Balaji Srinivasan

Simply not, right?

Even like, you know, a pendulum, you know, with a few different weights on it, there's different ways you can quickly get into chaotic or, you know, that kind of behavior.

So empirically right now, at least, and you can argue this point, whether AI is centralizing or decentralizing, but I think there's so many AI companies and there's so many AI models out there that it's pretty hard to put a lid on the thing.

And then finally, if you go to the end, the optimal amount of AI is not 100%.

If you scroll down a little bit further, right?

0% AI is low, but 100% AI is slop.

Just having that concept in mind means that in almost any process,

You do not want to have 100% AI, but you probably, and frankly, sometimes you want 0% AI.

You often want to just learn offline with pencil and paper and then speed up with AI because AI is a shortcut.

And like any shortcut, it can be overused to the point that you don't know how to take the long cut, right?

Okay, this is a laugher curve, but for AI.

And then, you know, I talk about how referential is this?

Whoa, we're referencing Torenberg on Torenberg.

Yeah.

A16Z podcast, that's right.

This is recursive, right?

And fundamentally, my worldview on AI is it's constrained AI.

It's economically constrained because every API call is expensive and now it's energy constrained, right?

It's

There's so many competing models.