Reply

Avatar

or to participate

Avatar

Tom Greenhaw1 month ago

Math is a language and we use it to describe reality and we call it physics. Math superintelligence doesn’t imply physics intelligence, it simply describes it. While math is an important tool for the scientific method, it is the observation of reality that define the terms. LLMs aren’t quite there yet, but with world models and the investment in robotics it will come quickly.

Avatar

Sean Flaim1 month ago

I think my answer is between yes and no. If we continue with AI as currently designed and iterated on, I think we're going to end up with systems that are incredibly sophisticated at pulling patterns out of huge sets of data. And that will both work well with math as well as a number of physical sciences. It may even exceed the ability of humans to pick out patterns, which is kind of the way we operate to begin with. But picking out patterns and the intuition that creates new ones are slightly different things.

True creativity will take designing AIs toward a different model, one I write about here: https://belowtheconversation.substack.com/p/the-hard-problem-has-a-solution

And one bonus: I actually wrote that post in response to something Adam wrote many months ago for Noema. I have something else in the works that I hope to get out soon which will dive into AI even more.