Dr. Peter Lebedev

If it was smart enough to pick up all of the items from the store, pick the stuff up off the shelves and put it in this basket and then come to you to wherever you parked your car, I'd be like, yeah, this definitely feels like AI at that point.

But if it's deterministic, I'm like, I don't know what this feels like.

This feels like marketing hype, you know.

Oh, man, I hate that.

I really, really don't like that.

Yeah.

Oh, I mean, I think this is a really, really lovely question.

I love the, when you look down at a river from an airplane and you see all the beautiful meandering shapes.

Yeah, it's just really, I mean, we talk a lot about AI stuff right now and I'm like, I really, the thing that I've been really trying to do is like ground myself and focus on like the beautiful parts of the world and yeah, appreciating rivers.

Yeah.

It's a good time to do it.

This is a really good question, and I really, really love this.

Thank you, Dr. Maui.

The first thing for me that I want to say is that the rate of progress in AI just on our normal, ordinary computers

has been so staggeringly high.

If, uh, there, there are a bunch of ways you can measure like the autonomy of systems or the intelligence of systems, but it's in, in, uh, along a bunch of metrics, it has been doubling every seven, seven months in terms of capabilities.

We are now at the point where AI systems can almost fully autonomously hack into governments.

This has happened in January of this year where, um,

Claude Code using Opus 4.5, which is a commercially available product that is available to you and me, was used to hack into the Mexican government and like three state governments and steal a bunch of really, really sensitive data.

Yeah, the rate of progress for AI with normal computers is very, very high right now.