It's not changing the world economically the way that these folks said. I think if we really did have AGI, here are some things you might expect. You might expect wages to drop. You might expect unemployment to go up. You might expect that we could rely on agents. Like, you know, one guy proposed on Twitter today the pizza test. You should be able to ask any AI system, order me a pizza, and it should figure out, you know, where you live, how to order the pizza, and who's open and what. And like, that's not that hard. But, you know, anybody who's played around with these agents know you can't really trust that.
Welcome to Prof G Markets. I'm Ed Elson. It is September 8th. Let's check in on yesterday's market vitals. US markets were closed for Labor Day, but stock futures fell as the US and Iran escalated their attacks. Brent crude climbed towards $98 per barrel. Copper hit an all-time high on the prospect that Trump will expand his tariffs on the metal. And finally, the Japanese yen rose to its highest level since February as the dollar fell.
Okay, what else is happening? OpenAI's newest and most powerful model is officially here. GPT-6 Astra released on Friday. Astra was trained on more than 100,000 Nvidia GPUs at the Stargate data center in Texas. That makes it the largest training run in OpenAI's history, and it costs two and a half times the cost of OpenAI's previous model. Astra represents a large advance in capabilities for OpenAI. And across the AI industry, leaders have been celebrating. In response to the model, Nvidia CEO Jensen Huang said that artificial general intelligence, or AGI, has, quote, arrived. Meanwhile, OpenAI's president Greg Brockman said, quote, "Welcome to the AGI era," describing Astra as a generational leap. Now, if Astra is indeed AGI, then that would be a very big deal, as many AI researchers have specifically cited AGI as the milestone that might change society forever. Sam Altman, for example, once said that AGI could, quote, break capitalism. Meanwhile, Stuart Russell, a prominent AI scientist, has said that when we achieve AGI, human wages will, quote, go to zero. Well, AGI has supposedly arrived, but capitalism and wages remain intact, which would imply one of two things: either the industry's predictions about AGI were wildly wrong, or this isn't really AGI. Here to give us his view, we are speaking with Gary Marcus, author, scientist, entrepreneur, and emeritus professor of psychology and neuroscience at NYU. Gary, thank you so much for joining us on the show. I'd like to just start with your reactions to the model itself, Astra. What do you make of it? Can you describe some of its capabilities, and then we will get into whether or not it is artificial general intelligence, AGI?
I mean, in terms of the model itself, it's a bit better than the models before. What we have now is an era where some people call it benchmark maxing or benchmaxing, where these systems are often, it would appear, trained on lots of benchmarks. They do really well on the benchmarks, and then they don't typically do as well in the real world. So every time we see the same phenomenon, a model comes out, everybody's super excited about it, and then after a few days people are like, yeah, but I tried it and it doesn't really work on this, it doesn't really work on that. It deleted my files. The code it writes is for... something I saw about Astra yesterday on Twitter. You know, some people will like it, some people won't. But every model is hyped as if it's this quantum leap forward, and it never turns out to be true. You know, each new model is better than the last one. There's no doubt about that. Astra is better than Fable, its competitor, on a bunch of measures. In real-world practice, it's better on some and not others, you know, based on people's different accounts. It used to be when a new model came out, I would personally test a lot of things about it, but now the scope of what people want to do with them is so vast, no one person can do that. And you sort of have to see what the world's experience with it is. And what the world's experience has been with all prior models is they don't live up to expectations. We've talked about this before. You know, the return on investment isn't there. Certainly, wages haven't dropped to zero. Instead, we have what I call the Klarna effect, which is Klarna said they were going to get rid of a bunch of humans and then they quietly walked that back and started rehiring humans. We've now seen that dozens of times. We may see a version of the Klarna effect with Astra, where a bunch of people say, "We're going to lay off a bunch of people and then quietly rehire them." So going back to your opening question, right? If this were really AGI, well, let's put an asterisk around that. What do you mean by AGI? We'll come back to that in a second. Then yeah, I guess you would expect wages to drop to zero. I mean, assuming that the AI itself were cheap enough, and we could argue about that. The thing is that there's what I would call the AGI bait and switch, right? There's a way that the term was originally defined, which was basically that you could do any cognitive work that a human could do. And that would indeed, assuming the price was right, be pretty valuable. I mean, even if it cost just as much as humans, maybe it would still be valuable because you could make them work all the time with less requirements and so forth. So bait and switch, and then people will panic about that in various ways, like we're going to lose all our jobs or they're going to take over humanity or whatever. All of these are premised on, let's call it, a generous version of AGI that really could do everything that people do. And then there's a less generous version of AGI. Like the least generous version is something like it can do most of the things that the drunk guy at the bar can do, and he's human. So let's just call that AGI, right? So you have a sort of lower bound and an upper bound. You know, maybe that's being a little bit crude about it, but you know, there's the kind of like it's verging on superintelligence. Oh my god, there's nothing I as an individual human being can do anymore that AI can't. And there's the version where it can do a bunch of the things that I can do, but it can't really do all of them. And it does a few things better than me, but I do a few better. And you know, I can't really trust it, but I just want to, for marketing purposes, call it AGI. That's what we have now, right? Is the marketing version of AGI. You know, there's a historical literature, and you can think about these things in different ways. There's a historical literature about what AGI meant when people like Ben Goertzel and Peter Voss, you know, first coined the term, and Shane Legg, who is a co-founder of DeepMind. And then we're talking about the strong version that I was talking about, that you know, that it could do anything. And then you have different people have tried to weaken it in different ways. People accuse me of goalpost shifting. And I'm always like, which goalpost did I shift? And they always leave the conversation that can't actually document the claims, even though I write all the time. But I can document how this goalpost has shifted, right? Because I can go and show you where people talked about it originally, or I can give you my own goalpost. For example, I said it should be able to watch a movie and understand what's going on. Nobody's ever showed me that like Astra can watch a movie and tell me what's going on as well as a high school student, right? That was my, you know, one of my target examples in 2014. It's 12 years later. We can't do that. I haven't shifted the goalpost. So, you know, one way to think about it is historically, has it met those historical things? No. Another way to think about it is how has the history changed? Well, after an original period all the way up until 2023, when people pretty much agreed that it meant like doing whatever people could do, in 2024 is about when I see it starting to happen. People started talking about it economically. So I know Vinod Khosla was interested in this. He was, I guess, on the board of OpenAI, or at least he was invested in OpenAI, and he, I think, he pushed for a definition which is like can do 80% of the work that humans can do. We haven't met that one either, right? You know, I mean, it can't do, you know, it can do maybe a bunch of tasks that people can do, but doing a whole job is harder, right? Eric Brynjolfsson has always made the distinction between doing tasks and doing jobs, right? There are lots of tasks that current AI can do. I don't doubt that. And maybe some of them even reliably, you know, certainly can do. It's not really a task, but it can write, you know, rhymes or whatever, make a song about such and such. It's not, that's not really somebody's job, right? You could do that, but it's not a job, right? A job typically involves a lot of things. Well, then you could say a different way to think about this is like, what is it doing for you? You know, if you have a term, that term does something for you, let's say scientifically, right? So you have a term like mass that's in the context of a theory, and now you can, you know, make some calculations relative to it. So, you know, you could have a theory about economic change. You just gave one at the beginning, right? The theory about economic change is that wages will go to zero once you have AGI. Well, if you want to call what we have now AGI, then that theory is wrong because really what you've done is you've bastardized the definition such that it doesn't mean anything anymore. What AGI now means is like the thing that they want to sell you on this week. And then the last thing I'll say about this is it's not just this week, but many weeks. So I started compiling a list today. I didn't finish. But of all the times that people have declared that AGI has arrived, going back to, I think, 2023, and like Tyler Cowen said, GPT-3, a model that nobody would take seriously anymore, was AGI. And Sam Altman said we know how to build AGI at the beginning of 2025. Jensen, everybody went nuts when he said that Astra was AGI, but he said on Lex Fridman six months earlier that AGI had arrived. So like, which is it, Jensen? Right? Like, so you can make these announcements. The cost of making these announcements is not just zero. It's that it drives up the stock price, but there's nothing, you know, intellectually coherent that stems from it the way that people are talking about it right now.
Yes. So you describe it as a bait and switch, which does sound quite apt because this is something that a lot of these people were emphasizing for years as an extremely important term, an extremely important definition that would be a moment that would change our society, change humanity forever. Sam Altman being one of them. However, we just heard him on a podcast with Alex Heath, and Alex asked him about AGI, asked him about what that means and how important it is, and his response was quite different. Here is what he said.
How are you feeling about AGI these days?
I mean, at best you could say it's a very poorly defined term. I was going to say it's like an irrelevant marketing term.
Well, last I checked, your charter defines it as a highly autonomous system that outperforms humans at most economically valuable work. I think there are many people that would look at current models and say, like, "Okay, it's there."
Um, do you think it's there?
Sort of. Close at least. I like...
I've heard varying versions of like what people on your team think.
I think it's... I think there are a lot of people who would look at our latest internal models and say this is like very AGI-like.
Um, I think there are people who would say, you know, here's something I can point to that it doesn't do or it's really bad at, and it's not.
Just listen to all the kind of wildness there. You know, first he says it can't be defined. Then he says maybe it's met it. Then he uses a term like AGI-like. He points to, you know, some people in his company say it. Like, you know, he never comes out and says, I think it's this and it is met this for this reason. He just sort of, you know, dances all over the place. You might have to listen to the clip twice to really hear that in it. But that is what he's doing. He's avoiding any commitment around it. He's pushing it off to other people, and ultimately he's taking this term AGI, which means nothing, right, and he's made up this new term, what is AGI-like. I mean, you know, that there's no boundary condition for that. I mean, like, you could say that a magic eight ball is a... or what is it, the six magic... eight ball, magic eight ball. You know, you shake it and it gives you answers, and sometimes those answers are helpful and sometimes they're not. Is that AGI-like? I mean, like, there's no there there. And again, it's only interesting if it has some consequence in the world, right? Either it fits in with an intellectual debate, then you have to stick with the terms of intellectual debate, or you make a prediction about economics that follows from there. None of that's happening. And again, it's not changing the world economically the way that these folks said. I think if we really did have AGI, here are some things you might expect. You might expect wages to drop. You might expect unemployment to go up. You might expect that we could rely on agents. Like, you know, one guy proposed on Twitter today the pizza test. You should be able to ask any AI system, order me a pizza, and it should figure out, you know, where you live, how to order the pizza, and who's open and whatever. And like, that's not that hard, but you know, anybody who's played around with these agents know you can't really trust that. I mean, even shopping has been hard for them. You know, they don't know which menu to click on and things like that. And so all of these systems asking for agents, then we do things that you don't need that, you know, you don't need a PhD to do those kinds of things. Those systems are not reliable. They're AGI-like if what that means is some of the time they do it and you're amazed that it does it, but they're not AGI itself if it can't do them reliably, you know, in the way that a human could. I mean, that was the definition all along, is you have to be able to do it, you know, like an intelligent human can. And they just don't.
Yeah, it seems that if they're using it as a marketing tactic to use this poorly defined term that can mean basically anything, and perhaps it could mean the entire world changing, and at that point maybe you start thinking about how incredible the AI future is going to be, etc., etc. They use this marketing tactic in expectation, as you said, that that is going to drive up the value of their companies. If we say that we have achieved AGI, then suddenly everyone's going to use our models, and suddenly we're all going to become rich. Is there a point at which people no longer buy it? Is there a point at which people hear the term AGI, they hear the term artificial general intelligence? The new one is recursive self-improvement that I hear a lot. And is there a point at which people say, "I don't think that that's interesting. I don't think that's meaningful, and therefore I'm not interested."
I mean, I guess yes and no. There's different populations of people, and different populations of people have different motivations. The average person, I think, in the United States is actually kind of fed up. The average person in the United States is kind of like, this is not actually doing that much for me, and I don't want a data center in my neighborhood, and you know, forget it. Right? So the views of the average people have changed. There is, however, a class of people that make investments, and I think a lot of them like these stories. And there's this whole meme stock thing that's been going on for a while now, right? This is not the beginning of it. And you can think of things like GameStop. I think social media has driven them. And in the meme stock world, these kind of pronouncements often drive things up. I mean, and so like every time that Elon has said that driverless cars are imminent, and he said it like 15 times in the last, or probably more than that, but he said it annually pretty much every year since 2014. Every time he says that, it drives the stock up. I mean, you could ask what is the stock market such that it still responds when he says those things, given that he doesn't really have credibility around them anymore. And I, you know, so there's some population of people out there. I don't know if they're trading on the notion that the cars are actually imminent or the notion that other people will believe that, so I should buy the stock, you know. So it's this, the stock market's become almost this meta thing. What might other people believe? You and I, I think, talked about SpaceX a little, and I loved your riffs on SpaceX on Twitter. There's no way that it's worth what it's valued right now, if you look at it in financial terms, but you could make sense of it in terms of, well, I think other suckers will buy it at this price, so I'll buy it in order to flip it, then maybe it makes sense.
But if other people believe that it's AGI, then perhaps it really is.
They don't even have to believe it's AGI. You could just say, "Well, I think that other people will believe this crap even if I don't, and so that'll drive the stock up." So blah blah blah. And like now you've lost any financial footing. And I think the problem with that is eventually it falls apart, and then, you know, somebody gets left holding the bag. And clearly the goal here is to have retail investors hold the bag.
All right. Gary Marcus is author, scientist, entrepreneur, and emeritus professor of psychology and neuroscience at NYU. Gary, appreciate your time.
Thank you. Pleasure to be here.
We'll be right back. And if you're enjoying the show, be sure to subscribe to the Prof G Markets YouTube channel at the link below.
Support for the show comes from Vanta. When you run security for a company that's scaling fast, the stakes just keep climbing. More compliance frameworks, more vendors, more risk, and a board that wants to see it all in one place. But your compliance data can be all over the place. Controls in one tool, vendor risk in another, customer commitments buried in contracts. Your team has to spend more time stitching it all together and less time running the program. And every quarter, you're basing decisions on last quarter's data. Vanta connects it all. Its market-leading agentic trust platform is built for enterprise scale with 400-plus integrations and continuous monitoring. Vanta automates evidence collection, surfaces the risks that matter, and organizes your audit around your auditor's requirements. And that Vanta agent works around the clock with full program context, so nothing slips through the cracks. The numbers back this up. Vanta delivers a 526% return on investment over three years and pays for itself in three months. Vanta is trusted by more than 16,000 companies, including Snowflake, Atlassian, and Ramp. Learn more at vanta.com/markets. That's spelled vanta.com/markets.
Support for the show comes from DeleteMe. Nowadays, it's so easy for scammers, stalkers, and hackers to try and harvest your data. But it's important to remember there are always ways to protect yourself. That starts with protecting your data. And DeleteMe can help you do just that. DeleteMe does all the hard work to wipe your personal information from data brokers' websites. It was named the top pick for data removal services by Wirecutter. Our producer, Jennifer Sanchez, has tried DeleteMe, and she had very nice things to say about it. Also, I've used it. I enlisted. I didn't even know they were a sponsor, and I decided I should probably create some digital hygiene across the ecosystem. Take control of your data and keep your private life private by signing up for DeleteMe now. At a special discount for our listeners, get 20% off your DeleteMe plan when you go to joindeleteme.com/profg and use promo code profg at checkout. The only way to get 20% off is to go to joindeleteme.com/profg and enter code profg at checkout. That's joindeleteme.com/profg, code profg.
We're back with Prof G Markets. The August jobs report came in, and it was much stronger than expected. The economy added 162,000 jobs, nearly three times economists' expectations, and the unemployment rate held steady at 4.1%. The gains were mainly concentrated in food services and drinking places. Yes, that is actually a technical term, and local government education. Still, stocks slipped, and the two-year Treasury yield jumped after the report as a stronger labor market gives the Fed cover to raise rates. Odds of a hike this month are now at 52% on Kalshi. So, joining us to discuss the state of the labor market, we are speaking with Kathryn Anne Edwards, labor economist and host of the Optimist Economy podcast. Kathryn, thank you very much for joining us. I want to jump right into this jobs report. 162,000 jobs added in August. The unemployment rate stayed steady at 4.1%. I saw an interesting analysis of this report that was on CNBC from Wendy Edelberg, who is a Brookings economist. She was not very convinced by these numbers. She was expressing some doubts about them. I want to play you this clip, and then let's jump into it.
My guess is that these numbers get revised down. I am very puzzled, actually, by the strength in the payroll numbers. So at 71,000 on average for the last three months, that's like seven times faster than the pace of job growth last year. I see nothing else in the labor market that suggests that kind of pick up in activity. In other words, too strong to be true, therefore must be revised down.
