There are time limits but to say they're primarily measuring speed sounds wrong.

Does it? Take it to the limit: would a single question from the IQ test with unlimited time be a serious challenge? Of course not. So time is the crucial part of the test.

My IQ is around 120, so it's higher than average, but nothing exceptional. But I'm pretty good at math, so I competed at the national level in my country when I was at school. And as a part of that, I got to attend a summer camp with people who have a REALLY high IQ. Several of them went to win the IMO.

They were able to solve problems much faster than I could during training sessions. Often by the time I started feeling that I had a way to the solution, somebody already had the complete answer ready to be discussed. But with unlimited time to think, I could solve most problems.

In real life if you want to solve a problem and you have unlimited time you also have access to resources such as books and other people.

Then there are some problems that some (almost all!) people couldn't solve with unlimited time and resources. Say prove Fermat's Last Theorem.

And then there are the problems that test special knowledge. I remember being in a waves and vibrations class where on Thanksgiving break the prof assigned us the homework to solve for the motion of two coupled harmonic oscillators. I didn't have anywhere to go but I did have a cold and had a fever but I powered through it anyway and managed to solve it with sin and cos. I was the only one.

Some of it was my general intelligence g and some was determination and hard work, not having distractions, etc.

When we had class next week the prof solved it in minutes on the board using the fact that

exp(iθ) = cos(θ) + i sin(θ)

which of course made an impression on all of us! An IQ test tries to ask questions which don't depend on your knowledge but it's tough to do that completely.

From time to time you see a press release that some celebrity took an IQ test got the maximum score on Raven's Progressive Matrices. [1] A lot of people are going to say "of course these people are really smart, this is proven by their success" and maybe that's right, I mean a good actor is a quick study who is good at learning the lines, easily adapts to whatever they have to do in terms of systems, technology and procedures on the set, etc. But it would not be hard at all to get coached on Raven enough to get a perfect score.

[1] see https://en.wikipedia.org/wiki/Raven%27s_Progressive_Matrices which are about as independent of culture and language as is probably possible but... verbal intelligence is it's own thing, like I maxxed out every test of verbal intelligence I've taken since I was 17 and my IQ score is respectable despite having a bit of Meehl's "cognitive slippage" which impairs my complex task performance in a test environment although I have a lot of strategies to correct for it in real life

> Then there are some problems that some (almost all!) people couldn't solve with unlimited time and resources.

This again comes down to the huge search space and specialized knowledge.

> Say prove Fermat's Last Theorem.

I'm not so sure about it. Can you prove it by working from the statement itself? Likely no. Can you do that if you know about the work on elliptical curves and modular forms? It's not unrealistic.

> And then there are the problems that test special knowledge.

So AI would totally crush any human on that. It has in its mind all the knowledge, including all possible methods.

So yes, AI likely has "IQ in millions". It can spot patterns really well, and it has access to all the specialized knowledge possible.

you're taking this on a personal tangent.

Back to the AGI/ASI:

Thanks to frontier models the playing field can accommodate slower thinkers.

Academia nowadays are dominated by people like Tao, fast, knowledgeable, but not necessarily deep thinkers. In physics, even mathematics, of the (g)olden age, before standardised tests, famous problems and whatnot---- deep but slower thinkers had an edge. Now every thinker has access to roughly one quasiTao worth of g and a worldclass library in a box.

[heterodox postKeynesian macro, maybe one could call it postpostAbundance. Where goes universities' spend? Internal social (media) moderation. Not PR/HR!!]

It's possible "we" will return to such golden ages, once the "intellectual arts" lose 90% of the prestige and all of the perceived usefulness. When tests are more than useless for curbing "talent oversupply"/"support undersupply". Watch out for an AI fuelled broom dancing resurgence.