Atmosphere TV is the largest streaming TV platform purpose-built for shared, social spaces like bars, restaurants, gyms, and airports. Atmosphere TV is the creator of ChiveTV and 30+ mesmerizing, audio-optional channels that entertain over 150 million monthly viewers.

Built for business, made for fun. Atmosphere TV. Learn more.

Marcus Johnson: Atmosphere TV is the largest streaming TV platform purpose-built for shared social spaces like bars, restaurants, gyms, and airports. Atmosphere TV is the creator of CHIVE TV and 30 plus mesmerizing audio optional channels that entertain over 150 million monthly viewers

Marcus Johnson: Hey, gang, it's Friday, September 18th. Yes, Grace Jacobell listeners, welcome. Uh, the only reason I know that is 'cause Danny just said it before we hit record. Yuri Jacobell listeners, welcome to Mind the Numbers, the marketing podcast made possible by Atmosphere TV. I'm Marcus, and joining me for today's conversation, we have, living in California, analyst Jacob Bourne.

Jacob Bourne: Thanks for having me today, Marcus.

Marcus Johnson: Of course, mate. Thank you for being here. And calling New Jersey home, principal analyst, Yuri Wanza.

Yory Wurmser: Hey, Marcus. How are you?

Marcus Johnson: Hey, fella. Very good. How are you?

Yory Wurmser: I'm doing great. Great. Yeah, happy to be here.

Marcus Johnson: Thanks for th- You're lucky. You're lucky, folks. I kept him on the episode.

Marcus Johnson: He's talking about Spain. Spanish football, not the country. The country's great. This football team infuriates me 'cause they're so good. They always beat us. Anyway, today's fact is where we begin.

Marcus Johnson: So which North American cities have grown the most in the last 20 years? So Visual Capitalist had a nice call with data from the US Census, StatCan, and the St. Louis Fed. They say... Any guesses, gents, on the city that's grown the most by percentage, not total people? This is a dark horse. I, I would never have guessed it.

Marcus Johnson: East Coast to give you a hint.

Yory Wurmser: Ooh.

Marcus Johnson: Not obvious at all.

Yory Wurmser: I was gonna go with Austin, but East Coast throws me.

Jacob Bourne: Yeah. I was not gonna go East Coast at all, so that really throws me for

Marcus Johnson: a loop. Go ahead

Jacob Bourne: Somewhere in North Carolina.

Yory Wurmser: Yeah.

Marcus Johnson: Oh my goodness. Yes. Genius. Charlotte.

Jacob Bourne: Okay. Wow.

Marcus Johnson: Well played. Sorry, Yuri, what did you say?

Yory Wurmser: No, I said genius. Like

Marcus Johnson: that is- Yes. Well played. Char- Charlotte, fastest growth, 93% grow- nearly doubling the highest percentage gain among major Canadian and US cities. Uh, western and southern metros like Phoenix and Orlando, Phoenix plus 33%, Orlando plus 53%, outpaced, uh, northern US counterparts such as New York plus 6%, just 6%, for the state, city, for the city, plus 6% in the last 20 years, and Chicago 0%, uh, underscoring a sou- uh, southward population shift.

Marcus Johnson: Uh, LA is, uh... it's only grown by 1%, believe it or not, Jacob. Uh, San Francisco by 12, that's where you are, close to where you are. San Diego by 14. Um, they didn't have... Well, we've got some stuff on New Jersey State in a second. Top three, Charlotte, Orlando, and Houston by percentage. What will happen in the next 25 years, the University of Virginia, Weldon Cooper Center for Public Service, and US Census expect, this is by state now, Idaho, Utah, and Texas, and Florida to grow their populations, uh, the most by between 30 to 40% each.

Marcus Johnson: The next 25 years, New York will lose 5%, that's the state. California 2%. And, uh, more people for you, Yuri, New Jersey will grow by five.

Yory Wurmser: I'll take it.

Marcus Johnson: They're probably gonna go to, um, uh... What do I keep? What's the town I always blank on, um... Where does Rob Rubin live?

Yory Wurmser: My town, Montclair.

Marcus Johnson: Montclair. Oh, you're Montclair too.

Marcus Johnson: I am

Yory Wurmser: Montclair. I

Marcus Johnson: bet they all go to Montclair. It's this paradise apparently. I can't wait to go.

Yory Wurmser: It's pretty nice.

Marcus Johnson: I'll see you soon, Yuri. Anyway, today's real topic. Yeah. Why OpenAI thinks AGI is here.

Marcus Johnson: Yes. We're entering the era, era of artificial general intelligence, writes Stephen Rosenbush of The Wall Street Journal. He notes that OpenAI recently proclaimed that the release of its GPT-6 Astra model may be regarded in the near future as the beginning of artificial general intelligence. We'll be referring to it, I'm sure, as AGI.

Marcus Johnson: Uh, he says, "The company is almost certainly correct in one sense. We're bound to look back on this period as the beginning of AGI, but in all probability, we've been in this technological stage for nearly a year, even though it was hard to notice given the relentless debate over the financial, economic, and social dimensions of AI."

Marcus Johnson: Uh, Jacob, I'll start with you. What would convince you that AGI had actually arrived?

Jacob Bourne: Well, I think there's a strong argument to be made that it's already been here, you know- Okay ... it's here and it's already been here. And it really depends on how you define it, which is the problem. How do you define it?

Jacob Bourne: It's, it's a very difficult to- Yeah ... you know, term to define. It's big. Um, I, I don't think we're currently in a world where, you know, AI can reliably perform everything that an average human can do across a broad range of tasks, even if AGI is here. And it's because it's not about the capabilities or the propensity for AI to hallucinate, it's that AI doesn't have real world, you know, real life experience that it then it can translate into under- it understanding the full breadth of human context.

Jacob Bourne: Mm-hmm. And so I think that that's where the failure part- point really still lies. I think that real world experience from robotics could help, but a, a recent, um, illustration of this is really, uh, Google's latest partnership with Accenture, where Google is training about 1,000 human engineers to go into various companies and help, you know, effectively integrate AI into the workflows.

Jacob Bourne: So what that tells me is that Google isn't really that confident that its own agents can't really do the work of going in and understanding the full context of work.

Marcus Johnson: Mm-hmm.

Jacob Bourne: So while there's plenty of room for AI to still improve, I think that that disconnect, uh, between human context, you know, AI understanding human context and, you know, the lived experience of the human being really to, um, being able to draw from that, you know, life experience and work experience and translate that into the, the, the workplace or whatever your task is, I think that that disconnect is gonna, you know, remain for some time.

Marcus Johnson: Yeah. Yeah, Yuri is- Yeah ... I love this full human context, uh, part of it, uh, Jacob's answer, and the, the word reliably because, um, Grace Harmon, who covers AI for us, we were talking to her a- about this topic before the episode. She was saying that AGI to her, if it had actually arrived, would be the ability to accurately nearly 100% of the time perform 80% of the intellectual tasks a person can do across fields with little, uh, supervision.

Marcus Johnson: However, she made the very important point that accurately and safely are different things.

Yory Wurmser: Yeah, and I, I, I think that- The, the ability to have this broad range of tasks is, is the key here. I mean, it's such an amorphous term, A- AGI, but if you think about, you know, the ability of ChatGPT, uh, Astra to, uh, solve the Millennial Prize, solve that, that, those math problems last week, um, that's just an incredible amount of intelligence in a fairly specific format, you know, with, with, uh, lots of inputs, lots of constraints given by the researchers.

Yory Wurmser: Um, still just an unbelievable achievement for them to do that. But at the same time, you know, if you asked, um, I would assume Astra, I haven't done it, but, you know, to solve relatively easy problems that, that we would know from experience, it's still not sure how good it is at planning and so forth. This is, you know, it's, it's gotten much-- You know, th- these programs have gotten extremely good at planning in many ways.

Yory Wurmser: Um, that's how they- you know, solved this problem.

Jacob Bourne: Mm-hmm.

Yory Wurmser: Um, and yet at the other hand, you know, sort of the whole scope of human experience, putting it all together into one, into one, one sort of coherent outlook and viewpoint. I'm not sure it's quite there yet.

Marcus Johnson: Mm-hmm. Yeah, broad

Yory Wurmser: range of tasks- It's getting close though in some ways

Yory Wurmser: So

Marcus Johnson: who's, who's that? Sorry.

Yory Wurmser: No, it's just me. I, it- Oh ... it's getting close in some ways though.

Marcus Johnson: Yeah.

Yory Wurmser: For sure.

Marcus Johnson: Uh, broad range of tasks is interesting 'cause OpenAI, so they've called their latest model, GPT-6 Astra, a generational leap in capability for areas like cybersecurity, professional work, software engineering, uh, science, and computer use.

Marcus Johnson: What can it do? In a video demonstration, Astra formatted a legal contract, built a 3D game whilst handling other tasks, including searching for food and booking a tennis court. Um, Jacob, why does OpenAI believe that GPT-6 Astra is a, quote-unquote, generational leap and could eventually be seen as the arrival of AGI?

Jacob Bourne: Yeah, I mean, I don't, I don't read this as belief necessarily. I think that, you know- Mm-hmm ... OpenAI obviously has, has a marketing incentive here to, to use this kind of AGI language. Yes. So I, I will- Especially this year given it's trying to IPO ... Yeah. And I would take this as, as a watershed moment for AGI.

Jacob Bourne: I'm not sure a watershed moment is ever going to happen because we don't know how to define it. So the situation is OpenAI has this new flagship model it wants to promote in this landscape of fierce competition between, uh, frontier, um, model developers, and also in this, this climate where investors are increasingly scrutinizing, um, AI companies.

Jacob Bourne: They're wanting to see when their re- return on investment is gonna come. Mm-hmm. So I think that OpenAI using this language helps, you know, get the right of it- right attention on its product. That said, I, I think that there is some substance behind the claim in the sense that with Astra, it's pairing increasingly powerful reasoning capabilities with agentic capabilities.

Jacob Bourne: So we're really moving further past the chatbot area, era into this more mature agentic era. So I think that that is the, the milestone to think about here. Um, really g- you know, this, this notion of this more fully autonomous general purpose system, whether you call it AGI or, AGI or not- Mm-hmm ... I think Astra is fairly, you know, it, it's fair to say it's an example of that.

Marcus Johnson: Yeah.

Jacob Bourne: On, on, on the terminology front too, though- Yeah ... I, I think OpenAI has also been part of this conversation, calling AGI a vague term a- and admitting that, you know, that, you know, different, different organizations, different viewpoints are going to, you know, see AGI differently. So I think- Yeah ... yeah, it, it, it's not necessarily, um, a, a moment in history.

Marcus Johnson: Yeah. No. And Yuri, it seems to be the answer is depends who you ask. Veteran tech entrepreneur Bill Nguyen, uh, who sold two companies to Apple, thinks that we began to enter the era of AGI months ago. The, the NVIDIA CEO, uh, Jensen Huang thinks that OpenAI has achieved AGI with, with Astra. Um, what say you?

Yory Wurmser: It's hard to say whe- when we actually will get there or have gotten there. Mm-hmm. I mean, it, Fable is- Has this model

Marcus Johnson: got here, do you think?

Yory Wurmser: Maybe. I mean, Fab- Fable, you know, Anthropic's Fable is better in some ways than- Yeah ... than, uh, Astra. It's worse in a lot of ways. Can't do the gener- you know, the image generation, the spatial, uh, reasoning that Astra has, but You know, it, it's very good at solving a lot of different problems.

Yory Wurmser: Maybe, you know, it, it really depends on how you define AGI, and I think we're, we're close. Um, we either have achieved it, you know, somewhere in this, uh, this timeframe from a few months ago to a few months from now. But we're- Mm ... we're getting pretty close to, to the point-

Marcus Johnson: Yeah ...

Yory Wurmser: where we can say that.

Marcus Johnson: Is it almost like saying who's the smartest person in the world?

Marcus Johnson: It's like, well, this person's the smartest at some things.

Jacob Bourne: It, it's a lot like that. I think that's a, a good way to put it, Marcus. I also wanna add that, you know, Astra was cha- trained on NVIDIA's chips, so NVIDIA certainly- Ah ... has some interest in, in calling it AGI.

Marcus Johnson: Yes. Yeah, definitely a horse in the race there.

Marcus Johnson: Um, Missy Nguyen, uh, though, was saying that Astra's most AGI obvious feature is the ability to combine computer control with its image and video generation ability. So he says, "You could imagine something and the model will find tools, other programs, to do the job." The breakthrough achieved by OpenAI's Astra and Anthropic's Fable 5.1, to what Yuri was saying, is that they can find their own resources.

Marcus Johnson: So going to find the tools to get the thing done. Even if we don't have ac- Yeah. Sorry, go on, please.

Jacob Bourne: Well, I just wanna point out, you know- Please ... human intelligence is really collective intelligence, and of course with AI, if you have agents working together, you're gonna get to AGI much faster. So.

Marcus Johnson: Yep.

Marcus Johnson: Yeah. Yep. Uh, Ina Fried of Axios cautioning that it remains to be seen just how well Astra can take on these highly advanced tasks in the real world without making critical errors or raising fresh safety concerns.

Yory Wurmser: These programs are still, you know, they're based on patterns. They still create pa- follow patterns, and, uh, it's kinda, it's a philosophical question of whether intelligence is just a really great pattern, uh, you know, recognition of patterns, in which case we're getting pretty close.

Yory Wurmser: But there's also the sense of, you know, this spark of origin- inspiration and originality.

Marcus Johnson: Mm-hmm.

Yory Wurmser: Um, you get a lot of incredibly interesting ideas out of these programs. Has it matched human creativity and originality? I think it's matched in some ways, but it's not there in many other ways yet.

Marcus Johnson: Yeah. You know- It's, yeah, it's a quest- uh, it's a question we've not...

Marcus Johnson: Sorry, J- Jacob.

Jacob Bourne: Well, I was gonna say, uh, my take is it's always gonna be different, right? Right. So ultimately comparing them i- is, is probably a flawed way of approaching it.

Marcus Johnson: Yeah. And, uh, Ina Fried of Axios pointing this out. AGI, she says AGI must, uh, well, basically saying AGI might just be a subjective measure.

Marcus Johnson: Greg Brockman, uh, co-founder and president of OpenAI, uh, saying he personally believes OpenAI h- uh, OpenAI has achieved AGI whilst leaving users to decide whether Astra meets that definition. Uh, Mr. Rosenbush from Journal saying, "There's no standard definition of AGI, and that makes it hard to assess when it has arrived."

Marcus Johnson: OpenAI president, Greg Brockman, who I mentioned, admitting, um, everyone has a different definition of AGI. He says, "It's gray. It's a gray and fuzzy thing." And when I looked to some of the definitions from some of the prominent voices in this space, Shane Legg, co-founder and chief AGI scientist at Google DeepMind, says, "The ability to solve general problems in a non-domain restricted way in the same sense that a human can."

Marcus Johnson: Uh, that's AGI for him. Sam Altman, OpenAI CEO, uh, "Highly autonomous systems that outperform humans at m- at most economically valuable work." A few more, Yann LeCun, uh, Meta's chief AI scientist, "Systems that can learn a huge variety of tasks, reason and plan, uh, and possess a world model which allows them to understand how the physical world works."

Marcus Johnson: And then finally, venture capital investor, uh, Vinod, uh, Khosla, uh, describes it as the point at which AI can perform 80% of the work involved, uh, in 80% of the world's economically valuable jobs. But gents, um, Yuri, I'll start with you for this one. If no one can agree on the definition, then is the term AGI even relevant or meaningful?

Yory Wurmser: Yeah, and I'm gonna just give my, uh, colleague Grace a, a hat tip here. I mean, it's really her idea that it doesn't ultimately matter that much to most people. I mean, it matters to investors- Hmm ... who are trying to sell the idea of these specific c- companies or buy into it, these specific companies. Um, it matters obviously to the companies themselves just for marketing purposes.

Yory Wurmser: Um, for the average user, what matters is what it can do rather than is it You know, how it does on these benchmark tests. Um, so I don't know how, how important it is to d- to decide whether we've reached it. What's certain is that in certain tasks, uh, these programs are now way exceed what humans can do, and that, and a lot of those tasks are very important economically.

Yory Wurmser: And I think that's ultimately what, what matters for us, you know, as people and as, uh, you know, analysts of what's gonna happen to these companies and to the economy.

Marcus Johnson: Yeah. Yeah, Jacob, I mean, Yann LeCun from Meta popularizing the term AMI, advanced machine intelligence. He doesn't like the term AGI. How relevant is AGI, in your opinion, if no one can agree what it means?

Jacob Bourne: Yeah, I mean, I think it's still something that it's worthwhile grappling with, maybe from a philosophical perspective. Mm. Ultimately, I don't think that human intelligence is really directly comparable to, to machine intelligence. And in terms of the, the term itself, I mean, in a lot of ways, AGI has ta- has taken a backseat to this term superintelligence, which is now even showing up in legal proposals.

Jacob Bourne: Um, I would say that sup- Mm-hmm ... the notion of superintelligence, um, has, comes with much more serious, uh, consequences. Um, and so especially if you consider, you know, recent events with OpenAI's agents, uh, going rogue during testing, I, I would say that some of the, um, you know, th- theoretical risks that we were talking about in years past have now become closer to reality.

Jacob Bourne: So I think that Ultimately defining what AGI is and what superintelligence are, you know, they, they could have some legal implications, so maybe in that respect they're important. But I think the more, more pressing question is really how do we effectively regulate increasingly powerful AI systems in a way to make them safe and useful, and how do we use the AI we already have in the most effective ways?

Jacob Bourne: And I think that those are really the most, the two mo- more important, uh, questions versus, you know, untangling these really s- messy definitions.

Marcus Johnson: Mm-hmm. Yeah, and, uh, Bernie Sanders, Senator Bernie Sanders, um, proposing a, a ban of artificial super- super- superintelligence, Artificial Superintelligence, uh, Act ban.

Marcus Johnson: Um, which superintelligence, just to, uh, for folks listening, it would come after AGI, correct? It's not a synonym for AGI.

Jacob Bourne: Yeah. I mean, I guess it depends on who you ask, Ian. Some people kind of fold the two together- Right ... and some people, you know, think about them separately. So-

Marcus Johnson: Yeah ...

Jacob Bourne: again, it, both are, are, are vague terms.

Marcus Johnson: Yeah. Yeah. Um, let's end with this, gents, and this is a q- a question that Grace had, which I thought was, um, was brilliant, and Yuri, it kind of speaks to what you, what you were just getting at. Um, so I'll start with you for this. What will ordinary Americans do with AGI? How necessary is it for us? What will it, uh, give us the ability to do if we, if we know that at this point?

Yory Wurmser: If it can help us do what we'd want to do anyway, um, help us with our tasks, um, then it's really relevant. It... The fact whether it's AGI or just really good at the task you want, I think matters less to the average person.

Jacob Bourne: Mm-hmm.

Yory Wurmser: If you can hire, you know, if you can go shopping and there is a shopping assistant that really responds to what you're interested in, finds things you wouldn't have thought of thinking about, and goes ahead and buys it reliably, um, you know, that might be appealing to a lot of people.

Yory Wurmser: Now, people still like shopping for, uh, the entertainment value, you know, as a, as a pastime. Um, but, you know, for tasks like that where- An agent, um, can replace what we do now much more, much e- easier, um, then it makes a difference. So I think for the consumers, that's where it makes a difference, if it helps them achieve what they wanna achieve, uh, much more easily.

Marcus Johnson: Mm-hmm.

Jacob Bourne: Well, I, I agree with, with what Yuri just said 100%. Um, I also think, though, that a lot of the efforts around developing AGI aren't necessarily d- directed towards consumers, uh, at least not all together. I think a lot of it is directed towards enterprise applications and research applications, including just more AI development itself.

Jacob Bourne: Um, but in terms of how people are gonna end up using it, I th- I think the sky is the limit, uh, as just as we've seen w- with other technologies, perhaps even more so. Um, as Yuri was talking about, the, the reliability is gonna be really important in terms of adoption. Um, but I don't, I don't, even with reliability, I don't see...

Jacob Bourne: I, I see this, I see the consumer adoption of agents, um, to be a, a gradual process. Mm-hmm. Um, and part of that is just pe- people like doing certain tasks. Um, people like con- control over certain tasks. Shopping is one good example. I think a lot of people enjoy shopping, so why would you want to, um, hand that off to an agent?

Jacob Bourne: Um, I also think that there's real, uh, backlash against A- AI right now. Right. Um, and of course, data centers are the focus of that, but it's not just data centers. It's, there's a whole host of concerns, um, in- including just human agency, right? So I, I think that While none, the backlash doesn't mean that this is not going to come to fruition in terms of people increasingly using, uh, agents to do tasks on the back end.

Jacob Bourne: I think it is gonna make the, the adoption more gradual.

Marcus Johnson: Yeah. Yeah, it's a real balancing act, isn't it? Because if you tell people that you've invented a car that can go, you know, 10,000 miles an hour, on the one hand you're like, "Great, that means I can commute in 10 seconds to w- you know, across the country."

Marcus Johnson: Um, on the other hand you're like, "Maybe that's too fast." Right. I don't actually need to go at that speed. Yeah. So yeah, it, it, the how this gets positioned, um, is not, uh, always a positive to, to rave about, especially with AI, how, uh, advanced your models are, um, especially as there's kind of growing distrust, uh, among some people around what AI means for society.

Marcus Johnson: Before we go, our takeaways for today, I think number one, even if AGI is good at quickly getting things done accurately, uh, Grace Harmon noting that accurately and safely are very different things. And number two, ask 100 people what AGI means, you'll end up with 101, if you're lucky, slightly to very different answers.

Marcus Johnson: AGI might instead be seen as more of a philosophical question or debate, uh, this from Jacob. And third, uh, most Americans won't care about the AGI race, uh, they just want to know, uh, if it can make their lives, or how, uh, it can make their lives better. Um, that's all we got time for, gents, for this episode.

Marcus Johnson: Thank you so much to my guests for hanging out with me today. Thank you first to Jacob.

Jacob Bourne: Thanks so much, Marcus.

Marcus Johnson: Yes, indeed. And to Yuri.

Yory Wurmser: Thanks, Marcus.

Marcus Johnson: Thank you for being here. And thank you so much to the production crew, uh, Luigi and Danny helping us out with this one. And thanks to everyone for listening to Mind the Machine Market Podcast, made possible by Atmosphere TV.

Marcus Johnson: Subscribe, follow, rate, review. We'll be back on Monday. Until then, have the happiest of weekends.