Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

IQ tests results for AI

trackingai.org

Post reply on HN
Print view View original

181–190 of 359 posts

Re: IQ tests results for AI

#181

Post by jacquesm »

hemabe wrote: post #73
brabel wrote: post #48

Earlier quoted context omitted.

Really? If not genetics then what is it? Just random??

IQ is largely genetic, even if some people claim otherwise. The evidence for this is now overwhelming: even when different ethnic groups grow up in very similar conditions in the same country, the PISA (which correlates r=0.9 with IQ) scores measured vary greatly. For example, among second-generation children in Germany, there are significant differences in PISA scores. Polish children achieve similar or even better…

Remarkable how you jumped from 'IQ is largely genetic' to 'Turkish children remain at the same poor level that their parents achieved in the tests'.

Don't you see your the mistake in your reasoning there?

Probably that too can be explained by genetics or maybe by a failing education system but the point is: there are very dumb Germans and very smart Turkish people who would still score different on an IQ test in German. Especially after going through the German school system (which of course would never discriminate against children with a different ethnic background /s) and so on. The confounding factors at play here make the whole comparison without accounting for those factors utterly meaningless.

Re: IQ tests results for AI

#182

Post by nerevarthelame »

leopoldj wrote: post #110
sigmoid10 wrote: post #22

Big caveat here: This website's method doesn't work at all for humans the way it works for LLMs. For humans, there is a strict time limit on these IQ tests (at least in officially recognised settings like Mensa). This kind of sequence completion is mostly a question of how fast your brain can iterate on problems. Being able to solve more questions within the time limit means you get a higher score because your brain…

The point of this is not so much to compare humans with AI. But to compare AI with other traditional software development approaches to solve this domain (IQ test, in this case). I believe, and I could be wrong, it will be nearly impossible, or too expensive, to develop deterministic software to beat AI in IQ test.

I agree that it's wrong to do so, but the maintainer of this site certainly thinks that the point is to compare humans with AI. He frequently compares the results to human IQ test takers without any sort of caveats: "Now o3 scores an IQ of 116, putting it in the top 15% of humans. The median Maximum Truth reader, for comparison, scored 104." [0]

0: https://www.maximumtruth.org/p/skyrocketing-ai-intelligence-...

Re: IQ tests results for AI

#183

Post by jacquesm »

robwwilliams wrote: post #177
froh wrote: post #91

Earlier quoted context omitted.

especially Germany is a very poor example for this claim as school performance (PISA) in this country correlates with the parents academic background more than anything. if your parents aren't educated you can be intelligent AF and still will fail the German school system while you can be dumb as a brick but still make Abitur if your parents drill you through. in Germany. and Welt is a media source with a right conse…

Yes. The smartest kid in my elementary school in Blankenese, Hamburg, Germany (a filthy-rich neighborhood) was an asthmatic son of the stable master of Alex Springer. By far the smartest kid. But he did NOT go to gymnasium. He was my best friend and I was furious at this social injustice. I was 10 and this was my first exposure to rigid class injustice. It still makes me mad. All of the other dumb rich kids from Blan…

This is pretty common. In the 80s (when I went to high school in Europe) the best way to predict whether or not someone would go to Gymnasium or Athenaeum was to look at how wealthy their parents were. It is still exactly the same today.

Re: IQ tests results for AI

#184

Post by d4rkn0d3z »

sejje wrote: post #176
d4rkn0d3z wrote: post #71

Human beings' IQ test results can vary significantly based on how much money is in their pockets. For example, if a farmer takes an IQ test before crops are harvested and sold they score lower than after crops are sold, in the same year. It seems fairly obvious to me that an LLM is the projection intelligence in the language domain. In other words, if you killed Intelligence and gave it a push in direction of languag…

What did your disclosure add to your point? They seem totally detached, the later being a brag.

The point is quite clear, if you think this is a brag you missed it though. Here it is again, pay attention; if a farmer has their IQ score change 15-20% by harvesting and selling their crop. Then it is very hard see a change in IQ score as a meaningful metric for intelligence, at least for humans that is.

Re: IQ tests results for AI

#185

Post by jacquesm »

mr_toad wrote: post #178
jacquesm wrote: post #139

Earlier quoted context omitted.

It used to be that society worked just fine with people of all grades of smarts. But we're rapidly getting to the point that to be able to earn a living wage you need to be above average, especially if you are sole income provider for a while. AI is further steepening that S curve's mid-section.

With the automation of agriculture and manufacturing economies have become highly service oriented. Low skilled service jobs have always paid poorly.

Poorly in the USA or Canada does not equal poorly in Europe and some other countries outside of Europe. Some societies never paid a living wage to begin with. The USA for instance has structurally blocked minimum wage increases on the flimsiest of pretexts for many years now. Meanwhile, inflation is through the roof. The end result of that combination is very predictable.

Re: IQ tests results for AI

#186

Post by pico303 »

Gimpei wrote: post #143

I’m surprised the score isn’t higher. What’s to stop an LLM from training on the complete corpus of IQ tests. I assume they’d get perfect scores

I was thinking that too. I wouldn’t even trust that the “offline” tests didn’t have the questions and answers posted online somewhere. This might really be an analysis of how extensive the dataset is for each LLM, not how much smarter one LLM is from another.

Re: IQ tests results for AI

#187

Post by holbrad »

cluckindan wrote: post #172
api wrote: post #155

Earlier quoted context omitted.

Throughout 99% of human history, the economy was mostly zero sum. If someone was rich it was because they stole it. If a group was wealthier it was because they stole it. By "stole it" I'm including theft of labor through slavery as well as the usual conquests and raiding. There was little to no innovation over time spans as long as thousands of years. Most of human culture and philosophy evolved during these periods…

I would hazard a guess that a lot of the so-called wealth is not actual, directly owned material wealth, but immaterial numbers: indirect, abstract proof of ownership — in Baudrillard jargon, simulacra of ownership. Numbers, which, due to their illiquid nature and their disconnect from material ownership, cannot be instantly and fully redeemed without sacrificing most of their purported value.

How can you possibly explain the world getting massively materially richer then ?

Are the new drugs we create immertial ? The better faster processor abstract ? The energy we produce unreal ?

Re: IQ tests results for AI

#188

Post by holbrad »

cedilla wrote: post #41
nsoonhui wrote: post #31

Earlier quoted context omitted.

Another component of this theory concerning g is that it's largely genetic, and immune to "intervention" AKA stability as you mentioned. See the classic "The Bell Curve" for a full exposition. Which makes me wonder what's the point of all the intervention in the form of teaching/parenting styles and whatnot, if g factor is nature and immutable by large? What's the logic of the educators here?

"The Bell Curve" is, let's say, highly controversial and not a good introduction into the topic. Its claim that genetics are the main predictor of IQ, which was very weakly supported at the time, has been completely and undeniably refuted by science in the thirty years since it's publication.

> has been completely and undeniably refuted by science in the thirty years since it's publication.

This is literally the exact opposite.

Re: IQ tests results for AI

#189

Post by Retric »

technothrasher wrote: post #117
gpt5 wrote: post #6

The way human IQ testing developed is that researchers noticed people who excel in one cognitive task tend to do well in others - the “positive manifold.” They then hypothesized a general factor, “g,” to explain this pattern. Early tests (e.g., Binet–Simon; later Stanford–Binet and Wechsler) sampled a wide range of tasks, and researchers used correlations and factor analysis to extract the common component, then norm…

> The way human IQ testing developed is that researchers noticed people who excel in one cognitive task tend to do well in others My son took an IQ test and it wouldn't score him because he breaks this assumption. He was getting 98% in some tasks and 2% in others. The psychologist giving him the test said it was unlikely enough pattern that they couldn't get an IQ result for him. He's been diagnosed with non-verbal l…

IMO g is purely an abstraction. As long as the rate you learn most things is within a reasonable bound spending more or less time learning/perfecting X impacts the time you spend on Y, resulting in people being generally more or less proficient in a huge range of common cognitive skills. Thus, testing those general skills is normally a proxy for a wide range of things.

LD breaks IQ because it results in noticeably uneven skill acquisition in even foundational skills. Meanwhile increasing levels of specialization reward being abnormally good at a very narrow sets of skills making IQ less significant. The #1 rock climber in the world gets sponsors, the 100th gets a hobby.

Re: IQ tests results for AI

#190

Post by mieubrisse »

thechao wrote: post #107
nialse wrote: post #81

Earlier quoted context omitted.

The training effect in test-retest is dependent on g as well. It is intelligent to learn from past experiences. Measuring g is hard and taking shortcuts is tempting. A reasonable repeatable g factor test takes hours, and is too often replaced by a single test. There are ways around the test-retest issues but they are roads less travelled.

My high school was right across from a branch of a university (UHD) where the PhD candidates developed IQ tests. We (the HS students) could take them for extra credit. My favorite example was a block-arranging test (there was a set of blocks & some pictures). Anyways, they printed the blocks "symmetrically"; once I figured that out, making the picture was limited only by how quickly I could move. (The test normally h…

A similar thing happened to me.

I once took a timed test with a section that had me translating a string of symbols to letters using a cipher, response being multiple choice. If you read the string left to right, there were multiple answer options that started with the same sequence of letters (so ostensibly you had to translate the entire string).

But if you read the string right to left, there was often only one answer option that matched (the right one). So I got away with translating only the last ~4 symbols, regardless of how long the string was. I blew through the section, and surely scored high.

I always wondered: did they realize this? Or did it artificially inflate my results?

And looking at the highest-entropy section felt natural to me, but only because of countless hours as a software engineer where the highest-entropy bit is at the end (filepaths, certain IDs, etc).

Is it really accurate to say I'm "more intelligent" because I've seen that pattern a ton before, whereas someone who hasn't isn't? I suspect not.

Post reply on HN