Sean Carroll

Right.

It's very, very well formed, clearly articulated problem spaces that are too big for humans compared to computers.

So it's just a matter of time before human beings became outclassed at that by the computers becoming more clever.

So I don't know how far that will go.

But I think that it's – this feeds back into Rob's question also.

I think it's important to recognize that we – one of the huge mistakes that people make and people make it at different levels.

So I'm sure that you, the person listening right now, doesn't make this mistake.

But many people still today insist on thinking of intelligence as a single one-dimensional thing.

I was having this conversation with Alison Gopnik, former Mindscape guest, a little while ago when we were both in Santa Fe.

We were talking about how frustrating it can be to get through to people who have an overly simplistic view of intelligence that leads them to think that, you know, any day now, AI is going to be just as good as PhDs or whatever.

The reality is that they're going to be just as good as PhDs at some things and not others.

And that's just not that sophisticated a thing to understand.

I think it should be very understandable.

What we kind of hit on was what we should emphasize is the fact that intelligence is not one-dimensional.

It is a nonlinear, high-dimensional, spiky, weird kind of space.

If you think about the space of, number one, problems you can solve, and number two, methodologies that you can use to solve those problems.

And LLMs do very, very impressive things, not in the same way as human beings.

And therefore, it is completely unsurprising to me that they're going to be really, really good at some things and

totally less good at other things.

I actually, like I'm pro AI in the sense that I think that this can be a wonderful tool to use.