Joe Lonsdale

You know, a lot of my smartest friends who run the AI labs still, when I talk to them the last few months, it's like, oh, yeah, it won't be until the mid-2030s that, like, we're not having people actually help on some of these things.

But what does that even mean at that point, you know?

So it's, and at what point is, like, is, like, scientific discovery pushed ahead by the computers?

I think we're just starting to get what look like scientific breakthroughs from computers, like, but they work really closely with people.

Like, at what point, is there a point where the breakthroughs are happening with the people not even involved?

Holy shit.

Right?

And none of us know the answer.

I'd say, like, when you talk to the smartest guys, like Dario, who runs Anthropic, like, as of a few months ago, he thought that, like, the next few years of this really fast advancement is kind of locked in.

Like, we used to have something called Moore's Law with chips, right, where the chips get better, double every year and a half.

And this is a lot to predict how the chips kept getting faster and faster.

And the guys running the chips companies, Intel, all the other ones, they'd have all sorts of new tactics and new technologies they were building that would let you make it twice as small, twice as fast over the next couple of years.

And they'd kind of always be able to see ahead

maybe two years, maybe three years.

And it's always like, wow, how are you going to make it smaller?

Like, well, we have to solve this and this and this problem.

And they keep going.

And they did it for a very long time.

And so similarly with Dario, with this advanced pace where it keeps getting better with new data, new techniques, new tactics, he could see like very confidently the next couple of years and probably a third year of how it keeps getting better, which is pretty amazing because the pace right now is like doubling every few months, right, in terms of what the agent is.

It's crazy.