David Eagleman

It's extraordinarily good at, for example, image recognition or categorization of things, but it can only tackle problems that are

discrete and rule-based.

So, for example, AI is great at chess and at Go.

It's beat the world champions at that.

But that's only because it's a constrained, rule-based system that doesn't have anything outside of it.

And the real world is nothing like that.

And so, by the way, even though

People often think, oh my gosh, AI can do anything.

It's taking over everything.

It can't even do any sort of strategy-based video game where you're running around with a gun and you're having to do strategies.

It can't do well at any of that stuff.

So that's the difference is that a live wire child can figure out all kinds of things in the world.

AI can only do these very basic things right now.

I mean, eventually, eventually we might.

Certainly not right now.

I mean, you can just, you know, turn the computer off.

I mean, there's a, yeah, it's still doing what it is told as in, hey, I want you to absorb a billion pictures of cows and horses and then get really good at being able to determine the difference between these.

So what it does is it trains on a training set of, let's say, a billion images where it's labeled, okay, this is a cow, this is a horse, this is a cow.

And then it's extraordinarily good, better than human at discriminating cows from horses.

But in real life, we don't have training sets with billions of examples.