Stephen Wolfram says that AI cannot predict future system behaviours because many systems perform complex computations that exceed AI's capabilities.

source

It did in fact confirmed my hypothesis that ability to speak does not necessarily mean higher intelligence. I've had cats that were smarter than some humans I've met that were saying seemingly sophisticated things.

Computational irreducibility is his thing, and I think he's absolutely right. It's going to take more than a simplistic LLM, no matter how big it is, to change that, if it's even possible at all.

@DanielleFong Future states of complex systems are inscrutable. It's kinda one of the things that makes the universe a fascinating place to live in! 😊

I’ll take his word for it. However, AI can probably help generate alpha without crossing the relatively high hurdle of “prediction”. (See RenTech, for example)

True. But that's true of us as well.

Is he just saying all science is NP? that's for sure wrong. Probability is always telling you what's likely based on what it knows. I think he's totally wrong in saying we didn't realize language was this simple. He's never read Shannon's 1948 paper or he forgot it.

an Ai told me he was gonna say that !

Man we need him

Do we need an Oracle instead of AGI? 👀