A single DGX Spark can service a whole department of mathematicians (or programmers), and you can cluster up to 4 of them them to fit very large models like GLM-5 and quants of Kimi K2.5. This is nearing frontier-level model size.

I understand the value proposition of the frontier cloud models, but we're not as far off from self-hosting as you think, and it's becoming more viable for domain-specific models.

> That would imply that a person with a high IQ score is uniformly better at every task compared to a person of lower IQ.

No it doesn't, that's why I said "tend to".

> On the contrary, it just misleads the educators to the student's real abilities. It ignores and hides the atypical talents that students possess.

The test simply doesn't measure atypical talents, so there's no "hiding". That's why I said the test is used to identify people who will face challenges. What happens next has nothing to do with the IQ test and everything to do with other methods of assessment that should be used. Every test has limitations that those applying the test should understand.

> That's far too narrow to be identified as an indicator of intelligence

That's why it's a (strong) correlation and not a strict equality.

Look, jumping height has been empirically demonstrated to be a good predictor of overall athletic performance, and you're coming here and saying that there exist some people without legs that can do more pull-ups than most people with legs. That's all well and good, but that doesn't somehow refute the idea that jumping height is a good predictor of overall athletic performance. It's just a complete red herring. If you happen to encounter someone without legs, then use a different test.

People with legs who don't exhibit good jumping heights but are very good at some specific athletic challenge could exist, but they tend to be good only at that specific challenge (and ones that are very, very similar). This is a completely different thing than the general factor of fitness, which is correlated with all sorts of physical abilities, just like general factor g is correlated with all sorts of cognitive abilities.

> The area where I see IQ failing miserably is in accounting for different types of intelligences.

We don't have any real evidence of "different types of intelligences". People with high IQ just tend to be better at all cognitive tasks than people with lower IQ. That's why it's considered a reasonably good measure of the g factor. People with higher IQ also have better memory recall.

While you're right that that IQ isn't the full story of a person's abilities, it's use for diagnosing people that may face challenges in a traditional learning environment well motivated. All of the examples you mention of people scoring poorly on IQ and facing challenges in school supports the use of IQ, it doesn't discount it.