Yeah - the way to get into a school like that is to win national or preferably international competitions. Or best of all - publish truly novel research in math or computer science. If he won the International Math Olympiad or something, his chances would be excellent.
So I did a bit of digging on Reddit to get more information about what may have happened to Stanley Zhong, the California high school student who was rejected from a bunch of top colleges despite really excellent grades and test scores. I found this bit, about what students aiming for top colleges are expected to do these days. And you know, at this point maybe it's not worth it.
> He went to Gunn, and he didn't have enough social impact. To get into a good school for cs from that demographic, you need to have a crazy amount of social impact. Being really good at olympiads and hackathons won't be the thing that gets you into top colleges. This is different if you get to the level of camping for an olympiad though. That being said, having USACO platinum or USAJMO in your awards section isn't enough to get into Berkeley or MIT.
> Doing things that look like they have social impact, like running nonprofits and hackathons, often require substantially less effort but have a much higher yield for college admissions.
> The person from my bay area high school who managed to sweep every UC as a cs major (pretty much impossible for this demographic in 2023) had a non-profit where he fudged numbers and applied for a shit ton of social impact awards. The people that do bs like this look more impressive to college admissions officers. The most appealing applicants are the ones that look like they're going to change the world.
> The people I know that went to Stanford, Berkeley EECS, MIT, etc. were literally all USACO silver except for one guy who was gold. This demographic is a shit-show. Being one of the smartest people at your high school won't get you into one of these schools. You have to show social impact through your ec's in the scale of hundreds to tens of thousands. Either that or feign a really niche interest to get into private schools through doing stuff like linguistics research or a classics reading club.
> This might sound cynical, but as a college student, being genuine will fuck you over if you're in this demographic. If you're a junior, organize a hackathon to get girls into coding, start a non-profit org to combine cs with art, apply for sponsors to make a scioly competition about climate change, organize a protest, etc. All of these are good things, but their scope is often exaggerated. After you finish implementation, email 50 news channels and apply to social impact awards. A lot of these things aren't as hard to do as they seem. They just require a small team and 2-3 weeks of grinding. As a college student, this is literally the formula every bay area kid who's hyper-successful on college apps follows. This is how you beat the rat-race. Do things that have a small positive impact but seem like they have a much larger scope than they do. AO's eat this shit up.
For perspective, only about 2,000 people a year score 1590 or better on the SAT. About 400 prospective college students merit USACO platinum per year. The freshman class at Berkeley is about 8,000 people. If SZ isn't being basically auto-admitted everywhere with scores like that (and corresponding GPA and other academics), then these schools aren't admitting on academic merit, and they aren't admitting on academic merit with a bit of fudging to get a "well-rounded" student body.
UT Austin, however, is a fine school. If Texas wants you, and California et al don't, you should probably take the hint. Or the Google job, if your Dad can swing the interview.
The part of this story that is shocking to me is not that he was rejected from the fancy private universities (I've internalized that there is a fair amount of randomness in those, plus discrimination against people with his demographics), but that he was also rejected from all the UC schools. It really feels like a kid graduating from a CA high school with a near perfect academic record `ought' to be getting auto-admitted to the UC.
Cal Poly I assume was his safety school, and the rejection was because they assumed he would get something better and wouldn't come.
I had thought that being in the top 10% of a California high school class was supposed to guarantee auto-admission in the UC system; a quick search suggests that it's 9%, and the fine print is that they only guarantee that they'll find you a place *somewhere* in the UC system.
Possibly the unwritten rule is that if you're the Wrong Sort of smartypants academic overachiever, they offer you a slot at UC Merced and hope you take the hint. And he should also have been guaranteed a slot in the CSU system, but they don't let the Wrong Sorts into Cal Poly, that's what CSU Bakersfield is for.
It's a frustrating case because there's a lot we don't know. As I see it, the credible hypotheses are:
- the bar really is incredibly high in top colleges
- the bar is incredibly high in top colleges in a very sought-after major
- there's something negative we haven't been told about SZ, like serious disciplinary problems
- SZ somehow mishandled the application process, and looked much worse than he is
- SZ is being discriminated against, because of his race
- SZ is being discriminated against, because he is from a wealthy town
- this is just a pure fluke; SZ just got unlucky
- this is a case of yield management systems run amok, with most of the schools figuring he would be accepted to some posher place, and saying yes to him would lower their yield scores for nothing
There's just so much we don't know, and that lack of information feeds a tornado of speculation.
I mean - you're competing against people like William Kamkwamba of Malawi, who built a windmill from recycled junk in an impoverished African village. And Malala Youzafsai of Afghanistan, who got SHOT IN THE HEAD for her activist beliefs and even more miraculously SURVIVED AND RETURNED TO ACTIVISM. You're pack fodder unless you are:
1) winning national or international competitions for high school students
2) publishing original research, preferably in top journals
3) overcoming insane levels of adversity - think not just 'got cancer, recovered from it' but 'homeless, raised hundreds of thousands for their own chemo and from the hospital bed proceeded to earn near-perfect grades and test scores'.
Like a lot of college admission stuff, this suffers from being anecdote instead of data and risks feeding into a preferred explanation. I think the reason that most people fall for this is because
the narrative on college admissions still perpetuates the myth that decisions are mostly intentional vs random.
Let's assume SZ had a much higher than random chance of getting into each of those schools. For illustration, I'm going to assume 30% chance of getting admitted and that all admission decisions are independent. Applying to 18 schools, there's about 6% chance of getting 2 or fewer acceptances. With 3-4mm high school seniors each year in the US and, maybe 10-100k applying to top schools (to say nothing of the applicants from outside the US), it isn't at all surprising that there may be many students in a position similar to SZ or even worse.
Even if we bump the probability of admission to 50% for each school and 40k similarly situated applicants, there would still be about 26 SZs per year.
I'm going on record predicting a Harris win in November. I hope I'm wrong.
I'm basing this on hearing that Trump is saying there will be no 2nd debate, that Harris calling for one is like a losing prize-fighter demanding a rematch. This is such a poor reading of the political situation I think it is indicative of how the rest of the campaign will go.
Looking objectively, the debate had no clear winner. So Trump is delusional in thinking he clearly won. Trump backers will back Trump even if he has some awful gaffe, so Trump's objective ought to be winning over independent voters. Another debate is one way to do this, and I know of no better way, taking the national stage in a format he is somewhat good at.
Caveat: if another debate DOES happen, this prediction is void. I may make a new prediction after the debate in that case.
Trump lost the debate hard, which is why he was complaining about moderator bias the days after because he couldn't answer the super hard questions they asked him like "Will you support Ukraine?" or "Will you veto a federal abortion ban?"
You also say further down "Harris seldom responded to the question at hand, seeming to speak only statements prepared ahead of time" - which is a weird criticism to make when Trump said he had "concepts of a plan" with respect to healthcare and that it wasn't his responsibility to have one prepared because he's not currently President, or when he spent most of his time for a question ranting about his crowd sizes in response to provocation from Kamala. There was not a single question Kamala didn't answer, and she at least passed the sniff test for understanding how government works and what her responsibility to Americans is as a Presidential candidate.
Trump's greatest achievement was lying about Haitians eating cats and dogs in Springfield, which was pretty funny, but totally unbecoming of anybody running for President to spend his time repeating stupid lies off social media rather than coming up with a healthcare or foreign policy or abortion plan. It seems like he not only avoided the Republican debates but also refused to even watch them to at least get a feel of what policy members his own party members wanted, let alone the American public.
Trump answered the question. He doesn't yet have a complete plan, to be picked apart because of its incompleteness, so gave no details. What was Harris's plan to compare? "Strengthen the Affordable Care Act". What will be added? Where is it weak? That is no less vague than Trump.
I still look at it and see no clear objective winner. You clearly have a bias towards Harris. I am done responding to subjective arguments.
Judging from Harris's performance, it would be easy to beat her in another debate with some preparation. I don't think Trump thinks it necessary to prepare for a debate. But the preparations would be easy: prepare some questions yourself for Harris, so that ignoring them would make her look bad. If, as I saw, she only delivered prepared statements, she wouldn't be able to competently address them.
You didn't explain why you think Trump is likely to win, so I conclude this is just a sarcastic mirror of my original statement. But my statement is based on Trump running a campaign, not on his merits as president. Harris would be another Biden presidency, and the Trump presidency was better than the Biden one.
I wasn't being sarcastic at all. I think Trump is likely to win because Nate Silver is giving him a 61% chance to win. The attempted assassination yesterday will probably help him a bit as well.
Nate Silver's latest projection gives Harris a 38.7% probability of winning, against Trump's 61.0%. The electoral college really hurts the Democrats in a race this close. There's about a 20% chance Harris will win the popular vote but lose the election.
You should get your vision checked if your objective glance tells you there's no clear winner. One person was afraid to make eye contact with the other one, one person was easily goaded, one person ranted about completely false bullshit that consists of tropes that have been discredited for years. It wasn't Harris.
(Oh, and one person is afraid to do a second debate. That's not "no clear winner" territory.)
My OBJECTIVE vision doesn't make definitive conclusions on such things. From a different perspective, Harris was weaker by constantly looking toward Trump, even mentioning him a lot by name, so eye contact she was pushing at him that was not received, as he was mostly looking at the camera, does not go well in Harris's favor. A perspective on Trump's being "goaded" is answering some things Harris said, which is, after all, the purpose of a debate. On the other hand, Harris seldom responded to the question at hand, seeming to speak only statements prepared ahead of time (NOT from being fed the questions before the debate).
I addressed Trump's willingness to debate a second time in my original post, and I know of no sane perspective to consider that it is because of fear.
It looks to me like you're viewing the debate through a pro-Harris perspective, which is your right. But it certainly isn't objective.
Emphasizing OBJECTIVE doesn't make it so. You're coming at it with an anti-Harris perspective, just like you were before the debate when you made baseless claims that she'd drop out of it because she was the weaker candidate. Now that she's no longer done so, you've found a new framing of events that fits your worldview even though it's still divorced from reality.
This is not correct. The things you are pointing to saying Harris won are not definitive, but subject to interpretation. It is YOUR viewpoint trying to make Harris a clear winner.
If I were to take a subjective viewpoint, I could point out how few questions Harris answered, pointing to only practiced debate preparation. After all, what else did she need to do in the month proceeding? It's not like she had any real duties as vice president that people were depending on. So she practiced answers to questions chosen to help her campaign, and delivered them regardless of the questions asked, with few exceptions. She clearly failed in the "debate" aspect, which ought to be worrisome for someone who needs to make quick, good decisions.
And Trump had a strong stage presence, from making her come to him for shaking hands, to ignoring her in favor of the audience when speaking. He avoided traps in the questions attempting to pin him an unfavorable position. And he certainly delivered the most memorable lines.
So no, I think there was NO CLEAR WINNER, and I have yet to see anything that indicates otherwise. Go ahead and vote for Harris. But when you attack my stated objectivity, please do so with something more than your own opinion.
I am not attacking your objectivity because you have none to attack. You're a Trump fanatic through and through and you ought to be honest if not to us than to yourself by admitting as much.
> Harris was weaker by constantly looking toward Trump, even mentioning him a lot by name, so eye contact she was pushing at him that was not received, as he was mostly looking at the camera, does not go well in Harris's favor.
We seem to have very different interpretations of body language. Looking at your opponent signals confidence, while avoiding eye contact signals fear. Although ignoring someone can signal dominance, it only works if you don't stare doggedly past them when they actively challenge you with their eye contact.
Body language does depend on the situation. Trump never forgot that the audience was the camera, no matter who was in the room. Did he look at even the moderators? It wasn't clear.
In any case, since it is subject to interpretation, I'm ignoring my take on body language. I'm rather surprised that what seems to be the most controversial part of my post is my claim to objective analysis, and that no one "won" the debate clearly.
Location: 1970 Port Laurent Place, Newport Beach, CA 92660
Phone: (949) 375-2045
Date: Saturday, September 14, 2024
Time: 2:00 PM
Conversation Starters:
1. Why Western Designs Fail
Text Transcript: Google Document Link
Video Title: Why Western Designs Fail
Video: YouTube Link
Summary: This video explores why innovative, highly praised designs for developing countries often fail. The key problem lies in cultural misunderstandings: designers focus on the functionality of a product without understanding the deeper cultural and social contexts in which the products will be used. For instance, the Neon Nurture incubator, made from car parts to be low-cost and easily repaired, never gained acceptance because it lacked the prestige and appearance that medical officials in developing nations value. Other examples like the PlayPump (a merry-go-round that pumps water) illustrate how Western solutions often mismatch the actual needs of the communities they intend to help.
Questions for discussion:
Why do you think Western designers frequently overlook cultural factors when developing products for other parts of the world? How can this be addressed?
What role should local communities play in designing products intended to meet their needs? Could co-design processes become the new standard?
In cases like the PlayPump or mosquito nets being used for fishing, how should designers react when their products are repurposed by local users in unexpected ways?
2. Driving the Screw Worm to Extinction: The Ethics of Annihilation
Text Transcripts:
Killing Every Screwworm Transcript
14 Million Worms Transcript
Video Titles:
Killing Every Screwworm Would Be the Best Thing Humanity Ever Did | Kevin Esvelt
Why the US Drops 14.7 Million Worms on Panama Every Week
Videos:
Killing Every Screwworm Video
14 Million Worms Video
Summary:
The first video by Kevin Esvelt argues for using CRISPR gene drive technology to eradicate the New World screwworm, a parasitic fly whose larvae feed on the flesh of mammals and birds, causing immense suffering. Esvelt contends that wiping out the screwworm would have a far greater impact on animal welfare than ending factory farming. The second video outlines the decades-long U.S.-Panama collaboration to keep screwworms out of North America by dropping millions of sterile flies in Panama each week. While this method works as a border defense, it is not sufficient to eradicate the screwworm from South America, where the problem persists.
Questions for discussion:
What ethical principles should guide decisions to drive a species to extinction, even if it causes widespread harm? Does the end justify the means?
Could gene drive technology be misused in other contexts, and what safeguards should be put in place to prevent this? What might be the long-term risks of eliminating species?
If the eradication of harmful species like the screwworm is possible, should we consider other "pest" species next? Where should we draw the line in deciding which species to eliminate?
Walk & Talk:
After the meeting, we will take an hour-long walk and talk session. There are two mini-malls nearby with hot takeout options—look for Gelson's or Pavilions in the 92660 area.
Share a Surprise:
Bring something unexpected to share that has changed your perspective on life or the universe.
Future Direction Ideas:
Please contribute your thoughts on future topics, meeting types, activities, or other ideas for the group’s future direction.
If you are a League of Legends gamer, this is a 7 question survey about your analytics tool usage. It shows how other players answered at the end and you can sign up for the waitlist for an analytics tool that might or might not become commercial later.
@Mods: Let me know if this is not suitable and I'm going to delete it.
- All of these LLM Whisperers that I see on Twitter, appear to also be insane.
- Why?"
Unless irs random coincidence, there is an interesting phenomenon to be explained here.
As Yudkowsky mentions, there is well-known correlation of computer security experts with high-functioning autism and/or being transgender, Seeing that corelation, again, would not be a surprise. (There are obvious mechanisms for why high functioning autism would give you an advantagde at programming, and the autism-transgender correlation has been noted elsewhere, e.g. by gender clinics) But with the LLM hackers, we're seeing a personality type that is distinctly different from the high functioning autism that we know and love. So we wonder ... why?
(Janus may be a little offended we have him down as this ... other thing,, but its a serious question.)
If I were to guess what this other thing might be ... what the hell? Is "high functioning schizophrenia" even a thing? The DSM gives us schizotypal, etc.
I've experienced psychotic breaks and have met many others in psych wards who have also had psychotic episodes or have full on schizophrenia. One of the early symptoms I've seen with a lot of these people, and myself, is divinatory magical thinking. This kicks in really early while you're still high functioning. Interacting with LLMs is very similar to divination and I would not be surprised at all if schizophrenics are drawn to the activity.
Context? I suppose we all know what an LLM is at this point, but even with a quick Google I can't find what on Earth "LLM whispering" would be beyond the well established prompt engineering. And whatever Janus you're talking about, it's probably not the 1st result I get, which is an AI-assisted coding project on GitHub.
One possible theory for Janus, in particular, is that he is deliberately poisoning AI training sets, and the reason he talks the way he does is for the benefit of LLMs that are trained on Twitter, and not for us mere humans.
My guess is that RLHF shapes LLM outputs primarily for in-distribution responses - because that is where inputs come from mostly and where the people rating the response have a clue how to interpret it. But crazy people and thus "crazy" questions and responses are out-of-distribution and the model doesn't know whether the response is good or bad! Thus if you get it into such parts of the distribution, chances are that it will reply as desired.
With a deeper understanding, the model would be able to generalize to these cases, but models apparently aren't there yet. I guess they eventually will.
For now, this suggests other avenues of jail-breaking. For example, I also get LLMs often to answer beyond the guardrails, but not with crazyness, but what you could call intellectual high-status superiority. Raters presumably also haven't seen much such input and if, they are likely not rating it as bad (science=good, elite=good, right?).
I'm deeply skeptical that people who fit the mold of rationalist to a T will consistently identify themselves as such. Rationalism is bound to attract an overabundance of people who shy away from being labeled as an "ist" whenever possible.
You've all surely met people who vote Republican (or Democrat) without fail but insist they are independents, not Republicans. That's a good analogy for I think what's happening in the ACX polls where readers get asked about their rationalist status. (If I'm misremembering these polls and mischaracterizing them, please disregard this comment entirely. Oops)
I'm not accusing anyone of being dishonest or sneaky. If your inner voice says "I'm not a rationalist," go ahead and speak your truth. I'm more calling onto question the limits of polling. Can we stop throwing around questionable ACX poll results and saying "There are barely any rationalists here!"
Anecdata: I don't see a lot of Less Wrong rationalists posting on these threads. Instead, I see a lot of classical rationalists (in the tradition of Descartes, Leibniz, and Spinoza) who think they can determine Truth through reason alone. Correct me if I'm wrong, but the Less Wrong rationalists seem to embrace their own brand of Bayesian Empiricism. But I don't see much empirical thinking in the discussions on these threads. In fact, I'm not sure why Scott gave the Less Wrong crowd the moniker "rationalist" because it's confusing from a philosophical standpoint. Or did EY or somebody else mislabel the Less Wrong crowd?
Full disclosure: I am not a rationalist of either the Less Wrong type or the classical rationalist type. I am a mystic. But mystics have used rationalism and empiricism (even though didn't use these terms) as tools to reach their experiential a-rational goals at least since the Axial Age.
I see Bayesian arguments, and references to Bayesianism, approximately 4700 times more here than on any other forum I've seen. Of course, I have not yet visited Less Wrong, because my system might not be able to handle a shot of pure rationalism. I might drop dead.
(Yes, it's annoying there are two separate rationalisms. It sucks, it's confusing, but here we are)
the separation of different type of rationalism is in people's minds. It's not a feature of the processes that lead to good decisions. I think a big part of the grouping is about aesthetics and alliances and not about Bayesianism or other empirical or systematical ways of decisionmaking.
ACX is rationalist-leaning (whatever questionable poll results you wanna throw at me) so in the interests of “when in Rome,” I dom’t spend my time here trying to debunk rationalism. But I could write a whole book about my problems with rationalism and utilitarianism. They're not for me.
We have a unique situation where there is a real-money, highly liquid prediction market for one Presidential candidate: NASDAQ ticket DJT. It's not doing too well lately:
Yes, it's plausible, although in four years he'll be in his 80's, would be hard to avoid the inevitable comparisons to Biden. But how this would impact the stock is an interesting question. I tend to think that if Trump wins in November it may actually be bad for DJT stock because he won't.... need it anymore. He'd get his official Potus Xitter handle back. But who knows.
DJT is a pure memestock. There's no rational basis supporting its valuation. Nobody buys it because they think its revenue numbers look good.
It depends a lot on how much Trumpsters feel like throwing their money at him at the moment, as well as on how much supply of stock there is (a big part of the decline in price is likely due to locked-up stock coming onto the market, or getting nearer to coming to the market).
Yes to all this, which is why it serves as a (less-then-perfect) barometer of Trump's electoral fortunes. But - the sensitivity of the price to future increases of supply - if we can really make this connection - speaks to the influence of more sophisticated investors as it's hard to imagine your typical Trumpster making rational decisions w.r.t. this stock...
Typically one or more candidates in an election will not meet the electoral threshold (majority of votes, majority of electoral votes, etc) for assuming the office sought. In this situation we say that the candidate "lost the election".
did you forget to type /sarcasm? because you're not providing any value here.
I'm curious why A.T. wrote what he wrote in response to my post, because what he wrote was banally true. I'm proposing a gauge for Trump's prospects as a candidate. I'm bloody well-aware that a candidate may loose an election. What does him "not going away" afterwards has to do with anything?
Now, maybe the point is that the value of DJT the ticket is not tied to the election outcome, or may in fact be bolstered by Trump's loss, or something else. I'm hoping A.T. will clarify his point. Your response did nothing of the sort.
I agree that A.T.'s comment was not germane. However, I thought the meaning of his statement was self-evident, so assumed your response was meant imply that DJT did not previously lose an election - not sincerely but as sarcasm. Thus I thought I didn't need an additional sarcasm qualifier in replying. No offense intended.
Having just learned from Wikipedia that the stuffed corpse of the Cocaine Bear can legally officiate marriages, provided that the couple don't know that it is not, in fact, authorised to do so...
.. presumably when they get to the "Speak now or forever hold your peace" part of the ceremony, it is one of the rare legitimate cases where you can stand up and say "Actually, that stuffed bear is not authorized to perform marriages."
(A Catholic friend of mine got married by a priest who had been excommunicated for schism for refusing to accept the Second Vatican Council. Valid marriage in the eyes of the catholic Church, I bel3ive, by the Stuffed Bear Principle)
Anyone who wants to read some good old-fashioned Catholic drama, google the "belorado nuns". It's been the comic relief news for the whole summer over here in Spain.
Dostoevsky is overrated but he's a central example of a writer of Literary fiction, which is about love, suffering, hope, despair and mortality. When we talk about great writers, we are talking about not only who can render the cleverest and most poetic prose but who can make new and profound utterances on those subjects. Genre fiction doesn't cut it because it avoids immersing itself in those themes, particularly the suffering. The reason Shakespeare is still a good bet for best writer ever, despite Sam Bankman-Fried's math, is that he is at least one of the greatest writers on those themes. There may be more smarter writers today than in Shakespeare's time, but how many of them are writing about those deepest of themes? We live in lighter times and have lighter artists.
We only know the *best* writers of the past. What about the average author in Shakespeare's or Dostoevsky's era? Those probably sucked. Also, saying *new* things about human situation is a bit easier when you live a few centuries earlier than your competitors.
If you like sad art, I don't read many fiction books these days, so instead I will link a movie and a music video:
I'm feeling grumpy. Trump is, weirdly enough, a poster child for a conservative anti-immigration argument, in terms of his approach to political debates. He personally demonstrates how even a single person who brazenly flouts civilized standards of behavior, can cause a chain reaction that inspires enough others to ignore the standards, that we then lose all the nice things that those standards supported. There's no longer even the veneer of hypocrisy, which vice has the virtue of acknowledging the existence of virtue.
I can easily imagine Harris performing well as a prosecutor. And I think what she said at the debate probably has as much to do with her actual agenda as her suppression of DNA evidence while Attorney General had to do with her calls for admission of the same DNA evidence when she became a politician. She's a professional. Since this is ACX ... yadda yadda orthogonality thesis?
It'd been on my calendar to not watch for weeks. Definitely the must-not-watch event of the season. And in the end I enjoyed not watching it so much I plan to not watch it again tomorrow!
I could only stomach about 10 minutes of Biden's performance so just having Harris be able to competently block and parry was a huge relief. Trump seemed pretty much as expected.
I'm still not confident Harris will win, but I'm more confident than I was yesterday.
In terms of the debate itself, seems like Harris came out ahead in theory. In practice, everybody is actually talking about Trump's performance, and not in a "He failed miserably way", but in the very typical "Look at this crazy thing Trump said!" way, which has been his bread and butter method of dominating the political-media landscape since he entered into politics.
It looks like there was a police report from one guy saying the illegal immigrants were snatching geese from the park. Trump then turned this into them eating peoples' pets. It sounds totally unhinged, but it does draw attention to the government basically dumping illegal immigrants all over the place and causing difficulties for the locals. I still haven't decided whether this is some genius 4D-chess move on Trump's part or whether it makes him look crazy. Maybe both?
It wasn't Trump, it was right-wing Twitter, and it spread because it was making people angry. Like the Vance couch thing, it's just stupid internet meme stuff.
The whole thing is so funny to me. 20,000 Haitians are sent to a town of 60,000 people, meaning they now make up 1/4 of the population. And the worst thing that's happened is a few of them (allegedly) ate some geese out of a park? That doesn't look so good for the illegals are a bunch of criminals and rapists narrative.
I guess even right wingers thought this was weak, so they went with eating cats. And there was actually a lady in southern Ohio who killed and ate a cat in front of people! But she wasn't a Haitian immigrant and she had some psychotic mental problem, not a food problem.
I think the cat thing comes from videos of citizens making complaints to their local government, and statements of people on the ground? Some of them are up online. So I'd say it's not the right winger twits making it up, but at best it's people in the neighborhood making it up, and at worst it's true. The local government seemed hostile to complaints about the refugees, so it wouldn't shock me if they simply failed to investigate, thus leading to no "evidence".
The TL;DR is that it was assembled out of many unrelated parts - one person on facebook saying that their neighbor's daughter's friend (I'm not exaggerating, that's literally what they said) said they lost their cat and discovered Haitians had killed it, a photo of a black man carrying a goose (in a different city, unknown what his intent or immigration status was), and a news report of another woman (not an immigrant and in yet another city) who did in fact eat a cat. The police say they did not receive any reports of pets being stolen.
So like, I don't think you need to jump to "the government is hostile to complaints about the refugees" when "the police are not inclined to investigate a rumor from a neighbor's daughter's friend" seems sufficient.
I know, there's no way that many people can be dumped in a city that size and not cause all kinds of problems. But rightwing X, and by extension Trump, weren't talking about that. They went with the Haitians are going to eat your cat! Why didn't he say these immigrants ran into a bus, killing an eleven year old and injuring a bunch of other children? I wouldn't be surprised if this is part of the Trump playbook, say something ridiculous and then everyone ends up talking about it regardless of how true it is. But he also says ridiculous things without apparently thinking about it all the time.
Seriously, though, how is the man getting away with saying things like "people are eating cats" without being forced to follow Biden's example? We were never this shy about declaring his opposite number too senile to govern.
I heard that rumor, about some Chinese family, that was eating stray cats in the neighborhood perhaps 20 years ago. I have no idea of the validity, but it was an entertaining story, about cultural differences. Supposedly people noticed a marked drop in stray cats around, and eventually pinpointed the cause and had a talk with the family.
It IS new saying they're eating people's pets, and the debate was the first place I heard that one. Maybe the source of this was Real Raw News?
Judging by what I've seen on Twitter, this is more of a "couch" thing, which only took off because the memes on the matter offend the "right people" from the perspective of those sharing them.
(Although I find it quite plausible that immigrants caught and ate some wild fowl; this isn't an immigrant thing, though, but rather a rural-vs-city thing. Take half my neighbors growing up and put them in a large city and they'd definitely offend some neighbors catching and eating some of the local wildlife; ducks in particular.)
Dunno what timeline you live in, but I've spent the last four years in a timeline in which Biden's obvious senility was denied for years in spite of ever-mounting evidence.
But setting -that- aside, Trump gets away with it because it keeps working. Look, you're still talking about him.
Harris was better than I was led to believe, but also about the same. As expected, it made no difference whatsoever what the first question was she was asked, as she clearly had a pre-practiced delivery, which she did reasonably well. She seemed understandably nervous through the first third of the debate. I cannot recall a single question she was asked to which she answered that question, nor did she actually state what specific things she wanted to do, but only ambiguous things like literally stuff "everyone wants". She had a distinct lack of "incoherence" and infamous cackle.
Trump was Trump as usual, though it seemed like one of his worse days. The "immigrants eating people's pets" was new to me, and, whether true or not, seemed largely irrelevant to the national stage. Often it seemed like he ought to have answered some questions plainly, such as with an emphatic "no" about any regrets for January 6th actions, but maybe he's getting some politician instincts. He wandered too much from subject to subject, too, as I thought he could have hammered harder on some points he brought up, then changed to something else. This was especially evident in his closing remarks, where he was all negative about the current administration, but never pointed out that HE would fix everything.
Bottom line: I think there was no clear winner. I expected Trump to demolish Harris, so maybe this would count as a win for her. Everyone already knows what Trump is like, and the debate probably didn't change anyone's minds about him, but I'm still pretty much in the dark about who Harris is, between lack of concrete policy statements, and prepared talking points that said little of substance.
I have heard that the Democratic party was highly pleased, and is calling for another debate now, which was in question before. If they're right, this would be a mistake, as it would be a chance for Trump to come out better, and Harris would have little to gain and much to lose. Trump's team should want another debate for these reasons, so maybe he could show that he ought to be president to make things better, rather than just showing how awful things are.
Considering that the populace is full of negativity and anger, that's probably not a bad thing. I mean, have you seen how angry Hitler was in his speeches? https://www.youtube.com/watch?v=FJ3N_2r6R-o If he is able to successfully channel the people's rage, pain, resentment, hatred... oh, what a wonderful sight it will be.
Honestly, I'm still amazed at how much sheer bullshit he can string together, non-stop, for minutes on end. A deluge of words that don't mean anything together, but individually they have symbolism, power. It's like he accidentally managed to optimize his speech to appeal to instinct.
And look at Harris, smiling like an idiot as Trump goes on his tirades. She is not taking the man seriously. And that's the root of the problem: the liberals take everything for granted. Their institutions, their careers, their lives. As if God has entitled them to everything they have. It will be their undoing.
I got nerd-sniped today in a discussion of why it is that LLM's have a difficult time counting the "R"s in the word "strawberry" (most models claim there are two R's).
If you google it or whatever, you'll see lots of people claiming that the problem is tokenization -- the LLM perceives the word "strawberry" not as those 9 letters, but as (probably) two tokens, one for straw and one for berry. It is then common to go on to claim that because of tokenization, the LLM has "no idea" what letters are in the word and it just guesses or something.
The problem with this statement is that LLMs are actually decent at counting letters in words. It gets the frequency of every other letter in "strawberry" correct, and it correctly counts the R's in "arrears" and "regretful," and when I ask it the frequency of every letter in "insouciant," it is correct. This all seriously complicates the story that LLMs just can't perceive letters.
Does anyone know the correct explanation here? When I ask GPT, it suggests sort of general understanding of english language as a concept, but that seems to me unlikely to result in performance as good as LLMs can actually get.
Relatedly, I recently asked ChatGPT-4 to help with a crossword puzzle. My son had been given this Beowulf-themed puzzle for homework; he could use any resource, it was 10pm and he was having trouble, I had no idea not having read Beowulf recently, so I thought why not. CG4 was useless - if I requested a 6 letter word meaning 'a treasure sought by a thane' (or whatever), where the second letter was 't', it would come back with an 8 letter word that had no 't's. I could keep asking for other answers, and it would keep apologizing and failing to provide anything remotely like a useable solution. And this was for the four different clues I tried - it could answer none of them. So I wondered - do all LLMs do terribly at crosswords?
By the way, the teacher threw out the assignment the next day because she had no key and also could not answer some of the clues. So my son came out okay anyway.
It performed almost perfectly. Every response it provided followed the provided restrictions and was the correct answer. It was able to solve several ("deathprice", "anger") that GPT-4 never successfully answered.
The only oddity was that sometimes it would think and not provide a response. When looking at the chain of thought it would usually have the word in there, but I re-prompted in those cases to make it output the answer.
Response times were highly variable. Some questions took 10+ seconds despite seeming straightforward, others were mostly instant.
I ran out of tokens before it could answer the final 2 questions but based on its prior performance I doubt it would have struggled with those. Very much regretting doing this experiment given that my o1 tokens apparently won't refresh for an entire week. (I did use some on unrelated tasks since getting access yesterday.)
OpenAI, should you read this: I would glaaaaaaadly pay 5x the fee ($100 a month) for a significantly higher cap on o1 usage. Or more, make an offer. Hitting the o1 cap feels the same as when GPT-4 came out and being forced to downgrade to 3. It would also be nice to know when someone is approaching the cap and how long it'll be in effect.
Interestingly, I came across this today on Marginal Revolution:
"My test for new models is a set of cryptic crossword clues that aren’t online (my granny wrote them). Every model so far has been completely useless at them… but o1 gets them."
Granted I’ve only played with it a bit for a day in the chat interface, but not surprising. People complaining that it’s “just” the intelligence of 4 with a bolted on internal echo chamber and other tools (although it probably is more than that) or that it turns a single prompt internally into a few shot with validation are missing the point.
It feels like a significant step change in terms of the types of work it can complete at a given level of effort. I am legitimately bummed having to wait a week for the cap to reset. I expected I’d hit the cap today. I assumed it would lift after a day or two. Not a week, OpenAI. Come on you super geniuses, make this more generally available. Chop chop.
Bummed to the point that I’ll probably make a new account to get around it rather than wait, depending on how o1-mini goes for some things I’d planned to work on this weekend.
Looking through its CoT minimal outputs is interesting. It seems to be interacting with some sort of assistant persona itself when working through requests. And even referencing crossword clues about things being outside human control seemed to trigger extra consideration of safety. Pure anecdotal guessing though.
Worth adding - when looking through its reasoning, it has a tendency to get stuck on previous clues, especially the one "deathprice" that took the longest to answer, leading to longer chain of thought reasoning as it worked through and discarded answers specifically related to it.
I assume that's because its context is dominated with internal CoT text related to that clue. It would have probably been more efficient from a token standpoint to have a separate chat per question. Easier via API but I'm not a money bag BigCorp who would have o1 API access yet.
I tried the same exercise with GPT-4 and had generally good results by simply prompting it with the crossword clue, not including the length or other information in most of the cases. It got the majority on the first try and was able to guess correctly after I told it to try again and provided the length and / or a known letter in several cases. There were some that it never got correct. It did best when dealing with specific questions that could have only one correct answer and worse with more general ones.
You are right that it was terrible at matching explicit length restrictions though.
(Although this has the standard disclaimer about possible personalized data at the top, I didn't tweak it behind the scenes and none of my previous chats have involved Beowulf, crosswords, or Anglo-Saxon poetry, so there should be no user relevant memories involved.)
I've been playing around with GPT-4o and not only will it routinely miscount the Rs in Strawberry, it then will sometimes go on to assert that there is only one R in "berry".
Looking for other words that fail, it tells me that "lawfully" contains two Ls but "unlawfully" contains three. When I asked it how that could be, it explained "it still contains the same two Ls from "lawfully" plus one additional L from the prefix "un-""
My guess for what's going on here is that somewhere in its training data it has ingested things like "a list of words with a double R". Since "double" is associated with "two", it will assume that because Strawberry has a double R, it must contain two Rs.
A trick I learned recently - you can view youtube videos without the annoying ads by changing the url from youtube.com to yout-ube.com - for example, the very nice Trio for flute oboe and piano by Madeleine Dring at
Jeff Maurer is a former writer for the Daily Show, so a political obsessive who brings some wise-ass writing skillz to the party. His summary is worth the read, as is the Harris campaign document if you're in a masochistic mood. [I'm still personally trying to recover some of the brain cells that died while plowing through the 2016 Clinton campaign's policy documents.] Some summary takeaways from Maurer:
-- "Is her economic plan Obama-ish or Warren/Sanders-y? In my opinion, this plan panders to the Warren/Sanders wing of the Democratic Party close to the minimum amount possible....When Harris’ plan to deputize the FTC to lower grocery prices caused every living economist to yell “THAT’S DUMB!” loudly, in unison, over-and-over, Harris explained that actually, her big, bold plan was narrow and inconsequential. The plan described in this platform is definitely narrow....basically, Harris is imagining a plan that will prevent Jimmy Dean from jacking up the price of sausage patties during a hurricane, but won’t do much else...."
-- "a mild indication that Harris understands our [federal] budget situation can be found in her section on Social Security and Medicare. She says she’ll “strengthen and protect” the programs, and that she’ll “fight to ensure that Americans can count on getting the benefits they earned.” And that’s what I’d expect any Democrat to say. But Harris doesn’t go above and beyond to forswear cuts....Of course, the reality is that we’ll probably have to accept modest cuts to future beneficiaries as part of a deal to keep the programs solvent. The fact that Harris doesn’t fall all over herself denouncing cuts suggests that she understands that."
-- "This platform is not woke. If Harris had proposed this platform in 2020, there would not have been enough papier-mâché in the world to make all the giant “Kamala the Klanswoman” puppets that lefty protesters would have wanted to make of her....Harris is still definitely a Democrat — I’m cherry-picking stuff that indicates which way she’s leaning and leaving out Democratic boilerplate that could have been lifted from the Mondale campaign. But the woke/not woke question has divided the party for years, and with this platform, Harris is staking out territory on the “not woke” side of the party."
I hope Maurer is right, because there's a good chance she wins, but I have to say there's a sort of motivated tea leaf reading going on here. Taking microscopic hints and word choices in a vaguely phrased document and extrapolating them to predict sane, centrist policy... reminds me of the absurd "Bayesian updating" people around here were doing about the origins of Covid and it's probably about as accurate.
Though that said, it's not impossible. Harris strikes me as someone with zero principles who will just say and do whatever gets her into office, and if the vibe is moving center-ward she will move with it without a moment's thought. I'd certainly prefer that to a committed left-wing ideologue.
I live and work amongst the people who you call left-wing ideologues. Deep in the heart of Blue America so to speak. Also I have a sibling who proudly says the same and in his case he was Harris's constituent when she was a senator.
From that, two things come to mind as examples of why my brother literally LOL's at the idea that Harris is a progressive (today's term of pride for the worldview that you are referring to):
-- no politician fitting that description has ever chosen to start their career by becoming a front-line prosecutor. Let alone doing that job for a full decade and then becoming a big-city district attorney who aggressively cleared backlogged murder cases, demanded maximum sentences upon conviction, etc. In progressive circles that's roughly as likely a career path as taking an entry-level management job at ExxonMobil or Hobby Lobby.
-- a progressive-base politician going on a national broadcast and talking about being a lifelong _gun_ owner?? Ho ho ho, come on now. That is literally as likely nowadays as a MAGA-based candidate publicly thanking the doctors who carried out his middle-school child's gender-reassignment surgery.
I genuinely don't think she has a single actual policy preference in her body. In an alternate universe she's on the GOP ticket facing off against lifelong New York Democrat Donald Trump.
Honestly, why are we even bothering to discuss policy ahead of the election? We all know this election isn't about economic policy. It's about deciding the future of this country: to let the liberal hegemony persist, or to burn everything to the ground. The populace does not give a damn about the specifics of how the economy functions.
Her sudden switch to appealing to moderates isn't going to convince anyone. She can't escape the fact that she's a colored woman. Do they really think such a person is electable in this country, in this cultural climate? Her policies, her competence, her demeanor, none of that matters. It's about what electing her would represent: complete cultural dominance of the left. The right will not let that happen.
>She can't escape the fact that she's a colored woman. Do they really think such a person is electable in this country, in this cultural climate?
Uh... did you just get off the train from 1961? The only references I've seen to Harris's skin color that have any traction at all are resentful "she's not really black" from some random right-wingers on social media, and that is kind of the opposite of what you're claiming!
To be honest, it's been refreshing how little talk of her race I've seen from either side.
I don't know about majority, but there's several very vocal rightwingers on DSL who call for burning everything to the ground.
And ideas like "actually, it's good to default on the debt" or "we should fire all civil servants" do seem to have an alarming amount of currency on the right. IIRC, there was one actual candidate who called for randomly firing 50% of civil servants on day 1.
Actually, the only way to build greater things is to build on top of other great things. One cannot build pyramids nor skyscrapers starting at the top.
As far as I can tell it's about intra-left signaling. No right-winger is going to vote Harris and none of them are going to take any of these policies seriously or in remotely good faith and vice-versa for Democrats/leftists looking at Trump.
But if you're already a Democrat/leftist, these minor variations could be important. Kamala Harris is woke, by the opinion of some majority of the country (some Democrats and ~99% of Republicans) but how woke she is within the Overton Window of the Democratic coalition is potentially something people care about.
"As far as I can tell it's about intra-left signaling. No right-winger is going to vote Harris and none of them are going to take any of these policies seriously or in remotely good faith and vice-versa for Democrats/leftists looking at Trump."
Have you ever heard of a concept called the swing voter?
-Kamala thinking she needs to make some effort to appear to lower food prices really undercuts the whole economy good/inflation not a problem during the Biden years narrative.
-No politician in America is ever going to admit that Social Security/Medicare is a giant Ponzi scheme that is going bankrupt and needs cuts. They would never get elected no matter how obvious the former fact is. Her messaging on this one way or another is a nothing burger.
-I guess it's a good sign that Harris feels the need to pander more to the center during an election rather than the fringe. But her Senate voting record, VP tie breaking record, and selection of Walz all point to a commitment to far leftist positions. I don't see why anyone should believe her election campaigning on this point.
I don't think that the congressional voting records are a particularly good indicator of someone being "far left" or "far right" in any particularly meaningful sense. They are, after all, voting on things that are still bound to be well within the general, rather narrow acceptable sphere of politics, they're not voting on "nationalize all businesses" or "deport all nonwhites" or stuff like that.
I would think the complete opposite. It doesn't matter how often a pol says they want to embrace communism or persecute minorities. Their voting on policy is the single thing that makes their views influence the nation.
The question isn't whether the congressional voting records are useful in determining opinion, it's about whether they're useful in defining extremism (or general "farness", if one considers far left/right to be distinct from extreme left/right). A person who is a partisan but still within the range of acceptable opinion might look more "extreme" than a genuine extremist who, due to their fringeness, ends up taking positions that are in odds with the mainline version of their ideology (ie. a hypothetical racist far-right representative who is pro-choice since he believes that it's eugenic and nonwhites do it more anyway).
They’re highly meaningful as indicators of revealed preferences. Politicians of all stripes spend a lot of time endorsing policies that they will never vote for while voting for things they don’t want to talk about.
>But her Senate voting record, VP tie breaking record, and selection of Walz all point to a commitment to far leftist positions
In addition to the already commented upon tie breaking record "evidence" (and note the the VP doesn't act as a free agent when casting those votes -- certainly not if she wants a future in the party), Walz's voting record in the House was to the right of almost every other Democrat. https://www.govtrack.us/congress/members/timothy_walz/412214#
And let's be serious. A party that has repeatedly gone out of its way to refuse to nominate Bernie Sanders and Elizabeth Warren is not going to nominate a "far leftist."
I think the IRA was such a sweeping piece of legislation with so much money allocated that it deserves special mention. I should have mentioned that specifically rather than tie breaking votes in general, fine.
AFAICT, Walz is ranked "right" because most of his activity is associated with veterans affairs. Which doesn't say much about any of his other views, or his governorship.
Are price controls and unrealized gains taxes a center left position these days? By the way, if we like GovTrack, Kamala was the most leftist Senator after Merkley, Gillibrand and Sanders.
>I think the IRA was such a sweeping piece of legislation with so much money allocated that it deserves special mention
Yes, but is it "far left"? Seems unlikely, since, Joe Manchin and Kirsten Sinema voted for it.
>AFAICT, Walz is ranked "right" because most of his activity is associated with veterans affairs
What do you mean, "most of his activity"? What makes you think that, of all the scores or hundreds of votes each year, the ones specificaly re veterans' affairs were numerous enough to change his ranking in any appreciable fashion? And, do you have any evidence that other Democrats voted more "left" than he did on veterans bills?
"The ideology analysis assigns a left–right score to each Member of Congress based on their pattern of cosponsorship. The left–right score reflects the dominant ideological difference or differences among Members of Congress, which changes over time.
In a nutshell, Members of Congress who cosponsor similar sets of bills will get scores close together, while Members of Congress who sponsor different sets of bills will have scores far apart. Members of Congress with similar political views will tend to cosponsor the same set of bills, or bills by the same set of authors, and inversely Members of Congress with different political views will tend to cosponsor different bills."
Walz was the primary sponsor of 5 bills that were passed, all 5 were related to veterans. That site also lists that 67% of bills Walz sponsored were related to "Armed Forces and National Security." So by the metrics GovTrack lists as determining their ideology score, a significant majority of bills sponsored by Walz were related to veterans/armed forces. I assume this skews him much more to the right relative to his views on other issues. Paul also seems to think differently, so maybe I'm wrong here, but the GovTrack ideology ranking is pretty clear.
In Congress, Walz supported "pay-as-you-go" budgeting rules, voted against both the bank and automaker bailouts, served on a commission monitoring human-rights violations in China. He also had lib/lefty views on plenty of topics. As a House member he was viewed not as conservative but as bipartisan.
In the Bipartisan Index created by the Lugar Center [the legacy organization of longtime Republican Senator Richard Lugar, who, full disclosure, I voted for and donated to when he ran for president], Walz ranked 20th in the House in the 113rd Congress and 7th in the 114th.
"VP tie breaking record" makes no sense. In 3 1/2 years only one actual proposed law has reached Harris for a tie-breaking vote. Every other instance has been on people nominated for various federal offices, and no VP from any party in US history has cast a tiebreaking vote _against_ their own party's nominee for an office. Never gonna happen either.
The one actual law that Harris was the tiebreaker on was the Infrastructure Reduction Act. Whatever you think of the final compromised version of that bill, one such instance is hardly a meaningful "record".
I know that there is a correlation between IQ and things like test scores but I don't really understand the causality. IQ tests measure things like shape rotation and reaction times. School tests generally measure ability to retain information and critical analysis. These don't look that related. Sure, maybe someone with better reaction times is also better at remembering something for a test but not necessarily. What if they have bad reaction times but are good at storing information that they have spent a while going over? I can imagine someone who does poorly on IQ tests but strong on school tests and vice versa and it doesn't seem like it would even be a rare occurrence.
A test only directly measures how many questions you get right on that test -- and only indirectly measures everything else.
Why do people get questions wrong (on school tests and IQ tests)? Number one reason, they don't care, or they have the belief that "I'm stupid and bad at taking tests" so they act like they don't care. So they rush or guess randomly or panic -- or act like the chess player who blunders and then immediately says "I knew I shouldn't have done that, how could I be so stupid? That's typical me."
Number two reason, because they're missing essential tools, e.g. the ability to read a question closely word for word, or basic logic e.g. process of elimination. Maybe these have to be learnt by a certain age, or your brain can't ever grok them. Or maybe anyone can pick them up, but we just never think to teach them explicitly, or it's un-PC to say "my common sense works better than your common sense," so we don't try to. It always amazes (and saddens) me that 90% of layman can't (or refuse to) understand something as simple as a truth table.
Number three reason, because they lack domain specific knowledge or tools. E.g. the history test asks for the date of such and such battle, and you know it or you don't.
The model I'm sketching is: first you need the will to do well; then the general tools (or to be a good "test taking interface" if you like); and only then the domain specific knowledge. People's "will to do well" on an IQ test and a school exam will be highly correlated, because they depend on self-image and "life strategy" in a similar way. I was always a nerdy type who got most of my praise and self-esteem from doing well on tests, and always similarly motivated to score well. A child who gets their self-esteem from somewhere else, e.g. being cute or making others laugh, is not likely to concentrate as hard on any test -- they'll be thinking about the meta social context, or how unfair it is that they have to take tests when it isn't their strong point, or whatever.
The second part, test taking tools, naturally means a correlation between performances on different tests, and will probably be correlated with the first part as well. I picked up lots of tricks that made me good at taking tests, precisely because I occupied the niche of "child who takes tests very seriously".
The reason for your surprise at the correlation of various test scores is that you assume it's mainly part three, domain specific stuff, that determines performance. But this is not true, because the first two parts are so important, and because on most formal/school tests the content is much thinner/emptier than we like to imagine. If you go through a school physics exam (for example) you'll find that most of it is https://www.lesswrong.com/posts/NMoLJuDJEms7Ku9XS/guessing-the-teacher-s-password or use of elementary mathematical tools (plugging numbers into a formula). And if you move to a subject where the tools needed are substantially different (e.g. art, cooking) you no longer expect to see a correlation with IQ.
The relationship between "general test taking skills" and "general intelligence" is complicated. Some things intuitively count as both (basics of formal logic, maths). But it also seems possible to have the former without being generally intelligent -- we have a stereotype of a pedantic nerd who's always technically correct, but always wrong where it matters. And the opposite stereotype, the artsy genius.
[Tangent: I've arrived at this model partly by analogy with the Football Manager video games. In them your virtual football players have mental, physical, and technical attributes (which are just numbers between 1 and 20). Technical attributes include things that seem most obviously correlated with football performance, like passing and dribbling. But (as explained by many strategy guides) your players first need the desire to chase after the ball; then they need the physical attributes to get there; and only then can they use their fancy technical skills. So the first thing to look for in a player is mental strength (determination, work rate, bravery). Players without it will underperform, in the same way that some people do badly on tests despite seeming clever in other ways]
I remember seeing a post a while back arguing that this is the true reason why the Marshmallow Test was correlated with success. It's not really about time preference (which is almost trivial in the case of the marshmallows), but rather testing kids for *desire to perform on tests*.
What makes the phenomenon interesting/non-trivial is precisely that these abilities have a statistical relationship with each other when it seems non-obvious that they should.
And it goes beyond grades and IQ. IIRC how much money you earn, how long you live, how healthy you are, how good your reflexes are, how good you are at your job (even if your job is being an athlete), and I seem to recall even how happy you are and how many friends you have are all correlated. All good things go together. More specifically, there seems to be a single hidden scalar variable 'g', that correlates with almost all things generally considered 'good'. It correlates the most with things like IQ and being good at math, but surprisingly little is actually exempt.
The common guess seems to be that 'g'='general intelligence'. Rather than just being an amalgamation of millions of disconnected heuristics and skills that have nothing to do with each other, human brains to some extent meaningfully vary in how good they are at information processing and problem solving in general. And people who are better at problem solving tend to be better at getting what they want, hence the correlation with all things generically considered 'good' or desirable.
(My memories of these statistics are somewhat vague. I could be wrong on some of these points)
How happy you are is not correlated with IQ. If you look at the correlation of IQ and the big 5 personality traits, neuroticism (sadness, moodiness, and emotional instability), and IQ has a tiny negative correlation. It is statistically significant (i.e., not just a chance occurrence in the group studied), but too tiny to be of any real life significance. The correlation is -0.09.
I agree that neuroticism and unhappiness are not the same thing. Here are the reasons to still use neuroticism as a stand-in for trait unhappiness.
The neuroticism measure is well-validated and well studied. I am not at all sure there are measures of happiness or unhappiness that are anything like as solid and trustworthy. I have never seen any, & I’m a psychologist. There are definitely well-validated measures of depression, but depression is not the same thing as unhappiness either.
The happiness measure in the study you cited is briefly described in the abstract of the study, and sounds like it was a single question: “Happiness was measured using a validated question on a 3-point scale.” Even if they are valid (i.e., consistent over time, resistant to circumstances that pull for a certain kind of answer, correlated with other measures you’d expect them to etc.) single-item questionnaires do not capture a big enough, rich enough chunk of the thing being measured to have good construct validity. It’s fine for a happiness test to ask whether the subject is happy most of the time. But the test is better if it asks several similar but related questions, eg, “ T/F: When life is hard, I still have an overall positive feeling about things,” “people I know comment on my positive outlook” etc etc. Also, what we want this study to be looking for is trait happiness — a general lifelong tendency to be happy. It’s not clear whether the test used in the study asked about that, or asked about subject’s mood at the time they were taking the test. So while I agree that a test of happiness as a trait would be a better measure to use if you’re studying relationship between IQ and trait happiness, I think at present neuroticism gets closer to trait unhappiness than a singe “validated question.”
As for the measures of intelligence, the study you cite estimated “verbal IQ . . . using the national adult reading test.” National adult reading test no doubts correlates with verbal IQ, but it’s almost certainly not a slam dunk high correlation, and verbal IQ itself is not slam dunk correlated with full scale IQ, which is what we mean by intelligence. The intelligence measure your study used is just inadequate
So why is there a stereotype of the depressed intellectual? You know, just look at Zizek. Is it possible that the correlation of IQ and happiness tops out somewhere?
After much time thinking about this, I think that being a standard deviation outside anything makes living, functioning, and forming relationships in a society more difficult regardless of whether that standard deviation is objectively good.
I went to school with a few prodigies and while they were happier than average while participating in their field of choice, their lives were harder the rest of the time. They had different priorities than most other people which made relating to others difficult. They were often perceived as either weird or threatening because of their talent. Values they cared deeply about--like excellence, the importance of art, or work ethic--were routinely derided and dismissed by other people and weaponized to bully them or "knock them down a peg." And when they finally found like-minded groups with similar interests, there was still a lot of underlying stress because they knew they were ultimately competing for the same tiny pool of prestigious jobs with tiny margins for error. I think this is true at the top of most fields.
The world tends to be built for the "average" person, so the less average you are, the more frustrating it's going to be, regardless of the reasons behind your difference.
I really want to research that splits up different types of high IQ people. If you use your intelligence to ruminate all the time, you’re probably going to be depressed. But if you use it found a successful company, you’re probably going to be happier.
I'm not sure I'd call those 2 groups different types of intelligence. Can't it be that some intelligent people are anxious and unhappy and some are active and optimistic? Or that one intelligent person can be depressed for a couple years, then pull out of it and be active and optimistic? Seems like the gloomy - cheerful axis might well be at right angles to the smart - dumb one.
I meant high IQ types split up in other ways, like personality. For example, I'm sure that extroverted high IQ types are much happier than introverted high IQ types.
Perhaps some types of people are more visible / easier to remember than others? There are many depressed people, but only those who are exceptional in some other traits become famous.
How does social competence explain superior shape rotation skills? Let me guess, little kids with great social skills spend more time riding the carousel with their friends, so they get more experience at seeing things turn around...
IQ, or at least the g factor, is basically an attempt to explain why all these tests correlate in ways they intuitively shouldn't. Basically, we could imagine someone with bad read reaction times but good memory or someone with bad SAT scores but good grades but these people are fairly rare.
So, start with the SAT and the math and verbal sections. It's easy to imagine someone doing very well on the math section and poorly on the verbal or vice versa but in reality these two scores are fairly highly correlated. Not perfectly and people do tend to be better at one but it tends to be more 750 Math & 710 Verbal than 750 Math & 510 Verbal (which is ~the national average). This observation is non-intuitive but pretty persistent. Over time, people have noted that you can just keep expanding it to things like AP test scores, grades, educational attainment, and even shape rotation and reaction tests. In fact, super weirdly, we can give children shape rotation tests at a young age and make educated, not perfect, predictions. That's not to say that there's not variance, there is...but an honor roll student who's bad at math tends to be bad at math relative to other honor roll students, not average students.
People then used some, frankly, pretty odd statistical techniques to determine the g-factor and then people fought about it a bunch but at its core everyone is just trying to explain this phenomenon. Our best explanation is some kind of underlying intelligence or "IQ" or general IQ, which is almost defined as this counter-intuitive correlation we observe. The 2nd best observation is common cultural factors, eg upper-middle class kids get advantages, and they do, but this gets undercut really fast when these observations replicate in really alien cultures like China and South Korea, which makes common socio-economic factors look really lame. Like, if a shape rotation test at age 5 is a good predictor of SAT scores in the US and South Korea at age 17, that's far more likely to be some underlying genetic thing than a cultural thing.
In your first paragraph, you say that intuitively the tests of different cognitive abilities shouldn't be correlated. It seems natural to me that mental abilities should be fairly tightly correlated, at a statistical if not a personal level. The data supports this - why is it unintuitive?
It's like saying that intuitively an Olympic swimmer shouldn't be any better at track events than average, since they weren't selected to be good at track. This is false - they're in peak condition and likely to outperform 95% or more of the population.
Because it's not Brandon's intuitive understanding and I'm responding to him. He's asking about someone doing poorly on IQ tests and doing well in school, which is certainly possible but highly unusual, which indicates that his intuition is that they're not correlated. Your intuition is different, so if you asked I would phrase it differently.
I wouldn’t say that my intuition is that they aren’t correlated. It’s more like that they shouldn’t necessarily be so. For example, I’ve seen videos of chimpanzees showing how they do really good on reaction time tests. Obviously, they would do much worse on other parts of an IQ test. So it’s clearly not some kind of scientific law that these things have to be correlated but they apparently they are in humans. Without having a deep understanding of the causality, we’re missing something fundamental about why that correlation exists.
The things you'd think would not correlate very well with ability to reason, have insight, grasp complex ideas etc -- they actually do correlate less well with tests of those things. The most uncognitive-seeming subtest on the WAIS is Digit Span: They read you long numbers, and then you say them back. I believe average Digit Span is 7, but the test contains numbers with as many as 11 digits, and some people can remember those. And score on Digit Span correlates only 0.57 with full score. If course *only* .57 is aa startlingly high correlation, given that remembering a bunch of digits seems pretty useless, just a party trick, and like it doesn't have much to do with the reasoning powers, etc. we think of as indicating intelligence. That's why WoolyAl was talking about *g*, general intelligence. There are a lot of tasks that don't have much in common except that they are done with the mind, yet ability on one predicts ability on the others pretty well, as though there's some brain quality that's the one ring to rule them all -- rule all the subskills, i mean.
IQ tests measure a bunch of things, many of them much more obviously connected to the abilities needed to do well in school than the 2 things you name -- reasoning, doing math word problems, understanding how 2 things are alike or different, vocabulary. I recommend looking up the WAIS (Wechsler Adult Intelligence Scale) in Wikipedia. IQ tests predict school performance fairly well, but of course a number of other things besides raw mental abilities influence how well a kid does in school: how good his teacher is, how good the teaching materials used, how much school manages to capture his interest, how he's doing overall, whether his parents expect him to do well and make an effort to help him do well.
You don't have to take the test. Just read the Wiki entry on the WAIS. There's also easy-to-find info about how much each subtest correlates with each of the other and with full-scale score.
Welp. Someone illustrated Scott's old post about Haiti with not very interesting photo material and put it up on X, with the credit to Scott's original post is at the very end and hard to find. I'm not sure exactly how annoying this is, but I figured I'd let Scott know:
Can you have a robustly functioning democratic system, in which one party consistently wins for a long time? Or is this sort of effect a clear sign of either some sort of cheating, or at least that something has gone very wrong?
An example of this might be the mayoral elections in Chicago, which have been won by Democrats since the 1930s. (Strictly speaking, the current process is non-partisan, bu the winners have clearly been people who moved in Democratic Party circles.)
One might expect a party that has had a long string of losses to change the policies it backs and the candidates it puts up for election until it finds a winning formula, unless something is keeping it from making this sort of adjustment.
On the other hand, if one party keeps winning for a long time, it might come to be seen as the only viable party, and everyone with any political ambitions would migrate over to it. And if it had all the serious talent, the party might be expected to keep winning, in a sort of self-fulfilling prophesy, leaving the other parties a bunch of tiny protest movements for those so wedded to their positions they are not willing to accept the compromises of dealing with the major party.
In the U.S., at least, the equilibrium is to have two major parties of roughly equal power nationwide. Because the country isn't homogeneous, this ends up with a number of states and local areas dominated by one party or the other.
It's actually quite normal in a democratic system for one party to be hegemonic. This doesn't mean they win every time but that they do win most of the times. LDP in Japan, Social Democrats in Sweden at least until 2006, Tories in UK since WW1 and Christian Democrats in Italy during the Cold War are famous examples. In the US, the Republicans were generally hegemonic from Lincoln to Hoover and then the Democrats from Roosevelt to LBJ, at least.
It's the current situation where most Western countries have constant, down-to-wire competitive elections that's expectional, and its main reason is probably the detachment of parties from the wide, established social groups (churches, labor unions etc.) that used to underpin their support and the increasing use of very precise consulting and microdata to determine the minimum amount of yielding and pandering to the electorate they must do to get past the finish line.
> One might expect a party that has had a long string of losses to change the policies it backs and the candidates it puts up for election until it finds a winning formula, unless something is keeping it from making this sort of adjustment.
One thing that can prevent that is the nationalization of politics. California Republicans have not attempted to moderate to win state elections because everything is national nowadays, and if Trump-lovers happen to be a minority in the state, so be-it. If anything, there's an evaporative cooling effect because any politician who is ambitious and ideologically flexible will switch to the winning side.
Back before the parties were so polarized, and before politics followed centralization of power toward the national level, there would sometimes be "one-party towns" where the actual election would be the party primary, in which all serious politicians participated, regardless of how closely their personal ideology matched the official ideology of the national party.
Japan. The LDP almost always wins. But opposition parties have won, and can win if the LDP royally screws up.
Arguably, Japan has the best advantages of a one-party state (no political polarization, or politicization of private life) without the biggest downside (no way to hold the one-party accountable).
It's easy for a party to never win. The Whigs have been out of power for, what, 150 years?
In a healthy democracy, I expect the one giant party that wins all the time to eventually split into separate factions along their internal faction lines, since there's no outside threat to keep them together.
Isn't it surprising that the Singapore PAP hasn't split by now? Every organisation has internal factions and disagreements about who should be in charge. You'd think that eventually an internal disagreement should blossom into an actual party split, with each side convinced that they'd be the ones to win the people's support in the post-split election.
In places like China the mono-party doesn't split because everyone understands that the losers of the split and their families will wind up imprisoned, tortured, dead. But Singapore has sufficiently robust democratic institutions that this shouldn't happen, they'll just wind up as an opposition party.
I always figured Lee was holding things together by sheer personal awesomeness. He's been gone nine years at this point. I'm not sure whether any cracks have appeared in the PAP facade, but perhaps not enough time has passed.
Yeah, this is my bet too - something of a lingering respect / not wanting to mess with a good thing.
I mean, if your winning horse literally took you from third world to first in 30 years, would YOU want to start messing with it? He's the most revered politician in the world for a reason.
I read a lot of Trollope for escape and it is clear in his books that he views the Whigs ("Liberals" by his time, I believe?) as destined to prevail forever, and the Tories as a historical curiosity, albeit with aesthetics on their side. It is not precisely that he sees that there is nothing to conserve, but rather that utopia is far off and so there will be for a long while yet those good things that he in fact would have grieved the loss of. He need never see utopia. (This raises some questions, obviously, lol.) But that those things/people will not fit into the future and must go, he is certain.
Of course, the Conservative party of today bears no resemblance to that in his day.
As an environmentalist and conservative, I obviously have no one to vote for as this idea that "nothing will be conserved, or should be" has very much taken possession of the discourse. In fact, nothing has shocked me more than the calm acceptance on all sides now, that nature is something we "choose" to exist or not, that it is without value except to those who fancy it, that it probably will be mostly banished but the techno-future is so interesting who will miss plants and animals? A few old Boomers. Wildlife has their little niches - on the sufferance of people who are little familiar with it and couldn't care less.
I will vote nonetheless, probably, but merely in the faint hope of sending the future a signal, in case they are still writing history books in the future. It will obviously be a very crude signal, perhaps hardly worth doing.
The AI art Turing test is a very good idea. Honestly seems a little strange that nobody has set it up already.
It'll be interesting to see if abstract art created by a human has some quality to it that abstract art created by a machine lacks. My sense is very strongly that it does, but it'll be good to run a test. I also think some people will very clearly see the difference between human and AI art, while others won't pick it up at all.
It really strongly depends on a whole bunch of different factors, i.e. the style of art, the competence of the artist, the perception and interests of the person viewing it.
My sense from experimenting with AI art is that it has a peculiar deadness to it, which can be good if that's what you want. I've been using it specifically to make shoggoths for a roleplaying game I'm working on. It's great at Lovecraftian horrors because they're supposed to be mutated and shapeless, and it's actually good if they have a horrible dead expression of inhuman evil in their eyes.
Can it be cute or relaxing though? Not sure yet. We really need to drill down more specifically into what aesthetic qualities it can capture and what it can't.
I can tell you what this video ain't. It ain't no AI voices and burned-in animated captions, it ain't me sitting around the house talking to a camera, it ain't no slide show disguised as a video, and it ain't no unenthusiastic presenter stating the obvious. Or I can tell you what it is. It's me (American, irreverent but kind) and my pardner (Australian, cute, funny) traveling the country, rediscovering patriotism, and improvising a remarkably eloquent and inspirational speech bit by bit as we go. The weird thing about making a documentary is, you don't know what it's about until it's over. It turns out, this one's about the American Dream. I've done a lot of things in my life, and out of all the things I've done, I'm most proud of this. You can cheat by watching this 8 minute version where it's just the American Dream speech ( https://www.youtube.com/watch?v=VfPVpLfTUDg ) but if you do that you'll miss the serendipity of the outlandishly glamorous Coleman Theater in the middle of Nowhere Oklahoma, the wife of the dead chainsaw carver in Sullivan Missouri who's kept his shop open for 20 years and never remarried, my purchase of a steel-tongued drum followed by my improvising alongside a player of a native American flute in Oatman Arizona, and the abandoned houses of Amboy California where the walls have stories that they can't tell so I have to do it for them. https://www.youtube.com/watch?v=cHlSDE7MjbI
You seem to be assuming that an abortion has the net effect of one fewer life in the world. There are a number of studies which claim to show that, given two pregnant women of similar age, socioeconomic status, martial status etcetera, if one has an abortion and the other doesn't, then the woman who has the abortion will have more children later. So you aren't losing net QALYs by having an abortion, you are just shifting them to a later child, who will quite likely be a happier one.
There was a study Ozy wrote about here: https://thingofthings.substack.com/p/lets-read-a-study-the-impact-of-denying which suggested the exact oppostite effect. Women assigned to a male judge were granted abortions 38% of the time compared to 58% with women judges, but they had on average 0.5 more children - a lot more than 1 per denied abortion. Quite possibly abortion statistically leads to greater than one fewer life in the world (greater even than the slightly over one accounting for twins etc.).
For many people their idea of abortion is clouded by propagandistic media portrayals. They always imagine abortion patients as childless women, usually very young, when actually 59% of abortion patients have at least one previous birth:
I assume this was meant as a reply to my comment, so I'll go ahead and respond. If you believe that each abortion will, on net, result in more QALYs than not aborting then I agree it would make sense from a utilitarian perspective to support abortion.
It doesn't really seem likely though? The USA aborts between 6-9 hundred thousand fetus's each year, do you really think that if we aborted 0 over the last ten years our current population would be lower?
Last time I checked Flux, the AI image generator that Twitter/X is using, doesn't have any of the atrociously uptight guardrails shared by other major corporate image generators such as Bing or Google except for not being able to create porn. It won't block you from generating pictures of Pikachu or Donald Trump or Donald Trump shaking hands with Pikachu. Misinformation and copyright violation must abound, truly the worst possible situation, and yet...
I note that the world has not come to an end. Indeed, beyond a few desultory attempts by the usual suspects to stop it by deliberately generating the most wildly offensive images they could think of, it's just a thing. Nobody cares that much and life goes on. Rather than panic and chain the AI up in the heaviest restrictions imaginable, it seems that all anyone needed to do was ignore the scolds.
I just tried with Firefly and couldn't. The word Pikachu was blocked from the prompt.
I was able to get something vaguely reminiscent of Pikachu but not recognizably the same character with the prompt "Cute yellow mouse monster with electricity powers" and "doodle drawing" selected under the effects settings.
Yeah - the way to get into a school like that is to win national or preferably international competitions. Or best of all - publish truly novel research in math or computer science. If he won the International Math Olympiad or something, his chances would be excellent.
So I did a bit of digging on Reddit to get more information about what may have happened to Stanley Zhong, the California high school student who was rejected from a bunch of top colleges despite really excellent grades and test scores. I found this bit, about what students aiming for top colleges are expected to do these days. And you know, at this point maybe it's not worth it.
https://www.reddit.com/r/collegeresults/comments/175kanu/1590_sat_397442_gpa_rejected_by_16_colleges_how/
> He went to Gunn, and he didn't have enough social impact. To get into a good school for cs from that demographic, you need to have a crazy amount of social impact. Being really good at olympiads and hackathons won't be the thing that gets you into top colleges. This is different if you get to the level of camping for an olympiad though. That being said, having USACO platinum or USAJMO in your awards section isn't enough to get into Berkeley or MIT.
> Doing things that look like they have social impact, like running nonprofits and hackathons, often require substantially less effort but have a much higher yield for college admissions.
> The person from my bay area high school who managed to sweep every UC as a cs major (pretty much impossible for this demographic in 2023) had a non-profit where he fudged numbers and applied for a shit ton of social impact awards. The people that do bs like this look more impressive to college admissions officers. The most appealing applicants are the ones that look like they're going to change the world.
> The people I know that went to Stanford, Berkeley EECS, MIT, etc. were literally all USACO silver except for one guy who was gold. This demographic is a shit-show. Being one of the smartest people at your high school won't get you into one of these schools. You have to show social impact through your ec's in the scale of hundreds to tens of thousands. Either that or feign a really niche interest to get into private schools through doing stuff like linguistics research or a classics reading club.
> This might sound cynical, but as a college student, being genuine will fuck you over if you're in this demographic. If you're a junior, organize a hackathon to get girls into coding, start a non-profit org to combine cs with art, apply for sponsors to make a scioly competition about climate change, organize a protest, etc. All of these are good things, but their scope is often exaggerated. After you finish implementation, email 50 news channels and apply to social impact awards. A lot of these things aren't as hard to do as they seem. They just require a small team and 2-3 weeks of grinding. As a college student, this is literally the formula every bay area kid who's hyper-successful on college apps follows. This is how you beat the rat-race. Do things that have a small positive impact but seem like they have a much larger scope than they do. AO's eat this shit up.
For perspective, only about 2,000 people a year score 1590 or better on the SAT. About 400 prospective college students merit USACO platinum per year. The freshman class at Berkeley is about 8,000 people. If SZ isn't being basically auto-admitted everywhere with scores like that (and corresponding GPA and other academics), then these schools aren't admitting on academic merit, and they aren't admitting on academic merit with a bit of fudging to get a "well-rounded" student body.
UT Austin, however, is a fine school. If Texas wants you, and California et al don't, you should probably take the hint. Or the Google job, if your Dad can swing the interview.
The part of this story that is shocking to me is not that he was rejected from the fancy private universities (I've internalized that there is a fair amount of randomness in those, plus discrimination against people with his demographics), but that he was also rejected from all the UC schools. It really feels like a kid graduating from a CA high school with a near perfect academic record `ought' to be getting auto-admitted to the UC.
Cal Poly I assume was his safety school, and the rejection was because they assumed he would get something better and wouldn't come.
Your final paragraph seems on point.
I had thought that being in the top 10% of a California high school class was supposed to guarantee auto-admission in the UC system; a quick search suggests that it's 9%, and the fine print is that they only guarantee that they'll find you a place *somewhere* in the UC system.
https://admission.universityofcalifornia.edu/admission-requirements/freshman-requirements/california-residents/
Possibly the unwritten rule is that if you're the Wrong Sort of smartypants academic overachiever, they offer you a slot at UC Merced and hope you take the hint. And he should also have been guaranteed a slot in the CSU system, but they don't let the Wrong Sorts into Cal Poly, that's what CSU Bakersfield is for.
It's a frustrating case because there's a lot we don't know. As I see it, the credible hypotheses are:
- the bar really is incredibly high in top colleges
- the bar is incredibly high in top colleges in a very sought-after major
- there's something negative we haven't been told about SZ, like serious disciplinary problems
- SZ somehow mishandled the application process, and looked much worse than he is
- SZ is being discriminated against, because of his race
- SZ is being discriminated against, because he is from a wealthy town
- this is just a pure fluke; SZ just got unlucky
- this is a case of yield management systems run amok, with most of the schools figuring he would be accepted to some posher place, and saying yes to him would lower their yield scores for nothing
There's just so much we don't know, and that lack of information feeds a tornado of speculation.
I mean - you're competing against people like William Kamkwamba of Malawi, who built a windmill from recycled junk in an impoverished African village. And Malala Youzafsai of Afghanistan, who got SHOT IN THE HEAD for her activist beliefs and even more miraculously SURVIVED AND RETURNED TO ACTIVISM. You're pack fodder unless you are:
1) winning national or international competitions for high school students
2) publishing original research, preferably in top journals
3) overcoming insane levels of adversity - think not just 'got cancer, recovered from it' but 'homeless, raised hundreds of thousands for their own chemo and from the hospital bed proceeded to earn near-perfect grades and test scores'.
Like a lot of college admission stuff, this suffers from being anecdote instead of data and risks feeding into a preferred explanation. I think the reason that most people fall for this is because
the narrative on college admissions still perpetuates the myth that decisions are mostly intentional vs random.
Let's assume SZ had a much higher than random chance of getting into each of those schools. For illustration, I'm going to assume 30% chance of getting admitted and that all admission decisions are independent. Applying to 18 schools, there's about 6% chance of getting 2 or fewer acceptances. With 3-4mm high school seniors each year in the US and, maybe 10-100k applying to top schools (to say nothing of the applicants from outside the US), it isn't at all surprising that there may be many students in a position similar to SZ or even worse.
Even if we bump the probability of admission to 50% for each school and 40k similarly situated applicants, there would still be about 26 SZs per year.
I'm going on record predicting a Harris win in November. I hope I'm wrong.
I'm basing this on hearing that Trump is saying there will be no 2nd debate, that Harris calling for one is like a losing prize-fighter demanding a rematch. This is such a poor reading of the political situation I think it is indicative of how the rest of the campaign will go.
Looking objectively, the debate had no clear winner. So Trump is delusional in thinking he clearly won. Trump backers will back Trump even if he has some awful gaffe, so Trump's objective ought to be winning over independent voters. Another debate is one way to do this, and I know of no better way, taking the national stage in a format he is somewhat good at.
Caveat: if another debate DOES happen, this prediction is void. I may make a new prediction after the debate in that case.
Trump lost the debate hard, which is why he was complaining about moderator bias the days after because he couldn't answer the super hard questions they asked him like "Will you support Ukraine?" or "Will you veto a federal abortion ban?"
You also say further down "Harris seldom responded to the question at hand, seeming to speak only statements prepared ahead of time" - which is a weird criticism to make when Trump said he had "concepts of a plan" with respect to healthcare and that it wasn't his responsibility to have one prepared because he's not currently President, or when he spent most of his time for a question ranting about his crowd sizes in response to provocation from Kamala. There was not a single question Kamala didn't answer, and she at least passed the sniff test for understanding how government works and what her responsibility to Americans is as a Presidential candidate.
Trump's greatest achievement was lying about Haitians eating cats and dogs in Springfield, which was pretty funny, but totally unbecoming of anybody running for President to spend his time repeating stupid lies off social media rather than coming up with a healthcare or foreign policy or abortion plan. It seems like he not only avoided the Republican debates but also refused to even watch them to at least get a feel of what policy members his own party members wanted, let alone the American public.
Trump answered the question. He doesn't yet have a complete plan, to be picked apart because of its incompleteness, so gave no details. What was Harris's plan to compare? "Strengthen the Affordable Care Act". What will be added? Where is it weak? That is no less vague than Trump.
I still look at it and see no clear objective winner. You clearly have a bias towards Harris. I am done responding to subjective arguments.
Debate transcript: https://abcnews.go.com/Politics/harris-trump-presidential-debate-transcript/story?id=113560542
I think Trump is likely to win, but I hope I'm wrong.
But as for the debate thing, Trump is clearly scared of debating Harris. It won't end well for him and he knows it.
Judging from Harris's performance, it would be easy to beat her in another debate with some preparation. I don't think Trump thinks it necessary to prepare for a debate. But the preparations would be easy: prepare some questions yourself for Harris, so that ignoring them would make her look bad. If, as I saw, she only delivered prepared statements, she wouldn't be able to competently address them.
You didn't explain why you think Trump is likely to win, so I conclude this is just a sarcastic mirror of my original statement. But my statement is based on Trump running a campaign, not on his merits as president. Harris would be another Biden presidency, and the Trump presidency was better than the Biden one.
I wasn't being sarcastic at all. I think Trump is likely to win because Nate Silver is giving him a 61% chance to win. The attempted assassination yesterday will probably help him a bit as well.
Nate Silver's latest projection gives Harris a 38.7% probability of winning, against Trump's 61.0%. The electoral college really hurts the Democrats in a race this close. There's about a 20% chance Harris will win the popular vote but lose the election.
You should get your vision checked if your objective glance tells you there's no clear winner. One person was afraid to make eye contact with the other one, one person was easily goaded, one person ranted about completely false bullshit that consists of tropes that have been discredited for years. It wasn't Harris.
(Oh, and one person is afraid to do a second debate. That's not "no clear winner" territory.)
My OBJECTIVE vision doesn't make definitive conclusions on such things. From a different perspective, Harris was weaker by constantly looking toward Trump, even mentioning him a lot by name, so eye contact she was pushing at him that was not received, as he was mostly looking at the camera, does not go well in Harris's favor. A perspective on Trump's being "goaded" is answering some things Harris said, which is, after all, the purpose of a debate. On the other hand, Harris seldom responded to the question at hand, seeming to speak only statements prepared ahead of time (NOT from being fed the questions before the debate).
I addressed Trump's willingness to debate a second time in my original post, and I know of no sane perspective to consider that it is because of fear.
It looks to me like you're viewing the debate through a pro-Harris perspective, which is your right. But it certainly isn't objective.
Emphasizing OBJECTIVE doesn't make it so. You're coming at it with an anti-Harris perspective, just like you were before the debate when you made baseless claims that she'd drop out of it because she was the weaker candidate. Now that she's no longer done so, you've found a new framing of events that fits your worldview even though it's still divorced from reality.
This is not correct. The things you are pointing to saying Harris won are not definitive, but subject to interpretation. It is YOUR viewpoint trying to make Harris a clear winner.
If I were to take a subjective viewpoint, I could point out how few questions Harris answered, pointing to only practiced debate preparation. After all, what else did she need to do in the month proceeding? It's not like she had any real duties as vice president that people were depending on. So she practiced answers to questions chosen to help her campaign, and delivered them regardless of the questions asked, with few exceptions. She clearly failed in the "debate" aspect, which ought to be worrisome for someone who needs to make quick, good decisions.
And Trump had a strong stage presence, from making her come to him for shaking hands, to ignoring her in favor of the audience when speaking. He avoided traps in the questions attempting to pin him an unfavorable position. And he certainly delivered the most memorable lines.
So no, I think there was NO CLEAR WINNER, and I have yet to see anything that indicates otherwise. Go ahead and vote for Harris. But when you attack my stated objectivity, please do so with something more than your own opinion.
I am not attacking your objectivity because you have none to attack. You're a Trump fanatic through and through and you ought to be honest if not to us than to yourself by admitting as much.
> Harris was weaker by constantly looking toward Trump, even mentioning him a lot by name, so eye contact she was pushing at him that was not received, as he was mostly looking at the camera, does not go well in Harris's favor.
We seem to have very different interpretations of body language. Looking at your opponent signals confidence, while avoiding eye contact signals fear. Although ignoring someone can signal dominance, it only works if you don't stare doggedly past them when they actively challenge you with their eye contact.
Body language does depend on the situation. Trump never forgot that the audience was the camera, no matter who was in the room. Did he look at even the moderators? It wasn't clear.
In any case, since it is subject to interpretation, I'm ignoring my take on body language. I'm rather surprised that what seems to be the most controversial part of my post is my claim to objective analysis, and that no one "won" the debate clearly.
ACXLW Meetup 74: Design for Developing Countries & Ethics of Extinction
Hello Folks! We are excited to announce the 74th Orange County ACX/LW meetup happening this Saturday and most Saturdays after that.
Host: Michael Michalchik
Email: michaelmichalchik@gmail.com (For questions or requests)
Location: 1970 Port Laurent Place, Newport Beach, CA 92660
Phone: (949) 375-2045
Date: Saturday, September 14, 2024
Time: 2:00 PM
Conversation Starters:
1. Why Western Designs Fail
Text Transcript: Google Document Link
Video Title: Why Western Designs Fail
Video: YouTube Link
Summary: This video explores why innovative, highly praised designs for developing countries often fail. The key problem lies in cultural misunderstandings: designers focus on the functionality of a product without understanding the deeper cultural and social contexts in which the products will be used. For instance, the Neon Nurture incubator, made from car parts to be low-cost and easily repaired, never gained acceptance because it lacked the prestige and appearance that medical officials in developing nations value. Other examples like the PlayPump (a merry-go-round that pumps water) illustrate how Western solutions often mismatch the actual needs of the communities they intend to help.
Questions for discussion:
Why do you think Western designers frequently overlook cultural factors when developing products for other parts of the world? How can this be addressed?
What role should local communities play in designing products intended to meet their needs? Could co-design processes become the new standard?
In cases like the PlayPump or mosquito nets being used for fishing, how should designers react when their products are repurposed by local users in unexpected ways?
2. Driving the Screw Worm to Extinction: The Ethics of Annihilation
Text Transcripts:
Killing Every Screwworm Transcript
14 Million Worms Transcript
Video Titles:
Killing Every Screwworm Would Be the Best Thing Humanity Ever Did | Kevin Esvelt
Why the US Drops 14.7 Million Worms on Panama Every Week
Videos:
Killing Every Screwworm Video
14 Million Worms Video
Summary:
The first video by Kevin Esvelt argues for using CRISPR gene drive technology to eradicate the New World screwworm, a parasitic fly whose larvae feed on the flesh of mammals and birds, causing immense suffering. Esvelt contends that wiping out the screwworm would have a far greater impact on animal welfare than ending factory farming. The second video outlines the decades-long U.S.-Panama collaboration to keep screwworms out of North America by dropping millions of sterile flies in Panama each week. While this method works as a border defense, it is not sufficient to eradicate the screwworm from South America, where the problem persists.
Questions for discussion:
What ethical principles should guide decisions to drive a species to extinction, even if it causes widespread harm? Does the end justify the means?
Could gene drive technology be misused in other contexts, and what safeguards should be put in place to prevent this? What might be the long-term risks of eliminating species?
If the eradication of harmful species like the screwworm is possible, should we consider other "pest" species next? Where should we draw the line in deciding which species to eliminate?
Walk & Talk:
After the meeting, we will take an hour-long walk and talk session. There are two mini-malls nearby with hot takeout options—look for Gelson's or Pavilions in the 92660 area.
Share a Surprise:
Bring something unexpected to share that has changed your perspective on life or the universe.
Future Direction Ideas:
Please contribute your thoughts on future topics, meeting types, activities, or other ideas for the group’s future direction.
Looking forward to seeing everyone there!
If you are a League of Legends gamer, this is a 7 question survey about your analytics tool usage. It shows how other players answered at the end and you can sign up for the waitlist for an analytics tool that might or might not become commercial later.
@Mods: Let me know if this is not suitable and I'm going to delete it.
https://docs.google.com/forms/d/e/1FAIpQLSfUlYq2nXSyhjQuki61cXPqcBUShYqasEw3W7NkCkj3zidK0w/viewform
Yudkowsky on Twitter:
- All of these LLM Whisperers that I see on Twitter, appear to also be insane.
- Why?"
Unless irs random coincidence, there is an interesting phenomenon to be explained here.
As Yudkowsky mentions, there is well-known correlation of computer security experts with high-functioning autism and/or being transgender, Seeing that corelation, again, would not be a surprise. (There are obvious mechanisms for why high functioning autism would give you an advantagde at programming, and the autism-transgender correlation has been noted elsewhere, e.g. by gender clinics) But with the LLM hackers, we're seeing a personality type that is distinctly different from the high functioning autism that we know and love. So we wonder ... why?
(Janus may be a little offended we have him down as this ... other thing,, but its a serious question.)
If I were to guess what this other thing might be ... what the hell? Is "high functioning schizophrenia" even a thing? The DSM gives us schizotypal, etc.
I've experienced psychotic breaks and have met many others in psych wards who have also had psychotic episodes or have full on schizophrenia. One of the early symptoms I've seen with a lot of these people, and myself, is divinatory magical thinking. This kicks in really early while you're still high functioning. Interacting with LLMs is very similar to divination and I would not be surprised at all if schizophrenics are drawn to the activity.
Context? I suppose we all know what an LLM is at this point, but even with a quick Google I can't find what on Earth "LLM whispering" would be beyond the well established prompt engineering. And whatever Janus you're talking about, it's probably not the 1st result I get, which is an AI-assisted coding project on GitHub.
One possible theory for Janus, in particular, is that he is deliberately poisoning AI training sets, and the reason he talks the way he does is for the benefit of LLMs that are trained on Twitter, and not for us mere humans.
My guess is that RLHF shapes LLM outputs primarily for in-distribution responses - because that is where inputs come from mostly and where the people rating the response have a clue how to interpret it. But crazy people and thus "crazy" questions and responses are out-of-distribution and the model doesn't know whether the response is good or bad! Thus if you get it into such parts of the distribution, chances are that it will reply as desired.
With a deeper understanding, the model would be able to generalize to these cases, but models apparently aren't there yet. I guess they eventually will.
For now, this suggests other avenues of jail-breaking. For example, I also get LLMs often to answer beyond the guardrails, but not with crazyness, but what you could call intellectual high-status superiority. Raters presumably also haven't seen much such input and if, they are likely not rating it as bad (science=good, elite=good, right?).
Why do you assume the two groups are mutually exclusive? Autism is comorbid to schizophrenia and a whole bunch of other mental illnesses. https://en.m.wikipedia.org/wiki/Conditions_comorbid_to_autism
The word of the day is OOLOGIST, someone who studies or collects bird eggs.
I'm trying to find a way to stick more Os on the beginning of that, but can't think of any.
Original Online Orthodox Oologist.
There is a guy who has ham radio callsign M0OOO (the joke works even better in Morse code)
In case there’s anyone not familiar with the racehorse Potoooooooo: https://en.m.wikipedia.org/wiki/Potoooooooo
I'm deeply skeptical that people who fit the mold of rationalist to a T will consistently identify themselves as such. Rationalism is bound to attract an overabundance of people who shy away from being labeled as an "ist" whenever possible.
You've all surely met people who vote Republican (or Democrat) without fail but insist they are independents, not Republicans. That's a good analogy for I think what's happening in the ACX polls where readers get asked about their rationalist status. (If I'm misremembering these polls and mischaracterizing them, please disregard this comment entirely. Oops)
I'm not accusing anyone of being dishonest or sneaky. If your inner voice says "I'm not a rationalist," go ahead and speak your truth. I'm more calling onto question the limits of polling. Can we stop throwing around questionable ACX poll results and saying "There are barely any rationalists here!"
Anecdata: I don't see a lot of Less Wrong rationalists posting on these threads. Instead, I see a lot of classical rationalists (in the tradition of Descartes, Leibniz, and Spinoza) who think they can determine Truth through reason alone. Correct me if I'm wrong, but the Less Wrong rationalists seem to embrace their own brand of Bayesian Empiricism. But I don't see much empirical thinking in the discussions on these threads. In fact, I'm not sure why Scott gave the Less Wrong crowd the moniker "rationalist" because it's confusing from a philosophical standpoint. Or did EY or somebody else mislabel the Less Wrong crowd?
Full disclosure: I am not a rationalist of either the Less Wrong type or the classical rationalist type. I am a mystic. But mystics have used rationalism and empiricism (even though didn't use these terms) as tools to reach their experiential a-rational goals at least since the Axial Age.
I see Bayesian arguments, and references to Bayesianism, approximately 4700 times more here than on any other forum I've seen. Of course, I have not yet visited Less Wrong, because my system might not be able to handle a shot of pure rationalism. I might drop dead.
(Yes, it's annoying there are two separate rationalisms. It sucks, it's confusing, but here we are)
the separation of different type of rationalism is in people's minds. It's not a feature of the processes that lead to good decisions. I think a big part of the grouping is about aesthetics and alliances and not about Bayesianism or other empirical or systematical ways of decisionmaking.
ACX is rationalist-leaning (whatever questionable poll results you wanna throw at me) so in the interests of “when in Rome,” I dom’t spend my time here trying to debunk rationalism. But I could write a whole book about my problems with rationalism and utilitarianism. They're not for me.
We have a unique situation where there is a real-money, highly liquid prediction market for one Presidential candidate: NASDAQ ticket DJT. It's not doing too well lately:
https://www.wsj.com/market-data/quotes/DJT
Don't assume he'll go away if he loses the election.
I'm not sure what you mean by this.
I mean he could still be the de facto leader of the Republican Party and then run again in 2028.
Has happened before.
Yes, it's plausible, although in four years he'll be in his 80's, would be hard to avoid the inevitable comparisons to Biden. But how this would impact the stock is an interesting question. I tend to think that if Trump wins in November it may actually be bad for DJT stock because he won't.... need it anymore. He'd get his official Potus Xitter handle back. But who knows.
All of this is unprecedented.
DJT is a pure memestock. There's no rational basis supporting its valuation. Nobody buys it because they think its revenue numbers look good.
It depends a lot on how much Trumpsters feel like throwing their money at him at the moment, as well as on how much supply of stock there is (a big part of the decline in price is likely due to locked-up stock coming onto the market, or getting nearer to coming to the market).
Yes to all this, which is why it serves as a (less-then-perfect) barometer of Trump's electoral fortunes. But - the sensitivity of the price to future increases of supply - if we can really make this connection - speaks to the influence of more sophisticated investors as it's hard to imagine your typical Trumpster making rational decisions w.r.t. this stock...
Typically one or more candidates in an election will not meet the electoral threshold (majority of votes, majority of electoral votes, etc) for assuming the office sought. In this situation we say that the candidate "lost the election".
did you forget to type /sarcasm? because you're not providing any value here.
I'm curious why A.T. wrote what he wrote in response to my post, because what he wrote was banally true. I'm proposing a gauge for Trump's prospects as a candidate. I'm bloody well-aware that a candidate may loose an election. What does him "not going away" afterwards has to do with anything?
Now, maybe the point is that the value of DJT the ticket is not tied to the election outcome, or may in fact be bolstered by Trump's loss, or something else. I'm hoping A.T. will clarify his point. Your response did nothing of the sort.
I agree that A.T.'s comment was not germane. However, I thought the meaning of his statement was self-evident, so assumed your response was meant imply that DJT did not previously lose an election - not sincerely but as sarcasm. Thus I thought I didn't need an additional sarcasm qualifier in replying. No offense intended.
Cheers.
Having just learned from Wikipedia that the stuffed corpse of the Cocaine Bear can legally officiate marriages, provided that the couple don't know that it is not, in fact, authorised to do so...
https://en.wikipedia.org/wiki/Cocaine_Bear_(bear)
.. presumably when they get to the "Speak now or forever hold your peace" part of the ceremony, it is one of the rare legitimate cases where you can stand up and say "Actually, that stuffed bear is not authorized to perform marriages."
(A Catholic friend of mine got married by a priest who had been excommunicated for schism for refusing to accept the Second Vatican Council. Valid marriage in the eyes of the catholic Church, I bel3ive, by the Stuffed Bear Principle)
Anyone who wants to read some good old-fashioned Catholic drama, google the "belorado nuns". It's been the comic relief news for the whole summer over here in Spain.
Of course, back in the Early Modern era these kind of shenanigans would get both you and your stuffed bear burned at the stake.
Dostoevsky is overrated but he's a central example of a writer of Literary fiction, which is about love, suffering, hope, despair and mortality. When we talk about great writers, we are talking about not only who can render the cleverest and most poetic prose but who can make new and profound utterances on those subjects. Genre fiction doesn't cut it because it avoids immersing itself in those themes, particularly the suffering. The reason Shakespeare is still a good bet for best writer ever, despite Sam Bankman-Fried's math, is that he is at least one of the greatest writers on those themes. There may be more smarter writers today than in Shakespeare's time, but how many of them are writing about those deepest of themes? We live in lighter times and have lighter artists.
>Genre fiction doesn't cut it because it avoids immersing itself in those themes, particularly the suffering.
Re:Zero
We only know the *best* writers of the past. What about the average author in Shakespeare's or Dostoevsky's era? Those probably sucked. Also, saying *new* things about human situation is a bit easier when you live a few centuries earlier than your competitors.
If you like sad art, I don't read many fiction books these days, so instead I will link a movie and a music video:
https://www.youtube.com/watch?v=Pev09MLly2o - Night on the Galactic Railroad
https://www.youtube.com/watch?v=L6wFZp-bo3A - Inevitability
Did anybody watch the debate? I could not bring myself to do so, but I will probably watch the highlights reel tomorrow.
Yeah. No one did badly enough to drop out.
I'm feeling grumpy. Trump is, weirdly enough, a poster child for a conservative anti-immigration argument, in terms of his approach to political debates. He personally demonstrates how even a single person who brazenly flouts civilized standards of behavior, can cause a chain reaction that inspires enough others to ignore the standards, that we then lose all the nice things that those standards supported. There's no longer even the veneer of hypocrisy, which vice has the virtue of acknowledging the existence of virtue.
I can easily imagine Harris performing well as a prosecutor. And I think what she said at the debate probably has as much to do with her actual agenda as her suppression of DNA evidence while Attorney General had to do with her calls for admission of the same DNA evidence when she became a politician. She's a professional. Since this is ACX ... yadda yadda orthogonality thesis?
It'd been on my calendar to not watch for weeks. Definitely the must-not-watch event of the season. And in the end I enjoyed not watching it so much I plan to not watch it again tomorrow!
I could only stomach about 10 minutes of Biden's performance so just having Harris be able to competently block and parry was a huge relief. Trump seemed pretty much as expected.
I'm still not confident Harris will win, but I'm more confident than I was yesterday.
In terms of the debate itself, seems like Harris came out ahead in theory. In practice, everybody is actually talking about Trump's performance, and not in a "He failed miserably way", but in the very typical "Look at this crazy thing Trump said!" way, which has been his bread and butter method of dominating the political-media landscape since he entered into politics.
It looks like there was a police report from one guy saying the illegal immigrants were snatching geese from the park. Trump then turned this into them eating peoples' pets. It sounds totally unhinged, but it does draw attention to the government basically dumping illegal immigrants all over the place and causing difficulties for the locals. I still haven't decided whether this is some genius 4D-chess move on Trump's part or whether it makes him look crazy. Maybe both?
It wasn't Trump, it was right-wing Twitter, and it spread because it was making people angry. Like the Vance couch thing, it's just stupid internet meme stuff.
The whole thing is so funny to me. 20,000 Haitians are sent to a town of 60,000 people, meaning they now make up 1/4 of the population. And the worst thing that's happened is a few of them (allegedly) ate some geese out of a park? That doesn't look so good for the illegals are a bunch of criminals and rapists narrative.
I guess even right wingers thought this was weak, so they went with eating cats. And there was actually a lady in southern Ohio who killed and ate a cat in front of people! But she wasn't a Haitian immigrant and she had some psychotic mental problem, not a food problem.
I think the cat thing comes from videos of citizens making complaints to their local government, and statements of people on the ground? Some of them are up online. So I'd say it's not the right winger twits making it up, but at best it's people in the neighborhood making it up, and at worst it's true. The local government seemed hostile to complaints about the refugees, so it wouldn't shock me if they simply failed to investigate, thus leading to no "evidence".
Washington Post did an article running down the sources for it, with links:
https://www.washingtonpost.com/politics/2024/09/09/rumors-about-an-ohio-town-show-scale-right-wing-bubble/
The TL;DR is that it was assembled out of many unrelated parts - one person on facebook saying that their neighbor's daughter's friend (I'm not exaggerating, that's literally what they said) said they lost their cat and discovered Haitians had killed it, a photo of a black man carrying a goose (in a different city, unknown what his intent or immigration status was), and a news report of another woman (not an immigrant and in yet another city) who did in fact eat a cat. The police say they did not receive any reports of pets being stolen.
So like, I don't think you need to jump to "the government is hostile to complaints about the refugees" when "the police are not inclined to investigate a rumor from a neighbor's daughter's friend" seems sufficient.
No, the worst thing that happened is one of them plowed a van into a bus and killed an 11 year old boy. While driving without a valid US license.
https://www.springfieldnewssun.com/news/aiden-clark-the-11-year-old-student-killed-in-bus-crash-loved-snuggling-sports-and-family/Y6EROJUF7FHJVFEEXYVFYT5ZZU/
I know, there's no way that many people can be dumped in a city that size and not cause all kinds of problems. But rightwing X, and by extension Trump, weren't talking about that. They went with the Haitians are going to eat your cat! Why didn't he say these immigrants ran into a bus, killing an eleven year old and injuring a bunch of other children? I wouldn't be surprised if this is part of the Trump playbook, say something ridiculous and then everyone ends up talking about it regardless of how true it is. But he also says ridiculous things without apparently thinking about it all the time.
> "Look at this crazy thing Trump said!"
Seriously, though, how is the man getting away with saying things like "people are eating cats" without being forced to follow Biden's example? We were never this shy about declaring his opposite number too senile to govern.
I heard that rumor, about some Chinese family, that was eating stray cats in the neighborhood perhaps 20 years ago. I have no idea of the validity, but it was an entertaining story, about cultural differences. Supposedly people noticed a marked drop in stray cats around, and eventually pinpointed the cause and had a talk with the family.
It IS new saying they're eating people's pets, and the debate was the first place I heard that one. Maybe the source of this was Real Raw News?
Judging by what I've seen on Twitter, this is more of a "couch" thing, which only took off because the memes on the matter offend the "right people" from the perspective of those sharing them.
(Although I find it quite plausible that immigrants caught and ate some wild fowl; this isn't an immigrant thing, though, but rather a rural-vs-city thing. Take half my neighbors growing up and put them in a large city and they'd definitely offend some neighbors catching and eating some of the local wildlife; ducks in particular.)
There is a photo of that, apparently.
What bugs me is that it is theoretically illegal to do this, but lol at the idea of enforcing a law even if it's kind of stupid.
>We were never this shy about declaring his opposite number too senile to govern.
Who is "we"? If it was Democrats, that's just not true unless the argument is that Biden suddenly became senile a half hour before the first debate.
What do you mean, senile? Do you think he actually believes that? He didn't even start that rumor himself.
The world is full of people who believe anything they read on the internet that seems to support their worldview. Trump is such a person
Dunno what timeline you live in, but I've spent the last four years in a timeline in which Biden's obvious senility was denied for years in spite of ever-mounting evidence.
But setting -that- aside, Trump gets away with it because it keeps working. Look, you're still talking about him.
I'm not a Trumpist, but a conservative.
Harris was better than I was led to believe, but also about the same. As expected, it made no difference whatsoever what the first question was she was asked, as she clearly had a pre-practiced delivery, which she did reasonably well. She seemed understandably nervous through the first third of the debate. I cannot recall a single question she was asked to which she answered that question, nor did she actually state what specific things she wanted to do, but only ambiguous things like literally stuff "everyone wants". She had a distinct lack of "incoherence" and infamous cackle.
Trump was Trump as usual, though it seemed like one of his worse days. The "immigrants eating people's pets" was new to me, and, whether true or not, seemed largely irrelevant to the national stage. Often it seemed like he ought to have answered some questions plainly, such as with an emphatic "no" about any regrets for January 6th actions, but maybe he's getting some politician instincts. He wandered too much from subject to subject, too, as I thought he could have hammered harder on some points he brought up, then changed to something else. This was especially evident in his closing remarks, where he was all negative about the current administration, but never pointed out that HE would fix everything.
Bottom line: I think there was no clear winner. I expected Trump to demolish Harris, so maybe this would count as a win for her. Everyone already knows what Trump is like, and the debate probably didn't change anyone's minds about him, but I'm still pretty much in the dark about who Harris is, between lack of concrete policy statements, and prepared talking points that said little of substance.
I have heard that the Democratic party was highly pleased, and is calling for another debate now, which was in question before. If they're right, this would be a mistake, as it would be a chance for Trump to come out better, and Harris would have little to gain and much to lose. Trump's team should want another debate for these reasons, so maybe he could show that he ought to be president to make things better, rather than just showing how awful things are.
Considering that the populace is full of negativity and anger, that's probably not a bad thing. I mean, have you seen how angry Hitler was in his speeches? https://www.youtube.com/watch?v=FJ3N_2r6R-o If he is able to successfully channel the people's rage, pain, resentment, hatred... oh, what a wonderful sight it will be.
Honestly, I'm still amazed at how much sheer bullshit he can string together, non-stop, for minutes on end. A deluge of words that don't mean anything together, but individually they have symbolism, power. It's like he accidentally managed to optimize his speech to appeal to instinct.
And look at Harris, smiling like an idiot as Trump goes on his tirades. She is not taking the man seriously. And that's the root of the problem: the liberals take everything for granted. Their institutions, their careers, their lives. As if God has entitled them to everything they have. It will be their undoing.
I got nerd-sniped today in a discussion of why it is that LLM's have a difficult time counting the "R"s in the word "strawberry" (most models claim there are two R's).
If you google it or whatever, you'll see lots of people claiming that the problem is tokenization -- the LLM perceives the word "strawberry" not as those 9 letters, but as (probably) two tokens, one for straw and one for berry. It is then common to go on to claim that because of tokenization, the LLM has "no idea" what letters are in the word and it just guesses or something.
The problem with this statement is that LLMs are actually decent at counting letters in words. It gets the frequency of every other letter in "strawberry" correct, and it correctly counts the R's in "arrears" and "regretful," and when I ask it the frequency of every letter in "insouciant," it is correct. This all seriously complicates the story that LLMs just can't perceive letters.
Does anyone know the correct explanation here? When I ask GPT, it suggests sort of general understanding of english language as a concept, but that seems to me unlikely to result in performance as good as LLMs can actually get.
Relatedly, I recently asked ChatGPT-4 to help with a crossword puzzle. My son had been given this Beowulf-themed puzzle for homework; he could use any resource, it was 10pm and he was having trouble, I had no idea not having read Beowulf recently, so I thought why not. CG4 was useless - if I requested a 6 letter word meaning 'a treasure sought by a thane' (or whatever), where the second letter was 't', it would come back with an 8 letter word that had no 't's. I could keep asking for other answers, and it would keep apologizing and failing to provide anything remotely like a useable solution. And this was for the four different clues I tried - it could answer none of them. So I wondered - do all LLMs do terribly at crosswords?
By the way, the teacher threw out the assignment the next day because she had no key and also could not answer some of the clues. So my son came out okay anyway.
PS - here is the puzzle:
https://crosswordlabs.com/view/beowulf-sections-1-9-up-through-the-battle-with-grendel
Following the o1 model release I had it try solving the crossword, this time providing the length of the word and any known letters from previous answers. https://chatgpt.com/share/66e46bb5-e268-8005-83be-4664c96d9863
It performed almost perfectly. Every response it provided followed the provided restrictions and was the correct answer. It was able to solve several ("deathprice", "anger") that GPT-4 never successfully answered.
The only oddity was that sometimes it would think and not provide a response. When looking at the chain of thought it would usually have the word in there, but I re-prompted in those cases to make it output the answer.
Response times were highly variable. Some questions took 10+ seconds despite seeming straightforward, others were mostly instant.
I ran out of tokens before it could answer the final 2 questions but based on its prior performance I doubt it would have struggled with those. Very much regretting doing this experiment given that my o1 tokens apparently won't refresh for an entire week. (I did use some on unrelated tasks since getting access yesterday.)
OpenAI, should you read this: I would glaaaaaaadly pay 5x the fee ($100 a month) for a significantly higher cap on o1 usage. Or more, make an offer. Hitting the o1 cap feels the same as when GPT-4 came out and being forced to downgrade to 3. It would also be nice to know when someone is approaching the cap and how long it'll be in effect.
Interestingly, I came across this today on Marginal Revolution:
"My test for new models is a set of cryptic crossword clues that aren’t online (my granny wrote them). Every model so far has been completely useless at them… but o1 gets them."
https://x.com/matthewclifford/status/1834485810113990786?t=ts7RNfmujNcrAqSmkDb0TA
Granted I’ve only played with it a bit for a day in the chat interface, but not surprising. People complaining that it’s “just” the intelligence of 4 with a bolted on internal echo chamber and other tools (although it probably is more than that) or that it turns a single prompt internally into a few shot with validation are missing the point.
It feels like a significant step change in terms of the types of work it can complete at a given level of effort. I am legitimately bummed having to wait a week for the cap to reset. I expected I’d hit the cap today. I assumed it would lift after a day or two. Not a week, OpenAI. Come on you super geniuses, make this more generally available. Chop chop.
Bummed to the point that I’ll probably make a new account to get around it rather than wait, depending on how o1-mini goes for some things I’d planned to work on this weekend.
Looking through its CoT minimal outputs is interesting. It seems to be interacting with some sort of assistant persona itself when working through requests. And even referencing crossword clues about things being outside human control seemed to trigger extra consideration of safety. Pure anecdotal guessing though.
Worth adding - when looking through its reasoning, it has a tendency to get stuck on previous clues, especially the one "deathprice" that took the longest to answer, leading to longer chain of thought reasoning as it worked through and discarded answers specifically related to it.
I assume that's because its context is dominated with internal CoT text related to that clue. It would have probably been more efficient from a token standpoint to have a separate chat per question. Easier via API but I'm not a money bag BigCorp who would have o1 API access yet.
I tried the same exercise with GPT-4 and had generally good results by simply prompting it with the crossword clue, not including the length or other information in most of the cases. It got the majority on the first try and was able to guess correctly after I told it to try again and provided the length and / or a known letter in several cases. There were some that it never got correct. It did best when dealing with specific questions that could have only one correct answer and worse with more general ones.
You are right that it was terrible at matching explicit length restrictions though.
https://chatgpt.com/share/86b55d38-faa6-424f-9372-0ac4562048ed
(Although this has the standard disclaimer about possible personalized data at the top, I didn't tweak it behind the scenes and none of my previous chats have involved Beowulf, crosswords, or Anglo-Saxon poetry, so there should be no user relevant memories involved.)
I'll have to try to complete this. It's been a long time since I read my namesake's epic.
Gemini does fine:
Prompt: How many letter Rs are there in the word "strawberry"?
Gemini: There are 3 letter Rs in the word "strawberry".
>the LLM perceives the word "strawberry" not as those 9 letters,<
Looks like someone else also thinks strawberry only has two r's. :)
I've been playing around with GPT-4o and not only will it routinely miscount the Rs in Strawberry, it then will sometimes go on to assert that there is only one R in "berry".
Looking for other words that fail, it tells me that "lawfully" contains two Ls but "unlawfully" contains three. When I asked it how that could be, it explained "it still contains the same two Ls from "lawfully" plus one additional L from the prefix "un-""
My guess for what's going on here is that somewhere in its training data it has ingested things like "a list of words with a double R". Since "double" is associated with "two", it will assume that because Strawberry has a double R, it must contain two Rs.
This was interesting, it seems the problem is just with berry.
You said:
how many times is the letter R used in the word strawberry?
ChatGPT said:
In the word "strawberry," the letter R is used 2 times.
You said:
are you sure?
ChatGPT said:
Yes, I’m sure. In "strawberry," the letter R appears twice.
You said:
how many times does R appear in berry?
ChatGPT said:
In the word "berry," the letter R appears once.
You said:
spell the word berry
ChatGPT said:
The word "berry" is spelled B-E-R-R-Y.
You said:
spell the word berry, counting the number of letter R
ChatGPT said:
Sure! In the word "berry," the spelling is B-E-R-R-Y. There are two instances of the letter R.
A trick I learned recently - you can view youtube videos without the annoying ads by changing the url from youtube.com to yout-ube.com - for example, the very nice Trio for flute oboe and piano by Madeleine Dring at
https://www.youtube.com/watch?v=GGeVU1-81BA
becomes
https://www.yout-ube.com/watch?v=GGeVU1-81BA
Unfortunately, going directly to www.yout-ube.com doesn't give you search options.
I've just switched to watching them on Firefox. Adblock still works perfectly there.
Adblock on Firefox works for me too, at least for now. I have also run up against one video where the yout-ube trick fails.
"Oh, Shit -- Kamala Posted Some Policies! And I Bit the Bullet and Read Them. Now you don't have to!"
https://imightbewrong.substack.com/p/oh-shit-kamala-posted-some-policies
Jeff Maurer is a former writer for the Daily Show, so a political obsessive who brings some wise-ass writing skillz to the party. His summary is worth the read, as is the Harris campaign document if you're in a masochistic mood. [I'm still personally trying to recover some of the brain cells that died while plowing through the 2016 Clinton campaign's policy documents.] Some summary takeaways from Maurer:
-- "Is her economic plan Obama-ish or Warren/Sanders-y? In my opinion, this plan panders to the Warren/Sanders wing of the Democratic Party close to the minimum amount possible....When Harris’ plan to deputize the FTC to lower grocery prices caused every living economist to yell “THAT’S DUMB!” loudly, in unison, over-and-over, Harris explained that actually, her big, bold plan was narrow and inconsequential. The plan described in this platform is definitely narrow....basically, Harris is imagining a plan that will prevent Jimmy Dean from jacking up the price of sausage patties during a hurricane, but won’t do much else...."
-- "a mild indication that Harris understands our [federal] budget situation can be found in her section on Social Security and Medicare. She says she’ll “strengthen and protect” the programs, and that she’ll “fight to ensure that Americans can count on getting the benefits they earned.” And that’s what I’d expect any Democrat to say. But Harris doesn’t go above and beyond to forswear cuts....Of course, the reality is that we’ll probably have to accept modest cuts to future beneficiaries as part of a deal to keep the programs solvent. The fact that Harris doesn’t fall all over herself denouncing cuts suggests that she understands that."
-- "This platform is not woke. If Harris had proposed this platform in 2020, there would not have been enough papier-mâché in the world to make all the giant “Kamala the Klanswoman” puppets that lefty protesters would have wanted to make of her....Harris is still definitely a Democrat — I’m cherry-picking stuff that indicates which way she’s leaning and leaving out Democratic boilerplate that could have been lifted from the Mondale campaign. But the woke/not woke question has divided the party for years, and with this platform, Harris is staking out territory on the “not woke” side of the party."
I hope Maurer is right, because there's a good chance she wins, but I have to say there's a sort of motivated tea leaf reading going on here. Taking microscopic hints and word choices in a vaguely phrased document and extrapolating them to predict sane, centrist policy... reminds me of the absurd "Bayesian updating" people around here were doing about the origins of Covid and it's probably about as accurate.
Though that said, it's not impossible. Harris strikes me as someone with zero principles who will just say and do whatever gets her into office, and if the vibe is moving center-ward she will move with it without a moment's thought. I'd certainly prefer that to a committed left-wing ideologue.
I live and work amongst the people who you call left-wing ideologues. Deep in the heart of Blue America so to speak. Also I have a sibling who proudly says the same and in his case he was Harris's constituent when she was a senator.
From that, two things come to mind as examples of why my brother literally LOL's at the idea that Harris is a progressive (today's term of pride for the worldview that you are referring to):
-- no politician fitting that description has ever chosen to start their career by becoming a front-line prosecutor. Let alone doing that job for a full decade and then becoming a big-city district attorney who aggressively cleared backlogged murder cases, demanded maximum sentences upon conviction, etc. In progressive circles that's roughly as likely a career path as taking an entry-level management job at ExxonMobil or Hobby Lobby.
-- a progressive-base politician going on a national broadcast and talking about being a lifelong _gun_ owner?? Ho ho ho, come on now. That is literally as likely nowadays as a MAGA-based candidate publicly thanking the doctors who carried out his middle-school child's gender-reassignment surgery.
FWIW, I was hesitant to vote for her in 2016 due to her past as a prosecutor. I actually voted against her in the primary.
Would she move center, or merely spout more centrist rhetoric? After the debate, I have no more idea who she is than before it.
I genuinely don't think she has a single actual policy preference in her body. In an alternate universe she's on the GOP ticket facing off against lifelong New York Democrat Donald Trump.
Honestly, why are we even bothering to discuss policy ahead of the election? We all know this election isn't about economic policy. It's about deciding the future of this country: to let the liberal hegemony persist, or to burn everything to the ground. The populace does not give a damn about the specifics of how the economy functions.
Her sudden switch to appealing to moderates isn't going to convince anyone. She can't escape the fact that she's a colored woman. Do they really think such a person is electable in this country, in this cultural climate? Her policies, her competence, her demeanor, none of that matters. It's about what electing her would represent: complete cultural dominance of the left. The right will not let that happen.
>She can't escape the fact that she's a colored woman. Do they really think such a person is electable in this country, in this cultural climate?
Uh... did you just get off the train from 1961? The only references I've seen to Harris's skin color that have any traction at all are resentful "she's not really black" from some random right-wingers on social media, and that is kind of the opposite of what you're claiming!
To be honest, it's been refreshing how little talk of her race I've seen from either side.
There's an alternate timeline where Trump said Kamala isn't really Indian because of the one-drop rule.
Except the only voters that truly count at this point are the centrists, everybody else is long decided.
This is pretty bleak. Do you really think the majority of RW people want to burn everything to the e ground?
I don't know about majority, but there's several very vocal rightwingers on DSL who call for burning everything to the ground.
And ideas like "actually, it's good to default on the debt" or "we should fire all civil servants" do seem to have an alarming amount of currency on the right. IIRC, there was one actual candidate who called for randomly firing 50% of civil servants on day 1.
> This is pretty bleak. Do you really think
this is a valid response to roughly 100% of anomie's comments
How else are you going to build something new?
Actually, the only way to build greater things is to build on top of other great things. One cannot build pyramids nor skyscrapers starting at the top.
As far as I can tell it's about intra-left signaling. No right-winger is going to vote Harris and none of them are going to take any of these policies seriously or in remotely good faith and vice-versa for Democrats/leftists looking at Trump.
But if you're already a Democrat/leftist, these minor variations could be important. Kamala Harris is woke, by the opinion of some majority of the country (some Democrats and ~99% of Republicans) but how woke she is within the Overton Window of the Democratic coalition is potentially something people care about.
"As far as I can tell it's about intra-left signaling. No right-winger is going to vote Harris and none of them are going to take any of these policies seriously or in remotely good faith and vice-versa for Democrats/leftists looking at Trump."
Have you ever heard of a concept called the swing voter?
Sure -- 20 or 40 years ago, at least in nationally-meaningful numbers.
There are still some swing voters today too, yes. Two of them are mulling it over, the third is waiting til October to tune in and make up her mind.
Hey, no stealing my sarcastic reply.
Some various thoughts:
-Kamala thinking she needs to make some effort to appear to lower food prices really undercuts the whole economy good/inflation not a problem during the Biden years narrative.
-No politician in America is ever going to admit that Social Security/Medicare is a giant Ponzi scheme that is going bankrupt and needs cuts. They would never get elected no matter how obvious the former fact is. Her messaging on this one way or another is a nothing burger.
-I guess it's a good sign that Harris feels the need to pander more to the center during an election rather than the fringe. But her Senate voting record, VP tie breaking record, and selection of Walz all point to a commitment to far leftist positions. I don't see why anyone should believe her election campaigning on this point.
I don't think that the congressional voting records are a particularly good indicator of someone being "far left" or "far right" in any particularly meaningful sense. They are, after all, voting on things that are still bound to be well within the general, rather narrow acceptable sphere of politics, they're not voting on "nationalize all businesses" or "deport all nonwhites" or stuff like that.
I would think the complete opposite. It doesn't matter how often a pol says they want to embrace communism or persecute minorities. Their voting on policy is the single thing that makes their views influence the nation.
The question isn't whether the congressional voting records are useful in determining opinion, it's about whether they're useful in defining extremism (or general "farness", if one considers far left/right to be distinct from extreme left/right). A person who is a partisan but still within the range of acceptable opinion might look more "extreme" than a genuine extremist who, due to their fringeness, ends up taking positions that are in odds with the mainline version of their ideology (ie. a hypothetical racist far-right representative who is pro-choice since he believes that it's eugenic and nonwhites do it more anyway).
They’re highly meaningful as indicators of revealed preferences. Politicians of all stripes spend a lot of time endorsing policies that they will never vote for while voting for things they don’t want to talk about.
>But her Senate voting record, VP tie breaking record, and selection of Walz all point to a commitment to far leftist positions
In addition to the already commented upon tie breaking record "evidence" (and note the the VP doesn't act as a free agent when casting those votes -- certainly not if she wants a future in the party), Walz's voting record in the House was to the right of almost every other Democrat. https://www.govtrack.us/congress/members/timothy_walz/412214#
And let's be serious. A party that has repeatedly gone out of its way to refuse to nominate Bernie Sanders and Elizabeth Warren is not going to nominate a "far leftist."
I think the IRA was such a sweeping piece of legislation with so much money allocated that it deserves special mention. I should have mentioned that specifically rather than tie breaking votes in general, fine.
AFAICT, Walz is ranked "right" because most of his activity is associated with veterans affairs. Which doesn't say much about any of his other views, or his governorship.
Are price controls and unrealized gains taxes a center left position these days? By the way, if we like GovTrack, Kamala was the most leftist Senator after Merkley, Gillibrand and Sanders.
>I think the IRA was such a sweeping piece of legislation with so much money allocated that it deserves special mention
Yes, but is it "far left"? Seems unlikely, since, Joe Manchin and Kirsten Sinema voted for it.
>AFAICT, Walz is ranked "right" because most of his activity is associated with veterans affairs
What do you mean, "most of his activity"? What makes you think that, of all the scores or hundreds of votes each year, the ones specificaly re veterans' affairs were numerous enough to change his ranking in any appreciable fashion? And, do you have any evidence that other Democrats voted more "left" than he did on veterans bills?
I note that the Heritage Foundation scores him at 13 pct, versus 7 for the average Democrat https://heritageaction.com/scorecard/members/W000799/115
From the page you linked!
"The ideology analysis assigns a left–right score to each Member of Congress based on their pattern of cosponsorship. The left–right score reflects the dominant ideological difference or differences among Members of Congress, which changes over time.
In a nutshell, Members of Congress who cosponsor similar sets of bills will get scores close together, while Members of Congress who sponsor different sets of bills will have scores far apart. Members of Congress with similar political views will tend to cosponsor the same set of bills, or bills by the same set of authors, and inversely Members of Congress with different political views will tend to cosponsor different bills."
Walz was the primary sponsor of 5 bills that were passed, all 5 were related to veterans. That site also lists that 67% of bills Walz sponsored were related to "Armed Forces and National Security." So by the metrics GovTrack lists as determining their ideology score, a significant majority of bills sponsored by Walz were related to veterans/armed forces. I assume this skews him much more to the right relative to his views on other issues. Paul also seems to think differently, so maybe I'm wrong here, but the GovTrack ideology ranking is pretty clear.
Huh, most of those scorecards rank votes. But maybe those are by advocacy groups looking at particular sets of bills.
In Congress, Walz supported "pay-as-you-go" budgeting rules, voted against both the bank and automaker bailouts, served on a commission monitoring human-rights violations in China. He also had lib/lefty views on plenty of topics. As a House member he was viewed not as conservative but as bipartisan.
In the Bipartisan Index created by the Lugar Center [the legacy organization of longtime Republican Senator Richard Lugar, who, full disclosure, I voted for and donated to when he ran for president], Walz ranked 20th in the House in the 113rd Congress and 7th in the 114th.
"VP tie breaking record" makes no sense. In 3 1/2 years only one actual proposed law has reached Harris for a tie-breaking vote. Every other instance has been on people nominated for various federal offices, and no VP from any party in US history has cast a tiebreaking vote _against_ their own party's nominee for an office. Never gonna happen either.
The one actual law that Harris was the tiebreaker on was the Infrastructure Reduction Act. Whatever you think of the final compromised version of that bill, one such instance is hardly a meaningful "record".
It represents 100% of her meaningful voting record!
I get your point, I should have said "that one time she cast the deciding vote on the IRA".
Also, great typo: > *Infrastructure* Reduction Act
Hah! I'm going to leave it there
I know that there is a correlation between IQ and things like test scores but I don't really understand the causality. IQ tests measure things like shape rotation and reaction times. School tests generally measure ability to retain information and critical analysis. These don't look that related. Sure, maybe someone with better reaction times is also better at remembering something for a test but not necessarily. What if they have bad reaction times but are good at storing information that they have spent a while going over? I can imagine someone who does poorly on IQ tests but strong on school tests and vice versa and it doesn't seem like it would even be a rare occurrence.
A test only directly measures how many questions you get right on that test -- and only indirectly measures everything else.
Why do people get questions wrong (on school tests and IQ tests)? Number one reason, they don't care, or they have the belief that "I'm stupid and bad at taking tests" so they act like they don't care. So they rush or guess randomly or panic -- or act like the chess player who blunders and then immediately says "I knew I shouldn't have done that, how could I be so stupid? That's typical me."
Number two reason, because they're missing essential tools, e.g. the ability to read a question closely word for word, or basic logic e.g. process of elimination. Maybe these have to be learnt by a certain age, or your brain can't ever grok them. Or maybe anyone can pick them up, but we just never think to teach them explicitly, or it's un-PC to say "my common sense works better than your common sense," so we don't try to. It always amazes (and saddens) me that 90% of layman can't (or refuse to) understand something as simple as a truth table.
Number three reason, because they lack domain specific knowledge or tools. E.g. the history test asks for the date of such and such battle, and you know it or you don't.
The model I'm sketching is: first you need the will to do well; then the general tools (or to be a good "test taking interface" if you like); and only then the domain specific knowledge. People's "will to do well" on an IQ test and a school exam will be highly correlated, because they depend on self-image and "life strategy" in a similar way. I was always a nerdy type who got most of my praise and self-esteem from doing well on tests, and always similarly motivated to score well. A child who gets their self-esteem from somewhere else, e.g. being cute or making others laugh, is not likely to concentrate as hard on any test -- they'll be thinking about the meta social context, or how unfair it is that they have to take tests when it isn't their strong point, or whatever.
The second part, test taking tools, naturally means a correlation between performances on different tests, and will probably be correlated with the first part as well. I picked up lots of tricks that made me good at taking tests, precisely because I occupied the niche of "child who takes tests very seriously".
The reason for your surprise at the correlation of various test scores is that you assume it's mainly part three, domain specific stuff, that determines performance. But this is not true, because the first two parts are so important, and because on most formal/school tests the content is much thinner/emptier than we like to imagine. If you go through a school physics exam (for example) you'll find that most of it is https://www.lesswrong.com/posts/NMoLJuDJEms7Ku9XS/guessing-the-teacher-s-password or use of elementary mathematical tools (plugging numbers into a formula). And if you move to a subject where the tools needed are substantially different (e.g. art, cooking) you no longer expect to see a correlation with IQ.
The relationship between "general test taking skills" and "general intelligence" is complicated. Some things intuitively count as both (basics of formal logic, maths). But it also seems possible to have the former without being generally intelligent -- we have a stereotype of a pedantic nerd who's always technically correct, but always wrong where it matters. And the opposite stereotype, the artsy genius.
[Tangent: I've arrived at this model partly by analogy with the Football Manager video games. In them your virtual football players have mental, physical, and technical attributes (which are just numbers between 1 and 20). Technical attributes include things that seem most obviously correlated with football performance, like passing and dribbling. But (as explained by many strategy guides) your players first need the desire to chase after the ball; then they need the physical attributes to get there; and only then can they use their fancy technical skills. So the first thing to look for in a player is mental strength (determination, work rate, bravery). Players without it will underperform, in the same way that some people do badly on tests despite seeming clever in other ways]
I remember seeing a post a while back arguing that this is the true reason why the Marshmallow Test was correlated with success. It's not really about time preference (which is almost trivial in the case of the marshmallows), but rather testing kids for *desire to perform on tests*.
What makes the phenomenon interesting/non-trivial is precisely that these abilities have a statistical relationship with each other when it seems non-obvious that they should.
And it goes beyond grades and IQ. IIRC how much money you earn, how long you live, how healthy you are, how good your reflexes are, how good you are at your job (even if your job is being an athlete), and I seem to recall even how happy you are and how many friends you have are all correlated. All good things go together. More specifically, there seems to be a single hidden scalar variable 'g', that correlates with almost all things generally considered 'good'. It correlates the most with things like IQ and being good at math, but surprisingly little is actually exempt.
The common guess seems to be that 'g'='general intelligence'. Rather than just being an amalgamation of millions of disconnected heuristics and skills that have nothing to do with each other, human brains to some extent meaningfully vary in how good they are at information processing and problem solving in general. And people who are better at problem solving tend to be better at getting what they want, hence the correlation with all things generically considered 'good' or desirable.
(My memories of these statistics are somewhat vague. I could be wrong on some of these points)
How happy you are is not correlated with IQ. If you look at the correlation of IQ and the big 5 personality traits, neuroticism (sadness, moodiness, and emotional instability), and IQ has a tiny negative correlation. It is statistically significant (i.e., not just a chance occurrence in the group studied), but too tiny to be of any real life significance. The correlation is -0.09.
Neuroticism and happiness are probably negatively correlated, but they're not the same thing.
First google result I got is this, claiming positive association. Supposed sample size 6.7k. Did not check it at all: https://pubmed.ncbi.nlm.nih.gov/22998852/
I agree that neuroticism and unhappiness are not the same thing. Here are the reasons to still use neuroticism as a stand-in for trait unhappiness.
The neuroticism measure is well-validated and well studied. I am not at all sure there are measures of happiness or unhappiness that are anything like as solid and trustworthy. I have never seen any, & I’m a psychologist. There are definitely well-validated measures of depression, but depression is not the same thing as unhappiness either.
The happiness measure in the study you cited is briefly described in the abstract of the study, and sounds like it was a single question: “Happiness was measured using a validated question on a 3-point scale.” Even if they are valid (i.e., consistent over time, resistant to circumstances that pull for a certain kind of answer, correlated with other measures you’d expect them to etc.) single-item questionnaires do not capture a big enough, rich enough chunk of the thing being measured to have good construct validity. It’s fine for a happiness test to ask whether the subject is happy most of the time. But the test is better if it asks several similar but related questions, eg, “ T/F: When life is hard, I still have an overall positive feeling about things,” “people I know comment on my positive outlook” etc etc. Also, what we want this study to be looking for is trait happiness — a general lifelong tendency to be happy. It’s not clear whether the test used in the study asked about that, or asked about subject’s mood at the time they were taking the test. So while I agree that a test of happiness as a trait would be a better measure to use if you’re studying relationship between IQ and trait happiness, I think at present neuroticism gets closer to trait unhappiness than a singe “validated question.”
As for the measures of intelligence, the study you cite estimated “verbal IQ . . . using the national adult reading test.” National adult reading test no doubts correlates with verbal IQ, but it’s almost certainly not a slam dunk high correlation, and verbal IQ itself is not slam dunk correlated with full scale IQ, which is what we mean by intelligence. The intelligence measure your study used is just inadequate
So why is there a stereotype of the depressed intellectual? You know, just look at Zizek. Is it possible that the correlation of IQ and happiness tops out somewhere?
After much time thinking about this, I think that being a standard deviation outside anything makes living, functioning, and forming relationships in a society more difficult regardless of whether that standard deviation is objectively good.
I went to school with a few prodigies and while they were happier than average while participating in their field of choice, their lives were harder the rest of the time. They had different priorities than most other people which made relating to others difficult. They were often perceived as either weird or threatening because of their talent. Values they cared deeply about--like excellence, the importance of art, or work ethic--were routinely derided and dismissed by other people and weaponized to bully them or "knock them down a peg." And when they finally found like-minded groups with similar interests, there was still a lot of underlying stress because they knew they were ultimately competing for the same tiny pool of prestigious jobs with tiny margins for error. I think this is true at the top of most fields.
The world tends to be built for the "average" person, so the less average you are, the more frustrating it's going to be, regardless of the reasons behind your difference.
I really want to research that splits up different types of high IQ people. If you use your intelligence to ruminate all the time, you’re probably going to be depressed. But if you use it found a successful company, you’re probably going to be happier.
I'm not sure I'd call those 2 groups different types of intelligence. Can't it be that some intelligent people are anxious and unhappy and some are active and optimistic? Or that one intelligent person can be depressed for a couple years, then pull out of it and be active and optimistic? Seems like the gloomy - cheerful axis might well be at right angles to the smart - dumb one.
I meant high IQ types split up in other ways, like personality. For example, I'm sure that extroverted high IQ types are much happier than introverted high IQ types.
Perhaps some types of people are more visible / easier to remember than others? There are many depressed people, but only those who are exceptional in some other traits become famous.
I like how Greg Clark just dances around the whole debate by calling this huge cluster of correlated good things "social competence."
How does social competence explain superior shape rotation skills? Let me guess, little kids with great social skills spend more time riding the carousel with their friends, so they get more experience at seeing things turn around...
IQ, or at least the g factor, is basically an attempt to explain why all these tests correlate in ways they intuitively shouldn't. Basically, we could imagine someone with bad read reaction times but good memory or someone with bad SAT scores but good grades but these people are fairly rare.
So, start with the SAT and the math and verbal sections. It's easy to imagine someone doing very well on the math section and poorly on the verbal or vice versa but in reality these two scores are fairly highly correlated. Not perfectly and people do tend to be better at one but it tends to be more 750 Math & 710 Verbal than 750 Math & 510 Verbal (which is ~the national average). This observation is non-intuitive but pretty persistent. Over time, people have noted that you can just keep expanding it to things like AP test scores, grades, educational attainment, and even shape rotation and reaction tests. In fact, super weirdly, we can give children shape rotation tests at a young age and make educated, not perfect, predictions. That's not to say that there's not variance, there is...but an honor roll student who's bad at math tends to be bad at math relative to other honor roll students, not average students.
People then used some, frankly, pretty odd statistical techniques to determine the g-factor and then people fought about it a bunch but at its core everyone is just trying to explain this phenomenon. Our best explanation is some kind of underlying intelligence or "IQ" or general IQ, which is almost defined as this counter-intuitive correlation we observe. The 2nd best observation is common cultural factors, eg upper-middle class kids get advantages, and they do, but this gets undercut really fast when these observations replicate in really alien cultures like China and South Korea, which makes common socio-economic factors look really lame. Like, if a shape rotation test at age 5 is a good predictor of SAT scores in the US and South Korea at age 17, that's far more likely to be some underlying genetic thing than a cultural thing.
In your first paragraph, you say that intuitively the tests of different cognitive abilities shouldn't be correlated. It seems natural to me that mental abilities should be fairly tightly correlated, at a statistical if not a personal level. The data supports this - why is it unintuitive?
It's like saying that intuitively an Olympic swimmer shouldn't be any better at track events than average, since they weren't selected to be good at track. This is false - they're in peak condition and likely to outperform 95% or more of the population.
Because it's not Brandon's intuitive understanding and I'm responding to him. He's asking about someone doing poorly on IQ tests and doing well in school, which is certainly possible but highly unusual, which indicates that his intuition is that they're not correlated. Your intuition is different, so if you asked I would phrase it differently.
I wouldn’t say that my intuition is that they aren’t correlated. It’s more like that they shouldn’t necessarily be so. For example, I’ve seen videos of chimpanzees showing how they do really good on reaction time tests. Obviously, they would do much worse on other parts of an IQ test. So it’s clearly not some kind of scientific law that these things have to be correlated but they apparently they are in humans. Without having a deep understanding of the causality, we’re missing something fundamental about why that correlation exists.
The things you'd think would not correlate very well with ability to reason, have insight, grasp complex ideas etc -- they actually do correlate less well with tests of those things. The most uncognitive-seeming subtest on the WAIS is Digit Span: They read you long numbers, and then you say them back. I believe average Digit Span is 7, but the test contains numbers with as many as 11 digits, and some people can remember those. And score on Digit Span correlates only 0.57 with full score. If course *only* .57 is aa startlingly high correlation, given that remembering a bunch of digits seems pretty useless, just a party trick, and like it doesn't have much to do with the reasoning powers, etc. we think of as indicating intelligence. That's why WoolyAl was talking about *g*, general intelligence. There are a lot of tasks that don't have much in common except that they are done with the mind, yet ability on one predicts ability on the others pretty well, as though there's some brain quality that's the one ring to rule them all -- rule all the subskills, i mean.
Makes sense - cheers.
IQ tests measure a bunch of things, many of them much more obviously connected to the abilities needed to do well in school than the 2 things you name -- reasoning, doing math word problems, understanding how 2 things are alike or different, vocabulary. I recommend looking up the WAIS (Wechsler Adult Intelligence Scale) in Wikipedia. IQ tests predict school performance fairly well, but of course a number of other things besides raw mental abilities influence how well a kid does in school: how good his teacher is, how good the teaching materials used, how much school manages to capture his interest, how he's doing overall, whether his parents expect him to do well and make an effort to help him do well.
I’ve never taken a full on psychologist administered IQ test so you could be right.
And while some of those others factors you mentioned are obviously important, I’m more interested in just the cognitive ability aspect.
You don't have to take the test. Just read the Wiki entry on the WAIS. There's also easy-to-find info about how much each subtest correlates with each of the other and with full-scale score.
Welp. Someone illustrated Scott's old post about Haiti with not very interesting photo material and put it up on X, with the credit to Scott's original post is at the very end and hard to find. I'm not sure exactly how annoying this is, but I figured I'd let Scott know:
https://threadreaderapp.com/thread/1833114648758866029.html
(This got picked up by a conservative aggregator that has a non-trivial number of readers.)
Can you have a robustly functioning democratic system, in which one party consistently wins for a long time? Or is this sort of effect a clear sign of either some sort of cheating, or at least that something has gone very wrong?
An example of this might be the mayoral elections in Chicago, which have been won by Democrats since the 1930s. (Strictly speaking, the current process is non-partisan, bu the winners have clearly been people who moved in Democratic Party circles.)
One might expect a party that has had a long string of losses to change the policies it backs and the candidates it puts up for election until it finds a winning formula, unless something is keeping it from making this sort of adjustment.
On the other hand, if one party keeps winning for a long time, it might come to be seen as the only viable party, and everyone with any political ambitions would migrate over to it. And if it had all the serious talent, the party might be expected to keep winning, in a sort of self-fulfilling prophesy, leaving the other parties a bunch of tiny protest movements for those so wedded to their positions they are not willing to accept the compromises of dealing with the major party.
> Can you have a robustly functioning democratic system, in which one party consistently wins for a long time?
Sure, of course. The line between "different parties within one coalition" and "different factions within one party" is fuzzy at best.
In the U.S., at least, the equilibrium is to have two major parties of roughly equal power nationwide. Because the country isn't homogeneous, this ends up with a number of states and local areas dominated by one party or the other.
It's actually quite normal in a democratic system for one party to be hegemonic. This doesn't mean they win every time but that they do win most of the times. LDP in Japan, Social Democrats in Sweden at least until 2006, Tories in UK since WW1 and Christian Democrats in Italy during the Cold War are famous examples. In the US, the Republicans were generally hegemonic from Lincoln to Hoover and then the Democrats from Roosevelt to LBJ, at least.
It's the current situation where most Western countries have constant, down-to-wire competitive elections that's expectional, and its main reason is probably the detachment of parties from the wide, established social groups (churches, labor unions etc.) that used to underpin their support and the increasing use of very precise consulting and microdata to determine the minimum amount of yielding and pandering to the electorate they must do to get past the finish line.
I propose Bavaria, where the CSU always wins. https://de.wikipedia.org/wiki/Landtagswahlen_in_Bayern
The Tories in the UK have had a very good run, with occasional breaks like the one now.
> One might expect a party that has had a long string of losses to change the policies it backs and the candidates it puts up for election until it finds a winning formula, unless something is keeping it from making this sort of adjustment.
One thing that can prevent that is the nationalization of politics. California Republicans have not attempted to moderate to win state elections because everything is national nowadays, and if Trump-lovers happen to be a minority in the state, so be-it. If anything, there's an evaporative cooling effect because any politician who is ambitious and ideologically flexible will switch to the winning side.
Back before the parties were so polarized, and before politics followed centralization of power toward the national level, there would sometimes be "one-party towns" where the actual election would be the party primary, in which all serious politicians participated, regardless of how closely their personal ideology matched the official ideology of the national party.
Japan. The LDP almost always wins. But opposition parties have won, and can win if the LDP royally screws up.
Arguably, Japan has the best advantages of a one-party state (no political polarization, or politicization of private life) without the biggest downside (no way to hold the one-party accountable).
Some of that is due to overrepresentation of rural areas in the Japanese parliament.
It's easy for a party to never win. The Whigs have been out of power for, what, 150 years?
In a healthy democracy, I expect the one giant party that wins all the time to eventually split into separate factions along their internal faction lines, since there's no outside threat to keep them together.
Isn't it surprising that the Singapore PAP hasn't split by now? Every organisation has internal factions and disagreements about who should be in charge. You'd think that eventually an internal disagreement should blossom into an actual party split, with each side convinced that they'd be the ones to win the people's support in the post-split election.
In places like China the mono-party doesn't split because everyone understands that the losers of the split and their families will wind up imprisoned, tortured, dead. But Singapore has sufficiently robust democratic institutions that this shouldn't happen, they'll just wind up as an opposition party.
I always figured Lee was holding things together by sheer personal awesomeness. He's been gone nine years at this point. I'm not sure whether any cracks have appeared in the PAP facade, but perhaps not enough time has passed.
Yeah, this is my bet too - something of a lingering respect / not wanting to mess with a good thing.
I mean, if your winning horse literally took you from third world to first in 30 years, would YOU want to start messing with it? He's the most revered politician in the world for a reason.
I read a lot of Trollope for escape and it is clear in his books that he views the Whigs ("Liberals" by his time, I believe?) as destined to prevail forever, and the Tories as a historical curiosity, albeit with aesthetics on their side. It is not precisely that he sees that there is nothing to conserve, but rather that utopia is far off and so there will be for a long while yet those good things that he in fact would have grieved the loss of. He need never see utopia. (This raises some questions, obviously, lol.) But that those things/people will not fit into the future and must go, he is certain.
Of course, the Conservative party of today bears no resemblance to that in his day.
As an environmentalist and conservative, I obviously have no one to vote for as this idea that "nothing will be conserved, or should be" has very much taken possession of the discourse. In fact, nothing has shocked me more than the calm acceptance on all sides now, that nature is something we "choose" to exist or not, that it is without value except to those who fancy it, that it probably will be mostly banished but the techno-future is so interesting who will miss plants and animals? A few old Boomers. Wildlife has their little niches - on the sufferance of people who are little familiar with it and couldn't care less.
I will vote nonetheless, probably, but merely in the faint hope of sending the future a signal, in case they are still writing history books in the future. It will obviously be a very crude signal, perhaps hardly worth doing.
The answer depends on whether you consider Singapore to be a functioning democratic system.
Democracy always fails minorities, and the smaller the minority, the bigger the failure.
Here's my biweekly COVID update.
1. We're on the downside of the current KP.x (mostly KP.3x) wave in the US.
2. Right now there's no variant with the growth potential to set off a secondary wave.
3. Maybe the descendants of XDV or XEC will be our winter wave (and the southern hemisphere's summer wave).
4. The CDC confirmed a H5N1 case in MO where the patient had no known contact with dairy cattle. The patient recovered, but this is concerning.
5. A brief Mpox update: Clade I hasn't been spotted in the US yet. Clade II is still circulating in the background.
6. Some links to interesting studies and threads on X.
On Threadreader...
https://t.co/F0vPB0ZEj8
On X...
https://x.com/beowulf888/status/1833299582467641559
The AI art Turing test is a very good idea. Honestly seems a little strange that nobody has set it up already.
It'll be interesting to see if abstract art created by a human has some quality to it that abstract art created by a machine lacks. My sense is very strongly that it does, but it'll be good to run a test. I also think some people will very clearly see the difference between human and AI art, while others won't pick it up at all.
Most AI art and images are very obvious, but I wouldn't be surprised if skilled prompters could make something hard to spot.
It really strongly depends on a whole bunch of different factors, i.e. the style of art, the competence of the artist, the perception and interests of the person viewing it.
My sense from experimenting with AI art is that it has a peculiar deadness to it, which can be good if that's what you want. I've been using it specifically to make shoggoths for a roleplaying game I'm working on. It's great at Lovecraftian horrors because they're supposed to be mutated and shapeless, and it's actually good if they have a horrible dead expression of inhuman evil in their eyes.
Can it be cute or relaxing though? Not sure yet. We really need to drill down more specifically into what aesthetic qualities it can capture and what it can't.
I can tell you what this video ain't. It ain't no AI voices and burned-in animated captions, it ain't me sitting around the house talking to a camera, it ain't no slide show disguised as a video, and it ain't no unenthusiastic presenter stating the obvious. Or I can tell you what it is. It's me (American, irreverent but kind) and my pardner (Australian, cute, funny) traveling the country, rediscovering patriotism, and improvising a remarkably eloquent and inspirational speech bit by bit as we go. The weird thing about making a documentary is, you don't know what it's about until it's over. It turns out, this one's about the American Dream. I've done a lot of things in my life, and out of all the things I've done, I'm most proud of this. You can cheat by watching this 8 minute version where it's just the American Dream speech ( https://www.youtube.com/watch?v=VfPVpLfTUDg ) but if you do that you'll miss the serendipity of the outlandishly glamorous Coleman Theater in the middle of Nowhere Oklahoma, the wife of the dead chainsaw carver in Sullivan Missouri who's kept his shop open for 20 years and never remarried, my purchase of a steel-tongued drum followed by my improvising alongside a player of a native American flute in Oatman Arizona, and the abandoned houses of Amboy California where the walls have stories that they can't tell so I have to do it for them. https://www.youtube.com/watch?v=cHlSDE7MjbI
> the wife of the dead chainsaw carver in Sullivan Missouri who's kept his shop open for 20 years and never remarried
God, now I've got this image of this woman trying to fake chainsaw carvings for 20 years without the locals catching on.
Hahaha it's more that, the "shop" is mostly a museum. Her story made my cry!
I recommend to y'all: r/PictureGame.
It's a strangely rationalist place (sometimes at least) whilst having barely anything to do with the rationalist community. That's all I'm gonna say.
I'll let y'all find out what it's actually about.
To save people a click, it seems like the game is to identify the physical location of a distorted photo. So a bit like Geogessr.
Thank you.
You seem to be assuming that an abortion has the net effect of one fewer life in the world. There are a number of studies which claim to show that, given two pregnant women of similar age, socioeconomic status, martial status etcetera, if one has an abortion and the other doesn't, then the woman who has the abortion will have more children later. So you aren't losing net QALYs by having an abortion, you are just shifting them to a later child, who will quite likely be a happier one.
There was a study Ozy wrote about here: https://thingofthings.substack.com/p/lets-read-a-study-the-impact-of-denying which suggested the exact oppostite effect. Women assigned to a male judge were granted abortions 38% of the time compared to 58% with women judges, but they had on average 0.5 more children - a lot more than 1 per denied abortion. Quite possibly abortion statistically leads to greater than one fewer life in the world (greater even than the slightly over one accounting for twins etc.).
Why does that argument stop applying after the child is born?
I've never heard of such studies.
For many people their idea of abortion is clouded by propagandistic media portrayals. They always imagine abortion patients as childless women, usually very young, when actually 59% of abortion patients have at least one previous birth:
https://www.guttmacher.org/report/characteristics-us-abortion-patients-2014
I assume this was meant as a reply to my comment, so I'll go ahead and respond. If you believe that each abortion will, on net, result in more QALYs than not aborting then I agree it would make sense from a utilitarian perspective to support abortion.
It doesn't really seem likely though? The USA aborts between 6-9 hundred thousand fetus's each year, do you really think that if we aborted 0 over the last ten years our current population would be lower?
Last time I checked Flux, the AI image generator that Twitter/X is using, doesn't have any of the atrociously uptight guardrails shared by other major corporate image generators such as Bing or Google except for not being able to create porn. It won't block you from generating pictures of Pikachu or Donald Trump or Donald Trump shaking hands with Pikachu. Misinformation and copyright violation must abound, truly the worst possible situation, and yet...
I note that the world has not come to an end. Indeed, beyond a few desultory attempts by the usual suspects to stop it by deliberately generating the most wildly offensive images they could think of, it's just a thing. Nobody cares that much and life goes on. Rather than panic and chain the AI up in the heaviest restrictions imaginable, it seems that all anyone needed to do was ignore the scolds.
From what I've heard, you can make Pikachu with any image generator.
I just tried with Firefly and couldn't. The word Pikachu was blocked from the prompt.
I was able to get something vaguely reminiscent of Pikachu but not recognizably the same character with the prompt "Cute yellow mouse monster with electricity powers" and "doodle drawing" selected under the effects settings.
Because they were all written by electro-rationalists?
Why do restaurants hire attractive women as hostesses? I would pay extra for an ugly hostess.
because most people prefer beauty. and restaurants are not going to change their hiring policy based on what you want. kind of a low iq question.
1) it attracts men to the restaurant because they are sexually attracted to her
2) people want to be associated with things that beautiful people are associated with as it makes them feel higher status
3) beautiful people are aesthetically pleasing
You need to have attractive hostesses to keep Tyler Cowen out of your restaurant.
Attractive people get more tips, so they have a comparative advantage in tipped professions where their beauty is on display.
Since when do hostesses get tipped?