#2311 - Jeremie & Edouard Harris

Joe Rogan (0:01)

Joe Rogan Podcast, check it out! The Joe Rogan Experience. Train by day, Joe Rogan Podcast by night, all day!

Jeremie Harris (0:12)

All right, so if there's a doomsday clock for AI, and we're fucked, what time is it? If midnight is, we're fucked.

Edouard Harris (0:21)

We're going to write into it.

SPEAKER_4 (0:22)

You're not even going to ask us what we had for breakfast?

Jeremie Harris (0:24)

No, no, no, no, no, no, no.

SPEAKER_4 (0:26)

Jesus, okay.

Jeremie Harris (0:27)

Let's get freaked out.

SPEAKER_4 (0:29)

Well, okay, so there's one, without speaking to the fucking doomsday dimension right off the bat, there's a question about where are we at in terms of AI capabilities right now, and what do those timelines look like? There's a bunch of disagreement. One of the most concrete pieces of evidence that we have recently came out of a lab, an AI kind of evaluation lab called METER, and they put together this test. Basically, it's like, you ask the question, pick a task that takes a certain amount of time, like an hour. It takes a human a certain amount of time. And then see how likely the best AI system is to solve for that task. Then try a longer task. See a 10-hour task. Can it do that one? And so right now, what they're finding is, when it comes to AI research itself, so basically automate the work of an AI researcher, you're hitting 50% success rates for these AI systems for tasks that take an hour long. And that is doubling every, right now, it's like every four months.

Edouard Harris (1:27)

So like you had tasks that you could do, you know, a person does in five minutes, like, you know, ordering an Uber Eats or like something that takes like 50 minutes, like maybe booking a flight or something like that. And it's a question of like, how much can these AI agents do, right? Like from five minutes to 50 minutes to 30 minutes. And in some of these spaces like research, software engineering, and it's getting further and further and further and doubling, it looks like every four months.

SPEAKER_4 (1:54)

So like, if you extrapolate that, you basically get to tasks that take a month to complete, like by 2027, tasks that take an AI researcher a month to complete, these systems will be completing with like a 50% success rate.

Edouard Harris (2:07)

So you'll be able to have an AI on your show and ask it what the doomsday clock is like by then.

Jeremie Harris (2:13)

It probably won't laugh.

Edouard Harris (2:17)

It'll have a terrible sense of humor about it.

SPEAKER_4 (2:19)

Just make sure you ask you what it had for breakfast before you starting.

Jeremie Harris (2:24)

What about quantum computing getting involved in AI?

SPEAKER_4 (2:27)

So yeah, honestly, I don't think it's, if you think that you're going to hit human level AI capabilities across the board, say 2027, 2028, which when you talk to some of these, the people in the labs themselves, that's the timelines they're looking at. They're not confident, they're not sure, but that seems pretty plausible. If that happens, really, there's no way we're going to have quantum computing that's going to be giving enough of a bump to these techniques. You're going to have standard classical computing. One way to think about this is that the data centers that are being built today are being thought of literally as the data centers that are going to house the artificial brain that powers super intelligence, human level AI when it's built in 2027, something like that.

Jeremie Harris (3:09)

So, how knowledgeable are you when it comes to quantum computing?

SPEAKER_4 (3:14)

So a little bit, like I did my grad studies in the foundations of quantum mechanics. Oh, great. Yeah. Well, it was a mistake, but I appreciate it for the purpose of the...

Jeremie Harris (3:25)

Why was it a mistake?

SPEAKER_4 (3:27)

So academia is this kind of funny thing. It's really bad culture. It teaches you some really terrible habits.

So basically, my entire life after academia and Ed's too, was unlearning these like terrible habits of... It's all zero sum basically. It's not like when you're working in startups. It's not like when you're working in tech where you build something and somebody else builds something that's complimentary and you can team up and just like make something amazing. It's always wars over who gets credit, who gets their name on the paper. Did you cite this fucking stupid paper from two years ago because the author has an ego and you got to be on... I was literally at one point, I'm not going to get any details here, but there was a collaboration that we ran with this, anyway, fairly well-known guy and my supervisor had me write the emails that he would send from his account so that he was seen as the guy who is interacting with this big wig. That kind of thing doesn't tend to happen in startups, at least not in the same way because everybody's-