Robot-mancers

Hello and welcome to The Attention Mechanism. I'm Justin Robert Young, joined as always by the one and only Andrew Mayne. How you doing, buddy? Fantastic. Lot to go over. uh we we you know every once in a while we miss some time but but trust us friends it's because we're doing stuff that you can also enjoy and like uh we definitely do this as as often as we can because

We could probably, in another world where we were less busy, make this a twice daily show considering how fast the news comes down the pike. This is just the stuff that's happened in the last... 12 hours, NVIDIA announcing a $100 billion announcement into OpenAI for several gigawatts worth of chips. What do you think this presages? Are any of these numbers surprising to you? What was your initial reaction when you saw the announcement?

Being an NVIDIA shareholder aside. Hell yeah, brother. And, you know, the former OpenAI employee who may have equity. Obviously thrilled to see this happen. I think that, listen, I think that this is beyond NVIDIA, beyond OpenAI. I mean, it shows that there is this belief that's not just Sam Altman. you know that we need a lot more compute we need to go into this future

that we're starting to see really good market fit for these things. You know, NVIDIA is an extremely valuable company. It's worth $4 trillion, although $100 billion is, even for a company like NVIDIA, is a big commitment. But again, it's a commitment over time. It's tied to compute and other things like this, so it's not like they just gave Sam Altman a steel briefcase filled with like $100 billion, which would have been my terms, but there's a reason Sam Altman is who he is.

I said this before that people get caught up on who is going to win. And it's like 1902 and like, who's going to win electricity? Like, well, we're eventually going to create utilities and we're going to produce it. But, you know, 100 years later, who's making the most money using electricity? I don't know. Microsoft? What's that? Well, that's a longer conversation. So I think that gets into the, you know, Alexander Graham Bell.

when he looked at the phone, he thought this might be good for delivering music. He wasn't thinking about the iPhone, you know, and the social implications of where this would go once you made these things wireless and attached them to computers, et cetera. So. What do I think about this is very, very, very smart money is betting that we are going to be moving a lot of work to AI. AI is going to be generating a lot of revenue, which means what? Well...

It's going to be doing a lot of current work. We know this is happening. I am, you know, you and I are both, we're long-term very positive about the role of human work. We think that every time there's a major upheaval, at the short term, you go like, what happened to my typewriter repair job, you know? Well...

There's a whole new world of using computers and other stuff. And, you know, we've talked about before about the typewriter. Was this a revolution? Because it allowed people who were from, you know... could not get into what are considered higher class jobs like clerical work or working in offices, all of a sudden they can do that. And so we're seeing this transformation.

So what we know on one hand is, yes, we're seeing that there's people betting that the future is going to involve way more AI than where we're using now. And that's where we're headed. To the people listening to here, what does that mean for you?

Learn to use AI. Learn to use these tools. It's like the 80s. Should I learn how to use a computer to do my job? Now it's should I learn how to use chat GPT? And then I still have these questions. Should I learn the internet in the 90s? And now should I learn? I'm like.

I'm having this conversation. It could be, it could be Jim and I chat. Let me make that very clear. I'm not trying to push, you know, one product over another. No. Yeah. But I mean, like obviously the tools are the tools. So. When you see numbers like this, or at least when I see numbers like this, it does feel different than technology build-outs in my lifetime. specifically in that I was not old enough to really kind of understand.

the build outs of the nascent internet and the World Wide Web and everything. And also, we didn't have a technological or business press that covered it in the kind of level that we're going to cover.

these announcements these announcements are very much at the absolute center of business they're not a fringe element like some of the internet build out was but i would compare it to like you know laying the undersea cables for like phone conversations or building towers for radio, that this feels really infrastructurally significant in a way that I can't remember as a 42-year-old man.

uh anything feeling like this it's we have to go beyond those analogies because the the challenge with this is that that In the 90s, there was like, for a period of time, they overbuilt too much fiber optic. We talked about dark fiber, all this dark fiber. And you could predict the upper bound limit of how useful fiber was by the maximum amount of...

Video you could be sending through a line right and prior to Netflix streaming prior to that There just wasn't a lot of demand for for video and then eventually companies existed to fill that demand, but it took a long time What's weird about compute? is if you built a data center, you know, somewhere in Texas, right? You built a one gigawatt data center, a huge data center. Okay. Forget the one and start, you just built this thing and you told opening eye, Hey, listen.

We've got this data. We just found this data center. Can you use it? It's ready to go. Yeah. Yeah. The demand, there are two sources for demand. One is, or three sources. One is consumer, just route more chat GPT things in there, give people more of those tiers, create a higher quality experience, whatever. Second is go deploy API in there, maybe lower the cost of that, create more demand.

The other thing is you can just take up that compute entirely by the needs of an AI company alone because they are bottlenecked by the need for compute. They're bottlenecked by their researchers sitting there going, I want to run this thing like we don't have the compute to do it. Yes, but this could cure cancer.

Maybe not. But that's the thing is a data center and a team of researchers, you can get tremendous amount of utility out of it. Dark fiber wasn't that way. I needed somebody on the line paying for this thing and I needed to have be paid for every time data went through. there a data center is different as i can run a thing and train a thing and do a thing and just create it's like a factory it's like literally saying i have a factory that can make any type of information about technology product

the limitations of the number of factories. I think we have to think about it as just like factories that could make anything digitally. And that's where I... I think you see the confidence, the business confidence to make these kinds of deals. And for, you know, beyond Nvidia and beyond OpenAI, you see a lot of.

hedge funds and stuff like that, have a lot of their balance sheet tied up in the building of data centers with just the simple assumption that everyone's going to need more, that we are not even close. to satiating the desire for compute in a way that is going to reshape a lot. In terms of energy generation, in terms of everything that's going to have to support it, this is an economy-wide commitment to reshape the world.

In a weird way, I think it's undercover. Oh, I agree. I think it's very undercover. I would say that the Dubai deal got some news because of the political implications of that. Stargate UK. You know, that was last week. Barely. There's a lot of other stuff going on in the news to be sure. But this was, okay, this was the second one that I believe that's been announced of building an external thing in partnership with another country. And we're seeing this.

idea of let's not just build these data centers in the US and have it all here. Let's build cooperative data centers with other countries because we see the demand for this. And there are some other topics and things like that we'll probably be getting into in the opening podcast in the future about this. But we can see that there is this large-scale cooperation. And there was a... There was an analogy I once heard, which I thought was kind of interesting. And that was the idea that a...

you know, 747 was like a cathedral of the sky. Because in the Middle Ages, a cathedral involved people working from all different levels to contribute and to build a cathedral, right? It was stonemasons, mathematicians, artists, classmakers, all these people. to build a cathedral. If you wanted to say, here is all the human expertise that you can see in one place, you go look at some amazing huge cathedral. Forget the religious connotation, just the idea of what it took to do that.

When you look at modern jets, like a 747 from the electronics to the seats to the engines, it's a global effort. Like literally you're involving all sorts of domains, people in there.

And I think that we're starting, you could say your iPhones like that too, but like you start to look at a data center and it's almost goes back to the cathedral analogy because it's a thing that is, you know, serves this global purpose and is also requires all of these different players to come in and make it happen. Is there a winner in this world?

You know, it sounds like everybody is rushing to build data centers. We are building data centers on a level that, again, when you talk about what needs to happen to support it, we're going to push. nuclear reactor permitting further in the next like two years than we have in the last 50, quite simply because there just needs to be a proven way to generate enough.

a power for this. I, I, I, I tend to agree with you. This does feel like something where look, there's a lot of people that are going to get, get going to get paid to build. a lot of these things and solve a lot of the engineering problems that go along with their operation. And you have to get several paragraphs down into, for example, the UK initiative. And it mentions that part of isn't just the idea that they're building out this massive data center in the UK.

But also that they're going to help support the UK government because they have a plan to upskill seven and a half million workers by 2030. It comes with this idea, too. How do we create an AI native population? Which I think is, that's the thing. Again, not only people not paying attention to the buildup, but also like there are these big global initiatives we're seeing. I think that was part of the thing with the Dubai deal was the idea of how do we upskill the population?

Governments are talking about that. It's one of the things I'm like, ah, we need to be doing more. I'm like, I agree. There's actually press releases and conversations about this. Is it going to solve it? I don't know, but step one is- you know, if you're complaining about it, which I am, but is to go pay attention to what's being done. Because if you're just like, why won't anybody? Well, go look, you know, and maybe you can be part of that and help and help that happen.

Do we want to talk about the pessimism? I saw another thing today. I sent you a link to a quote from Dario from Anthropic saying that 25% of jobs would be eliminated. by AI in the near future, to which seemed like a sunny day for Dario, that only 25% would be disrupted. And then today I saw the CEO of Stability. say that there would be massive job disruption within the next year. With those investors alone. What is the state of economic disruption?

So I have – I'm writing – I'm in the middle of writing a new piece because there are a couple economics papers that came out that make some fundamental mistakes in my opinion. I was in the middle of writing a piece we're just going to send to you. And I would say that –

Dario's a genius. Dario's absolutely a genius. And so, you know, I had the chance to work at Dario's opening. I could see this guy is very smart. And I'd say like many people. Saw my first television ad for Anthropic during football yesterday. So they are out there. Yeah, yeah. And by the way, hats off to Google Gemini passed OpenAI for the number one slot on the App Store right now. So it's getting very frothy out there. Yeah, so congrats to Google on that.

We'll see what's going to happen. Oh, but he's got some announcements. You know, they said today on Twitter, that's all I know. But I think these people are brilliant. But I think the thing is, is when it comes to trying to make complex predictions like this, we're all kind of in the woods. There's a paper that came out that said the title of this was

And they quote Daru in this is, we won't be missed work and growth in the era of AGI. And by Pascal Restrepo from Yale, who's making a claim. And I will get to the core problem with a lot of these premises. So I'm writing my rebuttal to this sort of thinking. The tentative title right now is, How Much Compute Does It Take to Replace Taylor Swift?

In the paper and in Dario, in that world, and again, if I'm putting words in Dario's mouth, I apologize, but as I understand it, they look at a formula. They create this formula. The formula is the amount of capability of compute it improves. And then the point at which it becomes more cost-effective to throw a computer at a problem than it is to have a person solve it, right? You could say there's a mathematical formula.

The problem with that is it makes a basic assumption that some degree of compute can actually replace every human job. My argument is that you can't. There are certain things and also... Sometimes the more you put compute at a thing, the more desirable humans filling the world is this. In Restrepo's paper, he gives a perfect example of I think of sort of this misunderstanding of even like how basic work works.

He talks about the idea of an AI therapist. What if you could have a team of AI therapists that were made more effective, way cheaper, whatever, had a way higher success rate than actual therapists, than human therapists? You know, it's an example of how even therapy, he's trying to show even a human touch job like therapy, how that can be eliminated. And I'm like, well, number one. Okay. Psychics in astrology.

Don't work. We've known this since the Middle Ages. You have no lower bar than outperforming psychics and astrologers, yet there's still a huge demand for this. So therapy was a very poor choice to say, what if the AI is better? Doesn't matter. It's not why people often go to it. Yeah. I would say that for people are aware that it's not great. It's why people use chat GPT. There's a demand for this, but it's not going to displace the idea of sitting in a room. And I use an example of.

Because this false equivalency, he doubles down to even more. I go, okay, you're presenting this scenario. Would I want to have the AI therapy team that's going to be 100% success outcome or the human? My option is the human with the AI. Because particularly for areas of therapy, like imagine grief counseling.

Do you want to talk to a robot? What is going to be the outcome of a grief counselor that has no idea what grief is? I think that there might be roles where we're going to be using AI. I'm a big believer in AI therapy. I like that very clear. But I also think that a lot of what we... want.

is a placebo effect. We want just the fact to talk to another human. I use an example, physical therapy. I believe that a robot physical therapist can perform all the things I need, know all the things to tell me whatever, but when it's telling me to feel... the burn, I'm going to hit it with a crowbar because there's no idea what it means to fuel the burn. And I think that's the problem is they get into these analogies of saying, well, with enough compute, I can replace it. Like, no.

We wanted the person. You know, you could have a robot marriage counselor. It's not really why people go to marriage counselors. They do it to get another human to agree with them that their husband's terrible. I don't speak from personal experience yet. No, no, no, no. Yet. So we'll see. I may be seeing it enough in tune. But the point is, is I think that completely misunderstood stands the question of why we use a thing and where the value.

comes from. He then says, you know, why would you send your child to human doctors if the AI could cure them and seem inhumane? I'm like, okay. Am I going to send my kid to the children's hospital that is completely run by robots and staffed by robots? the one with just humans that are inefficient or they're highly efficient ai and human robot hospital where my kid gets to talk to a human there you know it has human nurses that

Their job is just to explain stuff or be human or whatever. And I think that that's the thing is that for his idea to work, that, oh, you're going to get rid of, we're going to get human from labor, you have to be able to say that there's a one-to-one, that if enough compute, I can do this.

People make the argument. Well, what if I'm building Westworld robots? What if they're exactly like us? Whatever. I'm like, cool. We're building mechanical liars. Is that going to make me? Some people might. Like, listen, there is. There is a market for blow up sex dolls. I don't deny that. But I would say that often what we want is the real thing as opposed to the counter. Yes. And I get into an argument explaining that, like, listen, like I was just in New York City last week.

And the Louis Vuitton store, Louis Vuitton, five stories, looks like luggage, it's amazing, it's beautiful. Louis Vuitton is owned by LVMH. which is one of the most successful luxury brands on the planet. Bernard Arnault is worth $100 billion, incredibly successful. At one point, he was richer than Bill Gates. Within a block, I can buy counterfeits.

I can buy a counterfeit handbag, counterfeit Louis Vuitton, counterfeit any version of this, yet that building still stands there. Why? The only reason the counterfeit has any value is because the real thing has more value. The reason that I want this thing-

fake logos on there is because the one with real logos is valuable with perceived value. So much of our economy is driven by that. And it's a point to make like, I've, you know, I've got this back and forth with chat GPT. I gave it the paper. I'm like, listen. What is, what is the, how much computer plays Taylor Swift? And I can have an AI talk about teenage angst and, you know, and about this, but I've known like for many years as an artist.

Often the thing that wins isn't the best thing. It's the most relevant thing. Twilight novels, right? Not the most well-written novels, but when you're a 15-year-old reading what sounds like a repressed woman's idea of what it was like to be 15, it's more relevant. So my point is I think that you can match. I can make a counterfeit.

Louis Vuitton that looks exactly like it. But the moment you know it's not the real one, give one to your wife, see what happens. You'll be in marriage counseling wishing for a robot. I think... The further we get along this road and the more we get into understanding where we're going and less the abstract of what the... possible damage can be based on this technology, the more we complicate the idea of optimism of effectiveness.

equals almost one-to-one economic chaos and disruption. I think that was a model that happens in a very early phase in this, and a lot of people still have... you know their priors there where for me at least the farther we go down along this road the more i think that our idea of what the economy is going to be to your point

it's just not going to be what it is we're going to find a million different fractures in what we normally think of as a simple way to get what we want and now we're going to find ways especially if costs come down on some of these especially service-driven economy jobs. We're going to find reasons to buy and employ more people. We're going to figure out ways that more interesting services can be offered to us. And to your point about enhancing things.

I've spent a lot of time thinking and obsessing about the idea of how much of what my dream world of a general practitioner for my doctor. can I replicate with my wearable data, my scale ways, tracking my fitness, tracking my diet, and chat GPT. Plus then like having a schedule of lab work that I would just get done on a schedule or by the recommendation of the GPT. And really what I've found, the further I've gone down this road, both in terms of building the tools and...

just thinking about it, is that what I kind of really would like as a first minimum... uh case scenario is just to never feel bad when i leave my doctor for my annual physical which i have often done that i feel like like what am i doing here why why did i come here what you know this didn't feel satisfying. I don't feel like I have more of a grip on my health than I did before. But if I can give them a bunch of data.

And I can say, hey, look, here's where my data's at. I can break it down at a table in an executive summary for you to go and then match it with whatever data you grab from me now and what else you have with me on file. I can now have a more interesting conversation, a more fruitful conversation that just makes their job better. It makes, it makes me feel better. And it just, it, I don't know if I'm really.

I think the job law stuff is going to be interesting because there are going to be certain things that change and they are going to change very fast. But I also think that opportunities are going to be just boundless. Yeah. Yeah. I, I think, I think there will be, I think there will be massive disruptions. And I think the thing I think we have to either understand, have to come to have an idea. Are there, is there full employment on the other side of it or is there not? And that.

determines how we handle the disruption. And that's my frustration right now with Dario and some of these other people like, nope, you're done. And I'm like, no, you're, this doesn't make sense. Like you guys don't under, like you can't explain to be NASCAR, the Ortega is another big, you know, luxury brand. Like all the stuff like that doesn't factor into their calculus when they try to imagine things. And I'm like, no, like every time.

The plow ruined 99% of all jobs. And then what we continue to do is we find the things that humans are uniquely great at that we want, and we get more of that. And for a while, it was like, well, I was doing math and doing this. Like, now I have computers. Like, yeah, but... That wasn't a human skill. That was just a function. I think that we have to, I believe very much that there is a role of full unemployment on the other side of it. And we have to, once we acknowledge that- Full employment.

Yeah, full employment, full employment. Anybody who wants to have a job can have a job. The more we realize that, the more we'll be able to handle this the right way. And I think there are people advocating, and I'm getting more vocal, I think, for absolutely terrible solutions that could create an outcome where we... force ourselves into less good outcomes. And anecdotal stuff is like...

I haven't been camping much, but I have a feeling that I've spent more time in the presence of Bigfoot than a physician. You know, you look at the amount of time your doctor spends with you. It's like next to zero time.

People talk about like, oh, what do we do? Humans replace, you know, robots replace, you know, humans. It takes weeks for me to take where I live. It takes weeks to get somebody to come look at your air conditioner. Weeks to do this because there's not enough people doing this role.

If you want to do a kitchen remodel and you have a nice big home, you can buy a small house for the price of a kitchen remodel, which means people don't do kitchen remodels very often. But if the cost of a kitchen remodel was, you know. paying the guy, you know, the, the, the contractor stepson, you know, whatever a thousand bucks to come to my house, measure some things out there and send a bunch of robots in to do it. We do it every year and we would keep that guy in.

employment because they give you in three houses a day, like the scope and scale of the things and services that we would buy if it were easier and the things that we would offer if they were easier. There's a lot of work you and I don't do because of all the little busy work that it takes place to do it.

We've both run small companies together and whatever, and so much of our time was bottled up in stupid little tasks that we could have offered better services, more services, more to people through automation. And I think that when you're on that side of it, you go, oh my God, the robots can't.

come here fast enough because for entrepreneurs, it's going to be a wonderful time to do this. For somebody who's been working at the same job for 20 or 30 years, I think, yeah, it's scary. It's to think that your boss is going to come to you tomorrow and say, hey, listen, we don't need you. EPT just added this feature. But I'm telling you, learn to use ChatGPT because now you're the most valuable person in that org. Yeah. And that's what I...

I think that there's just such a tremendous world of opportunity. You know, even if you're in a lot of these, I mean, I think, you know, for me, if. I've been around the media and journalism since I was a teenager. A lot of that has already happened for reasons that go beyond AI. AI is closer to giving more journalist jobs than it is for taking journalist jobs. A lot of other media consolidation trends. And poor decision making from an industry perspective has done a lot to erode that headcount.

It doesn't mean that there is not a future for this. In fact, I'm more bullish on the idea. I'd be more likely... to want to go join a newsroom today than I would have 10 years ago and I would have 20 years ago when I actually had the chance to do it and decided to, you know, strike out on my own in the internet wilderness.

We, you know, the rumors right now that, you know, the free press, Barry Weiss's newspaper might be making a deal with CBS to take over CBS News. I don't know if that's true or not. But you go, okay. It has been widely, widely reported. Widely reported. People I've met close to it are like, yeah, it's not, they're saying it's like- Andrew's personal reporting of people that may or may not know. Yeah, yeah, talking to people who know people there. But anyway, who know people? Unnamed-

Unnamed sources. Right. And the point, though, is that what does that represent? You had Barry Weiss leaves the New York Times for reasons of, you know, she's I think she's a traditional old school newspaper person, wants to get to the story, regardless of her personal point of view, whatever, likes the story, really likes the story. New York Times was.

a news outlet that, um, One had to adapt to a world where you weren't supporting yourself by all the people paying for classified ads or ads in the New York Times, the print edition or then subscriptions had to move to a digital edition and then found themselves in a world where. they had to cater to a particular kind of audience and became less what their brand was before, you know, apologizing for op-eds.

You know, people lose their jobs over publishing op-eds. Like that's the craziest thing in the world. The op in the end is the op part. Like, but no, a little too op for us. Just crazy because they became basically this idea of thinking that their job was.

you know, we have to keep our subscribers happy. And even if that means bending our audience capture. Yeah. Barry Weiss is like, nah, well, I'm going to find the audience that maybe wants the broader thing. She goes off, uses tools like Substack, uses these platforms. I did an interview, by the way. with the CEO of Substack last week, which will be airing soon on another platform. And goes off, creates an entirely new publication, creates a new enterprise there.

is able to leverage technology, leverage audience understanding, and literally now maybe taking over CBS News? I mean, maybe even more than that, based on some of the rumors about how far the new CBS wants to spread. You never know. And really...

I don't know if we've talked about it here. I know that we've talked about it personally, but the reason why I'm bullish for news in the world of AI... is that I think that there's going to be a lot of these platforms, including Google and including OpenAI, which has GPT, where people by way of their personal assistance are going to be desperate for good. reliable news and that is something that either by mini subscriptions or instant payments or even just

These platforms wanting to be the place where you get your news every day will be willing to make deals with good news, with good newsrooms to surface it. And in a world where that happens on that surface level of just being able to gather it, you are. Two things are real. Number one, you're meaningfully able to decouple news from the advertising model in a way that has not been seen since, you know, pre-Gutenberg.

That would be a revolutionary idea in the world of news gathering and dissemination. And two, it would stop... The audience capture dynamics that currently happen in a world where the only relationship that you have are either comment threads or social media, which is kind of built to self.

corral certain kind of groupthink, I think that's incredibly healthy for news. A, divorcing itself from the advertising model, and B, not being... as beholden to the current audience that they hear the most from if you're able to draw a you know a mass source of income from reporting that kind of news i'm I'm very bullish on where it could go. Who knows whether or not it will.

Can I read, I was making a reference to like, you know, when I hear these descriptions of when compute does everything, how dystopian they sound without them realizing it. And I was pulling up an example from the book, The Machine Stops. by E.M. Forrester, which was, I don't know if you're familiar with this, was written in 1909. It was kind of a rebuttal to H.G. Wells wrote a book about his sort of utopia idea.

And everybody basically lives in these self-contained mechanical bubbles, and the machine supports them. It's written in 1909, right? There's this rolled machine, provides everything they need, whatever. And this was how they describe... The two main characters, Vashti and her son, Kuno, live on opposite sides of the world. Vashti is content with life, which is like most inhabitants of the world. She spends producing and endlessly discussing secondhand ideas.

What? Literally a world in which people sit around circulating secondhand ideas back and forth and just, just 1909. What is this? The machine stops. 1909. Machine stops. Yeah. Yeah. I really liked the idea of the framing of how much compute is Taylor Swift. Like what is the price tag, the compute price tag on Kanye? yanking the microwave from Taylor Swift at the VMAs that makes, you know, both of them in different ways infamous. mega stars that, you know, uh, exposes a, like.

How many chips do you need to produce the bottle of Hennessy that Kanye had on the red carpet that led to that moment? What is... Yeah, and I brought up the Bring the Force thing. This was 1909. They're talking about basically internet and meme culture. The idea there's no original thought. They're just sort of exchanging things back and forth. But, you know.

For there to be a meme economy, memes have to have meaning. And I can have an AI generate endless memes, but if they're disconnected to everything else, they don't have value. And I think that's... And I think I'll get to too, is it like... We hear this sort of description, which I sort of bought into, and then I step back and said, no, let's analyze this. Because people go like, oh, so we're all going to be influencers. Aristotle was an influencer.

Benjamin Franklin was an influencer. A really popular professor is an influencer. What is an influencer? An influencer is somebody who has something to say that other people want to pay attention to. And it can be a doctor who has got a TikTok account talking about really good medical advice. There are mathematicians. I watch people on YouTube who talk about mathematics and physics.

And would we say, oh, they're an influencer or are they a physicist? I don't know. It depends what they're talking into. If it's a lectern, they say a professor. If it's a phone, then they're an influencer. Yeah.

I mean, obviously a lot of these lines are kind of blurring, but so will the amplification. That's another thing that just kind of comes along with a lot of this being able to be solved is that now... the the rising tide for expectation from an audience and ease of production is going to make so much more available so here's here's the thing too i just thought of analogy for this was

You know, we've talked about this. So just tonight, we also run Interdimensional, which is our AI firm where we put together teams who work for companies. We're always trying to find people who are AI natives, people who understand this technology. And you could have six months experience and you might be highly employable because if you have those skills, they're highly employable. Oh, yeah.

And I have this – I was thinking about this before of like there's this analogy where the newer you are to it and the deeper – the more you embrace it, the more useful you are. When I was – In more involved in TV and I was, you know, having to doing TV shows and stuff when I'd go meet with either the networks or my agents. I would always get invited into a meeting where there'd be two blonde women. It was always two blonde women. I can never figure this out. Who are the social media team.

And they were 22, 23, because it was like the typewriter, the only ones who took the time to bother to understand the lingo and everything you had to do and how the tweet culture worked, all of this stuff. And this was this thing that should have been interesting because... The really good ones now are...

Very highly placed, running PR for big firms and things like this. And we see these upheavals. That was the joke. It was in the 90s was you hired your 14-year-old neighbor to build your company's tech stack because they were the only ones who spent the time to do it. The message.

The thread there is that as technology advances, having a four-year degree or an eight-year degree and a PhD aren't as relevant as just... deciding to pay attention and learn how to use the tool because a lot of get in the water yeah get in the water right now dude like that's all you need to do is And we've experienced this trying to build out the educational component to interdimensional is that it's like...

the the biggest lesson we've had to learn is the adaptability and what is constant in this field because you know the more the more you try to peg it into like here's what this thing is right now it's like holy crap if you don't if you don't sell it today if you don't teach it today then it's going to be something different you got to figure out what is what are the lessons that will always be there which is which is interesting i've been in we do we have corporate clients

And what we've seen now, a pattern that's emerged is when they're like, hey, we want to learn how to use ChatGPT, Enterprise, whatever. And the company looks around to see who internally knows how to use it. And often it's the power users they didn't connect us with to talk about it. And those people become.

extremely important within the organization. So that's like, what are the signs? Like, no, what is your, our company's going to automate, learn how to use it, you know, and then we'll all be AI generalists. We've had a lot of conversation about the role of human and robot. What we have had very, very little conversation about is the role of the robot lord.

the robot mancer like that that may or may not be 70 of our economy in five years will be will be will be robot mansers those who charm and deploy the robots in the exact configuration that that solves problems which by the way is what we would have said about the computer in 1971 yeah i i think that

I think that very narrow knowledge work jobs are going to have to rethink. If you're not thinking right now how to do your own job with AI, somebody else will start to do it for you with AI. But if you start to think about how to scale up like this, I think a lot of the things that we think... I think robotics will get better really quickly, but...

you know, if a robot comes knocking on my door, answering offering landscaping services, I'm probably not going to answer. I saw that movie by James Cameron, you know, but if a human comes up to my door and has 10 robots standing in a pickup truck and says, do you want me to come by every Monday and make your job? hard to look immaculate, you know, and also, you know, fix everything. Like, I'd be like, sounds great. You know? Yeah. You take Venmo? Yeah. So, I am...

Like I said, I think that if you're an entrepreneur, it's a wonderful time. If you're in the middle of a big org, learn how to use these tools. You'll find yourself being very much in demand for this. I think over time, there's a great future for everybody, but we are at a transition point. And I don't pretend to think that's going to be easier.

automatic or smooth. I just don't want us to buy into the idea that there's no jobs on the other side of it because then we go into the most nihilistic route and protectionistic route and we don't benefit from it. I think we're at a golden age of human jobs coming. Yeah. I just think that like service can be an entirely different world. in this scenario. If you have a great automation system and what can differentiate you is

how customers interact with you. Like there are going to be solutions that are more pleasing for a human that involve humans and not necessarily. a computer, you know, there's, you know, if I have like, you know, my, my Amex platinum card. which I'm, I'm, I'm glad that we're doing such a relatable podcast talking about kitchen remodeling and MX platinum cards, but like, I haven't done the remodel like that very clear.

I want to be able to talk to somebody. I want to have a bit of a white glove treatment. And I want that person to be using AI. I want that person to, to have the fastest experience that they possibly can. But I want to talk to people. You know, if I go, I don't know. I mean, you know, this gets us into the, the. in a projection world. But I do think that there's, there's a lot, the further we've gotten down the road and the clearer the future becomes, the less I am.

terrified about the worst case scenarios yeah more the more it feels to me that it's like man the fractures in what we assume the linear path is between one thing or another and how many times we're going to stop and want to pay for other little things to make this better or that we're going to have solutions that do work, but we're going to want to.

customize it or change it or somebody's going to say, hey, there's another thing because it's so easy to switch between various different things because they cost less. then we're going to be more adventurous and we're going to want to change different things. I don't know. It just, it feels exciting. Yeah, I'm optimistic. I think that our existential risks have dropped considerably, our understanding of them.

The advantage I think now is that like, you know, I can run a local LLM on my computer and look at all my network traffic. You're worried about some company trying to take over. We've got a lot of competition and people with eyes on everything. It's a much more, I think that.

The Doomsayers, you know, there was a book by one of them that came out, aren't getting as much. Not that I say we should totally ignore it entirely. I think you should ignore the cranks that aren't offering anything new. But I do think that we're now thinking about, okay. this is happening the progress there's still the people trying to argue that the models aren't better than they were five years ago or whatever and oh we had a wall cool yeah it's like all right cool yeah great awesome dude

But now we got to think about what happens next. I think what happens next is helping people adopt, embrace, make use of it, be entrepreneurial, plan for the future. Not be an MPC. One last question. Big meta.

announcements last week. New smart glasses. Some rough live demos. Now, I tend to subscribe. I like my meta Ray-Bans. I don't know if I'm going to get the in-screen version, but... i do think that you know for me they are a functional ai assistant away from being a killer product uh what what What did we make when you're watching those live demos, especially knowing how some of this tech works? Did you see the kitchen one?

Yes. The live AI. What was what was your aside from secondhand embarrassment? Because nobody wants to see anybody go through that, especially if it goes viral. What were your thoughts? Remember. I was six feet away from Greg Brockman when we did the live demo at GPT-4 of all the capability. Greg's like, I'm going to do it live. I'm like, no, no, Greg, don't. We can't. Let's not do it live. And every...

Every failed demo I'd ever watched ran through my head. I begged and pleaded and he agreed and we recorded his. the entire session the day before. So we had a complete run through of everything and it worked fine. Also, I'm like, if we had a problem bandwidth or whatever, we can at least run that or upload that or whatever like that.

And it worked great. Greg did a fantastic job and it became, I think, one of the most legendary live demos ever. But opening, I was presenting it as a developer product. This is for developers. This is what we're doing, whatever. And if there was slow or hiccups, whatever, that's to be expected because this is a tool that we're showing developers to get an idea of what's coming and how to build on top of it. Meta was in a tricky place because...

They're trying to show off hardware that's consumer hardware or whatever. It seems because other people tried it in demos, I think, afterwards and had some similar issues. I think it sounded buggy, but I don't think it's anything they're incapable of solving. I don't think they're not going to be able to fix the thing. I think it came down to probably the network, because it said Wi-Fi, but turn out it was probably something like that.

We know we can have voices understand the background. We know we can do these stuff. We know these are solvable things. We know we can do context-aware. We know that Meta has a lot of really smart people. that I used to work with. Oh, we know. That I used to, some of which someone got elsewhere. I... I'm more directionally interested. Do quote Jurassic Park, spared no expense. Yeah. I think that it didn't show there, but that's a different team.

I think directionally, I'm excited about the idea. I'm excited from you and me watching the Google Glass demo and getting so excited on what we thought it was going to be. Thank God I wasn't one of the idiots that bought it. Like me. Yeah. I'll have them in my closet. I say standing before my Apple Vision Pro. You and I are the first suckers. Let's put it this way. We're suckers. There is a world that we want to live in. Oh my God.

Yeah. I have, by the way, I've got like the in real that just the video glasses I use for watching movies. That's great. Like. My frustration with Apple was they didn't try to just build something like this. If you just built a simple pair of glasses that has a little display in there that I can see stuff, whatever, I could do a lot of things and I would start thinking about building applications for that world.

I don't want to put meta stuff on me. My wife last night wanted to check something in my account. I don't know my Facebook, my Instagram password. I don't use it. I'm not a fan of the product. I don't, you know, will I buy one of these? Maybe, but I will be ready to jump ship to somebody else who creates a version that is not meta, you know? And that's an example too. We talk about the future of like, no amount of compute.

it's going to make me want to just jump ship and stick to meta. Like reputation means a lot. Well, yeah. It's things that will never happen from me in my lifetime. building anything on a meta foundation at all, ever. Like, they generationally burn trust in terms of that. As a hardware thing, like, look, it's fine. They look like Ray-Bans. They take good pictures. They're great.

hands-free devices when you have a child and you know you can push your kid on the swing and take a video that's nice um the uh yeah the uh the live demo We'll see. It is narratively unfortunate, considering how much money they spent, that that's the first big thing that they do, and it doesn't work. I have faith that they can continue to push those products. And they're in a good place in terms of leading, in terms of market.

penetration of like where are the glasses the ai glasses people we do ai glasses uh now maybe if they were good be a different story but uh We will see. Mark is a very smart guy. Bosworth, they've got a very smart team there. As much as I'm not a fan of... the software stack and the the social platform and whatever um i've got well i haven't used my question forever the quest

Great hardware. Yeah. The Quest, $300. They delivered on VR in a way that nobody else could. Just nobody talks about anywhere he uses it, though. Wow. You know, I think it's because the guy who built that is now building. drones that shoot rockets that shoot another drone out of it. But even after that, though, I mean, like the versions that we have came out after. Yeah, no, they were good.

They were on that pathway. Hey, totally non-AI related thing on the way out of here is they downloaded iOS 26. Let me just say this. I don't know if I've ever been more bearish on Apple.

The iOS sucks. I mean, there's some bells and whistles there. The visual design on that is just ugly. It is just ugly and bad. The decisions that they made... I don't get why they made from the design language to move certain buttons from here to there. And this is my biggest thing. If this is their once every 10 years refresh. that they were pitching, like it doesn't move us forward in any way.

When it comes to the actual problems that we have with our smartphones, like it doesn't make apps easier to manage. It doesn't make alerts easier or more interesting to manage. It doesn't make my life any better. It makes it kind of weird and uglier. And I'm like, ugh, God. And that's before we even get into the AI stuff where they have missed the boat. They are so far behind at this point. I... I get that on one hand, every time something is new or a change, you often go, eh. But...

sometimes it really is a bad shift. And I don't like, yeah, I'm the same way. I look at this. There are places where I load up windows and things are outlined and I'm like, why? I look at the keyboard there as our last text. I look at the line. I go, is that supposed to fade like there? Is that line there? Why is the line there? Why is it outlined? It felt like... The glass was... I have a problem on my display where I can't see certain things because the glass is such...

There's when I have white things underneath it. And it's literally, I'm like, this is a thing that we've known forever. When you do translucent displays on top of objects that are light, it's really hard to see. I feel like with Apple, it's like a lot of other things. There's an inertia. They got an excitement around this. Apple likes glass. We know they love glass and glass, glass, glass. And that became the thing that they wanted to do. And it became like...

I think it's ugly. I don't think it's a really aesthetically pleasing thing. I've looked at and I say that like I've seen other things that go, oh, that's actually cool. There have been times they've updated stuff and I liked it. But I look around and I think that like. you know, uh, I, I don't, I, I, I feel I'm like, gee, I'm worried. I had a conversation with somebody in venture capital earlier today. We're talking about like, how do we feel about Apple? And I'm like, I don't want to be.

pessimistic but I don't have a lot of reasons to be optimistic I will say how do you like the hardware itself Did you get the new phone? I do not have a new phone. I did not get the new phone. I will say. And that's partly why I was like, I actually got a new phone. But like, I'm going to skip this version. The small pro. Oh, you got the new, the Air? I got the Air. How is it? I'm very happy at the form factor. Other than the ugly UI, very happy at the form factor. It's much lighter in the hand.

We'll see. I think the battery life is going to be terrible. I bought the battery pack for it. I think I will not go. I'm not going to venture outside into the city all day without bringing the battery pack with me. But I love the form factor, the thinness of it. I really, really.

Like I'm a fan of that. I was like, you know, like I give up like the super pro, super duper camera, which I use twice a year. I was okay with that. But for me, for having a slimmer thing, a lighter thing in my hand, that's what, that was my one, my one bit.

of complaint about Apple is always like, make a thinner phone. Make a thinner. I don't need more cameras. I want a thinner freaking phone. And they did that. Well, this is the other thing with Apple is that now we kind of know where they're going. Yeah. because the rumor had been the air this year and then next year they're going to put two airs together and fold them and they're going to have the apple fold and and you know we we kind of go from there so it's like

I was having a conversation with Will, our editor for this program and various other stuff. And he writes about Android and it's like, yeah, the phone market is just like... Everybody sort of copies each other. There's not really one like really bold pushing things forward leader in the space right now. And I don't know if there will be until there's stuff that is designed specifically AI in mind. Yeah, it's hard, because, like, you... The iPhone, when it came out...

Steve Jobs did a couple things and I think really does credit belong to him. He did a couple, he pulled a couple things from the future into the present that we didn't know we wanted and it made so much sense. Multi-touch, we knew at the time was amazing. People forget about how innovative it was to really touch your screen and have it move. And even for years, Android was janky. Android didn't feel real. But now we finally, all the processes are faster. That multi-touch.

was one thing. Even up until the announcement for it, Apple had filed patents for control systems using the edges because we're like, could you make a touchscreen that was really good? Because touchscreens up until then sucked. We're not good. Yeah.

And that was magical. That was really, really the magical thing was, I think was just the idea of that. Getting, you know, being convinced into the app store, which took some convincing for jobs, it all of a sudden made us realize how big of a new platform this was. But... Since then, the only innovation, there's been things like the way we charge it and some other things was just...

Better cameras, better cameras, better cameras. You know, processor, of course. And then it's like that became the joke was, oh, okay, how many cameras will this new iPhone have? And I'm like. One on the back, fine by me. You know, that's great because I don't need more cameras. All right. We have a call that we need to get to, so we got to wrap this show up. Andrew, where can people find you?

AndrewMain.com. And please check out the OpenAI podcast, which you can find on iTunes, Spotify, and YouTube. Go to YouTube.com slash OpenAI and you'll see us in there. Absolutely. Great, great, great episode about Codex. That was really fun. Talk to Greg. Justin Robert Young. Oh, yeah. Justin R. Young, everywhere you find me. See you later, folks. Diamond Club hopes you have enjoyed this broker. Dog and Pony Show Audio