Dr. Peter Lebedev

I'm not the famous one.

Yeah, thank you so much for having me.

But you will one day.

We're working on it.

We're working on it.

On AI.

Right, so I was on here five or so years ago, which is incredible how fast time flies.

I finished my physics degree, so properly a doctor, which is really exciting.

Everyone is a doctor on this show.

It's really exciting to have that PhD.

And then I worked as a documentary filmmaker for this YouTube channel called Veritasium, and we made a lot of stuff that was really cool and got a lot of attention, mostly about math and physics and all of that.

And then I got more and more and more concerned about how AI systems have been developed

developed over the past three, five years, something like that.

Well, I read a bunch of books about AI a decade ago, and they had all these very hypothetical sounding fears, right?

And then one by one, those predictions kind of started coming true.

And the progress in AI for the last three and a half years since the release of ChatGPT, but honestly for the last 15 years has been remarkable.

And the companies that are developing these systems are honestly not doing that good of a job of making sure that these systems are safe and that the good parts of AI, because there are good ways of using this, they're not really coming to fruition, right?

So yeah, I have been obsessed with that for the past couple of years.

Right.

So last year, Anthropic, which is one of the companies that are making an AI called Claude, they put Claude into the simulated environment and they were like, hey, at 5 p.m.