The Economics of Work In An Age of AI

Beth Lyons (0:00)

Hey, everybody. It is September 10th. Welcome. So glad to see you folks.

It is, I don't know what episode it is. It is a, it ends in four or nine. So, it's either 804 It's 809 It's 809, here we go. Yes, because it's Thursday. It's episode 809, awesome. I am Beth Lyons. With me in the studio today is Andy Halliday and Gareth Hood. And you are watching and listening to The Daily AI Show. Some things happened yesterday. Apple came out, showed us their no-foldy phone.

Two very mixed reviews. I don't know if folks are excited about that. I personally am not really excited about it.

But the story that we talked about yesterday, in terms of Jacob working for OpenAI and then Jason, maybe, the man who, a safety man who worked for OpenAI and then worked for Anthropic for six weeks, and then resigned very publicly and shot a warning about lots of reactions and discussions for that.

Andy Halliday (1:25)

Wait, wait. So this is news to me that Jacob Coxon, the fellow who created this firestorm of publicity, by leaving Anthropic, right? You're saying he was only there for six weeks?

And he came from OpenAI?

Beth Lyons (1:42)

That is the-

Andy Halliday (1:44)

No, that just opens a whole other possibility.

Gareth Hood (1:47)

That's an employee thing.

Andy Halliday (1:49)

Subterfuge, right? So you've got this plant that moves over from OpenAI prior to the IPO, goes over to Anthropic, only for six weeks, and then loudly leaves to Sully, Anthropic. But also, I mean, he did it in a clever way because he said, everybody's not paying attention, including OpenAI.

Beth Lyons (2:17)

Yep. There are other interpretations, but yes, the story got messier yesterday. Also, there was a potentially coordinated Wall Street Journal op-ed, or like those same-day kind of thing. There were some things that seemed outside to be arranged, and there are also a bunch of AI safety researchers saying, no, we agree, right? Whatever we interpret this specific thing, and people coming out defending him. But that story just got more complicated, and all sorts of other things. All right.

Andy, what did you bring for us today?

Andy Halliday (3:02)

Well, I think because it's sort of tangential to that whole question of AI safety alignment and ethics, and the bigger question that, will AI kill us all, or will it just have a devastating effect on the existing economy? Well, Anthropic released a thing called the Economic Scenarios Explorer, and I'll share what that looks like in just a moment. But what it lets you do is it's a web page basically, and it lets you click through and make adjustments to an underlying very complex economic model that was founded or grounded in a working paper and a survey of about 11,000 US adults.

So it creates projections, and they've kind of prepackaged projections for the sort of, let's see, it's the modest scenario where there is growth in the overall GDP of AI. There's a substantial scenario, and there's the extreme scenario where AI drives a completely transformed, unprecedented economy, lightly driven by recursively self-improving AI systems, which, harkening back to the Jacob Coxon thing in an NBC interview, he said full recursive self-improvement, he thinks, is about a year away. But now we've got this one researcher who quit Anthropic, becoming a major commentator on where the labs are behind the scenes. But anyway, let me share my screen, just show you what this looks like. It's kind of cool. I only have had a minute to take a look at it. But let's see if you will be interested enough to go take a look at it yourself.

This is it. You can see that you can scroll through this thing, and here's the substantial scenario, and this is prognosticating out to 2030, which is just, frankly, four years away.

Beth Lyons (5:26)

Currently, 2030 is a little too close for comfort for some of these.

Andy Halliday (5:30)

That's right. It's very close. And here, in the extreme scenario, they're saying this could be the economic impact of AI all the way up here at just four years away. So that's kind of crazy.

And, of course, it is the extreme thing. But this is a really nice UI. And it's, I think, well worth your time to just go through and see what they built for us to play with. And you can make changes to some assumptions, but they've already sort of laid it out in a beautiful graphic way to kind of give you a better framework than just our talking points about the possible impacts economically and societally of the advent of AI.