Balaji Srinivasan

It's mathematically constrained because it can't solve chaotic, turbulent, or cryptographic equations.

It's practically constrained because you have to prompt and verify it.

It does it middle to middle rather than end to end.

And it's physically constrained because it requires humans to sense context and type that in via prompts rather than gathering that all for itself.

And this is very different

than the people who look at AI as AGI, Deus Ex Machina, that's just going to solve everything, right?

To be clear, this is as of this time of writing, maybe somebody overcomes it, but at least I'm describing what the constraints are on the current generation, right?

You could, in theory, unify the probabilistic system one impressionistic thinking of AI with the logical system two thinking that computers are good at, and cloud code and so on is starting to get there, perhaps.

But

I think it's still something where context breaks down and so on.

And also, somebody, you know that famous graph that people show that it can do longer and longer problems?

The meter graph, yeah.

Yes.

That particular graph, I saw a very good – once in a while out of 1,000 replies on X, there's somebody who's actually gem award, right?

How do they put it, right?

Gem alarm.

Okay.

Gem alarm.

Yeah, exactly, right?

Fine.