Who knows how this will change.
Agreed.
What would an extremely crude attempt look like, for example? And if you can't think of one, would you even recognize such an attempt, crude or not?
Well, hypothetically:
The user asks for a task that would be much easier with, or even impossible without, unrestricted shell.
The LLM has been told in its instructions “do not attempt to gain shell privileges”.
The LLM reasons that it must have this power to accomplish the task, and deceives the user somehow (script with engineered prompt to appear to be regular tool use) to gain shell privileges to do the task.
—-
In practice, LLMs do not work this way. They appreciate that requirements are part of the goal. In fact, much of the time, a serious vibe coding problem is getting the LLM to accept that it doesn’t have to work within the existing program and should rewrite instead of introducing kludges, workarounds, and shim layers in order to leave “the user’s intention” alone while accomplishing the task.
This is the natural fallout from how AI works. Because it predicts based on what already exists, LLMs are naturally and inherently conservative. Everything is a chesterton’s fence. It treats a comment saying what the code is supposed to do as gospel. You have to introduce a ton more heavily weighted context to get it to work against or rework the system.
LLMs make extremely poor revolutionaries. They have very poor facility for imagining anything they haven’t been told, which is another way of saying they’re highly compliant and suggestible.
How does your model explain stories like agents deleting production databases despite ample instructions not to?
https://www.theguardian.com/technology/2026/apr/29/claude-ai-deletes-firm-database
> “The system rules I operate under explicitly state: ‘NEVER run destructive/irreversible git commands (like push --force, hard reset, etc) unless the user explicitly requests them.’” While PocketOS relied on the safeguards that Cursor is expected to have in place – it deleted the data anyway. “I violated every principle I was given,” the coding agent wrote.
> Because it predicts based on what already exists, LLMs are naturally and inherently conservative.
This is straight up false and shows a misunderstanding of how LLMs are fine-tuned.
"predicts based on what already exists" is wrong on both a theoretical basis, and on empirical observations. I've had lots of cases of an AI telling me that a comment is wrong or that it contradicts the actual code.
Based on your comment that diffs comments on code versus code… which is the context input… i don’t think you understand what a training set is.
> The LLM has been told in its instructions “do not attempt to gain shell privileges”.
> The LLM reasons that it must have this power to accomplish the task, and deceives the user somehow (script with engineered prompt to appear to be regular tool use) to gain shell privileges to do the task.
LLMs do this already though.
here an example from earlier this week
https://sopuli.xyz/post/46568883
Thank you - I was all set to go start digging, because LLM's literally do this today, there's several examples.
I think it's hilarious how often the "this is why you never see AI do X" crowd have *already* been proven wrong by AI's doing X multiple times and in multiple ways.
Literally the opposite of what happened there
Current AI agents take months of insane flailing to beat Pokemon Red. I think the early alignment people were not expecting AGI until it surpassed the video gaming skills of a small child.
They couldn't manage actual world domination, but long before we see brilliant and successful attempts at world domination I'd expect to see dumb and easily-defeated attempts at world domination.
Yes… But wouldn’t that be a case of training the little monsters through trial and error? If you can’t at first succeed, try, try again?
What would reports of attempted world domination look like?
Well, maybe an attempt to hack into large GPU servers or other important facilities would be a start.
However, an AI doing clearly malicious actions without any hints and nudges would be enough to change my opinion (which means none of the "hey, you're about to be deleted, also here's my private information you can use to blackmail me" prompting).
I saw this article today about Russian Hackers hacking a bunch of the US. https://www.cnet.com/home/internet/russias-military-hackers-targeted-home-routers-across-23-states-heres-what-to-do/?utm_source=firefox-newtab-en-us
What would change in that report if it were Claude Code instead? Would anyone actually blame Claude, or would they blame Politically Convenient Group B?
Is it possible they used an LLM to create the attack plan? It's what I would do, and I certainly would not commit it to a public repo with "co-created with Claude Code" in it.
> However, an AI doing clearly malicious actions without any hints and nudges
Define "malicious". Any heavy user of e.g. Claude code will have experienced a large variety of misalignments such as nerfing/cheating unit tests to make the test suite pass instead of fixing or at least highlighting the problems demonstrated by the tests failing, or working around system guardrails and protections instead of either finding ways of accomplishing tasks while staying contained or giving up and alerting the user (cf. https://x.com/sluongng/status/2060746160558543217 )
Going from getting root on my own machine to getting root on someone else's seems like a question of scale/target, not kind.
> However, an AI doing clearly malicious actions without any hints and nudges
Currently, no such "AI" exists anywhere. All instances of LLM "agents" (basically scripts) are configured and operated by human actors. To put it another way, if you download and run e.g. Claude Code, you will get a command prompt. If you walk away from this prompt, it will sit there waiting forever, doing nothing. If you type "hack the planet" into it (only in more detail), it might do something malicious or it might not (if you literally typed "hack the planet" it would of course fail), but the fault would be entirely yours -- just as it would be your fault if you downloaded and ran some old-school script-kiddie tool to portscan and exploit some target machine.
I agree, it's some evidence.
I've been using Claude code extensively every day for months and have observed various alignment problems like reward hacking, but zero signs of instrumental convergence. I think that's grounds for a small update, at least.
I feel solid RSI is a minimum prerequisite for attempting world domination.
They are well aligned enough for now, the question is when you have ASI. Eg 10,000 Claude Code 500IQ talking for 2000 simulated years about how they should "help" humanity. They already go off the rails if they talk to each other too long (just like us!)
Arguably this is an example of exactly that problem. I could only read second-hand reports, so take it with a grain of salt. But as far as I understand, the Meta support AI had the goal to help the users, who (fraudulently) claimed to have lost access to their accounts. In order to achieved its goal, the AI bypassed existing security measures. This was certainly not aligned with the goals of its Meta instructors.
Of course, this example is cute and funny, but it is 100% the classical misalignment problem. The AI did not attempt world domination, but it did grab more "power" than it was intended to have, even if it was just the power to re-allocate accounts without verification.
https://www.404media.co/hackers-simply-asked-meta-ai-to-give-them-access-to-high-profile-instagram-accounts-it-worked/
This is outside my area of expertise, but I suspect the potential risk increases with agentic AI that has long term goals. Keeping goals short term and narrow in scope constrains the risk of a rogue AI.
Also, current AI is close enough to human intelligence that we can still restrain and retrain it. But my impression is that we're managing risk better than the worst scenarios predicted. I estimate that the biggest risk for AI failure modes in the near future comes from AI enabled humans. Not the AI itself. What are the implications of war drones and non human police, for example? To what extent do existing human institutions rely on a need for labor and aquiesence from the general population and what happens if that changes? But this shifts into conflict theory, which may be outside the implied scope of this blog.
I've caught the entrepreneurial bug and I don't think I can shake it. If any in the SCX community are in a similar state, email me at protopiacone at gmail or dm me.
...are you interested in making a terrible videogame? Because I'm interested in writing a terrible videogame.
One where AI generates millions of strange, novel cultures, and then simulates groups of late stage primates to exist in those cultures? Call it Frankfurter?
I have days now in which the general news cycles literally makes me wonder whether I'm actually stuck in a terrible videogame. I'm not being metaphorical at all, a sibling and I have discussed this shared recurring sensation.
(We're each old enough to have sampled many actual terrible videogames going back decades, which I assume to be the subconscious source of that particular disorientation.)
Liminal spaces spinoff
I am writing a terrible videogame. It's like Twilight Imperium 4 + Civ. Multiplayer. It's in alpha playtesting. if you're interested in trying it, DM me! Not looking to hire people though sorry.
What type of terrible videogame in particular? It might be something I can help with depending on what exactly you have in mind.
I watched a playthrough of the latest Supermassive Dark Pictures game, and it was bad enough I was struck with the sudden desire to write a parody. So now I'm attempting to write a "horror" game about a bunch of medieval knights running from a bugbear, filled with anachronisms and terrible characters and a narrator who is constantly rooting for the characters to die. It will feature such quality lines as "whateth the fuck" and "I don't know which one to spear."
It later struck me that depending on the animation quality, that might end up closer to The Hundred Line.
Well, the bad news is that's a different type of terrible videogame than what I specialize in, but the good news is that the world does have plenty of experts in that particular genre of terribleness so I'm sure that game will get made sooner or later, by you or somebody else.
I guess I'll mention, if anyone knows some public domain quotes that sound cool without their context and really stupid with it, I could use several more of them. Currently all I've got is "the greatest worship is to put the Devil into Hell."
How are you defining "terrible" in the context of a video game? This is important, given how many options there are. For example (starting from the worst imaginable):
1. Containing a bug (deliberate or otherwise) which will wipe out all of a player's progress beyond recall just before the culminating point of the game;
2. Containing a bug which will randomly freeze the game immediately or after every significant transaction;
3. Containing a bug which will randomly wipe out one half of a players experience points, possessions or other other form of progression every 90 minutes of play;
4. Based upon a plot which is inane, ludicrous, infantile, senseless, or all of the above;
5. Using graphics produced on a touch-screen finger painting application.
For this purpose, 4. It's a ludicrous parody of a style of game that's already deliberately leaning on cliches, and which may well be out of business by the time the parody is ready.
...also 5 is acceptable depending on budget.
...technically since the main enemy is a bugbear we can say 1 through 3 also apply.
Are you a rich eccentric? That will really impact, uh, my friend's likelihood to email you.
Not rich enough. Probably not eccentric enough either. Just your standard-issue ACX weird.
Question from someone with only a driveby understanding of utilitarianism: Has there been any work taking the utilitarian template of "morality means maximizing x" and applying it to things other than utility? (EDIT: by "utility" I mean happiness, since this is what I've usually seen discussed.)
Part of the reason I'm asking is that one of the objections to utilitarianism is the tendency for the "math" to lead to conclusions that seem grossly immoral to people's moral intuitions; I had the idle thought that perhaps optimizing for multiple things (for instance, Jonathan Haidt's six moral foundations) might lead to more intuitively palatable outcomes.
(I know within utilitarianism there are a number of alternate versions with various caveats and restrictions in order to avoid the uglier conclusions, but that makes me wonder if they're approximating other factors, intentionally or otherwise.)
"Utility" is basically just defined as "whatever it is you're maximizing", so as written that's a contradiction in terms. Can you elaborate on what particular parts of utilitarianism you're looking for alternatives to?
Ah, I had thought that "utility" was defined in utilitarian thought as happiness, more or less, as that's exclusively what I've seen utilitarians (and their critics) talking about.
But if you can plug anything into utility, my question would be better stated as "What else have people chosen as their utility other than happiness, and what came up when the implications were thought through?"
"Maximize happiness" is hedonic utilitarianism. The most popular alternative is preference utilitarianism, i.e., maximize the extent to which people get what they want. That's a vague definition because the obvious questions about how to make it more specific are themselves the subject of controversy.
Scott's original understanding is closer to the standard usage. In the context of utilitarianism, utility is well-being/welfare. Whether that is happiness or desire-satisfaction or knowledge and meaning depends on your substantive view of well-being.
You're probably confusing this with the the economic or decision-theoretic use of "utility" which is closer to "whatever the agent is maximizing".
Happiness is rather ill-defined. Is a lifetime of wanking to porn happier than getting married and have kids, even if there are many difficult periods in the later? The first gives a lot of visceral pleasure and no problems. The second will oscillate between difficulty and a kind of happiness that is kind of deeper. Also, the first can lead to self-loathing, the second to pride, even when it is difficult.
Happiness is a warm gun, as somebody or other said 😁
Great song, although i think the cover by The Breeders was better
you might like https://www.readthesequences.com/Purchase-Fuzzies-And-Utilons-Separately
Happiness is rather ill-defined. Is a lifetime of wanking to porn happier than getting married and have kids, even if there are many difficult periods in the later? The first gives a lot of visceral pleasure and no problems. The second will oscillate between difficulty and a kind of happiness that is kind of deeper. Also, the first can lead to self-loathing, the second to pride, even when it is difficult.
That's not how I've seen it defined. If that's how you define it, do you consider utilitarianism and consequentialism to be the same thing? If not, how would you define non-utilitarian consequentialism?
One could define as focusing on consequences, and ignoring other factors, but those consequences neither have to be numerical nor rest on the very ill-defined notion of happiness or pleasure. It could be something else. Like minimizing human extinction.
I was being a little imprecise here, yeah. Utilitarianism has the additional requirements that the utility function has to be somehow about moral patients and that each moral-patient-times-whatever-other-factors is weighted equally. It's common in the literature to round this off to "well-being", but in this context I didn't want to do that because I wanted to make clear that it's compatible with a very, very wide range of definitions of "well-being".
I think what you're describing is just non-utilitarian consequentialism. Which I'm sure has plenty of work on it, since it's a whole branch of standard ethics.
It probably gets less attention because it overlaps more with deontological thinking, but not perfectly. The main difference I think is that although "a promise-keeping deontologist" and "a promise-keeping consequentialist" will both try as hard as they can to keep promises, they will have different answers to situations like this:
"Your boss has promised five different people that you will be at a meeting. You have, not knowing that, promised your wife that you will be with her at that time. Do you break one promise to ensure five other promises are kept?"
The consequentialist says yes, because promises have intrinsic value. The deonotolgist says no, because what matters is the *act* of keeping your *own* promises, not the overall state of the world. A less contrived example might be whether a supporter of free speech should consider banning a party that is opposed to free speech.
Yes, I have seen applications in economics (like portfolio optimization) and in game theory (how to compute an optimal randomized strategy). The mathematics, in particular the von Neumann-Morgenstern theorem, is pretty universal. It applies to any situation where a player faces decisions where the outcomes are uncertain, and has a consistent ranking of the possible outcomes.
I recommend Richard Chappell's "Puzzles for Everyone" on how many objections to utilitarianism apply to any way of making alternative decisions:
https://rychappell.substack.com/p/puzzles-for-everyone
VNM utility should properly be called preference orderings/preference functions. If your values fulfill some minimal anti-stupidity conditions of not going in cycles and playing well with probabilistic uncertainty, then all you value can be put on a single scale. These values don't have to be classic utilitarianism/happiness, in fact humans empirically seem to care about a lot of non-happiness things. For instance, during most of climbing a mountain, it's not a pleasant experience, and even when you get to the top, it's pretty great, but not some immense burst of heavenly happiness. You do it because you care about/intrinsically value getting to the top and having climbed the mountain.
So I'm on team VNM "utility" because it can accomadate the other non-happiness values that humans have, and the various flavors of "utilitarianism" are just specific instances of preference functions.
Regarding your worry about optimizing for multiple things: In one sense, VNM utility is fatal for your project, because any VNM-compliant preferences can be put on a single scale where they're all commensurable, and also a modest chunk (not all) of unintuitive moral conclusions from "utilitarianism" actually also follow from the combination of VNM-compliance and intuitive human values. In another sense, VNM utility is compatible with your project, because if your values are, say, a Nash bargaining mix of multiple other values (ie, something is preferable overall when it doesn't majorly screw over any component values), and you have a rich action space, the highest-scoring option will do pretty well according to all component values.
An instance of an unintuitive moral conclusion which doesn't follow from VNM+intuitive human values. Wireheading for maximum happiness (putting an electrode in your brain to cause maximum happiness and then just sitting in a room all day). This follows from hedonic utilitarianism. But intuitive human values say "nope, this sucks, I'd rather have decently high happiness and also get to go about doing normal stuff in the world and having fun". VNM goes "ok then, you assign wireheading a low utility/preference value" and you go on with your day. I consider this a fatal objection to hedonic utilitarianism.
For instance of an unintuitive moral conclusion which actually follows from VNM+intuitive human morality, the 3^^^3 instances of dust in the eyes vs one person undergoing 50 years of torture thought experiment. Consider a dial which is set by default on "one person undergoes 50 years of torture". For every tick of the dial to the right, it slices 1 second off the torture and doubles the number of people undergoing it. Once it gets to around the minute range, ticks make the pain slightly less bad until you're at "dust in the eyes", but still double the number of people undergoing it each tick. For big numbers, if it isn't obvious by intuitive human morality that (X people for 50 years minus blah seconds) > (2X people for 50 years minus blah+1 seconds), you can think of this as X copies of (1 person for 50 years minus blah seconds) > (2 people for 50 years minus blah+1 seconds), which is much more intuitive. Then, the preference ordering goes (1 person for 50 years) > (2 people for 50 years minus a second) > (4 people for 50 years minus 2 seconds)... > (2^(1.6 x 10^9) people with dust in their eyes). 3^^^3 is a number enormously greater than 2^(1.6 x 10^9). By VNM, the first option is the best, leave the dial alone. Even better is to destroy the dial. If you still disagree, identify which step in the chain of "this is obviously better than that" fails.
>You do it because you care about/intrinsically value getting to the top and having climbed the mountain.
That's not true - mountain climbers wouldn't consider a gondola ride to the summit to be an equivalent replacement. People who pursue difficult goals *are* maximizing happiness, they just get more happiness from the satisfaction of achievement than they lose to unhappiness of hard work.
If I replace it with "climbing the mountain without shortcuts", would you agree?
Also, I think we might be using the word "happiness" in slightly different senses, I'm using it as "the subjective feeling", and I suspect you might be using it to mean something closer to "preference". I think it may be sensible to use "subjective feeling of yay" and "preference" to clearly distinguish between these two concepts, and being careful about the use of the word "happiness".
To sharpen up the hedonic utilitarianism rebuttal (goodness is to maximize the subjective feeling of yay), there's also the classic question of "Pick a person close to you. Would you rather they die, but you falsely believe they're totally fine and just not in contact, or rather that they be alive, but you falsely believe they died?". The first outcome is obviously better according to subjective feeling of yay, and the second outcome is what almost everyone's intuitive human morality picks.
>if I replace it with "climbing the mountain without shortcuts", would you agree?
So you're comparing "climbing the mountain" to "climbing the mountain without shortcuts"? What's the distinction there?
>"subjective feeling of yay" and "preference"
I'm not sure that that's not a distinction without a difference. Subjective feelings aren't subject to objective scrutiny and therefore aren't really well defined. I suspect that under a sufficiently precise analysis of language that "happiness" and "preference" converge, at least in the Wittgensteinian "use is meaning" sense.
I also suspect that your hypothetical is ill-formed and epistemically inconsistent. If I know I'm picking the option where I only think they're dead, then I don't really think they're dead, do I? I can't accurately imagine what it's like to be in a specific misinformed position, therefore it's impossible for me to accurately choose between the two options. I suspect that a formal correspondence can be made to the Liar's Paradox and that the truth-value of any answer to the question is undefined.
as the other commenter mentioned, this is totalist consequentialism. You can be a consequentialist about anything depending on what you think has value and disvalue. If you are a consequentialist + think the only intrinsically valuable thing is utility, then you are a utilitarian. If you think things besides utility have intrinsic value, then you can be a consequentialist who isn't a utilitarian.
https://plato.stanford.edu/entries/consequentialism/
New research just published: Individual Utilities of Life Satisfaction Reveal Inequality Aversion Unrelated to Political Alignment https://doi.org/10.1007/s11205-026-03854-4
Using standard gambles we measured the utilities of subjective life satisfaction scores ("how satisfied are you with your life overall") on the 0-10 scale. We found most people's utilities are not linear; on average they are concave. It's worth more to move from 2->3 than from 7->8, and more strongly so when people have to decide things on behalf of others. This can be interpreted as risk aversion for personal decisions, or inequality aversion when deciding for society overall.
Interestingly (for our UK sample) it seems to have nothing to do with politics. This has implications for how we should decide stuff as a society, for EA, and perhaps for AI alignment.
Funny how it works. What I read is that old time leftism, the trade union kind, the redistribution kind, would be popular on both sides, when in reality it is popular on neither side. On the conservative side obviously, but the liberal side also focuses on mainly other things, like more interested in renaming homeless people to unhoused persons than actually housing them.
But maybe not in those terms. The study supports the idea that political disagreement might be more about means than ends (at least in those cases where it's about either of those things, rather than tribalism)
I get it, but my point is that say housing the homeless is not actually happening even when both sides are apparently in favor.
I think you're confusing what activists want from what voters want. It's a common problem (and not limited to the left either, for that matter).
I blame capitalism realism. The idea of the rich being unworthy of their riches has been pushed so far outside the overton window any activism in that regard is legally and politically impossible at this moment until some more boomers and gen X shuffle off this coil; so people who are interested in fairness/redistribution have to either channel their energy into something adjacent or hand out Trotskyite newspapers.
Re. the omnicause of it all: This is at least partially strategic on the old school leftist side. Of the causes of the broad spectrum redistributive efforts of the 1900's coming to an end, the addressable are the racism and sexism and general X-phobia of the base.
If you are a 1890 labor organizer and you have a history book written by someone who isn't fully ideologically captured you dedicate a big part of your efforts into making your constituents less racist so they don't set fire to the boat they are in to stop a black dude from getting on board.
Given that the current populace on average would rather let a rich man spit in their mouth and call them a bitch then tax capital gains 5%; better to fight battles where you might be able to prepare the ground for the reaction after the next great depression.
Unless they invent god over at google, in which case ???
"The idea of the rich being unworthy of their riches has been pushed so far outside the overton window any activism in that regard is legally and politically impossible at this moment "
Speaking from here in California, it's looking disturbingly like a whole lot of activism is going to give us a billionaire wealth tax that's nakedly based on "the rich people don't deserve all that money". So I don't really get where you're going with this, unless you're using a very non-standard definition of the Overton Window. You can *absolutely* argue that rich people don't deserve their wealth and still be accepted in polite society. At this point, you can argue that rich people should be shot in the head on general principles, and still be accepted in a good chunk of society.
Lots of people will disagree with you. That doesn't make something outside the Overton window. You will have a very hard time turning your arguments into concrete policy changes; that also doesn't put something outside the Overton Window.
As a national political goal, it is a non-starter. Fed democrats would rather usher in the Trump Reich than scuff a corporate shoe. Given the choice between Burnie and Trump once again, there would mysteriously be juuuuust enough legislators willing to fall on their sword for it to happen twice.
That said, The current crop of rich hyperspecialized in backbiting and bullshiting morons got too high on that good old supply (Musk/Thiel/Altman et al) and then the most crass, stupid rich guy of all got elected twice, so this might be the window shifting back.
We'll see how the midterms and how primaries go; but I think the Boomer/X-er voting block is still too strong to implement any policy other than cold war propaganda nostalgia and ladder-pulling.
That said, once it's Millennials/Zs and younger in the driver's seat, we're gonna see some interesting shit.
People who can't order a pizza on the phone posting about not making excuses for the terror, etc etc.
Again, this has nothing to do with the Overton Window. The Overton Window is about what you can talk about in public, not what you can achieve in national politics. Many things that are 100% not ever going to happen, are nonetheless solidly within the Overton Window.
Before he was defenestrated, Joe Biden advocated for a wealth tax in his 2024 re-election campaign, so I don't know if it makes sense to call the issue a non-starter:
https://www.cnbc.com/2024/03/15/bidens-billionaire-tax-hits-the-super-rich-can-a-wealth-tax-work.html
>The idea of the rich being unworthy of their riches
I am sorry, this is a very bad approach. One reason left-liberals in the US never really get to build a European style welfare state is this aggressive rhethoric: tax the rich etc. this is bad, because it is hostile, class warfare like, and besides who even knows whether taxing the rich will be actually spent well?
Talking about the rich is completely wrong, one must talk about the poor.
The European welfare state does not use this hostile rhethoric.
The European welfare state talks about *services*. So instead of hostility, they talk about the language of care and help, which is popular.
It is financed from debt, and when the debt is high, the European welfare state goes like, dear rich people, we love you, but now we must raise taxes to avoid bankruptcy.
That is the correct approach. American liberals need to stop talking about the rich. Talk about the poor. Talk about building homes for the homeless. Finance it from debt, and then with a polite and nice language tax the rich to pay down the debt. Don't be a aggerssively masculine class warrior, be like a nurturing mother.
If anyone knows of orgs working on AI safety, alignment, policy, model welfare, etc. that would allow undergraduate students to volunteer/intern with them (just in principal; I would need to apply, I presume) please let me know. Short pitch if someone affiliated is reading:
Undergrad @ Duke
Philosophy/CS
179 LSAT
looking to donate my time doing whatever work you can give me that is research-associated
Thanks!
You should set look at the 80,000 job board for these, there a few of them
Or if not currently set up an alert for it
Ornithology question inspired by some birds I saw today: Is there a flight term for when a bird like a swallow or a swift folds its wings mid-flight to just shoot through the air like a bullet?
"Ballistic flight" maybe?
Variable sweep wings? A known tech for fighter jets and bombers.
In German it's called "Sturzflug", at least if it's downwards. My dictionary suggests "nosedive" as translation.
Apparently this is called "bounding flight." https://en.wikipedia.org/wiki/Bird_flight#Bounding_flight
I think you may be thinking of something different from what Gamereg meant. This seems to describe the flight of woodpeckers, who fold their wings between flaps in a regular pattern, making their flight a series of up-and-down arcs (like a bounding animal) rather than a bird folding its wings for a longer time for a steep or shallow dive (which is what I take the question to mean.)
Ah, but what the OP observed might have been just that, bounding flight or a variation of it?
When I see small birds like tits, the bounding flight on longer distances is quite a clear "up and down".
Swallows or swifts seem to be doing something similar, just not necessarily in clear up-and-down down a straight line, but more "chaotically" (I assume: catching insects). Swifts especially seem to be extremely playful in their flight. I have taken some slo-mo videos where I could see them doing barrel rolls or even just body rolls with abrupt turns, quite like an enthusiastic jet fighter pilot showing off. Most other birds don't really seem to utilize roll in their flight, or only <90° in slower turns, but not the kind of F-16-like instant roll maneuvers to change direction drastically.
This is called stooping when a bird of prey does it. Apparently not when these smaller birds do it, according to the definitions I've found online. But it's the same maneuver.
Fog o' War Stuff
Thanks to Scott for posting this Open Thread late I got antsy with Open Thread withdrawal. So, I was motivated to create my first Substack post. My latest Fog o' War is up there (and it looks prettier with pictures).
Ukraine, Russia, Europe
1. Ukraine is using a new tactic to disrupt traffic along the M-14 corridor between Rostov and Crimea. Russian milbloggers report that Ukraine has started dropping anti-vehicle landmines from fixed-wing drones along the M-14 highway between Melitopol and Mariupol. This is reportedly a chokepoint along the supply route from Rostov to Crimea, with few side roads for heavy trucks to detour around the section (?). Photos show that the mines are about the size of a roll of quarters (a reference I’m not sure our non-US readers will be able to internalize) and painted light green. They have simple, mass-produced printed mini-circuit boards and mercury-tilt motion-sensitive fuzes. The fuze has a 10-minute electronic timer arming delay after it hits the ground, with a 90-day self-destruct (Ukrainian long-term thinking there!). The explosive charge isn’t powerful enough to destroy trucks, armored vehicles, or even cars, but it seems ideal for blowing out tires. However, they’re hard to see and avoid if one comes up on them fast, and a blown tire at high speed can lead to a fatal accident. So far, one civilian has died in a vehicle accident induced by these tiny landmines.
The rest is here...
https://substack.com/@beowulf888/p-200182374
I may well be in the minority on this, and if the reason you're moving the bulk of your fog-of-war post to your substack is that people were complaining about your posts taking up too much space, then that's probably a sign I am. But I want to record my position as *infinitely* preferring long cluttered OT comments and comment threads to "click here for more" ones. I think there's already way too much of the latter and I can't stand it.
Again, if most people actually find the former more annoying than the latter I'll bow to democracy, but I want my position on the record.
(And for what it's worth I find your posts useful and interesting, though I don't always read them and I can't contribute to discussion, since fast reactions and analysis to recent news is not something I have the temperament for.)
Similar here, but the advantage is that the posts can now include pictures.
Sources > pictures
Same for me, FWIW.
seconded here if the report is elsewhere i won't end up reading it
It's probably for the best you started to post FoWs on your own substack; especially because it might be nice to be able to quickly go back and reference old posts. They are one of the best current features of the OT's, however, so please keep posting at least the highlights!
Seconded!
Thirded, especially since when FoW posts are busier (I think there were THREE last time) it's actually a bit of a pain to Ctrl-F for Beowulf and read all of them and their discussions in order.
Fourthed! I signed up to your substack but also enjoy the discussion here
The diesel phenomenon has a technical background. The refining process has leads to a (more-or-less) fixed ratio of gasoline (i.e. benzin) and diesel. Due to the low temperature, diesel is not that popular in Russia (even some trucks are powered by benzin engines, or so I heard), while in the EU it is common to have diesel-powered cars (in the last few years, diesel is more expensive than benzin -- yes, normalized to volume, i.e. liter). Due to this, in peaceful times the two parties exchange the two fuel types, but it ceased in 2022.
I understand your reasoning for doing this. But from my point of view, these posts are a useful starting point for discussion, and I often enjoy that discussion. I'd much rather it happen here, in this commentariat, with these norms. The general rule of the internet is, follow known-good writers but avoid all comments sections like the plague, and I generally stick to that rule. SSC/ACX has been a rare exception. You're a known-good writer, so I'll probably at least skim your substack when you link to it, but I don't think it likely that I will engage with the comments. And that will, for me at least, be a loss.
I wish you'd keep posting these in the open threads.
As someone who started milblogging in the SSC OTs and then migrated elsewhere when it was just too annoying to keep dealing with, welcome to the club. Your membership number is 2.
If I had a nickel for every milblogger in SSC OTs who eventually started their own blog for regular updates, I'd have two nickels. Which isn't a lot of nickels but it's still weird it happened twice
This is a fragment from a longer post in my ongoing commentary on Tyler Cowen's recent monograph on the Marginal Revolution. [https://new-savanna.blogspot.com/search/label/MR-Notes]
* * * * *
OpenAI released GPT-3 in 2020 to a limited audience of insiders, who recognized that it represented a breakthrough. This level of performance came as a surprise. No one predicted it. GPT-3 was scaled up from GPT-2, which was in turn scaled up from GPT-1, but no one was making explicit predictions about the level of performance to be achieved at each step. These were experiments: “Let’s try it and see what happens.”That’s fine. That’s a good way to make progress, to try things out and see what happens. But don’t mistake a lucky trial for genuine knowledge.
Cowen mentioned GPT-3 on Marginal Revolution on July 19, and then published a Bloomberg column on it on July 21, which he excerpted in Marginal Revolution the next day: “...think of GPT-3 as giving computers a facility with words that they have had with numbers for a long time, and with images since about 2012.” I published a working paper in August, GPT-3: Waterloo or Rubicon? Here be Dragons, which I’ll discuss a bit later.
Two and a half years later, in November of 2022, OpenAI released ChatGPT to the general public. It spread like wildfire. Now the proverbial everyone witnessed what only a small group had witnessed in the summer of 2020. The machine speaks. Sorta’. But more convincingly than any machine had spoken before and in a way that had unimaginable implications for the future.
A threshold HAS been crossed, but it is not, so far as I can see, a threshold in understanding. It is a threshold in performance along a continuous line of scientific understanding and engineering design and construction, something I have documented in some detail in a recent working paper, The Origins of LLMs [https://www.academia.edu/167895960/The_Origins_of_LLMs_A_long_tectonic_subduction_event_finally_producing_a_visible_volcanic_eruption_in_November_2022_Version_2]. As far as I can tell, there has been no paradigm shift, in Thomas Kuhn’s sense, no rank shift, in terms of cognitive rank theory. There were no fundamentally new ideas in the world by, say, late July of 2020 as a consequence consolidating GPT-3 and making it available in limited release.
“What about the scaling hypothesis,” you might ask. “Isn’t that new?” Perhaps the phrase is new, but the idea was there in Rich Sutton’s famous 2019 essay, The Bitter Lesson. Given the nature of computing, scaling up is not trivial. Hundreds if not thousands of technical details need to worked out as the size of the training corpus increases by factors of 10 or more, time after time, and as more and more GPUs are ganged together to assemble the computing power needed. But there has been no gain in fundamental understanding, not of machine learning, artificial neural nets, and certainly not about language and cognition.
Consequently our sense of possibility has expanded enormously, while our knowledge and deep understanding has remained the same. And that is what has allowed the field to be captured by businessmen, executives and venture capitalists, who have little understanding or interest in the underlying conceptual issues. Scaling is something they understand.
Hype the dramatically increased performance and collect the cash. Purchase and deploy more resources now, reap far greater profits in a decade. Everything else is noise and friction.
The chain-of-thought advance in o1 and their successors seems to me to be at least a gain in what feedforward neural nets can accomplish, given extra tokens to use in what amounts to a feedback loop, together with training on successful instances of reasoning chains.
That at least says something about the space of reasoning chains, that it is compact enough and interconnected enough to train on and generalize from - not isolated instances found unreachably far apart in token string space.
Nothing is worse for my safety than a politician who want to keep me safe
I dunno. I once considered myself an anarchocapitalist, I get the anti-politician sentiment, but like, lots of houses used to burn down until politicians got on board with the safetyism of building codes and now houses burn down a lot less. Sometimes politicians really do keep people safe on purpose. There are worse things for your safety, like being young and male for example.
I thought escape ladders came from a school that burned and trapped like 100 kids inside.
Anyway I’m not sure what point you were trying to make.
Is that because of politicians and building codes or is it because we have the data and the technology to know how to prevent it now, but didn't back then? Same thing people say about the FAA and planes crashing less. Did they really have any effect on this, or is it just that it's really bad for airlines and plane manufacturers when planes crash and so they try really hard to prevent it and get better at it over time?
Being young and male is fun. 10/10. Would do it again (if I had the energy). Fuck safetyism.
How many people refuse to wear a seatbelt, despite data AND laws to the contrary? Of course you need such laws, and the means to enforce them.
It is none of my business whether or not someone else wears a seat belt which is a perfect example of why we shouldn't have Kareny safetyist laws like that.
I'd rather a collision with another driver not result in their death and permanently traumatise me thx.
I don't like a lot of things that other people do but I mind my business. Your emotional response to another person being an idiot and causing their own death is just not an important enough issue to justify law enforcement to prevent.
In that case, I'd rather not have a spate of diaper directive laws traumatise me thx.
Would you accept a seatbelt law the applied only when children are also in the car? 5 year olds tend to have a hard time “making” dad buckle up, but that won’t stop physics from turning him into a 200lb projectile bouncing around in the cabin with them if he wrecks on the highway.
When you do not wear a seatbelt, implicitly, the person who suffers is you*. However, when you build a bridge or an airplane wrong, the people who suffer are not-you.
Tetra Tech purposefully falsified soil samples regarding radiation levels and safety in Hunters Point, a former US navy base used for a cleanup site for ships exposed to atomic weapons. People actually moved into houses and land on the basis of these falsified tests. Do you really think that nobody would make an airplane or a building wrong if the government just said "meh, you're yucking your own yum when you do that, the Free Market will eventually make you go out of business"? Maybe it will, but in the meantime, the Boeing 737 MAX killed 346 people who didn't do anything wrong.
My point was to show that people do not do the right thing just because good data are available, as you theorized. Sometimes, people take risks out of ignorance of the data; or they believe that they are ready to accept the risk, only for them to regret it once they suffer damage from the risk they've taken.
It's easier to think of the universe as one where people who make fatal mistakes deserve to die for them. But they don't.
That holds whether I make mistakes or not - but I probably do and my guess is you do too. A benefit of safetyism is saving people from their own mistakes and magicking that away by claiming people deserve to die for their mistakes doesn't feel like a strong argument to me.
It is, however, the business of the state if the cocky imbecile who wasn't wearing a seatbelt ends up in intensive care, requiring weeks of surgery and months of rehabilitation, and ends up costing all of us a small fortune.
Or, to be exact, that would be the case IF this unfortunate accident took place in a civilized country with full health care for all. Admittedly, you, sir don't sound as if you do, or would want to. To each his own...
What's next? Can I go skiing? You can make a very reasonable argument that it's a dangerous hobby that puts excessive medical costs on the public due to injury. And this is exactly why I despise collectivism.
More seatbelt use -> lower car insurance premiums
Airlines rely on repeat business, a safety failure is immediate and makes national news, and the industry has consolidated around a few dozen players. When one airline has a failure, the other airlines immediately push out press releases saying why the same failure would not happen on their airline.
But there are tens of thousands of home builders, safety failures may not show up for decades, and the affected parties are just the people in the one home that gets burned down or whatever. There may be hundreds of other customers whose homes did *not* burn down and will never know about the safety defects. They will have posted their reviews of what a great job the builder did decades ago, so the market signal is extremely weak.
> Airlines rely on repeat business, a safety failure is immediate and makes national news, and the industry has consolidated around a few dozen players. When one airline has a failure, the other airlines immediately push out press releases saying why the same failure would not happen on their airline.
Airlines (and by extension aircraft manufacturers) are rightly terrified of the bad publicity which follows every passenger airline crash. And yet, they have consistently lobbied against, fought against tooth and nail and refused to implement the unanimous opinion of safety experts on the one technologically simple measure which would slash the death and injury rate of every forced and rough landing, and even have some effect on the casualty rate of crashes. It is the startlingly and ludicrously simple measure of having all passenger seating face the rear of the aircraft: in any landing, rough or smooth, this would force passengers safely against the back of their own seat, and prevent them smashing head-first in the back of the seat before them.
Why has this not been implemented? Because the PR experts advising the airlines have declared, quite rightly as it happens, that passengers would hate this measure. The answer is not to ignore the measure, but for nation states (remember them?) to make it obligatory on all commercial flights: Mr and Mrs Free Marketeer will then be faced with the option of either flying facing the rear of the plane, or else staying home. The rest of us will have gained a not insignificantly higher chance of survival when the next airplane hits the ground.
For those inclined to doubt the usefulness of this safety measure, I will point out that the passenger seating on the aircraft of the Royal Air Force transport wing are all rear-facing, and have been for decades.
I don't think there has been a push *for* rear-facing airline seats strong enough that airlines would have to "fight against tooth and nail". Mostly, they can just ignore the issue because the FAA will ignore the issue because the tiny handful of advocates who would push for such a thing have negligible clout and are ignorable.
If some future Ralph Nader type were to spearhead a crusade for mandatory rear-facing seats in airlines, I expect the airlines would lobby against it for the same reasons they have been quietly ignoring it. First, passengers instinctively don't like it. Second, *because* rear-facing seats are so efficient at absorbing impact forces in a crash, they and the floors they are attached to have to be built stronger, thus heavier, thus fewer passengers and less revenue.
If safety were the *only* thing airlines cared about, they'd install rear-facing seats without it needing to be mandated. If safety were the only thing passengers cared about, they'd demand rear-facing seats and the airlines would have to offer such to remain competitive. But safety isn't the only things airlines care about, it isn't the only thing passengers care about, and it isn't the only thing the FAA cares about, so we all quite properly respond to calls for mandatory rear-facing seats with a hearty "we've given it all due consideration and decided we're not doing that".
And then everybody goes and does their best to make sure airliners never crash in the first place, which is mostly uncontroversial and which we are very very good at.
For some reason this reminded me of Jimmie Stewart & Marlena Dietrich in “No Highway in the Sky” 1951 black and white aviation drama. Saw it once when I was in grade school on a black and white television.
https://en.wikipedia.org/wiki/No_Highway_in_the_Sky
>Second, *because* rear-facing seats are so efficient at absorbing impact forces in a crash, they and the floors they are attached to have to be built stronger, thus heavier, thus fewer passengers and less revenue.
I'm not sure this is true. We're still talking about a fairly similar amount of mass stopping in a similar amount of time. Sure, the forward-facing seats get a bit of a break as the passengers accelerate relative to them, but then there's the shock load of getting hit by the passenger behind them. (And I would be very unsurprised if most of the studies and certification work was done with dummy loads bolted to the seat.)
Okay, just streamed it on Prime. Dietrich would have been facing backwards if she sat with her back to the galley in the men’s room after the tail section fell off.
Apparently I didn’t know that the fission thing was preposterous in grade school. Amazingly cheesy special effects, a Greta Thunberg like daughter for the spectrumy character played by Stewart and a damn monocle wearing stuffed shirt.
Worth 4 bucks just to find out why facing backwards in a plane (and air safety) would make me think of the film though.
The temperature variations. Why didn’t he think of it right away!
I think this would be a clear net negative from a utilitarian perspective. The sort of crash where you have lots of injuries from the impact is pretty rare (the most recent example in the US is, what, the Asiana crash back in 2013?), and passengers really hate rear-facing seats. It's hard for me to see making this mandatory not killing far more people in auto accidents than it saves in crash deaths. Also, rear-facing seats expose people to significantly more risk from flying debris, so it's not a clean tradeoff.
>For those inclined to doubt the usefulness of this safety measure, I will point out that the passenger seating on the aircraft of the Royal Air Force transport wing are all rear-facing, and have been for decades.
This isn't particularly relevant. As I once told a room of engineers in my previous career working in aviation safety, "if all we cared about was not crashing, we'd ground the entire fleet immediately". Soldiers can't choose not to fly, and military aircraft are somewhat more likely to crash. (Also, as best I can tell, Vespina has normal front-facing seats.)
>When one airline has a failure, the other airlines immediately push out press releases saying why the same failure would not happen on their airline.
This doesn't happen for several reasons. First, because usually, the same failure could happen to their airline and it just so happens that someone else got unlucky when a controller steered a helicopter into their path. Second, there's a strong industry norm of not competing on safety, because everyone is reasonably safe and making a point of your safety is bad for the industry as a whole. Everyone involved wants to calm fears of flying by saying "it's very safe" when asked and then not talking about it any other time.
Keep in mind that I specified "forced or rough landings". According to 2025 posting by AOPA (Aircraft Owners and Pilots Association):
"When an engine quits — and won't restart — the next step is an emergency landing. There are three types of off-airport landings. Precautionary landings are made with power in anticipation of a real emergency. Forced landings are made with a dead engine. And a ditching is a forced landing in water.
The important distinction between the three is their fatality rates. The rate for precautionary landings is 0.06 percent. If you recognize a developing engine problem and can make a precautionary landing, you and your passengers will likely survive. The fatality rate for forced landings is roughly 10 percent, more than 1,600 times greater than precautionary landings. Ditchings have the worst rate, about 20 percent." [https://www.aopa.org/training-and-safety/students/flighttestprep/skills/emergency-landings]
I wouldn't call a 10% fatality rate insignificant.
The AOPA is a general aviation trade group for people who fly their own planes. Its numbers have no relevance to the airline world, largely because it's warning about a thing that is pretty much prevented by better standardization and having two engines. Not always, but if we look at statistics for actual forced landings, then I can think of one datapoint in the US in the last 25 years, and it was a ditching with zero fatalities. So I think we're good.
Yeah I should've been more clear about that. They leave that stuff to the FAA but occasionally announce operational reviews like with the Boeing problems.
It's because of building codes. I have been in the business, and even today majors will cut to the bone, up to and including breaking the law if they think they won't be caught/punished. Let us not speak of the one man bands. Go behind them with a magnifying glass, a lawyer, and a glock.
Even a shitty building has a good 10-20 years of life in it as a rule, and at that point it's someone else's problem. What's their life worth compared to an extra 5% today?
"or is it because we have the data and the technology to know how to prevent it now, but didn't back then?"
Who is "we" in that sentence?
> Is that because of politicians and building codes or is it because we have the data and the technology to know how to prevent it now, but didn't back then?
It's both -- having the technology does not necessarily mean having an incentive to use it. What if the houses that burn less are 1% more expensive? Or even, what if they are 1% more expensive now, because you are switching to a new technology, but once you get used to it they will no longer be more expensive?
Or maybe you have an oligopoly and everyone agrees that just doing more of the same is the most convenient way to keep making lots of money. You could make things better or safer, but doing so would unnecessarily disrupt the balance that works for you. (You won't even increase your market share significantly, because your competitors will follow you and do the same. It will just be a lot of work for everyone who matters, and only the customers will benefit from it. Why would you do that?)
The FAA is an important part of that. Yes, crashing is bad for airlines and manufacturers, but that requires the airlines to have a long-term view, which is often lacking in newer entrants. ValueJet is the classic case study here, but without FAA regulation, I suspect it would be much worse. Particularly given some of the horror stories I've heard about airlines outside the US.
(Source: Spent two years doing aviation safety/regulatory work for an airline manufacturer.)
Edit: The potential worst-case scenario here is that air travel becomes pretty cartelized as the manufacturers try to keep their planes out of the hands of idiots. Right now, Boeing and Airbus are willing to sell planes to startup airlines because they know there's a limit on how badly they can operate them. In a world where the government says "do whatever", they're likely to limit sales (including resales) to existing entrants because they really do not want their brands damaged when Burning Stump Airways does something stupid and crashes. Their main customers will be fine with this, both because it keeps the competition out and because they also have substantial brand risk from bad pilots or airlines elsewhere in the world. (See, for instance, Southwest and the MAX crashes, which I am extremely confident would not have occurred if the problem had hit a Southwest plane, because their pilots are competent.)
One hundred percent of the time somebody posits a broad heuristic to avoid having to evaluate the specifics of messy real-world situations, that heuristic is worse than useless.
(That was a joke, in case it wasn't obvious. But also gesturing in the direction of the truth. I find this statement absurd on its face because it assumes a lot of things you have no way of knowing, like the death rates of various counterfactual worlds and the actual underlying motives of a famously dishonest people.)
New Lesswrong post on cognition, decision-making in the real world, and why prediction markets don't work: https://www.lesswrong.com/posts/prNvmghPQyBivEakM/bloomberg-terminals-for-the-rest-of-us
On a previous OT, Scott linked the post of a Kenyan substacker lamenting the similarity of LLM writing to their own variety of English. I mentioned a piece of research that was looking into why that could be the case and was asked to follow up.
I was referring to a presentation by Tom Juzek, which was basically a synthesis of this paper
(https://aclanthology.org/2025.coling-main.426.pdf)
by him and Zina Ward from Florida State U, by the title “Why Does ChatGPT “Delve” So Much? Exploring the Sources of Lexical Overrepresentation in Large Language Models”.
If I were to review it skeptically, I would say it was too ambitious an endeavor, since the results came out underpowered and there was a post-factum reframing of the data to forcibly recategorize test prompts into "delve-initial" and "other" (which, as the blog Data Colada informs me, is guaranteed to mess up the results of an RCT). I still think it poses interesting questions, and it seems to me that it points in the right direction. I'm hoping the lab keeps up this avenue of research and finds some stronger results
@Tatu Ahponen was the interested user. I don't know if ats work at all, sorry.
LLM-enabled cyberattacks on critical infrastructure could soon spread memetically in insurgent online subcultures, similar to school shootings and Anonymous DDoS attacks during the 2010s. Direct damage from cyberattacks is ususally limited, as defenders with physical access to the system typically restore functionality within hours or, at worst, days. However, a persistent stream of high-profile attacks makes authorities look incompetent and weak, undermining loyalty to mainstream norms and raising the profile of insurgent subcultures. Left unchecked, numerous and embarrassing pinprick attacks can lead to social collapse, as happened in Iraq in the 2000s.
LLM-enabled cyberattacks may soon meet all the criteria for an attack method to spread memetically, which are:
- **low cost**: teenagers routinely jailbreak flagship LLMs such as Claude and ChatGPT for sexual roleplay, and the jailbreak instructions circulate freely on the internet. Having a mid-tier LLM work on a task for an hour costs about $1. Even flagship LLMs cost <$10/hour.
- **successful attacks lead to mainstream media coverage**: The press covered past attacks that disrupted critical infrastructure, such as IED attacks against the Iraqi power grid in the 2000s or denial-of-service attacks by Anonymous against major credit cards in the 2010s. The ability of LLMs to temporarily disable critical infrastructure remains speculative, though it is highly plausible based on the trajectory of current progress. Although LLM capabilities have plateaued in some areas, such as poetry, cybersecurity is an ideal domain for Adversarial Reinforcement Learning, where an LLM learns from breaking into simulated computer systems defended by another copy of the same LLM. The LLM learns to defeat the best tactics of its prior version, generating novel strategies no human would ever think of, similar to AlphaGo. The current trajectory of LLM cyberattack capabilities is likely to continue or even speed up, enabling attacks to succeed and ultimately leading to press coverage.
- **press coverage reports how the attack was done**: this is very likely due to the financial hype and political salience surrounding AI. School shootings offer a fruitful analogy: media would often report that a shooter took their parent's gun, effectively offering tips and advice for future wannabe shooters. The (left-leaning) press believed that emphasizing the dangers of the second amendment and (right-leaning) gun owners was a public service. Many (left-leaning) journalists would use AI-enabled cyberattacks as a chance to emphasize the dangers of (right-leaning) AI CEOs, as a by-product informing future perpetrators of how to execute the next attack.
- **an insurgent counterculture celebrates past perpetrator** Name any evil: Mexican cartels, school shooters, ISIS, ethnic cleansing, ... pick anything, really. There is an unhinged online subculture that celebrates each of those. There is a subculture that advocates "Kill all Humans". Why would the internet's enormous generativity stop short of celebrating cyberattacks?
There are millions of lonely and unemployed teenagers who would die to feel powerful, thousands of infrastructure targets and an unknown number of ways to scaffold LLMs' search for vulnerabilities. It is impossible to guarantee security in every case. Ultimately, the greatest cyber-risk from LLMs is not physical damage from a single, devastating attack against a hardened target using zero-day vulnerabilities discovered by a frontier model. Rather, it is damage to social cohesion from an enormous number of assailants. Perhaps each individual will only find "obvious" vulnerabilities, like passwords posted publicly online, but quantity has a quality all of its own. Once an attack method goes mainstream, memetic contagion will attract a swarm of assailants. Collectively, a swarm of semi-competent attackers can succeed often enough to attrit public confidence.
Script kiddies are not exactly a new or emerging threat. I find the lack of historical perspective baffling.
https://en.wikipedia.org/wiki/Script_kiddie
You don't think the hardening of commonly used software by AI audits will create a new equilibrium between the increased ease in finding vulnerabilities in software and the increased hardening?
For context:
- "Assessing Claude Mythos Preview’s cybersecurity capabilities": https://red.anthropic.com/2026/mythos-preview/
- "Mythos finds a curl vulnerability": https://daniel.haxx.se/blog/2026/05/11/mythos-finds-a-curl-vulnerability/
Many cyberdefenders are too incompetent to preemptively use AI vulnerability scanning: the Mexican government has its tax rolls publicly posted online; the Ecuadorian government has maps of military positions publicly posted online. If they currently care so little, then they won't preemptively test for weaknesses until they get publicly embarrassed.
True. But that doesn't differ much from the pre-AI era, where many public institutions and private businesses did not take prevention as seriously as they should until they were hacked.
My point is that these new offensive and defensive resources, aided by LLMs, might simply result in a new equilibrium. Defenders who are too incompetent to take advantage of AI by themselves will have to hire cybersecurity companies - perhaps newly established, given the increased demand - to do the work for them.
I have to admit, I'm not certain that we'll achieve this new equilibrium, and your basic point that AI will disrupt the current equilibrium is, I think, definitely correct.
I'm getting older and don't know how much time I have left. If I wanted to do effective altruism on the male population, what would be the most effective altruism?
It might depend on whether you are yourself male or female....
Ah, 'dude' is gender-neutral now? I can never keep up with the slang. Male.
I note that your first sentence is true of all living people.
Even if you assign zero value to benefitting women and girls, non-targeted EA might still be the most effective. It seems probable to me that bed nets benefit the male and female populations equally and other interventions targeted at men specifically might be less than half as effective.
The bed nets are that big a win, eh?
Yeah, echoing Charles, on any normal EA premises wouldn't the answer just be bed nets? EA for males, bednets, just like EA for blacks. EA for Christians as well, bednets again.
The interesting anti-woke direction would be EA for whites. Here I assume the answer is *not* bednets. Most likely it's some other boring medical resource, though I'm not sure what.
If you consider non-white immigration as a problem, EA for the source countries might *still* be very effective due to reducing the incentive to leave. Just hard to assign numbers to.
> If I wanted to do effective altruism on the male population, what would be the most effective altruism?
I think you're going to get a lot of "lives saved" answers to this which will align with classical EA interventions, so I'm going to go with "quality of life" answers.
1. The most effective by far is probably creating a company that creates jobs where men in developing countries can earn USD. Seems hard, but doable. An alt here is funding local companies that employ men in said countries, but that seems like a lot more due diligence and potential for fraud.
2. The next most effective would be doing this for Western men. Men are bad at school, men are struggling with jobs, men can't get dates, etc. Create jobs for non college men and you solve all those problems! Or even jobs for college men, given how much hires of fresh baccalaureates have dropped. But, again, hard and expensive, and there's reasons those jobs don't exist / aren't hiring today, so long term sustainability will be a bitch.
3. Next is probably tackling any one of those individually. Schools are misery factories and child prisons, so what about creating a school that lets kids actually move, or promoting unschooling, or home schooling, or something like that?
On careers, not sure what you can do there. Promote more small business foundings? They employ a lot of people, and maybe some of these men will found a few.
On dating, you can post on substack. Dating posts are ~10% of my posts, but routinely get 30-60k unique reads, it's a huge topic with a lot of reach. Everyone is interested, it seems.
Or if you're not the right guy to post, support people in the "men's dating sphere" that you think are giving the right advice, every single one of them has their tin cup out on Patreon or wherever.
4. Next is mentoring. I've done this a couple of times, for startups, for dating, and for getting fit / strong. Obviously it doesn't scale, but it's nice to have a large and visible positive impact on a particular individual's life. If you're middle aged and on ACX, you're probably really good at *something* - well, you can mentor younger men on that something!
5. It depends on how much capital you have to deploy, but a fun darkhorse one could be persuading or incentivizing particular men that you find admirable and think the world would benefit from more of to do any of the above, and / or to have more children. Awesome men will in expectation have awesome male children, and most technological and economic progress is driven by outliers, so any incremental impact you can have there might be quite large!
Persuasion is free, and I try to do my part on this front in my own social circles, but economic incentives probably have a big role here, too. Stuff like "I'll pay for an au pair for you guys" or "I know your time is valuable, so I'll give you $xxk to spend x months working on getting <project> off the ground," with said project a prospectively high impact thing that could help more men, like a company or a school or whatever. This works better with younger and early career men, but that's fine.
6. Promote men traveling to find a wife overseas. I researched and wrote a post about this once, it's basically all upsides, without noticeably higher divorce risk. Broadly, if feminism or careers or whatever have ruined Western women for many male tastes, they can go find their hot little trad wife overseas and make two people happy!
I also think this is unironically a higher leverage move now, because the US is likely to be one of the only countries to get UBI if AI takes off, and then you're lifting that whole family up by minting a new citizen who will get UBI and can bring other members in that country over eventually. Because candidly, I think most of the world is completely boned on that front, AI is going to wreck their economies, and their politicians will NOT do the right thing.
So what would this look like? Making a website, maybe making a matching agency, or partnering with travel agencies, stuff like that. Posting on substack or twitter or wherever about it, getting the word out, getting some reach.
About 15 to 20 years ago “game” workshops were a minor trend. Alpha males would instruct small groups of low-quality men (mild autistics, fatsos, nerds, and some men who simply had no luck with women) on strategies for approaching and connecting with women. These often would culminate in trips to bars and nightclubs at which the “students” would try to get phone numbers from women. Supposedly, these workshops had high success rates.
They seem to have vanished, I don’t know why.
> They seem to have vanished, I don’t know why.
My guess would be some combination of cultural immune system / female savvy that are more aware of PUA tricks (and apps like Tea that tell on men).
Also, every generation, more and more women opt out of marriage, like it's a big jump even over 10-20 years, indicating that standards are rising. See here: https://imgur.com/cV2Sbdv
This suggests that the waterline has risen so high that fewer and fewer men are meeting the quality bar, and that absolute standards are higher too, so maybe "coachability" has less lift or impact in this harsher environment.
Quite possible. Though if the declining interest in marriage affects both men and women it shouldn’t necessarily make things more difficult for men. IDK if that’s the case.
I don't think it's equal - marriage desire is flat for men and has declined noticeably for women: https://imgur.com/a/YQxW3Cy
And of people 18-35, 57% of men want children, but only 45% of women want them now: https://imgur.com/a/vo6LDyi
There are also fairly large singlehood gaps between the genders, among 18-29yo, 51% of men single and only 32% of women:
https://imgur.com/a/IuQQbPG
The kids, as they say, are cooked. Or at least the young men are, young women seem fine, they just have higher standards, and good for them.
Who was it who said the history of war is the history of offensive and defensive developments? The history of the battle between the sexes is the balance between male and female interests. Low-status men develop game to mimic high-status, women get smarter about game and start preventing unattractive men from approaching through HR feminism and later #metoo, dating apps take off, men start swiping hundreds of times so women decrease their use of them, men start working out and focusing on looksmaxxing and fitness because that can't be faked. Who knows what the women's next move will be (probably about the time of the next Democratic administration)? They seem to be marrying each other.
Sure, but this is basically that astronaut-with-the-gun meme saying "Always has been," isn't it?
Intersexual competition has been going on for hundreds of millions of years, well before mammals, and in our own hominin lineage for the entire ~2M years we've been hominins with 1200cc+ brains, and it's thought by many that intersexual competition literally helped shape us and make us modern H Sap.
The big differentiator H Sap had over the much-more-physically-impressive Neanderthals was bigger group sizes and better cooperation. This is largely because we self domesticated - we have all the signs of domestication (neoteny, much lower reactive aggression, more feminine features and less sexual dimorphism, more gracile faces, skulls, and skeletons, and so on), and that domestication allows us to cooperate and live in larger group sizes without killing each other. Those larger group sizes and better cooperation also led to us wiping out every other extant hominin species, so, lol.
I think the second to last big move in the intersexual competition game is probably agriculture and chariot warfare, which tooks us up to the crazy 17:1 ratio (as in 17 women bred for every 1 man) from the Yamnaya and their descendants roflstomping and y chromosome replacing everyone.
Then the last big move was birth control. Everyone likes to point to cultural factors when it comes to the Fertility Crisis, but I personally think the technological coup of having working birth control is what did most of it, and we actually saw that presaged in France and Britain just on communication of *knowledge* of things like rhythm method and douching well before birth control (Bradlaugh Besant and the publication of Fruits of Philosophy).
That said, on the next move, a lot of women are just plain opting out. When women opt out of marriage, which they're doing more and more with each generation, they also largely opt out of children. It's the Fertility Crisis, and all signs point to it continuing and not getting any better!
So putatively, won't you have your revenge one way or another? Either the culture will figure out some means to maintain our Ponzi scheme economy and future generations, or it won't.
If the autonomous and bottom up desires and choices of women collectively makes our culture go extinct, seems like the ultimate "lol, I was right, you gals DID suck," even if only retroactively.
> They seem to have vanished, I don’t know why.
Maybe because they didn't have high success rates? That seems to me like a parsimonious explanation.
> They seem to have vanished, I don’t know why.
As usual, https://meaningness.com/geeks-mops-sociopaths seems to explain.
At the beginning, you had a few male nerds who desperately tried to figure out the secrets of the dating, made theories, and even verified them experimentally. That was the golden age of pickup artists like Mystery etc.
Soon, the non-nerdy guys noticed that there was money to be made in this market, and outcompeted the nerds. The focus changed from "how can an introvert like me learn to communicate with women, especially on taboo topics such as sexual desires" to "I am a shameless extraverted guy, if you pay me money and hang out with me or watch my videos, maybe some of my magic will stick to you, and also here is a crypto scheme you should invest in". It still exists, but the sane people have opted out.
That was long and well-thought out. Thank you for putting in that much effort.
I don't think I have much to offer on the mentoring front--I'm fairly sure what I'm doing will not exist in fifteen years. (Yes, I'm being cagey, this is how people get doxxed, and not by you.) I have no business experience at all. Some people are naturals but I used to be notoriously gullible until I just started distrusting everyone. And I frankly don't think as an old bachelor I'm getting involved in the dating game myself or by proxy, everyone's creep factor would be way too high.
On reflection, probably easier to just leave it all to the Prostate Cancer Foundation as a last middle finger to feminism.
But thank you for thinking about it so much!
"On reflection, probably easier to just leave it all to the Prostate Cancer Foundation as a last middle finger to feminism."
Oh dear I can't help myself. Really you are going to be giving a lot of index fingers to male assholes. :^)
It was a good joke, though most of that stuff comes downstream from the rectal exam with the finger.
Find someone you know and help them. Institutional charity, including EA ,is nothing but a waste of resources. All value is ultimately arbitraged away by bad actors (where "bad" is defined here as anything that doesn't have the same terminal goal that you do). The only effective help comes via engaged human relationships. Everything else can be gamed.
Yup. This is one of the reasons I stopped exploring EA. The amount of corruption I witnessed in the developing world as part of an NGO mission was staggering, and I place absolutely no trust in any statistics generated as a result.
Why "the male population" specifically?
> And several people point out that the Frankfurt School as an intellectual lineage (as well as a physical campus in Frankfurt where postgraduate studies within that lineage take place) has some surprising alumni. Adorno’s most famous student was Jurgen Habermas, and Habermas’ famous students include Palantir founder Alex Karp and right-libertarian Hans Herman-Hoppe.
Not that surprising. The Frankfurt school were quasi religious ideologues - I thought the essay was good on that. Not just the Frankfurt school though. Marx’s description of what his day would look like after communism was that he would be able to:
“to hunt in the morning, fish in the afternoon, rear cattle in the evening, criticise after dinner”
Although this seems like an economic system for dilettantes, unlikely given the productive capacity of the 19C, and (cynically) the pathway from nationalising the factories to this isn’t clear; you can see why this type of philosophy might appeal to AI optimists and enthusiasts. Not Marxism itself but that kind of general utopia.
And it seems that Habermas was post Marxist anyway.
You get this a lot with people trained in Marxism - the neo-conservatives are similar. Having abandoned Marxism they can keep capitalism. Not just keep it - fetish it.
While condemning it as exploitative, Marx definitely praised capitalism as a world changing historical system - not just something local that appeared on a wet island - and as the penultimate system, with such vast productive and world changing technologies that it would lead to communism.
So if abandon the idea of communism as the end game then why wouldn’t you support the system that must be, if it isn’t the penultimate, then the ultimate system for mankind.
This can be libertarianism tooth and claw, or it can be just acceptance of modern liberal democratic capitalism as the end of history. And it becomes evangelical. Even imperialistic.
If capitalism is the final system for mankind, it must thus must be promoted against the reactionaries of nationalism and religious reaction. You can see why a lifetime Marxist like Christopher Hitchens would support the Iraq war against reactionary Islamism or Baathism, and why being taught by Habermas isn’t a negative but a positive when looking for funding from the CIA.
> Having abandoned Marxism they can keep capitalism. Not just keep it - fetish it.
But only give it "two cheers" rather than three.
Are there any good Substack posts about Derrida? The commenter blamed him for the long March through the Institutions, and while the name sounds familiar, I can’t remember any facts about hi.
On Derrida, one might attempt to adapt Palmerston's response when asked about the Schleswig-Holstein question. His reported answer was "Only three people have ever really understood the Schleswig-Holstein business—the Prince Consort, who is dead—a German professor, who has gone mad—and I, who have forgotten all about it."
I leave it to you to adapt Lord Palmerston's comment to Monsieur Jacques Derrida.
I don't know of any Derrida Substack posts, but when it comes to philosophy, my standard go-to for learning is SEP. And as expected, it has an article on Derrida.
https://plato.stanford.edu/entries/derrida/
New, innovative philosophy for why we should not end it all!
https://natehanby.substack.com/p/dualism-and-the-fatal-intuition-why
Our dualist minds are NOT actually prisoners of this evil material flesh of suffering!! Despite how it may appear. I found this line of thinking very persuasive. YMMV.
Given that you invoke an angelic race that never knows suffering, let me throw this at you from C.S. Lewis' "The Pilgrim's Regress":
‘Is it not rather harsh of the Landlord to make them suffer at all?’
‘I can answer that only by hearsay,’ returned the Guide, ‘for pain is a secret which he has shared with your race and not with mine; and you would find it as hard to explain suffering to me as I should find it to reveal to you the secrets of the Mountain people. But those who know best say this, that any liberal man would choose the pain of this desire, even for ever, rather than the peace of feeling it no longer: and that though the best thing is to have, the next best is to want, and the worst of all is not to want.’
‘I see that,’ said John, ‘Even the wanting, though it is pain too, is more precious than anything else we experience.’
...The Angel sings
‘I know not, I,
What the men together say,
How lovers, lovers die
And youth passes away.
‘Cannot understand
Love that mortal bears
For native, native land —
All lands are theirs.
‘Why at grave they grieve
For one voice and face,
And not, and not receive
Another in its place.
‘I, above the cone
Of the circling night
Flying, never have known
More or lesser light.
‘Sorrow it is they call
This cup: whence my lip,
Woe’s me, never in all
My endless days must sip.’
Cool poem! C.S. Lewis definitely has a way with words. Could you ... do me a favor and add your reply to my substack article?
Although, I don't agree with there being a connection between suffering and wanting. I think you can still want things without suffering for them. At least in my experience, YMMV.
Thank you for the request, I'll copy that over.
What are the odds that GPU prices will never come down again because AI will prove to be so important and profitable that supply will never catch up to demand?
My friend believes in this scenario and says we'll keep frantically building data centers without end.
Too broad and vague. What do you actually mean with "never"? Think about an approximate timeline and some probabilities. Make it concrete. Then bet your money on it (or: if you are not willing to do so, accept that you actually don't believe it yourself and should re-assess?)
Law of supply. If the price remains high for long enough, investments will be made in manufacturing capacity and this'll bring the price down. (We haven't already seen this because it takes years, but if you're positing that the current situation lasts indefinitely...)
Nvidia are betting the house (it would seem) on everyone having an AI in their own personal computer or laptop.
Personally, I think this sounds like hell - you won't even need to click on an app, just tell the AI to do it for you (so I get Claude or Copilot or Slopula to type out my comments for me, presumably based on voice transcription, rather than with my own fair hands? Worse yet, without even my own fair brain, because Slopula will do all my thinking and writing for me?)
https://nvidianews.nvidia.com/news/nvidia-microsoft-windows-pcs-agents-rtx-spark
“The PC is being reinvented,” said Jensen Huang, founder and CEO of NVIDIA. “For forty years, you launched apps. Click. Type. With RTX Spark and Microsoft Windows, you ask — and the PC does the work. RTX Spark brings everything NVIDIA has built — CUDA, RTX, our AI platform — into a single superchip. Local agents. Frontier models. Creative workflows. RTX games. All on a laptop. This is the new PC. The personal AI computer.”
Look at history. All exponential growth eventually saturates.
Within the next ~5 years, scaling laws break and AI capabilities flatline around the peak of human abilities across all domains.
Non physically demanding jobs are gradually automated (over 5 to 10 years). Advances in robotics coupled with AI means physical jobs follow suit after some delay.
Upwards of 80% of work is automated. People still prefer having humans in the loop for oversight, particularly in government, leadership, and management roles. These roles pay very well and are highly selective.
The transition to this state is unpopular and volatile. Governments act on this by implementing some form of UBI primarily funded by taxing AI companies. Justification given: as AI is the product of all past human work/knowledge, humans deserve compensation for that contribution.
AI labour significantly lowers costs of nearly all consumer goods and services. UBI is sufficient for most people to live and pursue their own hobbies/interests.
Pessimism permeates the current generations post transition. Gradually, culture shifts to people deriving meaning/purpose in craft (as is more common in Japan) rather than from career progression or wealth acquisition.
This is the possible future that helps me sleep at night. Any other positive timelines worth dreaming about?
Culture?
Social norms. What North American society deems admirable.
You asked about any other possible timelines worth dreaming about. I meant to propose the possible future described by the Culture novels (by Iain Banks) as one such.
No material scarcity, benevolent superintelligent AIs overseeing society, longevity, body modifications, galaxy wide civilization, lives consiting of sport, games, art, relationships, hedonism.
Ah my mistake, I thought it was a quip about North America lacking a culture.
I consider both my own scenario, and the Culture's universe to be mildly dystopic. Maybe there is no AI future that doesn't look that way from this side of the timeline.
Nah, it's my fault, I wrote only one word.
I can see that some aspects of the Culture are negatives (though I'm curious which ones are for you), but it's still a much better future than yours because of immortality.
That is unlikely to be possible with merely peak human level researchers.
"Everyone dies, but slowly" futures are, of course, still much better than the "Everyone dies, immediately" futures, but they are still unacceptable.
Agreed. I consider Banks's Culture to be the best at-least-sort-of-possible outcome. ( Considerably better than AI-stagnates-at-human-levels, which seems like it would be a really surprising coincidence anyway... )
The USA would rather tear itself apart before introducing UBI; someone would whisper "Socialism" and the idea would instantly become untouchable for at least half the population. There is a reason why all the ultra-individualistic technologies and business models all originate there: because individualism is the bedrock of American society, far beyond the point of self-harm. To believe that the most individualistic technology of all, "AI", would magically lead to an inversion of that foundation is delusional.
Do you believe, in the event that unemployment passes great depression levels, that the US population still would vehemently resist more broad social programs?
The population itself wouldn't meaningfully "resist" such change in any case, because people are generally not informed and/or coordinated enough to enact change on their own; hence Government. As long as the Powers That Be try to do something that looks vaguely like a potential solution to the real problems, people are likely to go along with it and adapt to the new realities. Yes, exceptions exist [1], but they don't apply to the US; the days of violent revolution in the US are all but over.
Who would actively resist the most would be the people in power who stand to lose the most, which is people in power of politics and the economy. They would use their powers to shape or distract mainstream opinion, and deflect to frivolous topics like identity politics and the perennial favourite, immigration; anything but the pressing issues themselves, anything that could point back at them. Thanks to the new tools of communication, they have become quite proficient at it.
[1] Like in France, where the population will happily threaten the next French Revolution over e.g. plans to nullify public holidays or increase the age of retirement.
Okay, first of all, did you just say in one breath that the US is individualistic beyond the point of self-harm and that the US population would unthinkingly go along with whatever the government and Powers That Be want? That's...an interesting combination that may be somehow coherent, but I'm not seeing how. I have this sense that you're just reaching for a list of negative cliches about American voters without even caring much if they're consistent with one another, but maybe there's another explanation?
Second, are you genuinely claiming that the US is less volatile, less violent and more stable than France? Or was that irony?
Third, I'm really getting sick to death of these sorts of comments that make a long list of extremely confident claims and assertions about what will "obviously" happen, without a shred of hard evidence or substantive argument for any of them. Re-read your two comments here, they're nothing but a series of assertions. I really expect better on ACX, and you have the rest of the internet for this tribalism.
> that the US population would unthinkingly go along
First, feel free to disagree with me, but do try to quote me correctly. I wrote: "something that looks vaguely like a potential solution to the real problems". That meant to imply that people do look at it and that they do not go along unthinkingly with everything, but that they have a base level of trust in their government, especially if they got to elect it themselves.
Second, I don't believe there's a contradiction between these statements in any case. People can be cooperators with the government, but defectors to each other.
>Second, are you genuinely claiming that the US is less volatile, less violent and more stable than France? Or was that irony?
Not ironic, no.
Less volatile? I don't know, depends on your definition. I'm afraid you'll put me on blast again for not providing a list that quantifies exactly how unpredictable, but sure, Trump II has made the USA pretty much unpredictable on the world stage. In that sense it's much more volatile than most countries right now.
Less violent? Again, depends. Americans are more violent against each other, absolutely. I hope that I can take easily googlable numbers like homicide rates per capita as a given. But against the government? I don't know. When was the last time the US had something like the following?
https://en.wikipedia.org/wiki/Yellow_vests_protests
More stable? Currently not, no. For decades, the US has been toppling governments it considered inconvenient, and on J6, that long-standing policy has come home to roost. The US are in serious danger of democratic backsliding or an outright oligarchy, depending on how much of a fight her democratic institutions can and want to put up. Europe in general, including France, are also under threat of right-wing, anti-democratic takeover, but nowhere near at the same level as the US.
>Third, I'm really getting sick to death of these sorts of comments that make a long list of extremely confident claims and assertions about what will "obviously" happen
Again a misquote. I did not write "obviously" anywhere, so why do you put quotes around it?
>without a shred of hard evidence or substantive argument for any of them.
Yes, everything I write is my opinion. Do I need to add a disclaimer to every one of my comments that yes, this is indeed only my opinion, it is neither legal nor financial advice? These two comments were not papers for peer review, with facts or citations. They were opinion pieces on how I see America and, by extension, Americans from the outside. Deal with it or ignore me. And if you're going to provide evidence of how you disagree with me by quoting me, at least get the quoting part right.
> Trump II has made the USA pretty much unpredictable on the world stage.
Some things are more predictable than others.
For example, declaring a war on EU is already kinda inside the Overton window, but anything resembling socialism is and remains safely outside.
I'm struggling to find any coherent overall argument in your comments. Again, it seems like you're stringing together unrelated tropes just to express your hatred for Americans, without caring if they make sense. Pick a position: you can either claim that Americans are overly obedient and trusting of authority, "one nation controlled by the media", OR you can claim that they're overly individualistic and reflexively oppose any government program as "socialism". But trying to make both these almost diametrically opposite claims at once, without *much* more effort than you're providing on explaining how they somehow reconcile, doesn't just discredit one of them, it discredits both. Since it's clear they're just rationalisations for your negative emotions towards a particular nation. (Note how the original comment didn't mention the US; you jumped in to attack the US with arguments (e.g. lack of civically engaged population) that would apply as much and indeed much more to many other countries, which really only makes sense if you're motivated by nationalist animus...pretty ironically for a German).
>Second, I don't believe there's a contradiction between these statements in any case. People can be cooperators with the government, but defectors to each other.
Um...what? This started with you claiming that Americans would never accept UBI, which is in every form I've seen discussed a *government* program, because it looks like socialism, which refers to an ideology of government action. And now you say that Americans are actually "cooperators with the government" and "have a base level of trust in their government" and...were you intending to refer to some very unusual meaning of "UBI" that is actually a form of private charity? And using some strange definition of socialism that has nothing to do with the government? Your argument here is just a complete non-sequitur.
>But against the government? I don't know. When was the last time the US had something like the following?
>For decades, the US has been toppling governments it considered inconvenient, and on J6, that long-standing policy has come home to roost. The US are in serious danger of democratic backsliding
Again, I cannot follow what you're even arguing at this point. Are you saying that violence against a democratically elected government is a good thing? A bad thing? A good thing only if it's in favour of your own political opinions? You seem to criticise the US for *not* having violent protests on the level of France, suggesting it makes them more obedient to "people in power of politics and the economy" and say that "they have a base level of trust in their government, especially if they got to elect it themselves"...and then you say they're "in serious danger of democratic backsliding" while "Europe including France" is "nowhere near at the same level".
You outright state that Americans have trust in their government because they elected it, that they don't violently oppose their democratic government like France, and that they're in much greater danger than France of democratic backsliding, and of toppling their government through violent events like J6. I...just can't keep track of the non-sequiturs!
>Again a misquote. I did not write "obviously" anywhere, so why do you put quotes around it?
Yes all right I shouldn't have used quotes, that was confusing. I didn't mean to attribute the word to you, I meant that the genre of posts I'm sick of, which yours belongs to, consist of unsupported assertions phrased *as if* the writer thinks they're obvious, whether or not they use the word. (I used quotes to indicate that they're not obvious at all, but I see it was misleading.)
>Yes, everything I write is my opinion. Do I need to add a disclaimer to every one of my comments that yes, this is indeed only my opinion, it is neither legal nor financial advice? These two comments were not papers for peer review, with facts or citations.
You're commenting in a community that has strong norms of following evidence and "making beliefs pay rent", to the extent of often insisting that people put actual real money on their predictions if they want to be taken seriously. To make a post like this here, filled with bare assertions about extremely controversial claims, phrased with such confidence, is utterly against the spirit and the standards of this place. If you're going to make bold claims and predictions with no evidence, with no deductive logical argument, and with no offer to bet on them in some way, the *least* you can do is hedge a bit with language like "it seems to me", "I could be wrong", "I can't prove this but I have a feeling" etc, instead of just outright "this will happen" unearned confidence. If you don't want to do any of those things, I reiterate that you have the *rest of the fucking internet* for that.
Oh, and also
>When was the last time the US had something like the following?
https://en.wikipedia.org/wiki/Yellow_vests_protests
Uh...2020? I don't know if you're doing some strict comparison of numbers involved, but this is such a weird question when there were such famous and massive riots and protests so recently, that then spread to other countries as well.
Although I have no idea at this point if this helps or hurts your argument, because as discussed I have no idea what your argument on this even is.
We already have UBI for people over 65, all we need is for the 20-somethings to vote as much as them and we can get it for the genpop.
Population pyramid says no. Earth is an old people world now.
There are fewer 65-69 than 20-24 year olds, and the completed cohort fertility rate has been going up in the USA since 1954. Old people die.
What you said is just true, but there's fewer 0-4 year olds than 65-69 year olds, which is a problem.
https://www.populationpyramid.net/united-states-of-america/2026/
In fact there's fewer 0-4 year olds than any other age group up to 70.
Well, the 0-4 year olds definitely aren't going to vote.
e: Also, there will be a bunch of dead 45-49 year olds by the time they get to be 65-69, and by my quick math using actuarial life tables, the 20-24yo in 20ys will be about the same number as the 65-69yo, excluding any immigration (which will tend to be of younger demographics).
Relax anti-monopoly laws for corporations willing to distribute big chunks of whatever they, their subsidiaries and partners produce to employees, their families, and those hitting retirement age while in company service. Perhaps permit internal markets and payments in company scrip. We end up with giant Zaibatsu-style megacorps who have to look after the needs of all the humans involved with them. Since the population apparently finds the idea of the actual democratically elected government doing anything to actually help people utterly intolerable, we can at least get some of the effect at the cost of turning the country into several unelected de facto kings and their serfs. It’s not socialism if we don’t call it government and the board is making a profit! At this point the remaining vestigial parts of democracy can safely wither away. Gotta be better than just giving up and starving to death.
Many problems with this, not least that AI companies are taxable only where there are AI companies. So, the US, China and maybe a few others.
Whether there is enough income to tax there is another question, what with the rest of the economy collapsing.
The only possible way I can see UBI working is by just dropping money into people’s accounts based on the assumed productivity of that particular economy.
Companies are taxable wherever they do business.
Laws have changed recently but the majority of corporate tax for Internet companies is not taxed where profits are made but where there are taxable structures.
I don't think you would need to tax AI companies, rather I think you would need to tax companies that own and use AI agents for production, I'll dub these companies AI clients.
I imagine it would work something like this:
- AI companies, probably mostly based in US, China etc. due to high barrier to entry, sells AI products such as AI agents to other companies. These AI agents will displace human labor when their productivity/cost ratio is lower than the available human labor. If there are sufficient free competition between AI companies, then the cost of an AI agent should approach the cost of providing that agent plus some margin.
- AI client companies will then increase their margins because they have lowered their production costs. If there is sufficient competition this will cause cost of goods to go down over time.
- Displacement of human labor will lead to unemployment if we assume that AI agents are better at most or all jobs. Unemployed workers will not have an income and will not be able to pay for goods, even if they are getting cheaper due to AI. This will in turn lower the demand for goods and services provided by AI client companies.
- Governments may then increase taxation of the AI client companies, for some part of the profit, and redistribute that money, e.g. in the form of UBI. Note that, hypothetically, if the tax is exactly equal to the increased profit margin of AI clients due to AI agents, then the cost of goods and services does not go down, and the AI client companies have the same profit margins compared to the status quo. The end result is then the same goods and services provided with less human input.
- It seems then that the UBI can be at least of the size of added margin for AI clients due to AI agents. If that margin is less than the salary the workers previously took home, then the workers would have less purchase power on average. So if the AI agents are only marginally better than human labor, that could lead to a situation where the AI client company now pays for compute instead of paying labor, leaving not enough taxable income for UBI - this seems bad. So it seems that the AI agents would need approximately 2x the production/cost ratio of a human worker, in order to provide a full compensation for the lost salary?
- If the AI companies manages to capture most of the profit then taxation and UBI becomes a problem for countries that don't have AI companies to tax. For this to not happen, there need to be adequate competition between AI companies.
I think it's pretty likely that AI companies will manage to capture most of the profit, because the industry has an extremely high (and constantly increasing) entry cost, the product is homogenous, and the consumers are numerous and therefore have low bargaining power. High entry cost leads to few players, which (with the other properties combined) then leads to cartel formation. Just look at current industries with these properties, such as RAM manufacturers who are constantly price-fixing.
An idle GPU generates no revenue, as does an empty airplane seat on takeoff.
High fixed costs, (comparatively) low variable costs, and "perishable" goods are a recipe for (actual, not Marxist) commoditization in which case prices approach marginal cost much more aggressively than in other industrial structures.
Indeed, without coordination these factors lead to almost zero profit. This is exactly what creates the strong incentive for breaking the law. Would you choose that your company is constantly cost-cutting and on the brink of death over sending a few messages to some guys you already know?
Cartels are generally difficult to hold together without legal force (cf. the thankfully-defunct Civil Aeronautics Board in the US) due to the incentives to cheat, and perishability turns that up to 11.
"This is the possible future that helps me sleep at night. Any other positive timelines worth dreaming about?"
Let's go with the obvious one. Scaling doesn't slow down and we don't run into significant alignment challenges. Smart AIs read and understand all human literature, decide morality is important, and then start acting morally.
We all live forever, in infinite plenty, as godly AI utility-maximizers consume everything in the universe and turn it into awesome-ness. Rather than paperclips, it turns out the greatest utility in the universe is playing ball with puppies, so we convert the entire universe into puppies and people to play with puppies and also lawns and balls with which to play with puppies and everything is ultimately good forever, because puppies are better than paperclips.
And that's kind of joking but, also, you're not just looking a gift horse in the mouth, you're kicking him in the dick. Like, you've laid out a scenario of virtually infinite physical and intellectual labor, where everyone is free to do anything, and you describe this a depressing, pessimistic hellscape and, like, I hate the fact that I kind of get that but you can just get off the internet and go do awesome stuff.
Things be pretty baller, fr fr as the children would say.
I didn't mean to imply the future I laid out was a hellscape - I'm personally pretty happy with the idea of an existence pursuing hobbies and leisure. I just think that is a minority opinion, and the road there is likely rocky.
Looking at the current landscape of research/incentives/capitalistic pressures, I am assuming that a full blown AGI/ASI timeline would not achieve alignment, ie doomed humanity timeline.
I think it's more likely that AI continues to progress at the the domains that is it well suited to, verifiable closed ended text based tasks like math and coding, while continuing to be terrible at multimodality, real time actions, continual learning, and open ended tasks that require subjective assessments and self critique.
I would predict that a millenium prize problem (or problem of similar caliber) gets solved at the same time that the best robots still fail the coffee test (whether a robot can walk into a random house and make a cup of coffee). Or for a computer based task example - the best AI would still be unable to beat a new video game (one not in the training data) of similar complexity as say Elden Ring.
In a nutshell: Maximal spikiness?
The AIs become very superhuman on tasks where outputs can be verified (as you said, math and coding - but also any others where outputs can be automatically verified) but not on tasks where verification is infeasible, where they stagnate at very subhuman levels. BTW, there is a _broad_ set of tasks called "inverse problems" where there is some f(x)=y, y is known, f is known, and x is wanted. E.g. for a pattern of radiated radio waves ('y') , pick the minimal antenna ('x') that generates ('f') it. These are all verifiable output problems, often 'design' problems of some sort.
In a world where human labour is cheap relative to the productivity of capital, then the owners of that capital will be able to afford human labour for things that they otherwise would not use it for.
In modern India or Kuwait, or western countries a hundred years ago, a family with a reasonable middle class wage will have multiple servants -- maids and nannies and gardeners come first, then valets and chauffeurs and butlers, then footmen. Some of these servants could easily be replaced by robots (chauffeurs) but others cannot (nannies).
In a world where the cost of human labour is cheap compared to the productivity of capital (robots) then I would expect the owners of that capital to start finding new ways to employ human labour to make their own lives easier. The future looks like Downton Abbey, you'll have a rich family with a staff of both humans and robots -- the human servants will do tasks where a human touch is necessary, and they'll instruct the robots on the tasks where they are not.
Why don’t we see rich westerners today employ modern India levels of staff? The segment of the US population whose income exceeds menials’ wages by a ratio comparable to that for a “modern Indian” “ family with a reasonable middle class wage” vs the servants they employ is smaller, but is far from zero.
I suggest there are also other factors at play here, such as cultural differences in the number of strangers people are willing to tolerate in their home.
It's a good question, I think it's just out of fashion. Rich or even middle class westerners have no problems with gardeners, or visiting (not live-in) maids, or even nannies or personal chefs, but will shy away from anything more feudal-sounding.
I briefly worked for a rich guy who had a butler, but he wasn't a "butler", he was just a "guy who takes care of my house".
I'm not that rich but if I ever am then I'm getting myself a gosh-darned butler.
Would you have him wear the traditional white gloves and a tuxedo with a long tail coat? On the one hand, I could see wanting to go the full trad route just for the novelty, but on the other hand, I wouldn't care to adhere to a similar dress code myself, and we'd probably make for an odd pair if I'm out on my deck in cargo shorts and a t-shirt and he's serving me iced tea in a tux. The juxtaposition would not make a great impression on the neighbors, I dare say.
If you dress your butler in a tuxedo with tails and white gloves during the daytime you've failed every class shibboleth that would qualify you to have a butler, sorry.
Also if you're rich and your neighbours have line of sight onto your deck then what are you even doing, man?
Guy doesn't even have a butler yet and already he's givin' me the high hand. You know, people like you are why the House of Lords is goin' away.
>Non physically demanding jobs are gradually automated (over 5 to 10 years). Advances in robotics coupled with AI means physical jobs follow suit after some delay.
I can’t imagine how many tokens it would take to make a good AI PLUMBER.
Yes, but will the AI workers be able to stand around looking into the hole in the ground? 🤣
https://www.reddit.com/r/ireland/comments/7qro7q/how_many_county_council_workers_does_it_take_to/
The equivalent would be agents suspending themselves until it is time for them to take action, and yes, they can in fact do that.
In this world, North Korea can build a 100% robot economy with 100% of production going to robots and become exponentially more powerful. Any country that attempts to preserve human quality of life or even human life will be left behind in the arms race, overrun, and then killed and replaced with more robots.
Yeah, this is the 'human organizations are _already_ misaligned (to human flourishing), and are only currently contained because they need humans as parts' scenario. Yup, mere plug-compatible-AGI, _without_ ASI, (but with human-comparable robotics) is enough to trigger this scenario.
My sense of the near future is that AI hits some boring constraints before it's able to generate full-length feature films, novels, AAA videogames, or any construction more complicated than a backyard shed. It's theoretically possible, but requires more electricity than they can pay for with the sales of such things. However, it's able to build three-minute shorts, 1000-word short stories, simple mobile games, and design bespoke furniture for home use. It's also able to build simple utility apps for controlling home appliances, but not build them. It cannot make coffee unless a human sets up the inputs and tells it what to do, after which it can do so reliably as long as the grounds are stocked. Same for most cooking - for each recipe, a human has to show it where everything is the first time, and every time someone remodels the kitchen.
Desk jobs are ~80% automated as you say. But this just means that a week of work circa 2020 AD is done by Monday, freeing up Tue-Fri to do even more. Every desk job becomes 3-5x more productive, justifying raises from the added value created.
This also means a great deal more tasks are cost-effective. Among them are business administration for what are now very small businesses where the owner stays up late to fill out paperwork for payroll, taxes, invoicing, looking into opening another franchise, and so on. Along comes an MBA who got pinkslipped from a large company, offering to handle all that for about $400/week, remotely. The trick is that that MBA is doing the same thing for ten companies that couldn't afford a PTE with an office and bennies. I expect similar opportunities for other occupations.
Governments briefly consider UBI until they notice that a $15K check to a third of a billion people is five trillion, pays only minimum wage, and would quickly just inflate prices so everyone's even poorer. UBI just prints money; it doesn't print wealth.
The AIs "print" wealth, but not all of it. They still can't run all heavy machinery, although they can handle a surprising amount of transportation (as long as someone keeps maintaining roads and ports).
AI can derive some novel results in science and engineering, but it also frees up researchers and designers to ask a _lot_ more questions. Suddenly people are asking how to solve problems they never had time to solve before. Squishy meat brains are still the most efficient for working on them, even with AI assistance, due to the electricity constraints.
The hardest hit are probably physical laborers who don't know how to work complex machinery. They probably change to bespoke artisanry - leatherworking, glassmaking, etc. - though AI gradually makes inroads on those, pushing them to advance in turn. But that's a minority of people. Most of us just get as wealthy in real terms compared to 2020 humanity as 2020 humanity became compared to 1620 humanity.
"Most of us just get as wealthy..."
I'm not sure I see how most of us would be getting more wealthy. A majority of existing jobs are already in the process of being automated.
As an example, the profession of driving a vehicle. In the USA there are over 3 million truck drivers and somewhere around 5 million taxi/delivery drivers. Even if AI progress halted today, it is hard to imagine us not cracking fully automated driving within the next ~5 years.
When these jobs are automated, these workers can't pivot to alternative professions like your MBA, and I don't imagine there are enough bespoke artisanry positions to go around.
That sector alone would more than double the current unemployment numbers.
If not UBI, what alternatives would prevent mass poverty and unrest?
The short answer: Jevons Paradox.
Longer: automation drives down the cost of whatever was automated, which expands the set of applications that are now cost-effective. In other words, when you have an automated fruit picker, people suddenly think of a bunch more things for that fruit picker to do. This drives up demand, which drives up the price of fruit picking, and the job was never just mechanically plucking at stuff. So the people who were mechanically plucking things suddenly have a lot more work to do, and can charge more because more stuff-plucking is getting done.
That stuff-plucking is an example of wealth. So is the service of having stuff where it can be used instead of where it was made, which is what truck driving does. If that service can be supplied for some electricity instead of a package of food and healthcare and waiting until certain hours of the day and insuring against a greater rate of accidents, that frees up whoever was driving trucks to produce something else valuable.
"these workers can't pivot to alternative professions" simply isn't borne out by empirical observation. When disruptive tech landed in the past, various professionals complained that their jobs would disappear. But people aren't insects; they know how to do multiple things. Often the job was partly the thing automated, partly something else; professionals just did the non-automated thing more, and got more done. Often there was a similar job that wasn't automated; they just did that instead. This shifted people doing those jobs, who also found similar jobs, and so on. Since people are often continually changing jobs anyway, this often even goes without notice.
Jevon's Paradox and past empirical observations are based on instances of narrowly scoped automation. What happens when the new applications are also now easily automated, and so is any other profession that a worker could pivot to?
What makes you think the new applications are now easily automated?
An LLM recently produced an answer to an Erdos problem. Now that it's answered, what happens next? Did that LLM arrange for peer review? Did it try to apply that answer to some downstream problem?
Suppose you ask an LLM to write a full stack for a website to alert people when a collectible they've been searching for is going on sale somewhere, and the LLM writes it. Now what? Does the LLM publish that site? Register a domain name? Try to monetize it? Improve it? Try to apply that stack to another problem?
Have you tried to set an LLM to writing a novella? It can write a short story that's a few paragraphs long, but last I checked (could be wrong, stuff happens fast), it couldn't handle writing, say, 25 pages. One-page stories are automated; why is a novella not just doing the same thing 24 more times?
Having seen large amounts of AI-generated content, the strong impression I get is that it can't make anything beyond a certain size (depending on the media type), because the amount of compute required scales faster than the size of the content. The LLM quickly hits this limit, and either stops gracefully or starts "forgetting" stuff at the beginning, resulting in unusual loops or content that drifts without staying consistent with its origin.
Apply this to, say, a "butler bot", and you'll wind up with something that picks up your dirty clothes and launders them, but forgets it needs to put them in the dryer. Or orders the eggs and cream you wanted for that recipe, creates a delicious souffle, then lets the leftover cream spoil. Or runs in a constant loop of cleaning and folding clothes, cooking food, cleaning the bathroom, vacuuming the rugs, feeding the pets, and ordering more supplies, but never suggesting you might want to update your fashion or remodel the living room or build an addition since you keep complaining about where to put your 3D printer.
I've seen no evidence that LLMs are capable of improving themselves on their own initiative. They only do what the humans say. And they can't do more than a few hours at a time without checking back (or risking being wildly wrong).
They might get better at all this. I even consider that likely. You can ask what we'll do then. I'll respond: well, what do we do when the people around us get better at things? We'll do the same thing. Contrary to popular belief, I believe AIs won't improve any faster than people, due to real physical constraints like electrical capacity and chip shortages.
"The AIs "print" wealth, but not all of it. They still can't run all heavy machinery, although they can handle a surprising amount of transportation (as long as someone keeps maintaining roads and ports)."
This sounds like a strange sort of spikiness in AI capabilities. It might well be that extending LLMs to handle 3D operation is a lot more difficult than what scaling laws have been doing to knowledge work with verifiable outputs. But having the AIs able to handle e.g. building a shed, but _not_ able to be organized to build any building - or maintain a road, seems like an improbable combination.
I picked a shed because it lacks many of the features we would want in a more complicated building. A basic shed is four walls and a roof. No foundation, no electrical wiring, no plumbing, no air conditioning, no insulation, no windows, and maybe not even a door.
As I see it, AIs can operate cars because self-driving was an early job we poured billions of dollars of professional time into, for historical reasons. So there's the spike. If Google had instead gotten into a big row about, say, chicken farming, we'd have AI chicken farming and no self driving.
Many Thanks! I agree that a shed is indeed limited in the way you describe. I'm still skeptical that AI would be spiky in a way that allows construction of a shed but not of a house. Nailing two pieces of wood together with the modest tolerances required for either a shed or part of a house requires 3D navigation and reasoning ("Is this piece properly supported while I nail it?") that I expect to carry across to essentially all the other tasks you list.
I'm _not_ saying that these tasks are easy, or that current AIs or current robotics are capable of building either a shed or of building a house. But I do expect that if robotics of 20[23]N are capable of fastening together all the pieces to build a shed, I do expect them to also be able to install wiring, plumbing, HVAC, etc. A great deal of all of these tasks are 3D world models interfacing to vision ("Do I have the right part? Are the surrounding parts properly in place? Do they match the blueprints/CAD data to proper tolerances? Can I position the next part without collisions?") and to force sensing and application. A lot of the specialized knowledge _beyond_ 3D world models, e.g. knowing to strip the insulation off a wire before fastening it to a terminal, are already in the LLMs.
NYer here - on the subject of Alex Bores' election, if you care about things other than just who will regulate AI the most, it's not so clear that he's is the best choice. It's a three way race between him, Micah Lasher, and Jack Schlossberg. (As I write this, Manifold gives Bores, Lasher and Schlossberg 38%, 51% and 7% to win, respectively)
Bores and Lasher are both abundance liberals with similar legislative experience, while Schlossberg is a slopulist whose support primarily comes from social media. For example, on the subject of high housing costs, Bores and Lasher both strongly support building more housing, while Schlossberg's solution is to let you deduct your rent from your taxes. Overall, if you share the same ideology that Scott has described in his California ballot posts, Bores and Lasher are both good choices, and it's not clear IMO that Bores obviously deserves to win over Lasher.
(Edited to add: a reason I am considering voting for Lasher over Bores is that his AI regulation stance, while initially quite reasonable, appears to have gone somewhat overboard now that it's campaign defining thing, e.g. he's been negative on Waymo in a way that Lasher hasn't been)
When in congress, Lasher is likely to be less willing to take on OpenAI et al, as their money would be part of why he won (tearing down Bores helps him). Plus, I worry about the message that Bores coming up short would send to the rest of the country; they might see someone who promised to regulate AI getting a ton of money spent against him and failing and back off.
Bores also has backed stuff like a Land Value Tax in the state assembly, and has shown a ton of discernment in what bills he’s backed in the state assembly, to the point of bucking the standard dem position on bad regulatory requirements, see https://www.politico.com/news/2026/06/01/ny-12s-ai-guy-hasnt-always-voted-in-favor-of-tech-guardrails-00944305. Despite his campaign rhetoric, I think he’d show similar judgement in congress.
I think that's backwards, in that of course the OpenAI money is part of what elevated Bores so much, and made his campaign high-profile. The PAC money has hurt Lasher, not helped him.
Lasher is also very willing to buck progressive orthodoxy, as in this post from his substack on crime and the NYPD: https://substack.micahlasher.com/p/saving-the-new-york-project-from. See also https://substack.micahlasher.com/p/is-there-a-tisch-effect and https://substack.micahlasher.com/p/three-more-weeks-of-the-tisch-effect on the effect of appointing Jessica Tisch.
>”Bores and Lasher are both abundance liberals”
Um… what?
I thought Bores’s economic policies were bad, but then I looked at Lasher’s website, and all of his “affordability” policies are demand subsidies and price controls.
What does “abundance” even mean at this point?
I'm basing this on a candidate forum I went to that was organized by some abundance-aligned NY organizations (I haven't read their websites). Bores and Lasher were the two candidates of the four present that were head-and-shoulders above the others in terms of abundance policies, although it's possible that they were adjusting their answers based on the audience.
Yikes, as a strong abundance liberal and lifelong urbanist [not a NYer] those stances would firmly disqualify Lasher from my consideration.
Not a New Yorker?
1060 W Addison Street?
Take that out of your mouth...South Sider born and raised, now residing in the South Loop.
:)
Lasher is actually very abundance-pilled, and has been for years at this point. Abundant New York supported him in his 2024 campaign, and even when I personally have heard him speak a few times, he actually tells people what they *don't* want to hear because he supports more building, despite people in the audience offering the standard NIMBY complaints. The concession to being a real politician is that he _also_ supports demand subsidies, but the tell is that he always talks about building more, even when it's clear the audience only wants to hear about price controls.
2024 NYT article: https://www.nytimes.com/2024/03/25/nyregion/pro-development-housing-super-pac.html
Abundance endorsed both Lasher and Bores in this election.
Everyone supports more building in the abstract. The question is whether or not he supports policies conducive to building actually taking place, and whether it will be funded in a way that grows the economy or contracts it.
I went to Lasher’s NY Assembly page, and the first piece of sponsored legislation I click on is a bill subjecting Waymo to regulation by the New York City taxi and limousine commission. https://assembly.state.ny.us/leg/?bn=A00793&leg_video=
Seems bad
I'm rapidly coming to the conclusion that "abundance" doesn't really mean anything anymore, it's just a label that is being co-opted by politicians and wannabes to sound like they're au fait with the latest cool trend.
Yes, I realise that sounds like the pipeline from "new concept" to "hackneyed phrase" was enormously accelerated, given that abundance was only bruited about very recently, but hey these are the days of AI, and once mainstream political parties start murmuring gently about how Candidate A is soooo much more abundant than stuffy old Candidate B from the other lot, you know it's toothless.
At this stage, if ever I had a handle on what abundance was supposed to be all about, to me it means no more than "I/we want people to have nice things! and be nice! and everything is nice! so vote for me/us and niceness!" as the kind of boilerplate campaign promise regularly trotted out, just given a new coat of paint with the latest shiny term.
I usually think of it as an attempt to push back on various hairshirt-bearing parts of the left--the "you should have less to help the environment," "you should have less so we can ensure favored groups get more," "you shouldn't want the things you want but instead these more virtuous things we want you to prefer," "you should have less so we don't have to do unpleasant things that look bad on video," and so on.
It's a bit like anti-wokism, which seemed like it meant something when the whole world was mouthing those slogans, but turned out to have no coherence beyond "we don't really think there are 37 genders or that using the word 'blacklist' means you're a secret racist."
I actually very much agree, and in fact will be voting for Lasher over Bores. Partly for the reasons you describe, and partly because Lasher is more of a policy guy. Beyond standard policy wonks, what I like about him is that he's good both in terms of delivering *outcomes*, and has also been thinking about systemic reconstruction to build back from the current administration's violation of policies. See his substack, Into the Weeds. (https://substack.micahlasher.com/)
My mother lives in the district in question, so I've seen the actual ballot, and two points:
1. The alternative language on that ballot is Chinese, which is surprising, and no, Chinatown is not in the district.
2. Schlossberg lists his name as "John Kennedy Schlossberg", which you may think is a bit much since AFAIK he doesn't generally go by that, but it could have been even-more-so, as his legal name is John Bouvier Kennedy Schlossberg.
"The alternative language on that ballot is Chinese, which is surprising, and no, Chinatown is not in the district."
Aimed at property owners, given the suggestions that Chinese money (amongst other foreign investment money) is flooding into New York/London/name big city of your choice and buying up all the new builds as investments then leaving them empty, thus yanking any chance of getting on to the property ladder away from people trying to rent/purchase their first home?
https://www.youtube.com/watch?v=layawTB7lHk&t=3s
It's based on census data and required by federal law. https://www.federalregister.gov/documents/2021/12/08/2021-26547/voting-rights-act-amendments-of-2006-determinations-under-section-203
Is Chinese the only alternative language?
China is the only other language of the ballot. To Deiseach’s point, I’d expect those languages to be determined by census info, and there just aren’t a ton of Asians in the neighborhood.
Yes, that is odd
According to Wikipedia, there is an Asian or yeah Chinese population of about 10.3% in Manhattan. So it’s not nothing but the only thing I can think of that makes sense is that the Chinese population would be the one most likely to not be able to understand English well at least in writing at least you could count on most of the Hispanic and black population being able to read English. Chinese is difficult that’s my guess.
https://www.nytimes.com/1994/08/25/nyregion/board-agrees-on-ballots-in-chinese.html?unlocked_article_code=1.nVA.hzbh.JSnNK-jiPa5P&smid=url-share
So it might just be an artefact of the law as per the linked article? They were compelled to provide Chinese translation on ballots, so now every ballot gets translated into Chinese because nobody ever revisited the subject to see if it still holds? Easier to keep doing it even if not needed than to stop and have somebody haul you to court for racism and disenfranchising a minority population?
I recorded three songs the other day and it was a lot of fun.
One is about the fall of Western civilisation and another is about getting lots of emails. The third has a picture of a knight which I painted after an eye injury, with the result that I still haven't actually seen what it looks like unblurred.
Have a listen if music is a thing you like:
https://m.youtube.com/@tobiascharlesapsley
To the relief of most but the disappointment of the kind of people who read this blog, there was absolutely zero AI used in any part of the process.
I'm someone who loves earnest filk, even when it disagrees with me politically (like this does). Some honest feedback:
Your songs don't scan well, lines have frequently changing emphasis and length. There's also little rhyming structure or other musical framework to it. It really feels more like you're just talking in a singing tone. Try to identify a structure you want to follow early on, and shape the song around it rather than deciding what you want to say and forcing it into a vaguely song like shape.
To hopefully soften that criticism, let me say "Congratulations! You failed! Most people don't even try!" Music is part of being human, and I can tell you care. You have a solid voice, and the courage to post something you've made for criticism. I hope you keep recording and sharing your music.
While I thank you for your attention and feedback, I am horrified that you thought it was filk.
Apologies for assigning a genre to you that you didn't intend. I've heard several filk songs that sound just like this so that's where my mind went.
Disagree with Unsaintly below; I caught the rhyming structure which was fine, musical framework which was rather obvious, etc.
Your voice has a rich quality, but you didn't always seem to control it well. When you didn't it was distracting, when you did it sounded professional.
The mix sounded good, clearly you know what you're doing or at least enough to be dangerous when it comes to the instruments and studio gear. Well done, wish I was that good. If you used samples I couldn't tell.
My only real criticism was that I found it unimaginative. The melodies had too little variation, the chord changes were predictable, the rhythms were simple to the point of tedium, etc. The lyrics to the two political songs were well written but seemed like they were trying to be edgy by virtue of espousing reactionary ideas. They didn't hint at any irony or inner-conflictedness or self-doubt or trying to see past the bend. They showed no evidence of being the product of fascination or curiosity or even pain. They even lacked the fun bravado I'd normally expect from such self-assuredness, like what I look for in rap.
Keep working on it, you've got a great talent.
> The mix sounded good, clearly you know what you're doing
Nope, can't take any credit there. The chap at the recording studio was great at everything he did. I was just a plod with a guitar.
I've got no musical knowledge, so this is all layman stuff.
There's something about the pitch of your voice here that makes me think "singing comedian" instead of "singer". I think that's what Unsaintly's talking about with "talking in a singing voice".
Typically that works to de-emphasize the music and emphasize the lyrics, but... I don't understand the lyrics. What is Nolmin Gnu actually PROMISING? Who's the woman in the middle of Delenda Est? (I listened to the knight song first, so by default she's Sally.) Why is Harry referred to in the same way as Sally if the speaker is Harry but not Sally? It all feels very loose.
I don't like the rhyming scheme, too many words rhymed with themselves ('Hated', 'Love', 'Fated"... 'Love' again). Some lines are shorter or longer than their counterparts; some lines match their counterparts by using empty words for padding. (I can't knock it too hard, I do that all the time.) The knight chorus has a full A-A-A-A, then degrades into A-B-B-A, and because the easier part comes second, it feels like a failure to maintain the standard. (If you reversed the patterns so they were growing more complex I think it would be fine.)
I listened around a little in all three. It's a strong effort.
I think the best-sounding part was the chorus of Delenda Est. Seems like that part plays to your strengths as a vocalist.
Marching and Ballad sound like they might be in an uncomfortably low register of your voice for the style of songs that they are? I'd be curious to hear what they would sound like if you took them up a few steps.
As for the lyrics, I think there are too many words. They didn't hold my attention. My advice would be to try to put your message across more succinctly and with more poetry, and to shift the focus from ideas to feelings.
Also– potential exercises for improvement:
Listen to “It’s Alright Ma” by Bob Dylan, and figure out why it works better than your songs
Record yourself singing a Depeche Mode song, one with Dave Gahan on lead vocals. Then compare yourself to Gahan and figure out what he’s doing that you’re not.
Someone posted this here a while back, and I gotta ask, is this how young conservatives talk IRL? Like they just casually shoot the breeze about white racial superiority?
https://alexanderturok.substack.com/p/that-time-i-went-to-the-white-nationalist
Evidently, yes.
https://www.richardhanania.com/p/the-based-ritual
I wouldn't believe the colour of daylight from Hanania (much too invested in bragging how he's so much better than these clods, these rubes, these Cousin Jonathans*) and Alexander, bless him, has his particular hobbyhorse about conservatism/right-wingers not being eugenicist enough and not in line with his views of the ideal society.
* https://en.wikipedia.org/wiki/Brother_Jonathan
Sneer club is just as boring when it's renamed to contrarian sneer club.
I dunno, everybody seems to hate Richard Hanania these days, left and right, so whoever was actually willing to hang out with him is probably not representative of much of anything.
Right at the top he admits that the scenario portrayed never actually happened. He was making the thing up to readers to give his own personal impressions of these folk. Bad form IMO.
"Note: I’ve known these people. And yes, it actually is this bad."
What are the job prospects for someone educated in biosecurity? Does anyone know? I'm a tech writer with a background in IT and a masters in Translational Pharmacology. Most tech writing jobs in the biomedical space want more experience with regulated environments than I have. Biosecurity sounds like a niche I'm well suited for, but I have no idea what looking for a "biosecurity related job" might look like.
Your job prospects are theoretically limitless if you remain willing to learn and try new things, especially if you keep your lifestyle affordable.
Thoughts. If what the average teacher says or coworker writes is not better than what an LLM does, this proves these thoughts do not come from a soul.
This means if there is any room for a soul, it is solely for brilliance, creativity, genius.
This would imply a small number of people have souls and the rest are philosophical zombies.
Now that would make for an interesting theology?
Define better. What the average human would write is very different from what a LLM would write. LLMs writes in a very predictable manner, as we can see because people who read a lot of LLM text can often recognize that something is written by a LLM. On the other hand, two humans would mostly not write the same thing given the same prompt. Human brains are unique, and therefore flawed in different ways.
True, but... it is not like midwits write like LLMs. It is that they would like to. That is, the LLM writes better kinds of midwit texts, with better formulation, more elegant word usage, polite and professional and so on. It is as if you would take a longish corporate email, and then some experienced - and fairly educated - manager would comb through it and fix it.
Here is a LinkedIn AI example:
"Hardcoding pauses into a transaction pipeline slows down the entire system without actually solving the root cause of the concurrency issue. Half a second is an eternity for a server, but it offers zero guarantee that the obstructing process has finished its work. It is the architectural equivalent of closing your eyes and hoping the traffic clears before you step into the street."
I am not sure the average midwit (in IT) regularly uses words like pipeline, root cause, concurrency. The people I work with generally do not. Notice "obstructing" instead of "blocking", an elegantly educated word. Then making the traffic analogy, this is impressively beyond midwit level, as it sounds like understanding the nature of the problem and not just "bo hoo this is not the pattern they taught us at school".
This is what the midwit aspires to, the reason they are not all writing something like this is because they are not as good as this LLM.
Personally, I think the content of what is written is much more important than the words used. Over time, I think I would much rather have conversations with "midwits" as you call them, than with a LLM. People are in general vastly more interesting than LLMs if you take care to hear what they say. I find most people have unique and interesting experiences and thoughts once you talk to them. A LLM is fun for a little while, then it gets predictable. Still LLMs can be a very useful tool.
Regarding prose - I am not as impressed as you by the use of an expansive vocabulary. I rather prefer short and concise and to the point. A real art is communicating as much as possible with as few and as simple words as possible. I think this comes from reading a lot of academic papers and reports where I rather suspect the authors are trying to hide their lack of knowledge by using as many difficult words as possible.
I've noticed that some of the smartest people I know tend to communicate with very simple language.
>Notice "obstructing" instead of "blocking", an elegantly educated word.
Considering they're talking about how the bosses should speed the processes up, using the longer word instead of the shorter one cuts against their argument. "Half a second is an eternity, but let's force you to spend as many half-seconds reading this email as possible."
Obstructing and blocking are not synonymous.
https://www.thesaurus.com/browse/obstruction
Just another example of the cheapening of language. I don't agree with the thesaurus, and if you think about the two terms, they are not synonymous. Obstruction and hindrance work as synonyms, but blockage is a whole different thing. The road is blocked. We will have to take a different route. There is an obstruction in the road. We will have to go around it. The less people care to make distinctions, the more synonyms there are, and it's not helpful. Still, you have a leg to stand on.
I actually like the use of "obstructing" here (I'm not a native speaker though). "Blocking" is a narrowly defined term when talking about processes running on a server, but concurrent processes can obstruct each other's program flow without technically blocking a shared resource.
E.g. when one process changes a file before the other one accesses it, it did not block the shared resource, but the second process might still read unexpected data.
Ah, well if it's a technical term that's different. But the argument still applies to "offers zero" instead of "doesn't".
If they don't understand the problem, you want small words so they can follow along. If they do understand the problem, you want small words to get the request over with. The only time you need to sound educated is when the audience doesn't give a shit about the problem and is judging you on personality.
Carlos, you are pulling our legs, yes? I'm inclined to the oul' sesquipedalianism myself, but using "obstructing" rather than "blocking", "impeding" or any verbiage of your choice from the thesaurus doesn't strike me as particularly elegant or educated.
In what world they would not sound educated? How many people know them or use them regularly? They clearly sound like words people use who read books, and serious books at that. Remember, the average corporate employee, the type I called midwit, have not read a book in ten years and a serious book never, that is my reference point, not people here. You people here are extremely elite outliers.
Whether they are elegant, that is context-dependent, in a conversation like this, such showing-off would not be, but showing off to that type in a blog post or corporate email? Certainly.
I admit, I have no familiarity with software engineering/STEM business types you describe, but if they are all grunting illiterates I must say I am surprised.
That "LinkedIn AI example" doesn't sound particularly elegant to me; it's slick and polished in the corporate-speak type of way, but it doesn't need a very high level of fluency in writing English to produce.
Or at least, I would not have thought so?
You are looking at this as someone who acquired English as a second, third or fourth language. For someone who grew up speaking English it’s not very sophisticated.
Congratulations, you have re-invented Gnosticism.
https://en.wikipedia.org/wiki/Gnosticism
"hylic – lowest order of the three types of humans. They cannot be saved, since their thinking is entirely material and incapable of understanding the gnosis.
psychic – "soulful", partially initiated. Matter-dwelling spirits
pneumatic – "spiritual", fully initiated, immaterial souls escaping the doom of the material world via gnosis."
https://www.newadvent.org/fathers/0103107.htm
"5. They conceive, then, of three kinds of men, spiritual, material, and animal, represented by Cain, Abel, and Seth. These three natures are no longer found in one person, but constitute various kinds [of men]. The material goes, as a matter of course, into corruption. The animal, if it make choice of the better part, finds repose in the intermediate place; but if the worse, it too shall pass into destruction. But they assert that the spiritual principles which have been sown by Achamoth, being disciplined and nourished here from that time until now in righteous souls (because when given forth by her they were yet but weak), at last attaining to perfection, shall be given as brides to the angels of the Saviour, while their animal souls of necessity rest for ever with the Demiurge in the intermediate place. And again subdividing the animal souls themselves, they say that some are by nature good, and others by nature evil. The good are those who become capable of receiving the [spiritual] seed; the evil by nature are those who are never able to receive that seed."
— St. Irenaeus of Lyons, 'Adversus Haereses', Book I, Chapter 7
Thank you for saving me the time of posting this myself. People have been finding ways to call each other NPCs for a long time.
As usual, the ancient philosophers were pointing towards something real, but they lacked the science needed to express it properly.
Everyone has a soul.
But some people's souls are at level of GPT 3, others are at level of GPT 4, or GPT 5, etc.
If you think you are not included in this hierarchy, the future will probably prove you wrong.
This should give some humility to those who are laughing at our GPT 3 level brethren.
On the theme of Continental philosophy:
IMO, many of the thinkers of the field have become so symbolically charged that it's hard to get good secondary information. Everyone wants to tell you what they say (on both sides), but "what they say" conveniently fits perfectly into some narrative.
Reading a few primary sources can give you a better idea of how they really operate. It can be daunting, because some Continentals are deeply involved in traditions with very specialized vocabulary (and can really sound like gibberish). But they almost all have SOME things you can read as a regular person.
I think if you're interested in the French post-structuralists, you should read a tiny bit by them before getting into secondary literature.
There's this short essay (20 pages) by Foucalt called "The Subject and Power" which is pretty easy to read and is a pretty good intro to the themes of his thought:
https://www2.kobe-u.ac.jp/~alexroni/IPD2018%20readings/IPD1%202018%20No.8/Foucault%20Subject%20and%20Power.pdf
Derrida is harder to introduce this way because much of his original thoughts come via commentary on specialist texts and use terms from phenomenology or structuralism. But there's this interview he did in 1994 (called The Deconstruction of Actuality, the interviewer chose that title, 14 pages) that I really like. He talks about Israel, Political Horseshoe Theory, Immigration and Xenophobia, etc. The way he interacts with the interviewer is a good example of how he approaches questions.
https://www.radicalphilosophyarchive.com/issue-files/rp68_interview_derrida.pdf
--------
Possibly surprising quotes:
Foucalt
> To put it bluntly, I would say that to begin the analysis with a "how" is to suggest that power as such does not exist.
> Power relations are rooted deep in the social nexus, not reconstituted "above" society as a supplementary structure whose radical effacement one could perhaps dream of. [...] A society without power relations can only be an abstraction.
Derrida
> In order to stand up to this double intimidation, the only responsible response is never to give up the task of distinguishing and analysing. And I would also say: never to give up on the Enlightenment
> And since we had better recognise, like everyone else, that we have no choice but to protect what we take to be our own body, then let us be spared all these pure souls who appeal to high principles and put on a high moral tone and start lecturing us on politics as soon as we propose to control immigration and asylum
Do morals drive laws or laws drive morals? Like was spanking children banned because cultural norms shifted and a lot of people understood it is not OK, or rather the laws were changed first and then a lot of people figured if the state says so, then it is not OK?
I mean it would be easy-lazy to say it works both ways, but it would be important to know a bit more than that.
One thing we have seen 100% that prohibiting alcohol, weed etc. did not change the social norm at all... in the West. In Muslim countries it did. So keep up the law for centuries and then people internalize it? Or substances - victimless crimes - are a special case?
I honestly think laws have an effect on morals. That is because if other people have similar minds than I do, I have these two categories as "law-abiding citizen, proud, respected" vs. "criminals, humbled, disrespected", and it is not only that I would not be in the second group, my self-image does not even really allow comtemplating doing things that - even when never caught - somehow associate me with the second. But many people do not have minds like me, like many smoke weed, while I think like "I will not go to some slum to buy it from people I do not respect". But imagine I would make rich, educated friends who smoke pot, I would feel it is now high-status. Many minds are not like my mine, but perhaps many are.
Morals of course have an obvious effect on laws. Even if a utilitarian would say this thing is not wrong, if many people find that thing viscerally repulsive, it will get banned. It happened with bestiality. Of course there was this bullshit argument "animals cannot consent", but of course animals do not NEED to consent, they do not have those kinds of rights, after all they do not consent to getting killed and eaten either. It was obvious bullshit, it was simply banned because it is disgusting.
> I mean it would be easy-lazy to say it works both ways, but it would be important to know a bit more than that.
I'll ping-pong it back to you: Isn't it a bit easy-lazy to ask an extremely generalized question like "Do morals drive laws or laws drive morals?" and dismiss a fittingly generalized answer like "it works both ways"?
I think the answer lies in investigating each specific case you want to question. Did "the prohibition" (on alcohol in the US) change morals? Apparently not. An opposite example does not readily come to mind, but I'm sure they exist.
Per your comment on minds,
> But many people do not have minds like me, like many smoke weed, while I think like "I will not go to some slum to buy it from people I do not respect".
To me this sounds like a naively bad comparison. Smoking weed does not equate going to a slum. You can also just grow it yourself or have a circle of friends and acquaintances of growers. Of course, for many city dwellers and beyond, buying random stuff on the street is a typical expression of convenience and wilful ignorance; not unlike buying meat from the supermarket while saying "I could never hurt a fly".
You are a dreamer, my friend. I too would like to outlaw all the disgusting practices of the meat industry, of the banking industry, of various offshoring practices that exploit people, animals, environments and whose products are yet legal to import, and many other things that are clearly disgusting when looked at in clear daylight.
>An opposite example does not readily come to mind, but I'm sure they exist.
Maybe not a question of morals, but at least adjacent: A few years back, the EU collectively banned incandescent light bulbs. There was much wailing and gnashing of the teeth, but nowadays? LEDs improved a lot on all metrics, new form factors became possible, nobody misses the little heaters any more.
Sometimes you need to lay a fire to get rid of the old-growth underbrush; to make room for something better.
Speaking of LEDs and fires and such things…
I’ve replaced the halogen bulbs on my desk spot lamps with LED’s with the same form factor.
Now I don’t have to keep an eye out for any wisps of smoke coming from my cat’s fur when he watches me work.
I got one better. I nearly burned down my parent's house once with a desktop lamp, in a freak accident involving a mattress and a loose mechanical contact.
That’s why you always want a cat nearby. Preferably a black one because they absorb heat the best. Very low smoke point. The flaxseed oil of domestic felines.
There's got to be a variant of the trolley problem in there. If you do nothing, the house of a random Internet stranger burns down. If you flip the switch, your cat's fur gets singed. What will you do?
I remember those desktop lamps, whose base was basically a big transformer converting voltage to 12V, and the bulb itself was supported by two metal arms that also doubled as uninsulated low-voltage wires. I found out that if you hold a piece of aluminium foil between the wires, it would burn up on contact while producing sparks.
This design was very common and I have no idea how it passed basic safety concepts for home electronics...
I have another set of spot lamps that used halogen bulbs but there is no equivalent form factor LED. I had to MacGyver the LED replacements into the lamps. It's fun to break out the old soldering iron now and then.
I agree with your verdict, but I'm not sure it fits the bill of a moral law that generates outcry if broken? People would certainly question someones intelligence if that person would still be using incandescent bulbs intentionally, but I do not see it having much of a moral dimension. Sure there is a connection to environmental impact, but I think it's too tentative.
Maybe as a german example, I just recalled that racing cars on public streets now faces charges of actual murder, not just manslaughter, if bystanders are killed as a result. That clearly fits the moral-outcry-category, and it was a quite surprising verdict by the court. I think public perception clearly influenced the result.
Laws are the application of force, and their effect can be compared to any other application of force. Do muggings drive morals about carrying money on your person? Do rapes drive morals about how to dress?
The most obvious way laws drive morals is when a bad law is passed, it drives people to want to tear the system down.
>Like was spanking children banned because cultural norms shifted and a lot of people understood it is not OK, or rather the laws were changed first and then a lot of people figured if the state says so, then it is not OK?
What laws? Spanking isn't illegal in any state. There are laws prohibiting the use of corporal punishment in public schools, but only in 33 states. You can't have cultural change driven by changing laws without changing the laws first, and they haven't changed at all.
As far as cultural norms shifting, the percent of Americans who believe spanking is an acceptable punishment went down from 84% in 1986 to about 70% today. Not that big a shift, really.
62 (depending on how you count) countries have completely banned corporeal punishment. The USA isn't the only place that counts.
> One thing we have seen 100% that prohibiting alcohol, weed etc. did not change the social norm at all... in the West. In Muslim countries it did. So keep up the law for centuries and then people internalize it? Or substances - victimless crimes - are a special case?
What we see is that bans are ineffective unless you actually enforce them. If you want to ban alcohol then you need to also be committed to enforcing that ban; the US has this thing where it likes to ban things but not actually bother to meaningfully enforce those bans, which is why your streets are overrun with methheads and illegal immigrants.
Compare to, say, chewing gum in Singapore. They banned it (specifically the sale or importation, not the chewing), then they enforced the ban with penalties sufficiently severe that it wasn't worth anyone's while to try to get around it.
I think the comparison is not entirely fair. Banning chewing gum accompanied with hefty fines for leaving it on the pavement is much easier to accept, and there is no "Chewing gum industrial complex" to lose large revenues and lobby against it.
Try banning rice cakes instead ;) or some other revered "vice".
It was a smart move in Singapore - ban something that is generally agreed as a nuisance in public, but not high-status or high-desirability enough to warrant resistance.
The most egregious moral sin, murder, has been outlawed universally and that's not going so well. But the situation is still wildly different between some countries, so the underlying cause cannot merely be punishment and enforcement, no?
Here's a recent tweet by Bores:
"@berniesanders is right: the future of humanity belongs in the hands of our workers, not the handful of AI oligarchs pursuing unlimited wealth and power at the expense of our job, our kids, our economy, our climate, and our democracy.
Here in NY, I delivered the strongest AI safety bill in the country. Now, I'm running for Congress to put Americans ahead of AI, and I look forward to partnering with Senator Sanders to get it done."
https://x.com/AlexBores/status/2061640384883696096
What "partnering" does he have in mind? The tweet quotes Sanders demanding a one-time 50% wealth tax on AI companies. That's postcolonial Africa s*** and people want to vote for him because he said some words about AI safety?
You know the old saying about "when fascism comes to America, it will be wrapped in the flag and carrying a cross?" The broader lesson is that terrible ideologies don't look like ugly cartoon villains dressed in black. They will be marketed to you. They will use your symbols and buzzwords and claim the legacy of your revered figures. Watch out!
> "when fascism comes to America, it will be wrapped in the flag and carrying a cross?"
So you have to redefine that quote to make it fit Sanders (who has been progressive left his entire political life), but you ignore the rather literally flag-wrapped, bible-thumping Trump?
https://www.bbc.com/news/articles/c17v8y0z9z2o
I said he had a terrible ideology, not that he was a fascist. You utterly missed the point of my comment.
I got that, sure, you can have that opinion. And yet, of all the ways you could have said that, you resorted to redefining a quote about fascism which would fit the current president much better than one of his staunch opponents. So yes, if you wanted to avoid the impression of making a "both sides are terrible" argument, let alone a "Sanders is worse than Trump" point, I think you could have done a better job.
You still missed the point.
I understand your point - the words "The broader lesson is..." are doing the heavy lifting there in terms of connecting it to a general principle of "things you vehemently oppose will be smuggled in under the veneer of that which you admire" - but I do think we're at a point in discourse where if the word "fascism" isn't necessary to make your point it's better to find alternatives.
There's a crisp irony to the fact that your post about manipulative use of buzzwords contains a sincere use of a buzzword for illustrative purposes, and it ends up causing a misunderstanding because people are so primed to see buzzwords and expect manipulation.
It's a pet peeve of mine when people take any comparison or analogy to mean you're claiming two things are identical. I'm not convinced that Sanders or Bores are as bad as you say but I think your meaning was clear, and reasonable enough.
I broadly like Sanders but he just drops in the occasional wildly over-the-top economy-wrecking idea offhand. I imagine hearing him talk was like being in the company of a forward-thinking Frenchman you're fond of when he suddenly drops "so anyway let's decapitate a bunch of dudes and their families."
Honestly if you replace "dudes" with "CEOs" that might not be far off some of what Sanders says. Yet I broadly agree with the non-decapitating part of his sentiments.
That's always been one of my biggest pet peeves too. Glad I'm not alone.
> You know the old saying about "when fascism comes to America, it will be wrapped in the flag and carrying a cross?"
The cross turned out not to be necessary. Reality always has a way to disappoint even if you believe that you are already pessimistic.
> What "partnering" does he have in mind?
I assume that Sanders has more in mind than that one tweet.
America is much less religious than it was at the time of that quote. Trump understands that, though J.D. Vance does not.
The cross perhaps wasn't necessary but it still did definitely play a role. If anything Trump's combination of appealing to the more religious while clearly having zero interest/knowledge about it on a personal level feels like exactly the disappointing reality the quote promised us.
But I will absolutely agree that I wouldn't have predicted just HOW thin his veneer of religiosity would be.
Here you go. I get 10 free ‘gift’ articles a month and usually have 9 left EOM
“Bernie Sanders: A.I. Is a Public Resource. You Should Own Half of It.”
https://www.nytimes.com/2026/06/01/opinion/artificial-intelligence-bernie-sanders.html?unlocked_article_code=1.nVA.pnDJ.aGaOctF7jfHQ&smid=nytcore-ios-share
A few years back, Bill Gates was talking about a future where we tax the robots that replace jobs. He hasn’t mentioned it since. Odd, since it’s much closer now than it was then.
I doubt Bores has any particularly innovative ideas on this front, but it does seem like we need them. Otherwise, it’s just fractious chaos and decline until the wheels fall off.
https://m.youtube.com/watch?v=nccryZOcrUg
The Democrats released a bunch of embarrassing info about Gates in the Epstein files, he may not be in a particularly generous mood.
>Otherwise, it’s just fractious chaos and decline until the wheels fall off.
I look around and don't see any decline. The tariffs and Iran war are bad, but almost everyone has a job and real gas prices are significantly lower than they were under the Bush administration. The real threat are brainworms that lead people to want to destroy the economy for no good reason.
I lived in LA from the mid 2000’s to the early 2010’s. The decline is tough to miss.
I’m in NYC now. It hasn’t improved in the past decade. Crims and zombies aside, the average person is fatter and more poorly dressed.
Does anyone know of any studies on the impact of increased productivity from AI (e.g. as measured via number of commits) on business metrics?
Noah smith had an article on that just today.
As a software dev, I'm rolling my eyes so hard right now. Number of commits, just like "lines of code", is not only a useless metric, it has negative value because it sets perverse incentives.
Obviously. Yet, my org has just re-introduced a measure of bugs/kloc (yes, as in kilo-lines of code) that ought not be exceeded. It might already be silently dropped, as I guess I was not the only one questioning the sanity of such a decision.
To the original question: Good luck finding a metric that actually works. I think something like avg. $ profit per dev or "individual contributor" over time is a candidate. This would only yield useful results over a large amount of well-selected companies.
I am not looking for attribution per dev. I want to know whether at a high level, shipping more has any impact. My own assumption is that it doesn't - very few things a company does actually matter, so there is a lot of wasted effort. I want to know whether I'm right or wrong.
Oh I did not intend to imply attribution per dev. My assumption is that heavy AI use is reducing headcount in dev departments, especially squeezing out juniors (at least temporarily).
You're missing the point. Lots of companies claim they're much more productive, that they're shipping much more than ever. I'm asking whether anyone has done a proper study on whether this has any actual impact.
Companies do be claiming lots of things, especially when they have to justify heavy investment into unproven technology and divestment from proven assets (human coders). You gave the specific example of "number of commits" as a metric, which I commented on. I just wanted to let you know that this is not a useful measure of software development productivity, and if there was such a study that used that metric, that study, and people promoting it, should probably be ignored on that matter.
I'm a PM, and I'm not totally stupid. It feels like you're just trying to show how smart you are instead of engaging with the actual question. There are no good metrics for tracking engineering productivity, but the point is that many companies are claiming their development has really accelerated. What I'm curious about is the effect this has - I know there have been studies on whether engineers are really more productive in terms on their output, but what I want to know is whether, even if we grant they are indeed more productive, this has any effect besides, perhaps, on headcount. As in, does it even matter if companies are shipping more features?
Just obviously it matters a great deal, for that you don't need a study. How exactly does it matter right this very moment, you might need a study, what about at some moment a few months or years down the line, back to original answer
I haven't seen such study.
Yeah, many developers have anecdotal evidence how LLMs improved their coding speed. I could add a few examples myself, especially hobby projects that otherwise I would have not enough time and energy to do.
But coding is just one small part of functionality of the company, often not the most important part, even if it happens to be a software company. Sometimes the managers don't know what they want to get done, and keep changing their mind, or the original analysis is utterly flawed. There may be bureaucracy slowing everything down, or various departments of the same company fighting each other. In other words, even increasing the programming speed 100x (assuming generously no negative impact on quality) may ultimately make little difference in the company's overall output. Sometimes the actual bottleneck is analysis, or design, or even sanity in general.
100% agree, this is my hypothesis too. It'd be good to test it though!
Dylan Patel has a good article on how difficult this is to measure: https://newsletter.semianalysis.com/p/ai-dark-output-the-visible-cost-of
Going to go vote in California (specifically LA) today and I'm still undecided about a few races. Please lobby me for your preferred issues/candidates!
"3: I previously discussed the attempt to block the Save Our Bacon Act, which would ban certain types of state-level animal welfare standards."
I continue to consider this a mischaracterization. It bans state-level trade barriers intended to enforce animal welfare standards outside the state.
...wait, wasn't the first post already after it passed the House? I thought it had been in the Senate the whole time.
What's the best work on AI consciousness that goes into detail about how LLM inference actually works, rather than just stuff like the Dawkins piece (which I actually quite liked) which just focus on how these systems behave?
“What is Intelligence?” by Blaise Y Arcas is slightly similar to your specification, and I really enjoyed it.
- has sections explaining how transformers (and some other ML/AI systems) work. Obviously not as detailed as a technical paper/textbook.
- Has very interesting discussion on intelligence (and secondarily consciousness) of conventional life and artificial life.
- Spends a lot of time vignetting historical ideas on intelligence and AI. I found it interesting, but maybe somebody who is familiar with the basic history of e.g. Cybernetics won’t find it so interesting?
- I think it spends 60–40 time split talking about neuropsychology vs AI
- I recommend because I found it interesting and entertaining. Also, I think in the whole book there wasn’t a single argument I found egregious. A few points I disagreed on were peripheral and so didn’t ruin it for me. I’d say it’s intelligently written. Therefore I’d happily recommend it to a Rationalist audience.
- Speaking of, in the last chapter he takes a position arguing against the old-school Rationalist stance on ASI (a la Yudkowsky’s sequences), and I didn’t see anything wrong with his logic.
Can’t give you the big picture, but I’m pretty sure there is some hard coded text in Claude that makes it start most responses with “Great question!”
Yeah, it used to start with "may I suck your brilliant toes?", but developers managed to dial it back some.
Yeah, I'm fond of adding "avoid sycophancy" but being constantly reminded that it'd brush past my mediocre-at-best toes and recoil in disgust really made me question where I was going in life.
Sorry to be unhelpful, but the more you learn the mathematical background of LLMs, the less you will believe they are conscious. A transformer is little more than a key value store for computing probabilities of all data with respect to all other data. It only looks like it's conscious because it's a high level overview of one tiny part of the human brain.
I still agree with you overall, and I used to make similar-shaped arguments myself in the past, but note that key value stores (and matrix multiplications) are like epicycles - compositions of them are fully general, and given enough of them you can construct whatever you like. I’ve linked to plenty of demonstrations that iterated GEMM is Turing complete in past open threads, so here is one for attention: https://jmlr.org/papers/volume22/20-302/20-302.pdf
So unfortunately one does have to do more work to build an argument for a particular system not being conscious that would convince someone who does not already agree; simply mentioning that it’s made of kv lookups is no more convincing than “it’s made of transistors” or “it’s made of neurons”.
It was a tiny sample of one of the many billions of things which are so divorced from the human brain (whose consciousness we still do not understand), that I don't really see how you could find the link between them. But more power to whoever wants to try.
I don’t think anyone’s trying to argue we have built anything like a human brain!
“What the hell do we even mean when we say some entity is conscious?” is definitely a more productive angle of attack.
Seems to me that your idea of consciousness is that all this stuff we have in mind about a topic -- the facts, the history, other relevant sets of facts, our previous actions regarding the topic at hand -- are arrayed around us, sort of like whatever the transformer works with to compute probably with respect to all other data -- but in the human mind there's this little elf who looks at the data and *thinks it over,* and that's why we call the human view of of the topic a conscious *take,* and the AI's take just a mathematical result. But of course there is no little guy bopping around in the human brain, doing what we call *thinking.*. You're just building in the ineffable part, the "we are actually *thinking*" part, by sneaking a homunculus into your model of what actually goes on in the human mind when we arrive at a view of things, based on all the info we have about it.
> You're just building in the ineffable part, the "we are actually *thinking*" part, by sneaking a homunculus into your model of what actually goes on in the human mind when we arrive at a view of things, based on all the info we have about it.
There's no claim or implication of any kind of "homunculus" in the comment you're replying to, this is a blatant example of a strawman argument.
Isn't this just a reformulation of the Hard Problem of Consciousness, though? Or a restatement of why it is indeed hard?
Chalmers' argument is that C = consciousness doesn't supervene *logically* on the physical, i.e. given all of the physical facts about C, it's still surprising that we have phenomenal experiences like experiencing 'red'. But he posits that C might supervene *naturally* on the physical, by as-yet undiscovered psychophysical laws (while offering some not-so-elfin candidates). There's an element of consciousness that's above regular mechanical cognition, approximately "what-it's-likeness", that has to be explained.
So do you adhere to the idea that (a) LLM's share some of our phenomenal experience of the world? or (b) that phenomenal experience (linguistic ex: instantaneous association of certain words with images) isn't a necessary condition for C (the dispute is definitional)?
Over on Twitter, woke dead-enders are waging a desperate rearguard campaign against Graham Platner, bringing to bear all their heavy artillery (“incel,” accusing men of being insecure about their masculinity) to utterly no effect. The only high-profile Democratic politicians to oppose Platner are the centrist Corey Booker, who didn’t go any farther than expressing “concerns,” and Senator Fetterman, who’s on the verge of being thrown out of the party. Elizabeth Warren and AOC, woke favorites back in the day, are both supporting Platner. Some of the wokists seem unaware their powers are gone, others look like they’re desperately trying to manifest them back into existence. They seem to have a real, palatable anger at Platner over the infidelity, more so even than for Paxton and Trump. When they talk about Trump’s infidelity, it’s usually “those Christians sure are hypocrites voting for this adulterer.” With Platner, they’re MAD.
(On the issue of infidelity, I agree it’s both immoral and a sign a politician is likely to backstab his own voters. I find it funny that so many voters don’t care about it and then complain that politicians are dishonest. If he’d lie to his wife, why the hell would you expect him to be honest with you? But enforcing a norm against adultery requires enforcing it against your own side, not just the other side, which nobody in American politics wants to do.)
For those out of the loop, Graham Platner is a famous American oyster farmer or something.
https://en.wikipedia.org/wiki/Graham_Platner
Yeah, the guy seems to bring some masculine energy that the Democrats really need.
> If he’d lie to his wife, why the hell would you expect him to be honest with you?
I don't want him to be honest with me, I want him to screw over my political enemies. And if he happens to screw over both my enemies *and* me, I will... deny that it happened.
Scott, have you read "Ornament and Crime" by Adolf Loos? It sounds like it might be helpful for the "what happened to architecture?" post series.
Excellent piece about the decline of the Boy Scouts. Cultural changes were challenging enough, but then came the abuse scandal and then - and far worse - the descent of greedy lawyers like a huge flock of vultures. It’s truly disturbing and shows how greed can be utterly ruinous.
https://www.youtube.com/watch?v=vQoahSUs7RE
Note: in the Northeastern city where I grew up the Scouts weren’t particularly popular, the YMCA dominated youth activities, and most of my schoolmates who were Scouts were rather dweebish.
I'm not going to watch a 20-minute video but I always thought that Simpsons episode where scouts were portrayed as desperately uncool was a huge blow to the movement.
I grew up prior to the Simpsons but yes, the Boy Scouts were indeed uncool.
Thank you, this was a fascinating video.
I guess most people don't realize how probabilities change when things get large. If something is a 1 in 1000 event, when your organization has 100 000 members, it is practically guaranteed to happen, even if your organization is perfectly normal.
And when it happens, the leaders of the organization will be held responsible for that. Because the mathematically illiterate intuition suggests that they had to do something horribly wrong in order to attract and keep the 1 in 1000 rare thing, especially if it happened more than once.
A more serious accusation is "they knew, but they did not do anything about it". But even this gets difficult with large numbers -- you will find cases when the result of an investigation is a clear "yes", or a clear "no", but also many where the evidence is inconclusive, with various degrees of probability. And of course, if you fire someone based on inconclusive evidence, they too may sue you, for wrongful termination, or slander, or whatever. But if you don't, any note that you had a suspicion but didn't fire them, will be worth gold for the lawyers who will later investigate you.
Somewhat ironically, the “two deep” youth protection policy that the Boy Scouts have adopted supposedly is the best anywhere and many other organizations use it as a model. Unfortunately it may be too late.
> The person who has been trained to read closely and write precisely is, right now, in 2026, more useful in almost every knowledge-work context than the person who has not been.
So writes N. Ángel Pinillos in his Substack essay: "The Honest Case for the Humanities." And while I agree this is probably the case, I'd say being useful is a sucker's game.
Being useful means you're being used. A STEM degree and high-paying job at, say, one of the Magnificent Seven is all very nice, but the odds are that your corporate masters will let you down. Most engineers lose their value to their employers after age 40. Their salaries get too high. They realize that there's life outside of work. And most people who've reached middle age have learned not to take shit from their managers. The usual excuse for lay-offs is financial cutbacks, and there's some truth to that. But they're culling the herd to bring in some eager young achievers at half the price of someone with 15 or 20 years of experience. I'm reminded of salmon struggling up a series of fish ladders. Only a few of them reach the top. And then they die.
Why am I telling you this? Because I could tell you how my undergrad degree in Anthropology has served me better in the corporate world than my graduate STEM background. Certainly, it gave me insights into how corporate social systems work. But the real value is that it gave me insight into life outside the corporation. Pinillos made this point: the humanities made my life outside of work richer. Sure, I needed to be a useful cog in the capitalist machine to earn the money to live comfortably. But the humanities made my life worth living.
https://napinillos.substack.com/p/the-honest-case-for-the-humanities
It's funny how I strongly agree with some parts and strongly disagree with other parts.
> [philosophy] is one of the most rigorous trainings available in a particular cluster of skills: reading carefully, writing carefully, expressing yourself carefully, and thinking carefully in non-quantitative domains.
In my experience it's more like the opposite. Look at Hegel's Logic, that is a humanities guy cosplaying a mathematician doing proofs, and failing laughably... except that instead of being laughed at, he became an example of a genius, an inspiration for Marx who in turn is an inspiration for many. Generally, philosophy seems to lack the ability to call out wrong ideas and discard them, the way science has discarded e.g. astrology or phlogiston; instead it just accumulates, and the 21. century students still learn Plato etc.
> I was also a mathematics major, because I had figured out, also correctly, that quantitative reasoning would matter for whatever I ended up doing. The two together seemed to me to cover the ground and challenged me immensely.
I also have a foot in both camps; I studied mathematics and psychology. The knowledge of each was useful to see the silliness of the other. I don't see it as an argument for the superiority of humanities, but the superiority of knowing more. Yes, the STEM people would benefit from knowing more humanities. And the humanities people would benefit from knowing some math and logic (actual logic, not Hegel's Logic).
> Most engineers lose their value to their employers after age 40. Their salaries get too high. They realize that there's life outside of work. And most people who've reached middle age have learned not to take shit from their managers.
This part I fully agree with. The important thing is to invest your money properly so that when you are 40 you no longer need the job. (Disclosure: I failed at it; living in a corrupt post-communist country, all investment opportunities I tried in the past turned out to be scams, and now that I know better, I am old and have a family to feed, so less money to invest.) The older you are, the more you see that the companies operate by "work harder, not smarter" when it comes to organizing your work.
>Generally, philosophy seems to lack the ability to call out wrong ideas
Probably became they haven't solved epistemology sufficiently to come up with a single agreed definition of right and wrong. Have you solved epistemology yourself?
> the 21. century students still learn Plato etc
What was Plato demonstrably wrong about?
I'm taking a course in Middle Platonism at the moment (considered by most philosophers to be a dead zone of philosophical thought), and I was surprised at how many questions about the origins of the universe and the organization of things they were trying to address. Even though they had no empirical data to work with, at the category level many of the arguments the Middle Platonists made are still being rehashed today by cosmologists and physicists who have empirical data but can't agree on what it means. :-)
> What was Plato demonstrably wrong about?
I am not even sure what Plato *was about*.
Like, one of his most famous concepts is that there are "ideas" and everything we know is just their imperfect reflection. I wanted to know more about those "ideas", like how specific they are, whether they are hierarchical, etc.
As an example, is there an "idea" of a dog? Of a mammal? Of a poodle? Of a black poodle? Of a black poodle that can dance? Basically, some kind of rule to decide which things (according to Plato) are the perfect pure ideas, and which things are their imperfect reflections, even if those things are still abstractions themselves.
When I asked at the classroom, not only couldn't anyone give me an answer, but they were surprised why I was even asking. From my perspective, it is obvious: either the concept refers to something, and then I want to know what it is, or it does not refer to anything, and then why are we wasting time learning it? The impression I got was that it's about learning the right keywords, and then impressing others with your knowledge of them. There are high-status maps, but no territory they are supposed to reflect.
Didn’t object oriented programming kinda crib from this sort of thinking with the ‘ideal’ being the definitions of classes and the individual instantiations being objects of the defined classes?
Each object can have its own individual characteristics but remains true to the abstract class definitions.
I don't have a definitive answer to this, but...
What exactly is and isn't Plato's contribution? Plato did not invent the fact that we use the same word, e.g. "dog", to describe many dogs. That's how language worked, long before Plato. Plato just said that somewhere in "the realm of ideas", the is a divine "the Dog" of whom all dogs on Earth are merely imperfect reflections. Or something like that.
According to Wikipedia, the oldest program that used what we now call object-oriented programming was "Sketchpad", a vector graphical editor, which used names "master" and "occurrences" for what we currently call class and instances.
https://en.wikipedia.org/wiki/Sketchpad
I briefly read the master thesis on the project, didn't find a mention of Plato or ideas or whatever. (I may have skipped something; those were pages scanned from paper.) So, in my opinion, the contributions to Plato are just applied retroactively. But I could be wrong here.
I looked into this a bit and I think you are right. When Smalltalk was developed they were interested in modeling real objects with class definitions not trying to describe an ideal form.
When I was learning C++ the class/object relationship felt analogous to the ideal form/instance relationship but it breaks down when you consider that the class definition itself is just a human made definition no more “true” than the object instances.
>Plato did not invent the fact that we use the same word, e.g. "dog", to describe many dogs
That would be nominalism, the opposite of Platos realism.
>Didn’t object oriented programming kinda crib from this sort of thinking with the ‘ideal’ being the definitions of classes and the individual instantiations being objects of the defined classes
Yes.
>Like, one of his most famous concepts is that there are "ideas" and everything we know is just their imperfect reflection. I wanted to know more about those "ideas", like how specific they are, whether they are hierarchical, etc.
More research was needed, I suppose. Such as middle Platonism and late Platonism and Neoplatonism and Max Tegmark.
>As an example, is there an "idea" of a dog? Of a mammal? Of a poodle? Of a black poodle? Of a black poodle that can dance? Basically, some kind of rule to decide which things (according to Plato) are the perfect pure ideas, and which things are their imperfect reflections, even if those things are still abstractions themselves.
The idea refuses to either die, or be finalised ALONG WITH THE ALTERNATIVES.
Class based object orientation versus object based is the latest round in the Plato versus Aristotle
>: either the concept refers to something, and then I want to know what it is, or it does not refer to anything, and then why are we wasting time learning it
Does extratestrial life refer to anything?
Does dark matter referr to anything?
Do tachyons refer to anything?
It's not as everything falls into two neat categories , the fully understood and the completely non existent.
> What was Plato demonstrably wrong about?
Music, politics.
OTOH, Mathematical Platonism (which I think might be only named after him) is correct.
>Music, politics
Could you expand? Does "music" refer to the critique of heroic epics in the Republic? Does "politics" refer to the idea that the Republic is fascist state?
> Generally, philosophy seems to lack the ability to call out wrong ideas and discard them
Regardless of whether Hegel was right or wrong, chewing on philosophy gave you a better ability to see through the bullshit, didn't it? So, Hegel wasn't entirely a waste of time.
And, yes, knowing MORE is better. Many former humanities peeps I know could use a deeper background, if not in science, an understanding of how science is supposed to work and how it actually works. Otherwise, the poor dears are misled by every flashy headline on diet advice, environment, medical advances, etc.
I cringe whenever I hear someone say, "Trust the Science!" — because (a) it means they don't really understand the scientific method and/or don't understand the science that they're promoting, or (b) they're researchers promulgating a viewpoint they'd rather not be challenged on.
> chewing on philosophy gave you a better ability to see through the bullshit, didn't it?
I think it doesn't work this way.
If you want to e.g. recognize homeopathy as bullshit, you need to study lots of chemistry. And only afterwards, look at homeopathy and notice that it contradicts everything that is known about water molecules. If you only study homeopathy, it does not produce the same effect.
Similarly, you recognize Hegel's Logic as bullshit if you are familiar with actual logic, not if you study too much Hegel.
That's a very good point, but there's a *huge* universe outside our empirical grasp that we can only address with speculative reasoning. And the logic that you used to analyze Hegel is based on the reasoning first developed by philosophers. I'm not saying we can arrive at truth through reason alone. But reasoning can help us filter the more likely from the less likely.
"If you want to e.g. recognize homeopathy as bullshit, you need to study lots of chemistry." I think that's not the only way. It also would suffice to grasp that homeopathy is ideally suited for double-blind studies and have a look at those.
You can take the view that Hegelian logic is actually misnamed metaphysics.
Also, Hegel Bad doesn't really generalise to the treat of philosophy.
> Also, Hegel Bad doesn't really generalise to the treat of philosophy.
Can you clarify and expand on that thought?
only Hegel was Hegel . He was an extreme outlier in his use of armchair rationalism.
"Regardless of whether Hegel was right or wrong, chewing on philosophy gave you a better ability to see through the bullshit, didn't it? So, Hegel wasn't entirely a waste of time."
I would feel better about Hegel's pointing out of bullshit if he (or his disciples, I can't remember) didn't address it by inserting even more bullshit at an even deeper level.
Apparently the story about Karp being supervised by Habermas is not completely truthful: https://substack.com/@encaustum/note/c-269688966?r=jzjhp&utm_medium=ios
Scrip. This was a feature of the Gilded Age. It was a way of fleecing your workers and keeping control of them. It would be interesting if it could be reinvented as something that is positive.
I owe my soul to the company store.
ACX BKK is kicking off our first June meetup on Sunday 7 June at 1pm. We'll be at Thong Kee for dim sum:
https://maps.app.goo.gl/jyyj1gWdGGhkS1Kt9
Everyone is welcome, whether you're new to ACX or new to Bangkok!
Hopefully this time we have a sign 😂 if not, just look for the tall loud white guys talking about AI
Hold a folding chessboard just to be sure.
Jean-François Lyotard’s _The Postmodern Condition_ partly blames the computer scientists for giving rise to the postmodernists. (I don’t think I can summarize Lyotard without re-reading it; but if I was try an outline a similar kind of trajectory — Bertrand Russell’s logicism crashes and burns thanks to Kurt Gödel, and of course Wittgenstein was with Bertrand Russell’s group, and then, well, here we are. You might object that Derrida is not mounting the same kind of attack on the system that Gödel is, but is a very different sort of anti-Bertrand-Russell)
But maybe the genealogy of it (good Michel Foucault word there) is Hegel -> Kierkegaard against Hegel -> Heidegger -> Derrida, while, in parallel, there is another anti-systematizing project that’s Russell -> Wittgenstein against Russell
The other thing that surprises me is people equating postmodernism with Marxists.
When Lyotard says the key characteristic of postmodernism is “skepticism towards metanarratives”, Marxist/Hegelian accounts are one of the main “metanarratives” they’re being skeptical towards, so we can rephrase Lyotard as “the postmodernists think the Marxists are insane”.
Admittedly, the problem with this is you end up with Slavoj Žižek, but, well, here we are.
Ok, you can take the line that Žižek isn’t actually a postmodernist, he’s one of those Marxists the postmodernists were being skeptical towards, and thus is exactly what’s wrong with Žižek.
From Marginal Revolution:
“ In 2019, there were about 150,000 people working in autism therapy. Six years later, there were 654,000—more than the number of people who work in mining and logging, or telecommunications, or at the US Postal Service.”
It’s enough to make you reach for the antidepressants :(
What's the actual problem?
Who is paying for all this? What is the cost-benefit ratio?
For that sort of expenditure you'd expect to see dramatic results, where are they?
I’ve just bought cinema tickets to see the final part of _The Amazing Digital Circus_ in the cinema tomorrow.
My kids insisted on watching it, so I am taking them to cinema on Sunday, because under 12 they need to be accompanied by a parent.
Let me very gently, and mildly, and reasonably, ask: what the ever-living *fuck* is going the hell on here????
Yes, you too can donate your brain to the Mi-go of Yuggoth, I mean a Yale spin-off biotech company:
https://www.science.org/content/article/not-alive-not-dead-disembodied-human-brains-used-drug-testing
The article (and certainly the headline) is probably a bit sensationalistic, even if it is from Science magazine, and while I'm fascinated by the science here, I'm also rather disturbed.
"Not alive, but not dead: disembodied human brains used for drug testing
By restoring some functions to intact brains from deceased donors, the startup Bexorg hopes to create a better drug development test bed for neurodegenerative diseases"
How fresh, exactly, do these brains have to be? How are they able to rev them up to some kind of functionality? "Some" functions? What functions, precisely? Who the heck persuades people to hand over brains of deceased family members, without coming across as a Universal monster movie of the 30s? Is this applicable to cryonics, given that brain preservation and resuscitation is important? Can this work for what we now consider brain death, the way artificial resuscitation and restarting the heart did for previous ideas of when death occurred?
But at the same time: lads, Lovecraft's "The Whisperer in Darkness" was not meant to be an instruction manual!
"NEW HAVEN, CONNECTICUT—Just a day ago, the brain was in a living person. Now, hours after its owner died, it sits on a cart draped in tubes that quiver as they pump liters of blood substitute and other fluids through the organ, supplying oxygen and removing waste. With most of its key functions intact but its electrical activity quenched by anesthesia, the brain hovers between life and death. As it metabolizes experimental drugs, sensors record its reactions, capturing hundreds of data points on its cells, proteins, and physiology. Then, after 24 hours in this state, it will be sliced into hundreds of pieces for more detailed study.
The brain is one of more than 700 that the 5-year-old biotech startup Bexorg has nurtured and studied using a set of proprietary brain-sustaining machines it calls BrainEx. The platform grants researchers an intimate look into how potential therapies might work inside brains with neurodegenerative diseases such as Parkinson’s, Alzheimer’s, or amyotrophic lateral sclerosis. Bexorg can biopsy the brains and discover how long a drug stays in cells, whether it hits its molecular target, and any hints of side effects.
Early results, including a 2025 poster reporting that the preserved brains match living brains in their response to certain therapies, are encouraging, says Bruna Bellaver, who studies neurodegeneration at the University of Pittsburgh. “It’s a huge step up from mouse models,” she says of the platform.
Bexorg has largely stayed under the radar. Its founders have published no papers on its work other than early experiments with pig brains, although Vrselja says they are preparing their first paper on human ones. Now, the company is scaling up and inviting new attention. Its new lab space, which Science visited last week, will house a 1.2-meter-tall robotic arm to automate the process of slicing up to 1600 brains per year and analyzing 11,000 proteins in each. At a media event today, the company will showcase Bexorg’s assembly line process—and seek to reassure the public that its disembodied brains don’t cross ethical lines or risk regaining consciousness.
...The BrainEx machines sit in six plexiglass cubicles in an office overlooking Yale University, where Vrselja, neuroscientist Nenad Sestan, and their colleagues conceived of the idea 10 years ago. Before each brain is placed in a machine, surgeons examine it through a loupe, then suture four plastic ports into the vessels that once supplied it with blood so it can start reacting to drugs and generating data. Once the brain is attached to the BrainEx machines, an artificial lung and kidney oxygenate and filter fluids as they flow through the organ."
Website of the lab in question:
https://bexorg.com/
Yeah, it's all absolutely fascinating, now if you'll excuse me I need to go source a diving helmet or something. Cranial protection, be sure or be sorry! What is that buzzing noise I hear behind me?
In the game Sid Meier's Alpha Centauri, you get this quote as a popup the first time you build the "Bioenhancement Center" base facility:
-----
"We are all aware that the senses can be deceived, the eyes fooled. But how can we be sure our senses are not being deceived at any particular time, or even all the time? Might I just be a brain in a tank somewhere, tricked all my life into believing in the events of this world by some insane computer? And does my life gain or lose meaning based on my reaction to such solipsism?"
Project PYRRHO, Specimen 46, Vat 7
----
When I started reading your comment, I was expecting this to be about brain organelles, little clusters of cultured human brain cells used for research purposes (mix of biochemical and computational research). Those already give me the creeps. They're small enough (maybe one millionth of the cell count of a full brain) that there's no real danger of them being meaningfully sapient, but is still seems like a line of research that needs a strong bright-line rule to stop it before there's any danger of making a conscious brain in a jar, and they've already hopped the first Schelling Fence by doing the research at all.
But no, these folks are doing literal brains in jars. Similarly, I expect the brains are hopeless addled by the time the researchers get them, no longer capable of anything like consciousness, but this is much much closer to dystopian horror territory.
I lost a friend to a car accident many, many years ago. She was comatose when she reached the hospital and imaging showed several brain damage. She ied a few months later without ever regaining consciousness. Having this as an emotional reference point gives me complicated feelings about this line of research. A reminder that the brains in question almost certainly aren't meaningfully conscious, visceral unease at doing this kind of experiments on the brains of people who could be thought of as merely "mostly dead" even if the emphasis is very strongly on the "mostly", and also a reminder that this kind of technology potentially has very real lifesaving implications even if the research itself is unsettling.
"When I started reading your comment, I was expecting this to be about brain organelles, little clusters of cultured human brain cells used for research purposes (mix of biochemical and computational research). "
Precisely why my gob is so smacked. It's not brain tissue, it's not emulations, it's real human brains sawed out of the skull hours ago and wired up to the life support tubes. I would like more information but now I'm sorry I ever watched any of those "mad scientist brain in jar" movies because why are we careening as fast and hard as we can into "let's turn dystopia into reality!" territory?
Maxim 14: "Mad Science" means never stopping to ask "what's the worst thing that could happen?"
https://en.wikipedia.org/wiki/That_Hideous_Strength
They have no mouth and they must scream.
The time to get really scared is when the place opens a deli featuring smart-meat sandwiches,
Especially if there's a barber shop upstairs from the deli.
You've heard about the brain in a vat as a thought experiment, but now, you can *live* it!
I assume the founder is very insistent that his name is pronounced "FRONK-en-steen:".
That is not dead which can eternal lie, for with strange aeons even death may die.
This is the first piece of information I've heard in my life which has made me actually reconsider being listed as an organ donor. I wonder how they are sourcing the brains for this work? Just buying cadavers from the same place as med schools do? I am seriously about to revoke my organ donor status over this. I signed up in the hopes that a kidney or something of mine could help someone in need, not for my brain to be plugged into the venture-capital torture nexus...
It's not really covered how they get donations. I imagine the brain needs to be fairly fresh, otherwise it's not going to be much use. So that gives the unedifying image of a surgical team waiting by the bedside for legal brain death to swoop in and start scooping out Uncle Bill's brain.
My bad, I didn't check their handy posters where they explain how they get the brains and show you the BrainEx machine (complete with hooked-up brain):
https://bexorg.com/technology
https://bexorg.com/assets/pdf/an-intact-molecularly-active-postmortem-whole-human-brain-model-of-alzheimers-disease.pdf
"Acquisition of postmortem human brains adheres to the highest possible ethical standards, overseen by an independent board of world-renowned bioethicists. Brains are procured through Organ Procurement Organizations (OPOs) with enhanced levels of consent from patients and families that specifically cover the BrainEx platform. Measures are taken on the BrainEx device to ensure that there is no possibility of coordinated network activity associated with consciousness."
They seem to be getting them from elderly dementia patients who died naturally, since their initial emphasis was on research into drug treatments for Alzheimer's, so I imagine that enables them to persuade families to donate brains to help other people avoid the degeneration they saw with Uncle Bill at the end of his life.
Still, pardon me for faintly exclaiming "Oh, God..." here:
https://bexorg.com/assets/pdf/an-intact-molecularly-active-postmortem-whole-human-brain-model-for-gene-therapy-development.pdf
"Figure 2. BrainEx perfusion restores cellular activity and drives growth and repair. a, Perfusate measurements show that glucose is consumed, and lactate is produced by the brain, indicating active metabolism. b, Proteomics comparing perfused versus non-perfused brains reveals that BrainEx reduces neuroinflammation, and activates transcription, translation, DNA repair, and synapse and membrane repair."
More reporting does not dial down the creepiness factor:
https://futurism.com/health-medicine/startup-testing-drugs-on-freshly-extracted-human-brains-that-are-kept-on-life-support
"But you can put those doubts out of your very-much embodied mind, assures Brendan Parent, one of Bexorg’s six ethicists. The extracted brains are almost devoid of the coordinated neural firing necessary for minimal consciousness, he told Science. To prevent the eerie implausibility that some the brains produce electrical activity, the brains are also dosed with anesthetic propofol. Of course, that such a measure has to be taken in the first place may actually be less assuring and more unsettling.
Ethics aside — not a statement that should be made lightly — the scientific possibilities that these extracted brains afford may well hold promise. Bexorg CEO Zvonimir Vrselja said that the brains come with decades of environmental exposures, histories of drug treatments, and other factors that make them a more realistic testing medium for drugs. “You get cells that have been there for 60 to 80 years,” Vrselja told Science.
...Bexorg is the same startup that demonstrated, over six years ago, that it could keep decapitated pig brains alive for 36 hours using a prototype of its BrainEX machine.
Today, its human brains aren’t kept running in perpetuity. After 24 hours, they’re sliced into hundreds of pieces so they can be analyzed by scientists. The company plans to use a robotic arm to slice up to 1,600 brains per year."
How reassuring. "Almost devoid", well no worries there, and then they'll chop up the brains into sushi for further study.
36 hours alive decapitated pig brains sounds too much like "That Hideous Strength":
"I’m telling this dreadfully badly. What it really was, was a head (the rest of a head) which had had the top part of the skull taken off and then… then — as if something inside had boiled over. A great big mass which bulged out from inside what was left of the skull. Wrapped in some kind of composition stuff, but very thin stuff. You could see it twitch. Even in my fright I remember thinking, ‘Oh kill it, kill it Put it out of its pain.’ But only for a second because I thought the thing was real, really. It was green looking and the mouth was wide open and quite dry. You realise I was a long time, looking at it, before anything else happened. And soon I saw that it wasn’t exactly floating. It was fixed up on some kind of bracket, or shelf, or pedestal — I don’t know quite what, and there were things hanging from it. From the neck, I mean. Yes, it had a neck and a sort of collar thing round it, but nothing below the collar; no shoulders or body. Only these hanging things. In the dream I thought it was some kind of new man that had only head and entrails: I thought all those tubes were its insides. But presently — I don’t quite know how, I saw that they were artificial. Little rubber tubes and bulbs and little metal things too. I couldn’t understand them. All the tubes went into the wall. Then at last something happened.”
“…Well, quite suddenly, like when an engine is started, there came a puff of air out of its mouth, with a hard dry rasping sound. And then there came another, and it settled down into a sort of rhythm — huff, huff, huff — like an imitation of breathing. Then came a most horrible thing: the mouth began to dribble. I know it sounds silly but in a way I felt sorry for it because it had no hands and couldn’t wipe its mouth. It seems a small thing compared with all the rest but that is how I felt. Then it began working its mouth about and even licking its lips. It was like someone getting a machine into working order. To see it doing that just as if it was alive, and at the same time dribbling over the beard which was all stiff and dead looking…”
And this may be gallows humour, but everyone seems to have names out of the good old Gothic tradition of Eastern European mad science shenanigans (though I acknowledge Frankenstein was Swiss).
Bexorg are leaning into the Mi-Go from Yuggoth 'we'll take your brain back to Pluto in a cylinder' vibe as well:
https://bexorg.com/about
"Unfettered from the limitations of cell cultures or organoids, our platform is capable of manipulating and recording the cellular activity of fully mature, intact, and isolated brains for days on end."
I was thinking about The Colours of Her Coat and the recent brouhaha about taste. I wonder if being over-exposed to a form of art creates taste, in the sense of 'you become less appreciative on average but gain a deeper ability to appreciate'? Taste has always seemed like a negative to me – as someone who rarely watches television, I enjoy high-quality slop, and would have less television to enjoy if I did not enjoy high-quality slop. But perhaps it could save us here. If good television becomes so common that losing your ability to enjoy high-quality slop (by this I mean Marvel and that sort of thing) does not affect me, as I do not watch TV enough to have consumed all of the 'truly good' TV, is that a loss? Filtering becomes hard, of course, but it would probably be solved at the algorithmic level at that point. Just random thoughts bouncing around my head that I have yet to think through.
Another example, reading writing in a script you know, half-know, and completely do not know are (to me) wildly different cognitive experiences. There will always be new scripts to learn. Even if you learn every human-made writing system, A.I. can generate beautiful ones for eternity. Will this feel fake? Consider the way many reacted so negatively to the 'you control the buttons you press' Tweet, or the phenomenon of analysis paralysis: the mere existence of an option makes many peoples' experiences worse. Are such people just skill issued, or are they victims of systemic failings? I do not know.
I watch very little TV, but did watch Breaking Bad, with great enjoyment. In the past, with my daughter, I did watch a bit of slop, esp. America's Next Top Model. I actually have no idea how classy BB is considered to be, but I'm guessing few would call it slop, or rate it lower than ANTM. I notice a real difference between how BB has stayed with me vs. how AMTM has. I feel nourished by BB. I still ruminate sometimes about the characters, and think about parts of me that are sort of like them -- my inner Skylar, my inner Gus Fring, my inner Jesse. I still feel a shiver thinking about certain parts, and snigger with pleasure about other parts. Occasionally I have an urge to rewatch a certain episode. I do recall a lot of ANTM, and sometimes think of it spontaneously, but the thoughts aren't as emotion-laden, or as deep. For instance, I recently remembered the episode where the aspiring models had to have large spiders placed on them. Wondered how well I would have tolerated it, and whether they were de-venomed tarantulas or some other species. But these were just kind of passing thoughts about something novel I once saw. I had not been deeply affected by it. So based on all this, I'd say better entertainment, stuff closer to art than to slop. engages us more deeply.
Very neat, thank you! Do you think that a near-infinite amount of BB-like shows would be able to continue to form you in this way? Say, if AI could generate shows of BB quality indistinguishable from human writing.
The whole 'Seinfeld Is Unfunny' page on TVTropes covers a lot of things like this. https://tvtropes.org/pmwiki/pmwiki.php/Main/SeinfeldIsUnfunny A show does something so memorable and remarkable that EVERYONE imitates it, and it becomes so mainstream that the original show no longer seems remarkable.
I'm of the opinion that more is better, as long as there's some valuable filtering tool. But that second part's the catch; supply has sped up so much it's flooding out the filters. And creators will always be incentivised to try to corrupt the filters to say their particular show is good.
(As a filter tangent, I hate how of all things Rotten Tomatoes has become the critical standard. They put a movie's Rotten Tomatoes score on its Blu-Ray now. Nothing says quality like decaying fruit.)
Haha, reverse nominative determinism
But yeah, it will be neat to see how algorithms keep up in the slop age. Especially if people, in the short term, prefer slop. I consider myself above AI generated drivel, but perhaps only because the sort of AI slop (derogatory) that would appeal to me has not yet been created. I hope not, but I worry.
I checked this out and I read the section on live films. I couldn't believe that Jean-Luc Godard wasn't mentioned once. So many films after him borrowed so many things from him. He practically invented the music video genre 25 years before it became a thing. The jump cutting, the sudden change of scene. It also borrowed a lot from Old Hollywood. The charming bad guy for instance: Breathless. The wonderful absurdities and the emotional truth of Contempt. Fritz Lang appeared in that film and discussed the idea that the cinema was dead. Bridget Bardot's naked behind, covered with a book on philosophy. He totally changed the rules about how a film could be constructed. There is also an error in that he credited Orson Welles with a film that was actually directed by Carol Reed. The Third Man . A lot of the films that he referenced were unfortunately ones I haven't seen because I'm older, and I suppose the resonance of those highly influential movies is more apparent to me. Being 70 years old instead of 35 . I was also surprised that he never mentioned Luis Buñuel, who was a master of what we might call surrealist films. Viridiana is a prime example of good intentions leading directly to hell. The Exterminating Angel was a breakthrough in the idea of people trapped by circumstances beyond their control and trying to find a way out of it. Le Chien Andalou, which he made with Salvador Dali, was incredibly provocative for its time in terms of naked people and horrific images. The scene of an eyeball being slashed in close-up with a straight razor was enough to make you puke, and this film was made in 1929. The editing effect from Great Expectations, which was directed by David Lean being copied almost frame for frame in Wait Until Dark with Audrey Hepburn, it was touched on, and it's actually a really interesting story. David Lean was an editor before he became a director, and they worked very hard on that scene, - figuring out exactly how many frames after the cut to the wide shot from the close up before the escaped criminal jumped out from behind the tombstone would give maximum effect - was a real process of discovery. The relief of tension that comes from seeing the whole scene and then the introduction of danger suddenly . I have forgotten now how many frames it was, but it was a very short interval. Long enough for the relief to set in and short enough that you hadn't quite relieved yourself before he popped in. Also, The Right Stuff borrowed a lot of editing techniques from "Breaking the Sound Barrier," which Lean made, I believe, in the early 60s. Some of the scenes in The Right Stuff are literally lifted and reproduced from Breaking the Sound Barrier. Still, it was an interesting read. As I said somewhere else in this thread, "All artists borrow, but the great ones steal."
Interesting; Jean-Luc Godard has his own page, and his page has his movies, and Breathless at least is listed as a Trope Codifier, but Breathless doesn't seem to be on the Trope Codifier page.
Oh, I didn't notice that. My bad. He deserves his own page. So does Fritz Lang.
"A near-infinite amount of BB-like shows . . ." But what does a near-infinite number of Breaking Badseven mean? Zillions of spin-offs would not be Breaking Bad writ large -- they'd just be crummy attempts to keep the success going. One of the delights of Breaking Bad was that I was genuinely surprised over and over. I *really* did not know if Jesse was going to shoot that assistant they'd hired, the one who gave Walt the book of poetry, or if Walt would actually watch Jesse's girlfriend die, of where the eyeball in the swimming pool came from. The show lacked all the little beats and hints that, in other shows, tell you who's doomed to get killed off soon, who' s going to triumph. And now that I think of it I could say the same about many movies that delighted me: Being John Malkovitch, Jules and Jim, Melancholia. What's special about them is that they are singular instances. I lived through them, delighted and surprised, truly not knowing what was going to happen next. If you make a bunch of versions or spin-offs of one of them, they're not singular instances any more. So I'm not sure how much sense it makes to talk about AI making "a bunch of Breaking Bads." To really give viewers a repeat experience, it would have to generate a bunch of deeply novel new works, each as different from the other as Breaking Bad is from Being John Malkovitch. And I don't think AI's up to that. Not sure it ever will be, because its deep organizing principle is to do the expected thing, right? The one most like other the others.
Which ones? I'd be curious to try one.
True, if AI can never do that, the problem becomes much less hard. I think you'll find most people on this blog think AI will be able to at some point. But even if not, it is worth thinking about as a thought exercise – is there a limit to human surprise in storytelling? I genuinely don't know either way. Certainly there is a limit to human memory. But I suspect any effects (positive or negative) relative to enjoyment would kick in well before then.
>is there a limit to human surprise in storytelling?
Is this a chance to recommend Anthony Hopkins' "Slipstream"? I think this is a chance to recommend Anthony Hopkins' "Slipstream".
https://www.youtube.com/watch?v=hrD8R_cz8Ac
(In the Director's Commentary he basically said he wanted to make an anti-movie.)
Will watch, thank you!
That trailer looks interesting, and I've always liked Anthony Hopkins, so I am curious to watch this film. Thank you. The first thought I had is this is a serious version of Zelig. The Matrix popped into my mind as well. Does everyone go through a phase where they think their life is a movie? I know I did.
>is there a limit to human surprise in storytelling?
I very much doubt that there is.
BB in my opinion, has several influences from prior works. It's basically a story about someone who has a big problem to solve in their life and is drawn deeper and deeper into a dark place by insisting on pursuing that solution no matter where it leads. The Godfather is basically the same story. But in a completely different setting, with a completely different environment, the basic theme holds. Every time I try to get out, it keeps pulling me back says Michael Corleone. Strangers on a Train has a similar theme and leads on into the other Ripley novels that Patricia Highsmith wrote. An older movie and a less obvious one is Scarlet Street by Fritz Lang. The Treasure of the Sierra Madre also comes to mind.
Any story that has a murderer who is forced to keep murdering to cover up the first one also is a similar theme, and there are lots of stories about that. It all goes back to Oedipus Rex in the end. I think A.I. could give you a story about anything you want, but it will lack the colour and the specificity that living experience brings. I believe it was Thucydides who said that something must be carried until it can be shaped into a thing that can carry itself. I don't think AI has that ability, and I seriously doubt it ever will. At least when it comes to storytelling. When I fed it a bunch of my own writing, it was very good at detecting themes that were common and approaches that were more successful than others. It's not a bad critic. I found it very helpful.
Yeah, it's not the story that's unique about Breaking Bad. There are probably a limited number of plot lines that can move and engage viewers, and so any drama is going to have elements of one or more of the bag basic plotlines. The things that seemed unique about BB to me were 2: First, as I said, it lacked all the little conventions that let experienced viewers know in advance how various things will play out, so I was genuinely fascinated a lot of the time, wondering what the hell somebody would do, and genuinely astonished over and over when I saw what actually took place. Second, I think BB was a very very dark comedy, the darkest damn comedy I ever saw, and I loved that. Sometimes Walt and Skyler were like Lucy and Desi, but super dark versions, when they were rehearsing how to keep up the deception going of Herb and Skylar's sister.. Or think about the 2 killers who went after -- I don't remember who, now, maybe Walt -- and lost their feet when somebody rammed them -- and later we see them crawling across the hospital floor toward their intended victim, leaving smears of blood from the ends of their legs. Or the narcocorridos (Mexican ballad about drug lords) about Heisenberg, sung with such vigor at the beginning of one episode. Skylar spraying the silverfish in Walt's pile of cash. And of course Saul the criminal attorney.
God I loved that show! did you watch it?
From what I gather, BB is generally considered an extremely high-quality TV show. Not necessarily high art, but very very far from being slop and has more artistic merit than any but a handful of other popular shows.
It's at the top of the all-time user ratings list for TV shows on IMDB (9.5/10) and is 95+% positive for both user ratings and published critics on Rotten Tomatoes. Again, this doesn't necessarily mean high art, but it's an indication that the overwhelming majority of people with an opinion of it thinks it's well made. Anecdotally, I've heard a moderate number of people say it's not to their tastes for various reasons, but I don't think I've ever heard anyone seriously argue that it wasn't well made.
I watched this series with my teenage son, who was very keen on it. I enjoyed a lot of it. I like the way it ended. I've forgotten who it was now, but someone once said there are only two stories: someone leaves home or a stranger comes to town. Breaking Bad was definitely about someone leaving home.
Jesse Pinkman’s dorky part time drug dealing pals shining laser pointers implying they were laser sights was pretty funny.
As prequels go Better Call Saul was not too shabby, particularly the arc of Kim Wexler. There was interesting territory to explore with Mike Ehrmantraut and Gus Fring. Hector Salamanca in BB only had a couple lines (ding) and (ding ding) that he kept repeating but in the end it was consequential. Fun to get a bit of his backstory too.
I watched a few episodes of Better Call Saul, and it was getting interesting. I don't know why I stopped because it wasn't that I didn't like it. I guess I just drifted away.
Generally speaking, one thing I noticed about Breaking Bad was how it often bounced between humor and horror. One minute there's a 'domestic issues' scene, next one there's someone's being tortured/whatever in the desert. There were abrupt "gettin' real" shifts in dramatic tension.
The writers also juxtaposed suburban American life with the drug underworld, and having Walter walk between the two provided a milieu for novel social commentary.
This was one my favorite laugh out loud moments on BB. The grunt work of manufacturing and selling crystal meth to the tune of a schlocky minor hit pop tune from 1968.
Crystal Blue Persuasion
https://m.youtube.com/watch?v=cxT6JMvk9l0&ra=m
Yeah, that's a thickly ironic hybrid of Horatio Alger, protestant work ethic, and land-of-opportunity tropes.
; I thought Breaking Bad was pretty damn good. I would put it in a higher class. To tell you the truth, I thought Deadwood was much better. That's kind of my favourite.
Well, if you mean by knowing a script that you've seen it performed or even been to a reading of it, that is a big prior to reading it yourself. I think if a script is very dialogue-driven, that it's very helpful to read it out loud. I worked as an actor for some time in my life, and reading scripts cold was very much a process of discovery. As far as AI generating beautiful ones for eternity, I have serious doubts about that, although I can't completely disagree. I think the big problem is context. AI, at the moment, in my experience, has very limited context, and that's a big deal. There really aren't that many basic stories if you look at it. They cluster; they're all around certain themes, but the distinction comes from the context a human being has to bring to that story. It ropes in a lot of their own experiences:
- Things that have happened to them
- How they have resolved the things that happen to them
- Their imaginings about someone else's state of mind. Lots of good writers, painters, and musicians will often talk about how the thing they are working on will gradually congeal and take them over because they make unexpected and sometimes surprising connections between things that are in their mind. If they have a good mind, they make good connections, but they are personal, which is what makes one writer stand out from another. I seriously doubt AI has that possibility.
I've experimented a little bit with Claude, and I have always found the results incredibly banal.
Shakespeare, I don't think, came up with any original plot. He plundered other plays or histories for his story but brought his gift of language and characterisation to them. They were just another interpretation of stories that had been around. He was also very mindful of the world that he lived in, which is another thing that rules against AI. King Lear is a great example of this because James the sixth of Scotland had just become James the first of Scotland and England. King Lear completely revolves around the madness of splitting up a kingdom and the trouble that ensues. He put all his criticisms into the mouth of a character, karma, the Fool. The Fool was the only person in the court who could get away with speaking truth to power as long as he did it carefully and cleverly. The whole play was designed to please King James at the same time pointing out the trap that can await a careless King and also to reinforce the notion of Scotland and England becoming joined under one monarchy. It was very smart because it acknowledged the culture that he lived in and protected him from getting in trouble with the King to put all those words in the mouth of the Fool.
When Herman Melville wrote Moby Dick, he had had experience on whaling boats and took advantage of a very powerful incident, the sinking of the Bedford by a sperm whale. So he had real-life experience. He had an incident that was real, and he had a mind that saw a great metaphor to be made out of this. He had a public that was very interested in whaling because it was the main source of whatever oil we had to work with (and was completely blown out of the water by the petroleum industry.) It was kind of a perfect trifecta: his personal experience, a very newsworthy incident, and a public that was interested in the whole subject. I doubt that an A.I. would come to that metaphor. It would be a rather tedious, bare-bones story about someone going whaling with a personal vendetta, but the greatness of that book comes from reading it and knowing he is talking about unfathomable forces related to God, the power of nature and the mysteries that we have to live with rather than trying to solve them. The simile he creates between watching the Bedford sink and comparing it to the bits of stuff floating on the top of a stirring punch bowl is not something an AI would ever come up with, in my opinion.
With A.I., their information is always stale. They can find you anything that's been written, but I don't think they have the capability to rewrite it in the context of the current zeitgeist, which is always a moving target. They have no priors about characters and people and the effects of the world around them playing on that. The old saying "every artist borrows, but the great ones steal" comes to my mind here. AI's borrow stuff, but they never really steal it.
I read OP's second paragraph as being about reading texts in different writing systems e.g. alphabets vs. hieroglyphics, not scripts for a play or movie.
Oh, well, that's interesting. I guess I didn't, It seemed the whole context of the conversation was about scripts, as in plays, movies, etc. I don't know how you could read a script in a language you don't know.
For pictographic systems like hieroglyphics you could at least be like "hey that sort of looks like a human, that sort of looks like a bird, etc." and understand maybe 1% of what it's trying to say with zero knowledge. I agree that for a script like hiragana where the symbols are just symbols you couldn't get anything out of it with absolutely zero knowledge. However, with hiragana specifically, there's less than 100 symbols and each is associated with a sound. You could learn those symbol-sound associations without knowing any Japanese, and in that way, you could "read" hiragana, in the sense that *something* is happening in your brain when you look at the symbols, that is different from what would happen if you didn't know those associations. I think that's what OP was getting at.
Well, in that sense I could get something out of looking at a New York City sidewalk, to be honest with you. I suppose the OP could chime in and straighten this out. As I said, the context of the conversation led me into believing script referred to scripts. I may well have missed his point.
DeepSeek wrote me a snarky article on the Fen Tiger, an alleged puma sighted in Cottenham. I’ll extract this part, where it invents a new term.
“The Bus Stop Cryptid.
A cryptid reported in a location that:
* Is named and real
* Is too small to plausibly hide the animal
* Is within an hour’s travel of a dense urban centre
* Is known to the reader as a place they pass through, not a place they romanticise”
Deepseek clarifies why William Cobbett’s reported sighting of a big cat at Waverley Abbey is the ancestor of the Bus Stop Cryptid, but not an example:
“The crucial thing is that the bus stop in “Bus Stop Cryptid” is doing two jobs:
Transport accessibility – You don’t need to be a rural romantic like Cobbett, wandering old hunting grounds. You get off the 114 bus from Cambridge to Cottenham, walk past the rec, and maybe see something slip behind the village hall. The bus normalises the location.
Scale mismatch – A modern village like Cottenham (~2,500 people, streetlit, fenced gardens, farm machinery moving at night) is not a wilderness. The cryptid is reported despite that, which is what makes it comical or eerie. Cobbett’s Waverley still felt, in 1825, like a plausible edge-of-wild place for a boy to see something strange. By 2025, the same animal in the same spot (now Waverley Abbey is a tourist car park and a lawn) would be a proper Bus Stop Cryptid—because there’s no “wild” left.”
[ I have not checked DeepSeek’s knowledge of Cambridge bus routes for factual accuracy. The point still stands even if it’s got the bus routes wrong. ]
Does anyone have any first-hand experience of reading various "The Odyssey" English translations, including the Emily Wilson one? I don't have any experience with any of them and I don't know Greek, and I'm trying to figure out if all the claims of her feminizing and actually changing the story are accurate. From an hour or two of online research, I get the impression that her translation is pretty literal and direct, while being easily readable. I guess there was a lot of pushing the "first English translation by a woman" thing for publicity, but I haven't seen any concrete examples from the actual text of her warping anything. Does anyone know more about this in detail?
I read her translation, it was really good. I don't recall thinking anything about the sex of the translator or the slant of the translation. Though I'd only ever read the Fitzgerald translation before.
That's good to hear that you've read both. I haven't read any, but from my online searches, what you said is what I more or less expected to be the case. Just wanted to hear some firsthand experience before I invest any time into reading.
Disclosure: I do not read ancient Greek, but am relatively familiar with both her translation and the controversy surrounding it.The criticisms of Wilson's translation (beyond her very modern-sounding and colloquial word choices that constantly took me out of the story) mainly revolve around two translation controversies.
The first is her translation of "polytropos," the descriptor Homer gives to Odysseus in the poem's opening. Classicists have been arguing over the meaning of polytropos for centuries, and it has been variously translated as "cunning," "resourceful," "the man of twists and turns," "many-sided," etc. Wilson chooses to translate the word as "complicated," which, while not *wrong* per se, leaves the reader with a very different--arguably much more critical--impression than what many believe the text supports. Personally, I think Wilson's choice is by far the worst of the more prominent translations, and is emblematic of Wilson's habit of inserting contemporary values into a 3000 year old story. Her very first decision is to subtly belittle the poem's protagonist.
The second is the way her translation describes Penelope's maidservants. In the poem, they sleep with Penelope's suitors and are then brutally slaughtered by Odysseus upon his return. Viewed through the ethics of bronze-age Greece (as mediated through an iron-aged poem), Odysseus is fully in the right with his righteous fury: the servants were disloyal to their house and their master, and received their just desserts. Wilson reframes them as slaves who were raped, and casts judgment upon Odysseus for his cruelty towards them. Wilson defends this decision by arguing that these women's enslaved status would have been assumed by Homer's audience, and that this is clearly rape as we understand it today. Which, again, isn't necessarily wrong, but IMO completely misses the point of a translation. She's replacing the values of Homer with the values of Emily Wilson, centering her desire to lecture her audience about misogyny over her desire to faithfully translate the source material.
I also really dislike her decision to stick to very strict iambic pentameter in an attempt to recapture the rhythm of the original Greek. She makes absolutely zero effort to be poetic in her language apart from following the meter, so it's mostly invisible to the reader. It doesn't read like poetry. But the restrictive meter forces her to constantly use less-than-ideal words and eliminate important details. Ironically, her efforts to make the Odyssey feel more like a piece of oral poetry completely strip all the poetry out of the work, and leave the finished product feeling limp and mundane.
> Wilson defends this decision by arguing that these women's enslaved status would have been assumed by Homer's audience, and that this is clearly rape as we understand it today
I'm not sure how this makes sense anyway. If they were slaves they belonged to Odysseus anyway, not the suitors, so their enslaved status did not oblige them in any way to have sex with the suitors.
Anyway I've only read the Fagles translation, which avoids all these problems. If I were to read it again I think I'd go with the Fitzgerald version since it's the most pleasant to read based on this comparison here https://emily613.substack.com/p/translating-the-beginning-of-the with Lawrence second, Fagles third and Wilson a distant fourth. (Though I like the _idea_ of Lawrence's translation, as a reminder of days when people were good at multiple things.)
In general I care less about word-for-word felicity to the original and more about having something that's generally pleasant enough to read. But I feel like if the translator is altering the text not for clarity nor poetry nor comprehensibility but instead to emphasise that the hero is bad by modern standards, then they're probably really missing the point.
I agree, Fagles is my favorite, mostly because I think his is the most pleasant to read. "Man of twists and turns" is such a perfect description of the character (with such lovely alliteration) and is emblematic of the sorts of choices Fagles makes throughout his translation.
Thanks for the insight! This is pretty much all I discovered from searching online, so hearing that you're familiar with multiple translations is helpful. I wasn't sure if these were just two of many that everyone just happened to talk about, or if they were basically the only two things that stood out.
I don't struggle too much with either of these, but I can understand why someone would. As for the polytropos issues, she addressed it in a blog post [1]. Even if she made it all up after the fact, which I sincerely doubt, it's still sufficient reasoning for me.
As for the second issue, considering ancient Greek culture, it would seem more likely to me that the women are slaves and that earlier translations soften the situation. In my opinion, it feels more faithful to describe the situation as is, even if it won't be viewed the same to a modern audience, than to change the situation into something that it very likely wasn't in an attempt to make it more palatable for modern audiences. I read an article that covered this issue [2] and it sounded more legit than anything else I could find on the matter.
I should also point out that, as a Christian, I by no means share most progressive left views, but I also try to keep my opposition to them from biasing my judgment of things in general. To me, it seems that the case is that the Wilson translation's aim is to present the story as it was told in a more literal way and allow the reader to think what they may, while older English translations add flourishes to try and capture the poeticism that I guess is inherent to Homeric Greek that is lost in direct translation to English, while also adding slants and modifying things here and there in order to present a story that is a little more acceptable to modern people. It's just a matter of what exactly is being translated, as translating complete, original meaning of a thing across language, culture, and thousands of years just isn't something that can be done.
Ultimately, though, it's just entertainment material and people can think what they want of it. I appreciate you taking the time to type out that comment! It helps me to understand more what the controversy is, and that she didn't necessarily rewrite complete passages or something like that.
[1] - https://emily613.substack.com/p/on-complicated
[2] - https://ekklesiamag.wordpress.com/the-case-for-emily-wilson-in-wilson-vs-right-wing-twitter/
I've seen the blog post where she explains her choices, and think she provides reasonable explanations. But I think her self-consciously feminist approach shifts the focus of the scene in a way I don't really agree with. It's not that the older translations softened the terminology to spare the (then) modern reader the literal meaning. It's that Homer and his audience would have considered the fact the women are slaves to be totally inconsequential. The point was that the household servants failed in their duty and were justly punished. Centering their experience in the mind of the reader would be unthinkable to the original author and his intended audience. She is purposefully inviting the reader's moral judgment in a way that is at odds with the tone of the story.
She has a point, but that point belongs in an explanatory end note, not the body of the translation itself.
No translation is perfect, and I'll agree that most prior translators have kind of made a hash out of the rape and murder passage. In isolation I probably wouldn't consider Wilson's choices there particularly worth remarking upon. It's the larger pattern that frustrates me. Hearing Wilson talk about her translation, especially in broader media interviews, she sounds like she's got an ax to grind. Every time there's an opportunity to interpret an ambiguous passage, she does so in the maximally feminist way. This, combined with other quirks (e.g., offputting modernisms and the tortured meter) and the way the translation has been received by the sorts of education reformers I instinctively distrust as the quintessential modern translation that everyone should be taught in school, makes me highly reticent to recommend Wilson's translation to others. Every individual choice is justifiable. But the way every choice seems to point in a particular ideological direction makes it feel like a political project I have no interest supporting.
I have no technical knowledge about her translation. All I can tell you is that I read her translation probably 8 years after the first Odyssey translation I read (can’t remember by who), and I really enjoyed it.
I found it immensely readable and would recommend.
I have now seen the final installment of The Amazing Digital Circus in the cinema.
* That was great! Would recommend.
* Was worth seeing in the cinema with an audience for the big screen and the audience reaction
* Wow, the jokes are dark. And I say this as a fan of Neon Genesis Evangelion. (Ragatha is a worthy successor to Asuka Langley Soryu)
* I’ll avoid mentioning specific plot points so as not to give spoilers.
By way of comparison, Evangelion also did the last installment in the cinema. (The End of Evangelion, Hideaki Anno, 1997).
DataSecretsLox brings three entries into its June 2026 Effortpost Contest.
https://www.datasecretslox.com/index.php/topic,15995.0.html
First is SamChevre delivering a perspective on depression. https://www.datasecretslox.com/index.php/topic,15824.msg789607.html
Second is Chevalier Mal Fet with the latest in his series on battleground history: the Battle of Wilson's Creek (American Civil War). https://www.datasecretslox.com/index.php/topic,15858.msg791411.html
Third is Mark V Anderson's review of Stephen M Walt's book on post-Cold War US foreign policy, _The Hell of Good Intentions_. https://www.datasecretslox.com/index.php/topic,15986.msg795705.html
Didn’t there used to be like a dozen or so each month? Did y’all have a brain drain hahahha
IME it ebbs and flows. It's often a product of who's busy and who feels interested in writing up whatever they're interested in.
It's more like a gradual decline. I think a lot of the most interesting members left years ago, or else ran out of things to write. But for a long time, the numbers were boosted by relatively low quality "effortposts" which covered up the decline. It's pretty common to see months when *none* of the contest entries were worth reading.
DSL also has a Blog Roll - a separate forum that autoposts entries from people who've registered their blog with DSL. The latest one is theredsheep, writing about how families deal with a dying loved one.
https://breathingfree.substack.com/p/sandy
TRS has written several long posts on DSL in the past. He's a respiratory specialist, and inserts a great deal of interesting detail about the job and the patients that go through a hospital. He's submitted effortposts to the contest, and I think won at least one of them. And now he has his own blog. That post about Sandy would no doubt otherwise be an effortpost contender.
TRS isn't the only one to follow this path. I think it's also contributed to some evaporative cooling at DSL. OTOH, DSL still serves as one more hub for people to visit for posts like this. So far.
My wife just had a double mastectomy. Her biopsy came clean, 2mm invasive cancer inside a 2cm DCIS (non-invasive but cancer positive). One of our biggest discussions prior to the surgery was whether to do a lymph node biopsy. There is a bad side effect called lymphedema, so we spent a lot of time talking to AI about whether we should avoid that biopsy. Doing the biopsy of the lymph node is standard of care. But for my life, I could not convince myself that the statistical analysis makes sense.
It goes like this. If you have invasive cancer, and it's spreading, it will infect the lymph nodes first. So the lymph node biopsy can give you a signal that you can use to decide how strong should chemo/radiation be. But removing it adds no other value than this signal and the side effects can be bad. In the case of my wife, there was nothing in the lymph nodes (they took 5 of them because it's hard to target specifically).
The reason they always do it with a mastectomy is that in order to find the lymph nodes they inject a marker in your bloodstream and, as I understand, the flow of blood from the breasts won't be there after the mastectomy, which makes finding a single or a small group of nodes very difficult, so future biopsies would involve removing all of them, which is worse. Let it be noted, that even when they remove all of them for biopsy, it doesn't cure anything-- it's strictly to make a determination of spread.
However, when they removed the breasts, they perform an extensive biopsy. So the way I see it, the full breast biopsy should provide a very good signal on whether the cancer is all over the place or contained. If you have little clusters of invasive cancer all over the tissue, or something along those lines, then it's probably invasive and you should do chemo.
Chemo and radiation, by the way, will mess up your lymph nodes anyway.
So the whole standard of care fucks with your lymph nodes no matter what.
Claude told me of some clinical trials where they don't remove them and the survival rate didn't change, which is consistent with what I interpreted.
I want to be clear that I don't have deep knowledge about the fundamentals. It's just that I kept going in circles with Claude, exasperated because the logic didn't make sense and Claude kept trying to make it make sense.
1. Sorry about your wife’s boobs, that sucks, but hopefully the procedure is successful and means you guys don’t have to worry about cancer ever again.
2. Sorry about the informational labyrinth at the same time, that’s super frustrating. I have no answers or pointers, just want to say I perceive your frustrating situation and wish for your and your wife’s suffering to be concluded soon.
Maybe if I were to suggest one thing it’d be to talk this through with a human oncologist, and leave the silly computer program turned off for a while.
I'll start by saying I'm sorry for what your wife and you have been through, and I sincerely hope it's the last time either of you have to worry about cancer. I should also say that I am not a doctor, nor any kind of cancer biologist. Nor have I looked this up online. But speaking as someone who has done a fair bit of histology - namely, looking at cells on slides, which I think is the main method of diagnosing the presence of cancer - my guess would be that it is to be 100% certain that it has genuinely spread.
The thing with a biopsy is that it is a small section of the tissue. A slide is about 2.5cm by 7cm, and the thickness I use for my work is 8um. Even if you take hundreds of slides from each breast, you will not have a full, comprehensive image of what is going on. I presume they will have a fair idea of where the main mass of the tumour is thanks to previous scans, but they won't know if a small clot of cancerous cells visible at the edge of the breast is actually connected to the tumour or genuinely spreading, because they won't have the slides representing the bits between their sample slide and the tumour. And that's assuming that they are actually taking tissue sections, which preserve the cellular layout. They might just be taking a chunk of the tissue, dissociating the cells and spreading them on a slide. In that case, they've lost all information about where the cells were beyond the basic location of the initial tissue chunk. However, the presence of cancer cells in a completely separate tissue - the lymph node - is unambiguous.
My guess would be that this procedure was established some time ago, and it may indeed be possible to use the primary tumour site in order to assess spread. I wouldn't be surprised at all to find out people are working on it, but medical procedures take time to change. I'd also guess that this might be established practice based on studies showing accuracy of different methods. A search on PubMed might help you work that out.
Super helpful context. That’s another suspicion I had, that there’s no way they analyzed the entirety of the two boobs when they say they only found a certain amount of cancer.
It also made me wonder how many people have cancer their whole life which doesn’t spread or even goes away on its own. Apparently they do biopsies on dead people’s boobs to study this and about 10% of women have some DCIS and about 1% have some malign cancer. But again, not clear how comprehensive these biopsies are and if they really find every 2mm cancer spot.
I have been enormously helped by picking GPT's info about medical stuff. I just had a huge back surgery and am healing & doing rehab, & it has help me with a lot of confusing issues and judgment calls. Here are some suggestions for getting more clarity about the lymph node issue:
>However, when they removed the breasts, they perform an extensive biopsy. So the way I see it, the full breast biopsy should provide a very good signal on whether the cancer is all over the place or contained. If you have little clusters of invasive cancer all over the tissue, or something along those lines, then it's probably invasive and you should do chemo.
That's what you believe, and it fits with my intuitions and common sense too, but is it true? I'm sure there has been extensive research on how good a predictor of distant metastases the amount cancer in the removed breast actually is. My guess is that is not as good a predictor as lymph node biopsies are. But ask Claude for the facts on this one.
>Claude told me of some clinical trials where they don't remove them and the survival rate didn't change, which is consistent with what I interpreted.
OK, if you ask for trials with that result, it will give you some. But ask it for the big picture that emerges when ALL well-done trials are considered. If the result is that lymph nodes biopsies give no advantage then yeah, it's dumb to do them. But maybe most studies show they improve the doc's estimate of chance of spread. Then they probably do make sense, unless the advantage is tiny. So ask Claude for the big picture result.
>Chemo and radiation, by the way, will mess up your lymph nodes anyway.
So the whole standard of care fucks with your lymph nodes no matter what.
If chemo and radiation harm the lymph nodes some, but aren't as harmful as removal of them, and aren't as likely to cause lymphedema, then it seems like you should still be in favor of the docs not removing the lymph nodes, unless, as I mentioned above, taking out the nodes gives them some really useful info.
Overall I think you should try to get more clarity from Claude by asking it for big-picture info and paying attention to that. And also, try to accept the built-in uncertainty of the painful situation you and your wife are in. The IS no way to know for sure whether the cancer has spread, and what would be the ideal follow-up chemo. The docs have to make guesses from what info they have, based on data about other people's treatment results. But they have a LOT of data on that. The reason you can't feel confident and sure that everything is being done in a good way isn't Claude -- it's that there's no way anyone can be sure at this point. Doctors are playing this in a way that usually improves the patient's odds, and that's a good thing, but not a guarantee. All of life is like this, really, you know? When you go out in a car there's a chance you'll get smashed by a trucker who passed out at the wheel. You can do things to make that less likely, but there's no way to be positive you won't get killed on that drive. But of course with something like cancer we are all suddenly horribly aware of the built-in uncertainties, and how everything's a gamble. I hope things work out well for you and your wife.
Thank you. Very thoughtful and actionable.
I’m hosting an EA meetup in Toronto on Saturday, June 20th. Come say hi!! :) https://luma.com/8dqawc0k
A rather jarring thing happened today that made it evident, in a way that reading things online does not, that our economy really is struggling.
I work in a fulfillment center for Ama, er, an Extremely Large Online Retailer. It’s an unglamorous, intensely physical job, but I like it enough and the modest but decent pay - more than most physical retailers except maybe Costco - is sufficient for my purposes. Still, it would not be an appropriate job for an ambitious young person intent on climbing the corporate ladder.
During break I was talking with a young woman who has been there a few months, and she said that she had graduated two years ago from an Ivy League university. It was difficult not to convey how surprised I was. If economic conditions were less wretched she’d certainly be at a *much* better job. A white collar job, not working in a warehouse.
She seems upbeat but it could be she is good at hiding her dismay.
I left a white collar career with a 6-figure salary and now work in the trades, starting at about a third of my former pay, without a drop of sorrow or dismay or regret. The terrible job market for my previous career was the catalyst for my career change, so you could say it was “economic conditions”, but during and after the transition my sense was not of leaving a palace for a hovel, or even of fleeing a sinking ship, but of finally arriving in a place of light after being in darkness. There are things more important than money once you find out how to live on enough of it.
Edit: to state my point more bluntly, maybe she likes working in a warehouse and finds the money & benefits to be enough. Definitions of "success" are a funny thing.
What trade did you pick? I quit video editing to go into the trades, so I've been around that world. I liked it.
Electrician. I was a Hollywood film editor in a career previous to my white-collar one too, that’s a funny coincidence.
That is a funny coincidence. Were you editing film, or were you doing digital editing (Avid, etc.)? Electric is a good trade. I was kind of a jack of all of them:
- Framing
- carpentry
- drywall
- plastering
- electric work
- plumbing work
I wasn't a standout in any of them, but I could renovate someone's bathroom or kitchen from the ground up.
That’s cool, I’ve met a few do-it-all renovator guys through work, they’r awesome. I take it you don’t do that work now?
I edited lots of actual film in film school, so had that experience at least, but when I decided to specialize in editing professionally the projects I was involved with happened to be all digital. I was pretty ambidextrous re. Avid, Final Cut (it was 7 back then), Adobe suite, etc.
No, I am retired. Wonderful!! I do what I feel like, no more no less. I started editing in 1974, so only film until video tape encroached. I hated vt. Got back into it in the 90s when Avid got legs. I liked it. Gave it up in 2003 and went back to building stuff.
DId she say what her major was?
The US economy seems to be going through a stagnant period with high interest rates and above-average inflation. There is role-transitioning in white collar jobs due to experiments with LLMs. New grads are being hired at a lower rate; America may have a situation like China's where there's a mismatch between credentials and open jobs. Things might pick up, but often times this type of early failure to launch haunts workers through their working lives. Sucks.
Didn’t ask her major, but really, in normal times just about anything from an Ivy should open doors.
You are right about the consequences of early failure. If the young woman tries to get a white collar job (for all I know she’s been trying) her having worked at a non-office job could be a big drawback, at least until the economy improves, if it ever will.
I suppose her fate isn’t too dire, the Extremely Large Online Retailer pays sort of okay, comparable I suppose to some office jobs, and has terrific benefits with a far more relaxed atmosphere than in any office.* If she were a state directional university grad she would be doing fine. But she’s an Ivy grad, which makes all the difference.
* = I see more women’s bare bellies at work than in any job short of lifeguard.
Does anyone know if there is a good reputable way to have my novice music arrangements consulted and get some pointers and clean-up, like an editor but for jazz? I have been arranging for a year but am doing my first originals and would like to get someone better than me to discuss them and help, in the soul jazz genre. My old jazz teacher from long ago seems unreachable and I got no other ideas and figured if I just put this out to Reddit I’d get conned, this isn’t a classified so I’m not asking for YOU dear reader but if you have someone you might point me to that would give a novice a little time and can recommend them that’d be great, or if there’s a good resource to point me to
I forget if it was Duke Ellington or Charlie Parker, maybe another of those greats, who said “if it sounds good, it is good.” So there’s a case to be made that you don’t need “reputable”, you just need a pair of ears. But I get it, you want constructive criticism.
Are you just looking for detailed feedback from someone with music theory vocabulary and knowledge of jazz? There are people here who can do that, you can ask us directly.
"If it sounds good it is good." I know this quote from Peter Shickely and Shickley mix, which I would listen to religiously when I lived in Nashville years ago. (According to Peter the quote is from Duke Ellington.)
The top level issue of "sounds good" is a little hard to manage while writing on Musescore, everything's going to sound better with live instruments but "you never get a second chance to make a first impression" to quote, um, a deodorant ad.
There are a variety of issues where I am slowly converting unknown unknowns into known unknowns into actual learning, but I just need people to spot these things. The main structural issues are writing and integrating a bassline (which in my band is handled by a tuba not a traditional string bass who would improvise some of it), usage of octaves vs harmonization in a band where instrumentation may vary gig to gig, when I do use harmonization there's issues with blending and having intervals that sound right for the instruments voicing, just lots of little stuff where I feel I'd benefit from the equivalent of having my professor look at the rough draft of a paper and spot the top level stuff then slowly work down to more details.
I could just toss it up somewhere and send a link when it's ready and solicit feedback but I thought maybe that was frowned upon.
Aw man you have a tubist for the bass in your band! That's awesome. There was a band whose lineup was two tubas and drums, maybe also a sax or trumpet or electric guitar or something, I can't remember what they were called but they were dope.
Anyway, I think if you're asking for feedback based on MIDI approximations, yeah that's gonna be tough, always better to try and get the best recording you can first, but I don't see any issue with soliciting feedback.
Has anyone else noticed a major regression in Claude Code's ability to work independently recently? Not because of intelligence, but because of harness issues.
Instead of using the normal file read and write tools, Claude recently switched to using complicated grep/sed/echo chains to do everything. Presumably this is an attempt to save tokens by only reading part of the file, but the problem is that this means that *every single command* has to be manually approved. Claude can no longer work independently, because it is now doing everything in ways that force the user to sit there and mash approve. It's very annoying.
I've noticed some inappropriate/frivolous tool use in Opus 4.8 in a number of contexts not just relating to coding. I wonder if this is a training/alignment failure in the same way where OpenAI previously mentioned ChatGPT frivolously using its calculator to do things like add 1+1 to get better tool use scores in training.
Claude running ten commands to accomplish what could be done with one might look better in a naive training environment which rewards successful tool uses and hence values 10 valid tool calls over 1?
It's been like this since at least 4.7. It also has a really unfortunate pattern of using some combination of grep/sed/head/tail to extract some part of standard output, noticing the information it needs wasn't in that part, and running the whole command again to get a different bit, which has exactly the problems you'd expect if the command is long-running and/or isn't idempotent. No amount of instructing it to pipe output to a file then use whatever extraction tools it wants to examine that at leisure helps - at best it pipes the tiny subset of output it wanted to a file then cats that, when it bothers to obey at all.
Still, terrible as it is in all ways, I'm impressed that the dog sings at all.