Beth Lyons

And we've talked about the levels of AGI.

Carl, I think you have talked about this sometimes, where there were like five levels, kind of like autonomous driving.

If it can do this, it's at this level, right?

There was a corresponding... It was like Nate Jones who said that.

Or, as we might call it now, a codex.

Because that is almost fully... You still have to talk to it, I think.

Oh, okay.

Yeah.

So maybe that's level five.

But there was a corresponding thing that I don't remember who put it out, but it was a research paper on can smaller models be...

safety patrol or that sort of thing for a super intelligent larger model.

And that was a couple of years ago, maybe a year ago, because what I remember thinking of it is, is that it's like the, the trainers of elephants who get them when they're young.

Right.

And, and so that relationship is established when it's more equal and,

And then as the intelligence grows or the size of the animal grows, that pattern is still established and it doesn't break out necessarily because that's a thing.

And I'm curious about that for this.

Does that start to come into play again?

Because the idea of superintelligence has been around the whole time.

It just seemed like AGI was going to be the more, I don't know, more beneficial, more across the board.

Maybe it was just the original idea when people got together.