> What is preventing AI from continuing to improve until it is absolutely better than humans at any mental task?

No matter how much compute time it's given to combine training samples with each other and run through a validation engine it will still be missing some chunk of the "long tail". To make progress in the long tail it would need to have understanding, and not just a mimicry of understanding. Unless that happens they will always be dependent on the humans that they are mimicking in order to improve.

It depends on if AI can invent cold fusion before running our of all the energy on Earth.

That’s one possibility. If it fails to convince a critical mass that it’s a net improvement in their lives, then the impediment to continual improvement will be sabotage.

> everything we care about

One qualitative distinction that remains for the time being is that humans care about things while AIs do not. Human drive and motivation is needed to have AI perform tasks.

Of course, this distinction isn’t set in stone.

I think there's been natural but steady progress with since 2024 with the release of the o1 model, which showed impressive reasoning capabilities. But I think it's wrong to look at the magnitude of the accomplishments and assume that will be field independent. We don't know the range of problems reasoning techniques are useful for. What we see here is refinement of capabilities that have been noticeable for years.

Maybe after decades. 2022 models were microscopic compared to latest models.

You should really look up a video about what GPTs fundamentally are.