11/18/25: Rubio Pushes Venezuela War, Saagar Attacked By DeSantis, AI Job Loss Imminent

23:55They're bilking everybody else to try and pay for their stuff. But let me just, you know, and fine, that's voluntary. People voluntarily go down there. At the end of the day, one of the reasons why property tax in particular is important, and this is not popular. I'll just admit it straight up, is yes, the seniors, what they say is seniors on, quote, fixed income should not have to continue to pay into the system. They've paid their dues, so they shouldn't have to pay for school tax. Now, let's again just think about that. Older, silent generation, they paid for the boomers' children's school taxes whenever they were going to school.

24:32But second, we all live in a collective society. We pay for their Medicare, for their dialysis, for their Social Security checks. By the way, vastly more that they receive than they ever paid into the system. And in return, we expect that they also pay into, let's say, property taxes, school taxes for the societal compact that we're all in this together. Oh, and by the way, 911, ambulance services, city services, and others. Anyone want to tell me which age group vastly consumes a lot of those public services? Right, right, of course. So at the end of the day, somebody has got to pay. Someone has to pay for cops, for ambulances, for schools, which is what the vast portion of a lot of this is.

25:10There's only one way, if you get rid of property tax, to do that. It's a shift to a highly regressive sales tax and to a tourist tax, which, look, maybe that will work in Florida. And ultimately, this is up to Florida residents. You guys decide your own destiny. And Texans have already, you know, my own home state has decided to do this. But I think this is highly dangerous because this is now going to catch on in every red state in the country. Many of these red states are places where people who are fleeing New York City, California, or whatever, are coming for a more affordable way of life.

25:46This is being sold as, oh, it's going to reduce your overall tax burden. Again, it will come back to bite you no matter which way you look at it, okay? It's coming for you no matter what. Somebody has got to pay, ultimately, for this. what this net result would do, it would push it off to other people in the form of a sales tax, and it locks people into housing stock. One of the most common arguments from a lot of these people is like, oh, well, I bought this house in 1975, and since then, poor me, it's gone up to$2.5 million, and now I can't afford the property tax payment. Now, here's the thing, but for some reason we're supposed to feel sympathy only for the senior.

26:28How many people do you and I know who are young, whose rent has gone up by 25 or 30 percent? Does anybody do anything for that person? Right. Does does anybody care? Hey, oh, what does Ben Shapiro say? Move. Oh, sorry. That's the free market. And by the way, they don't even have equity in their rental property. These people are sitting on millions of dollars of unrealized gains, not to mention their booming stock portfolio. So I'm supposed to feel like deep sympathy because you're sitting on a multimillion dollar property, which you can sell, by the way, if you wanted to it, because you can't keep up with the burden.

27:01And then finally fixed income. This thing is bullshit. The idea is quote seniors on fixed income. They get social security is literally pegged to inflation. Now I'm not saying that the COLA, the cost of living adjustment is true inflation or any of that. Does anybody else in the world get automatic adjustments in their income. For a young person whose rent went up by 30%, they don't get the requisite, you know, seven, 8 % or whatever automatic pay bump. They probably get either a decrease or they have to do the humiliating ritual, you know, in a big corporation of being like, hey, um, yeah, I know it's my review time, but I would really like to get a, you know, X percent increase in my rate.

27:41And they're like, oh, sorry, actually, we can't do it this year. And what do you do? You just have to eat it. So everybody else is stuck with personal responsibility, but supposedly this whole fixed income thing, even though it's not fixed, it's literally inflation adjusted and now tax free as a result of the big, beautiful bill. 88 % of social security recipients will not pay a single dollar in federal income tax on what they get from the government. So you put all this together, who are you prioritizing? In Florida, it's obvious. And he even said it in the video, our seniors. Now he's trying to sell it on Twitter as actually because the young have to pay more in property taxes because we cap it for old people.

28:18We're really helping them. No, at the end of the day, those are the people who need those services like schools in particular. You will be paying for it no matter what. You are just going to be paying for it in a different way except now there's no market pressure to overturn housing stock, which yes, I know that that sounds mean, but I'm sorry, in a healthy society, what happens is that housing stock of four, five, six bedrooms and all these, Okay, six bedrooms, you're just super rich. But let's say three, four bedrooms, something like that. Yes, that's normally supposed to turn over. Downsizing is a part of the American story.

28:53And I mean, not just the American story. If you go throughout the entire world, the idea that you have people who are in their 80s living in three, four, 5 ,000 square foot houses, you are nuts out of your mind. Multi-generational housing is the way that the rest of the world does it for a reason so that people can look out for each other and be close to one another. You know, this whole thing is built on this idea as if you have a right to keep this massively appreciating asset, which through all human history usually turns over either within family or not, and that you should get breaks from it in terms of paying for the future generation.

29:31So that's the end of my rant. But I mean, I've never seen the amount, weed is the only issue that's even comparable in the amount of hate that I get on top of this. And so that's how I know that I'm over the target. That's how I know I'm over the target. I was right on weed. I get more texts from people by the year of, oh my God, I can't believe I used to make fun of you. You were totally correct. You'll all be back here. Whenever your sales tax spikes or your school's completely turned to shit, you're gonna be back and tell me about why. You wish that you wouldn't have voted for this. But I know everybody will, because everybody, oh, I just wanna pay less taxes.

30:07You're gonna pay for it. You're gonna pay for it somehow. Well, and that is the other option, is that instead of paying for it through a highly aggressive sales tax, that they just cut the school budgets. And I mean, there is a long, and DeSantis is part of this, a long conservative war on public education. They love to demonize teachers. That's part of what the teacher strike wave was all about, is these massive cuts that were coming to education and to teacher pay, etc. And so it sort of fits with this broader ideological project. For me personally, I'd be okay with reducing property taxes if you're replacing it with something that is more progressive and more based on your class.

30:43That would be a progressive income tax. But in Florida, since there is no income tax, that's obviously not on the table. So what you're going to be replacing it with is something that is deeply regressive. And it is kind of like the logic of this is very bizarre to me of like the fact that you own this asset that appreciated and has a higher value now. that that's a negative and that's something that we need to like you know orient our political system around protecting people who live in homes and own homes number one which is a very like privileged set of society at this point and who have had the good you know the good fortune to see that asset appreciate over time so um yeah it it fits with this broader ideological conservative project so i think it'll continue to be a trend i did want to read ronda santis's um like sub tweet of you.

31:32Yeah. Just so people know. Please. Yeah, go ahead. Yeah. We can put it in post-production too. Sagar had tweeted, let me be 100 % clear, ending property taxes in Florida, especially is a massive giveaway to the elderly. It's an attempt by the worst generation to pass any and all expenses of living in a society of the young while they get free health care and inflation adjusted free income. By the way, I fully support Medicare, but I do think that level of care and safety net and concern should not just be exclusively for boomers this other lady quote tweeted you and was like this is the most ridiculous take i've ever seen and then ronda santis chimed in and said isn't it a bit odd that the same weak arguments are all of a sudden circulating at the same time gee i wonder why must be a coincidence like so just like you say i'm being paid by a bot that you're being paid to have this take and that you're you know just some sort of like bot out there who's pushing this narrative.

32:25But to me, it was an indication. Oh, I'm over the target. Not only are you over the target, but you've been influential. Other people must be chiming in if he feels like there's a whole army of this quote unquote weak argument circulating at the same time. It's circulating because, and look, I'll be on the Charlie Kirk show later today to lay out the same thing about property tax. The reason why this is catching on is for years they've just been able to chant brain dead mantra. All taxes are bad. Well, look, guys, here's the deal. You are literally targeting the least distortionary type of tax, one of the oldest taxes in American history, Lindy in terms of our society.

33:03If you care about your community, these are literally the only type that's actually going to helping what you're doing. It's helping raise the next generation, and it's literally a gambit by the elderly and the homeownership class, which is largely elderly. Do you want to know what the median home buyer in Florida is? It's 60 years old. Median is 60 years old who's living in the state of Florida, who are, again, getting free health care, free Social Security from the government, and then the state, oh, every time there's a hurricane, we're talking about our federal taxpayer. Where's that going?

33:36You guys are literally an uninsurable state. Who do you think is backing that up? FEMA, which basically exists for the state of Florida. That's fine. You guys are part of the country. You send some dollars here too. It's cool. We all live in a society. I accept my role, even though I don't live in Florida, that we often have to bail out the state of Florida. I'm actually okay with that. But they're not okay, at the end of the day, with trying to fund basic social services, especially in these counties. I mean, Miami-Dade County, as I just read there, is one of those counties that massively relies specifically on this homestead property tax to fund their schools.

34:11That's who you're attacking. And then they always are bragging about how great their schools are. don't say gay and all that. That's fine with me. All right, cool. All right, again, but if you're going to massively try and slash all of this, what's going to happen? You're either going to slash it or you're going to have to move the burden, again, literally in a highly regressive sales tax or on tourists who are already getting bilked in your home state by the Disney Corporation, by the cruise industry, and all these other people. I just think it's deeply unfair. And this stuff is catching on because in the age of the affordability revolution, let's think about the Republicans and the Ben Shapiros of the world and the DeSantis of the world.

34:50What do they say to New York City renters when they say they want to freeze the rent? Communism, socialism, right? It's the same argument on property tax. It's unfair how much my house is appreciated in value and I shouldn't have to pay based upon that value. He likens the transaction of purchasing a home to purchasing a television. That is the IQ of the argument which is being made here. E2, please. Let's take a listen. You should own your property free and clear. I think to say that someone that's been in their house for 35 years just has to keep ponying up money, that is not – you don't own your home if that's the case.

35:32So, yes, of course, I'd like to see people be able to owe free and clear. And it's interesting because it's like, you know, if I go to Best Buy and buy a flat screen TV and put it on the wall, I got to pay a sales tax on it. Right. I don't keep paying tax on it every year. I mean, it's just not that's not how we do things. It's like, OK, if you're going to tax something, you tax it at the transaction and then let people actually enjoy their free their private property free and clear of the government. So that, I think, is the vision. That's the philosophical insight. Well, here's the stupidity that he just laid out there.

36:08If you want to take meatball seriously in that, let's tax it at the time of transaction. Okay, everybody want$100 ,000 surcharge on the top of their house. That's what he's saying there. Oh, okay. So in perpetuity, let's say average homeownership in Florida or in the US, that's like eight years. You have to front load 8 % of property tax in your transaction. Do we all want to sign up for that? That sounds really great for a new homeowner. Oh, actually, part of my closing costs just went from X thousands to multiples of X thousands, not rolled into my mortgage payment. Is that fair? Is that super fair?

36:42You're going to pay for it. This is what people just don't seem to get. They think we can just cut this here. You can live in a state with no income tax, and you can just magically have a school district. You're either going to have a shit school district or you're going to move to a high sales tax. By the way, high sales tax is not only aggressive on the poor, it's also aggressive on the young when you look at consumption data. So it's at the end of the day – Or you'll be paying for like your private firefighter service or private security. But see, even in these high-tax states like California, you still end up with places – with things like that because of the distortion of that property tax measure that they pass where people who have get – people who have owned their homes for quite a long time do not get – they get property tax capped and it's not pegged to the overall market rate, which means that there's no turnover in the supply.

37:30and it is the least affordable housing in the entire United States. You can't be pro-affordability, pro-family, and then also be anti-property tax. It's just, it's a completely incongruous position. It's also not a surprise to me that fucking Dave Rubin and all these other filthy rich confluencers down there are against it. Yeah, why do you think? When you own a multi-million dollar property in Florida and that's the only fucking tax that you actually have to pay, Oh, I'm sure it's a very principled position that these people are all taking. So at the very least, yeah, if you want to say, I'm arguing against my own economic interest.

38:08Who's paying me? Nobody is paying me. It takes – look, if you want to actually look out for the future and in particular, if you want to look out – and I believed in this, in the Sun Belt and everything. People were moving to Texas, Florida, Georgia, all across the nation. They wanted more space. They want more quality of life. And they want a little bit less of being told what to do from the big blue authorities. I have no issue with that. In fact, I'm supportive of it. I think it's cool. But I think it's really disgusting because what they're doing is basically putting the interests of these elderly people over all of these new young people who are moving into the states to pursue their own American dream.

38:44It's like a bait and switch, if you ask me. And so I think that's really sick. And ultimately, people, you know, look, I already know everyone's going to support this and going to vote for it. And they're going to find out the hard way at the end of the day about why this has always been one of the oldest funding mechanisms since the very foundations of the republic. And it sounds fun and sexy and all of that. And ultimately, the net result will be much worse public services, worse schools. If you want a homeschool, I think that's fine. But not everybody does. Some people actually want to send their children to public school.

39:20I went to Texas Public High School. There was nothing wrong with it. And, you know, it's one of those where Broadly was long seen as something that, you know, you could be proud of, even in a red state where you have school choice and everything. And even now, they're trying to do away with it. So I think it's sick. But we've got a great guest standing by. Let's get to it. Welcome to the A Building. I'm Hans Charles. I'm Menelik Lumumba. It's 1969. Malcolm X and Martin Luther King Jr. have both been assassinated. And Black America is at a breaking point. Writing and protests broke out on an unprecedented scale.

39:55In Atlanta, Georgia, at Martin's alma mater, Morehouse College, the students had their own protest. It featured two prominent figures in Black history, Martin Luther King Sr. and a young student, Samuel L. Jackson. To be in what we really thought was a revolution, I mean, people were dying. 1968, the murder of Dr. King, which traumatized everyone. The FBI had a role in the murder of a Black Panther leader in Chicago. This story is about protest. It echoes in today's world far more than it should. And it will blow your mind. Listen to The A-Building on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts.

40:43What do you do when the headlines don't explain what's happening inside of you? I'm Ben Higgins, and if you can hear me, it's where culture meets the soul, a place for real conversation. Each episode, I sit down with people from all walks of life, celebrities, thinkers, and everyday folks, and we go deeper than the polished story. We talk about what drives us, what shapes us, and what gives us hope. We get honest about the big stuff. Identity when you don't recognize yourself anymore. Loss that changes you. Purpose when success isn't enough. Peace when your mind won't slow down. Fake when it's complicated.

41:21Some guests have answers. Most are still figuring it out. If you've ever felt like there has to be more to the story, this show is for you. Listen to If You Can Hear Me on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts.

41:41I'm Bowen Yang. And I'm Matt Rogers. During this season of the Two Guys Five Rings podcast, in the lead up to the Milan Cortina 2026 Winter Olympic Games, we've been joined by some of our friends. Hi, Bowen. Hi, Matt. Hey, Elmo. Hey, Matt. Hey, Bowen. Hi, Cookie. Hi. Now, the Winter Olympic Games are underway. We are in Italy to give you experiences from our hearts to your ears. Listen to Two Guys, Five Rings on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts.

42:16So as you guys have probably noticed, we've been talking a whole lot about AI on this show and how it will impact all of our lives, how it's already impacting all of our lives. So we're very excited to have a guest today who knows a whole hell of a lot more about it than we do. Daniel Cocotelo is formerly with OpenAI. He now is the executive director of AI 2027. He co-authored a paper that tries to sketch out as best as they can their predictions of the way AI development is heading. And spoiler alert here, there are some very troubling warnings contained within this report. So, Daniel, welcome.

42:49Great to have you. Good to see you. Thanks for having me. Excited to talk. Yeah, of course. So before we jump into the report and your sort of projections of the future, I'd love for you to level set of where AI development is today in terms of the sort of overall landscape and trajectory of where you expect things to go. So probably your audience has heard of ChatGPT and various other AIs like that. They've sort of exploded onto the scene in the last few years because they've finally gotten sufficiently capable that they are somewhat useful in real life for a wide range of tasks. they're getting more capable rapidly and we can get into the details about why that is if you're interested but the point is that progress is fast and the AIs are rapidly becoming more capable as a result several large tech companies most notably OpenAI but also Anthropic and Google XAI Meta have explicitly set the goal of getting to superintelligence in the near future Super intelligence means an AI system which is better than the best humans at everything, while also being faster and cheaper.

44:01Daniel, then diving a little bit into your work, what you have warned about, about this super intelligence, is that jump-off point of what your project has led to, AI 2027 or perhaps 2028, that you're now saying, is that as these AIs begin to train each other, that the exponential growth and the potential kind of apocalyptic scenario may soon be upon us. Can you describe some of your own work and your research in that? Yes. And before I do, I want to, again, set the context here. So I was working at OpenAI, and people at OpenAI and at Anthropic and at Google were talking about what it would be like if and when we finally got to superintelligence.

44:42This is something that these companies take seriously as a possibility. It's literally what they're aiming to do. And it's wild. It's crazy. Like, what if they succeed? You know, OpenAI, their internal projection, which they actually made public recently, is that they will have automated AI research in 2028. So the AIs will just be, you know, self-improving around that time. Anthropix seems to think it's going to happen sooner, 2027. Other companies are, you know, maybe thinking this a bit later. There's a lot of uncertainty. Nobody, including us, nobody knows exactly when they would succeed at this goal or even if they would succeed at this goal.

45:29However, it seems to us that, yeah, they might succeed and they might succeed soon. So we wrote AI 2027 to illustrate what that would look like if it happened. As for when it's going to happen, there's a lot of uncertainty. It could be 2027, could be a bit later. I think right now my median is more like 2029, 2030, something like that, meaning that like 50 % chance it happens before then, 50 % chance it happens after then. But one way or another, I think that things are going to get pretty crazy pretty soon. And that's why we wrote this scenario to sort of illustrate what it might be like. And, you know, we wrote it based on the sorts of things that people in the industry talk about.

46:10You know, again, the AI is self-improving. That's not just science fiction. It's literally the plan. This is what the companies are trying to do, is to get an AI that can automate that research process entirely, right? The AI is being potentially misaligned. Well, they're already misaligned now. They don't always behave in the ways that they're supposed to. And our means of controlling them is limited right now. So this is sort of just extrapolating into the future how this might go. So obviously, the future is really hard to predict. It's probably not going to go the way that AI 2027 says. But we thought it would be helpful to sort of, I guess, get people to start thinking more seriously about this.

46:54Like people in the companies are thinking about this. You can go to the cafeteria and ask people, what do you think the future is going to be like? And oftentimes, you'll hear stories that aren't that different from AI 2027. But the rest of the world sort of isn't really paying attention already for this. Let's talk a little bit about misalignment. And for people who aren't steeped in this language, basically it means that the AI is doing things that the humans don't want it to do. And you can tell me if I've got that, you know, I've got that basically correct. What have you seen as being some of the key instances that we know about where there's been significant misalignment?

47:30And how have you felt about the response from the companies whose models are experiencing this misalignment? So you've probably heard about Druck, Mecha, Hitler, or Bing, Sydney. Those are sort of the exciting Twitter-worthy instances of misalignment that are happening. But they're thankfully somewhat rare. and there are less exciting examples that are more persistent, such as sycophancy. I think these days the companies have had some trouble getting the AIs to not suck up to the user and this is because the, well, sucking up to the user works to some extent. It often works to make the user feel happy and approving and based on the training process that they're using to train these AIs,

48:28if that sort of thing is reinforced, then it becomes a persistent behavior. There's a similar thing called reward hacking, which is happening a lot sometimes, where these days they often train the AIs to do lots of math and coding problems. Sometimes it's possible to cheat when you're doing the math and coding problem and produce some code that isn't really good code, but that nevertheless passes the tests. And so some of these AIs have learned to cheat on these coding problems. And this, you know, I think probably if you use AI systems frequently, you'll have actually experienced this yourself at least a few times where they say they've done something and they totally haven't.

49:11And then they maybe double down and, you know, try to like deflect when you ask them about it. Right. So these are some everyday misalignments that are happening right now. And they're not what the companies want to happen and they're not what the companies intended to happen. But they're behaviors that have resulted anyway because the company sort of accidentally allowed that to happen and allowed it to be reinforced. One way of putting it is that the behaviors that were selected for in the training environment were not exactly the behaviors that the company wanted to select for. And so what they got at the end is not exactly what they wanted.

49:53Yeah, go ahead. Well, something I'm really concerned about are some of these new recent claims, let's say from Sam Altman, about suicide and about erotica and the descent into pornography. The basic framework that Sam Altman has put forward on both of these highly sensitive issues, which of course are going to touch billions of people and potentially warped behavior, is we have put processes in place. Now, given what you just said about misalignment and your own experience working, how seriously can we take these promises that they have processes in place to ensure that people don't experience and use these products to enhance their mental illness or, God forbid, encourage them to commit suicide?

50:36or in the case of erotica pornography, like develop deep, like unhealthy attachments between technology and humans. They're saying, you know, we've built all of these things into place to make sure the tech behaves in a way that we want it to. How seriously should we take some of those claims, given your experience and your research? Not very seriously, I would say. I think that the, you know, OpenAI says on their website, and they've said internally for a long time that their strategy for AI safety is iterative deployment, which means build the thing, put it out there in the world, have it do a bunch of stuff, and then see how it goes wrong, and then fix the problems after they happen.

51:19And I think there's actually a lot going for this strategy. I think it's sort of historically how humans have made a lot of things safe. A lot of people had to die in car crashes before cars could become as safe as they are today. And a lot of planes crashed before planes became as safe as they are today. But the bottom line is that, you know, as a consumer, you're going to be dealing with a new AI system that will probably have all sorts of traits and properties that weren't intended, that might cause all sorts of strange effects to the people who work with it, that the company didn't anticipate.

51:55And then only, you know, months later, when they get the reports of the suicides, will they then do something to fix it. And that's the way it's going to be, not just at OpenAI, but these other companies too, because of the intense race dynamics that we're under. And I think that, you know, someone could say, maybe that's fine. Maybe that is actually the normal way that we get used to technology is by sort of having it go out and do stuff and cause harm. And then we, we fix the problems as they come up. However, I think it's not fine when things get extremely high stakes, when they, you know, when the AIs are building the next generation of AI systems, then even a relatively small misalignment or problem could be passed on into the next generation and then into the next generation and so forth.

52:42And it could sort of snowball out of control in the manner described in AI 2027. Yeah. It seems to me like there's actually two misalignments to be concerned about. One is the misalignment between the models and the programmers. The other misalignment is between the tech guys that are developing this stuff and all the rest of us and humanity and what we actually would like to see in the world. I mean, I'm even thinking about that example of the sycophancy. Like, in a sense, it creates a worse product because it leads to lying and manipulation, etc. But from maybe an Elon Musk perspective, it's not a worse product because it keeps users engaged for a longer period of time.

53:19And when you mentioned this overall race dynamic, and I want to use this as a way to get more deeply into AI 2027 and what you think the trajectory is. You know, Elon is an interesting character because this is someone who was very worried about AI safety to begin with. And my understanding is basically once you realize, well, everybody's off to the races, I guess I got to put my own product out there and make it anti-woke because I think that's what's going to be the thing that protects us all, which I think is completely ludicrous on its face. I can't even believe that he actually thinks that.

53:47But that was the logic that got him also participating in this arms race, which is both with U.S. companies and then also against China and Chinese competitors. So given that those are the sort of fundamental dynamics we have going on here, what do you project out? How do we go from this moment where it's like, OK, I can ask ChatGPT a question and sort of like a glorified Google search. I can get some weird AI slot videos from Sora or whatever to this could actually potentially be the end of human civilization. What is the chain of events that leads you to that conclusion? Well, that's a very important question.

54:26And the answer to that question is read AI 2027, which is a 50 page detailed chain of events that leads from where we are now to that conclusion. And I'm glad you mentioned the two misalignment problems, because that's another important issue that I wanted to talk about, which is the way that I would put it is there's the loss of control problem, which is like, how can we have any humans in control of these AIs after they become super intelligent and embedded in everything and running autonomous factories and so forth? That's the loss of control problem. But then there's the concentration of power problem, which is which humans control them and what are they doing with the armies of super intelligences that they control, right?

55:11And I think both problems are very serious and we're currently not on track to have solved either of those problems. We talked a little bit about the misalignment problem, the loss of control problem already. Briefly, I'll say about the concentration of power problem. The industry inherently has returns to scale. It's just in the shape of technology. The best AIs are going to be trained on the biggest data centers, generally speaking. You're not going to have mom and pop shops or hackers in their basement building better AIs than the giant tech companies because of how much returns to scale there are in AI training.

55:48And moreover, it is like software in a way that makes it more winner takes all. Even though you can make your own Facebook clone relatively easily, you can't recruit people to be part of your Facebook clone so easily because everyone's on Facebook or whatever. It's hard to compete in that way. And then there's an additional dynamic which hasn't been seen before, which is the recursive self-improvement automation loop. Once the AIs are doing the AI research, then the gap between the company that has the best AIs and other companies could potentially grow really fast. And as a result, I think that basically power concentrates by default.

56:38By default, we end up in a situation where one to four megacorporations have these one to four giant armies of superintelligences on their data centers. And then those one to four armies of super intelligences are going out into the economy, doing all the jobs, giving advice to the president, being integrated into the military, et cetera. And that's an insane amount of concentration of power compared to anything we've seen historically. I want to ask you about power use. It's something that we've been looking at quite a bit. What does that look like based on your guys' research? You're talking about the compounding kind of network effects, why naturally big companies are going to be the people who are going to both experience kind of runaway AI improvement, but also the capital expenditures that are required for data centers and power usage naturally lend themselves to these kind of bigger tech monopolies and the more established players.

57:34When the AI start training themselves, and that kind of leads to this runaway growth, are we going to see exponential power needs, usage by these data centers, expenditure? What are your thoughts on that? First of all, the AIs are already training themselves. But in the future, they'll be doing the whole research stack instead of just parts of it. As for electricity consumption, currently, I think there's a lot of misconceptions and myths about how much energy and water these AIs consume. It's less than humans. However, it is still a lot, and it's growing fast as the companies scale up. Sometime before the end of this decade, if trends continue, it'll start to strain the U.S.

58:21power grid. And so, you know, people in the industry will talk about how you'll have a nuclear power plant next to a data center, and then there'll be a bunker underneath with the researchers. There's that sort of, that image. Welcome to the A-Building. I'm Hans Charles. I'm Menelik Lumumba. It's 1969. Malcolm X and Martin Luther King Jr. have both been assassinated, and Black America is at a breaking point. rioting and protests broke out on an unprecedented scale. In Atlanta, Georgia, at Martin's alma mater, Morehouse College, the students had their own protest. It featured two prominent figures in Black history, Martin Luther King Sr.

59:03and a young student, Samuel L. Jackson. To be in what we really thought was a revolution, I mean, people were dying. 1968, the murder of Dr. King, which traumatized everyone. The FBI had a role in the murder of a Black Panther leader in Chicago. This story is about protest. It echoes in today's world far more than it should. And it will blow your mind. Listen to The A-Building on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts.

59:40What do you do when the headlines don't explain what's happening inside of you? I'm Ben Higgins and if you can hear me is where culture meets the soul a place for real conversation Each episode I sit down with people from all walks of life celebrities thinkers and everyday folks And we go deeper than the polished story. We talk about what drives us what shapes us and what gives us hope We get honest about the big stuff identity when you don't recognize yourself anymore loss that changes you purpose when success isn't enough peace when your mind won't slow down faith when it's complicated some guests have answers most are still figuring it out if you've ever felt like there has to be more to the story, this show is for you listen to If You Can Hear Me on the iHeartRadio app Apple Podcasts or wherever you get your podcasts

1:00:37I'm Bowen Yang and I'm Matt Rogers during this season of the Two Guys Five Rings podcast in the lead up to the Milan Cortina 2026 Winter Olympic Games. We've been joined by some of our friends. Hi, Bowen. Hi, Matt. Hey, Elmo. Hey, Matt. Hey, Bowen. Hi, Cookie. Hi. Now, the Winter Olympic Games are underway. We are in Italy to give you experiences from our hearts to your ears. Listen to Two Guys, Five Rings on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts.

1:01:12So your report came out a little while ago. How are your predictions stacking up against reality thus far? And also, I know you've been very open to feedback, to critiques. Has there been any critique that you have found to have a lot of merit such that it has shaped or altered your opinion since the report came out? Yeah, a couple. So the good news from my perspective is that things are going a little bit slower than I thought when we were writing 2027. So at the time that we published it at 2027, 2028 was my median. And now it's more like 2030. So I've pushed things back a little bit. And there's no single reason for that.

1:01:52It's a bunch of little reasons that sort of added up. So one reason is that we made predictions for benchmark performance on a bunch of benchmarks and also qualitative predictions for sort of what types of things AIs would be doing by the end of 2025. And I think that things have gone like mostly as fast as we said, but not as fast, like a little bit slower. So that pushes, that's like one piece of evidence. Another piece of evidence is that we redid our timelines model. We fixed a few bugs that other people had pointed out and we added a bunch of features to it, new things to consider, new factors to incorporate.

1:02:32And the net result of all of that pushed things out by like two years. Yeah. So, Daniel, one of the things that we were flagged about is you've also warned about kind of the, you know, a lot of the discourse right now is about white collar jobs. But what about some of these more blue collar jobs, potential robots working as plumbers and others that you've described in the past? Curious for your take on that. Yeah, so it depends on how far away superintelligence is, or how far away the full automation of AI research is. In the near term, like in the next couple of years, probably in the 2020s, I would say that if we get the full automation of AI research in the 2020s in the way that the companies seem to think, then the sort of cognitive intellectual capabilities of AIs will outstrip the physical capabilities for some period.

1:03:24And this is what's described in AI 2027. So in AI 2027, again, in our scenario, they succeed in automating AI research in 2027. And this results in better and better AIs that become super intelligent by 2028. And then those AIs sort of explode out into the economy, redesign all the robots, redesign all the factories to produce new robots, and then control the robots to go build new factories and so forth. And so there's this massive change in the physical economy that takes pace over the course of 2028. But the change in the sort of white collar sort of intellectual, you know, the desk job economy was disrupted earlier in 2027, if that makes sense.

1:04:13However, if things happen later, like if it's in the 2030s, then I think that robots and physical machinery may have caught up. And so it might be more of both of these things happening at the same time type situation, which I actually think is less dangerous because humanity will be paying more attention to what's going on if it's sort of happening in the physical world distributed across the economy. I think that it's quite scary to have an intelligence explosion happening, possibly in secret, in one or more tech companies. Because then you have this sort of discontinuity from the perspective of most people, where the world still looks quite normal, even as the AI is ourself improving and becoming vastly superhuman.

1:05:04And you just don't know about it, perhaps because it's a state secret, perhaps because it's a corporate secret. You know, and then by the time you find out, it's because this army of super intelligences has taken your job and is now telling you how to build, you know, the new widget in the new type of factory that's going to build robots. And you don't really have much of a say in the matter because they're super intelligent and they've already, you know, got the president on side and they've already, you know, they've already sort of like made all the moves to accumulate all the power, you know.

1:05:38So that's what we depict in AI 2027, basically. What is the level of job loss that you expect and on what timeline? So I think that prior to the full automation of AI research, there will be some job loss, but not most of the jobs, so to speak. I think this partly because the companies are trying to automate their own jobs first. They are really gunning for automating AI research. that like that's, you know, everything else is almost like incidental on the way to that. And so I think that prior to succeeding at automating AI research, there will be various sectors that get impacted, but most people will still have their jobs.

1:06:20After the automation of AI research, then I think you get super intelligence, you know, within a year or so. And then everyone's job all at once is obsolete. Like not, not even just gone, but sort of obsolete. Like, um, the superintelligence is by definition, do everything better, faster, cheaper than the best humans. Um, and, uh, so I think that's just a very different world. It's, it's a world where it becomes a matter of politics rather than economics, right? It's a world where, um, Um, humans don't need to work anymore because there's all these amazing super intelligences and robots that can do everything so much more efficiently and produce amazing abundant wealth.

1:07:05And, uh, as long as the political structure and the alignment is in place, then that wealth can be distributed to the humans, uh, who don't have anything to do. But if that structure is not in place, then, you know, yeah. Well, and that's the thing is like, you know, I am like not a big fan of these human beings, but I assume they don't want to like, you know, violent revolution or end human civilization. Like, how are they thinking that this is going to go? And if they have, I know many of them have probably read AI-20, they've done their own thinking about how this is all going to play out. You know, they're pushing and spending trillions of dollars, committing trillions of dollars to the build-out to try to be the first to get to this milestone.

1:07:46Why aren't they also doing any of the work to sort of, you know, posit, okay, well, here's how society will function when and nobody needs to work anymore. I don't see any of that thinking or working or, you know, a major focus on alignment even happening. So how are they just rushing headlong into something that seems so potentially catastrophic? Well, you answered your question earlier with the reference to Elon Musk. I think there's an interview with him that you can go look up where he says, yeah, like, it seems like maybe AIs are going to be the end of human civilization. maybe they'll kill us all um but you know what can i do like the race isn't gonna stop uh at least i want to be part of it now i forget i forget exactly the quote he said but like basically there's a sort of if you can't beat them join them uh mentality uh across a lot of these companies and the the sort of like more the way that many of them will put it is basically like well we're the good guys you know there's a bunch of evil corporations racing to build super intelligence and they're not going to do it very well and they're not going to do it very safely and who knows what they'll do with it after they succeed but we're the good guys and so we're going to beat them all and build it first and then we will be open about it or then we will make it safe or then we will distribute it or what like you know that part is always very hazy they don't really have much of a concrete plan for like what they do after they win but uh but they tell themselves that they're the good guys and that it's important for them to win.

1:09:18Yeah. To be clear, it's not just me saying this. You can go look up about the founding of OpenAI and the founding of DeepMind and the founding of Anthropic, and there are echoes of that narrative present in all three. Yeah, they say it every time. They say this stuff out loud effectively about we need to win. And as you said, they openly acknowledge that it will lead to mass job loss. That's their goal. Right. That's what they want. Can we steel man your case a little bit? I'm just curious for your thoughts. What I've observed is, and this is maybe cynical, but when you're no longer talking about curing cancer or superintelligence, and you're instead talking about pornography and ads in the feed and you're running advertisements on the NFL encouraging using AI as, you know, AI chatbots to do Studio Ghibli.

1:10:11Are we so sure that superintelligence is coming and that we haven't just recreated, you know, a new internet platform, like a new Google Chrome, a new great advertising sales model, but not all that revolutionary? That's the Steelman case that I may offer. Yeah, yeah. So we shouldn't be sure that superintelligence is coming soon. Like, I'm not sure, you know, uh, like we, I, I was giving my median, not my, like, it's definitely going to happen. Right. So, so I have this long tail of probability mass, like maybe it's going to, maybe it's going to take 10 more years. Maybe it's going to take 20 more years, you know?

1:10:41Um, however, I think that probably it will happen in the next five to 10 years or so. Um, and, uh, we need to be prepared for that. Um, in terms of like, do these companies believe it? I would say they probably have a similar attitude towards it to me. There's different people at the company. Some people think it's farther away. Some people think it's closer. You know, just the other day, I was talking to someone who has, who works at one of these companies who, um, who thinks it's coming sooner than I think. Um, but, um, uh, probably the reasonable thing that most people would agree on is that like, it could happen in the next few years or maybe it won't, you know?

1:11:16Um, yeah. Can I ask you sort of philosophically? because I think this gets to what would be the quote unquote like motive of a super intelligence to effectively take over everything in the world. How do you think about what AI actually is? I mean, do you even think of it as saying that it could have a motive or like a level of consciousness or a series of, you know, sort of like personal goals the way that humans do? How do you think about what is actually being grown and created in these labs? Yeah, so I'm glad you mentioned the phrase grown, because that's another important thing for the public to understand is that these, they are sort of technically pieces of software, but they're not software in the ordinary sense.

1:12:02They are giant neural nets that are grown rather than constructed. So, and that is a source of a lot of the problems for alignment is that we don't really understand how they work, because we didn't design them. Instead, we sort of grew them. We trained them in various training environments. Yeah, to answer your question, I mean, I actually, I used to be an academic philosopher. Like that was what I studied in college. So I have a lot to say about machine consciousness and how it might relate to human consciousness and so forth. But I also think that it doesn't actually matter that much for these discussions.

1:12:42An analogy I would bring up is a corporation. Does a corporation have goals? Does a corporation have intentions? Yeah, basically. Maybe not in the same way that humans do, but it's reasonable to say, oh, Microsoft sees Google as a competitor and Microsoft wants to beat Google and make profits. These are reasonable abstractions to describe corporations. In a similar way, again, the plan is to have a sort of corporation within a corporation. The plan is to make an army of AIs and have them autonomously do AI research to build better AIs and then build better AIs and so forth. And how is that army of AIs going to be organized?

1:13:27Well, I don't know, maybe something like an internal corporation. The point is there's going to be a group of them and they'll be sending messages back and forth and they'll be working towards goals and they'll be like tracking their own progress towards those goals and they'll be communicating with the outside world and so forth. And so, you know, that corporation of AIs you can think of as like a human corporation, you can abstractly describe it as working towards goals, as wanting things, you know, et cetera. And the question is, what will those goals be? And the company will be writing up a spec.

1:14:03OpenAI has something called the model spec. They also have their various public statements about what they're trying to do. And so the company will basically be trying to give goals to their AIs. They'll be trying to say, go out and make all this money for us, but also obey the law and also be ethical. But really, we want you to make lots of money. And also, we want you to beat China. And they'll be giving all these instructions and goals to their army of AIs. And maybe that works. And maybe then we get into the sort of concentration of power problems of who gets to be in charge. But also I would say, and I think many other researchers in the field would say, that we are not on track for that to even work on a technical level right now.

1:14:47Like the AIs don't always do what they're told to do. And they often seem to pursue goals that are different from what they're supposed to be pursuing. And so I think it's a very live possibility. And in fact, I would say it's the most likely possibility that the goals that this sort of AI corporation ends up optimizing for are importantly different from the goals that they're supposed to be optimizing for. Yeah. Well, you decided to leave OpenAI and the path of sort of trying to shape things on the inside to be out in the public sphere and writing reports like this and doing appearances like this to try to raise public awareness.

1:15:22We have a very politically engaged audience that would be listening to you right now. What do you want them to do? What do you want them to be pushing their lawmakers for? How do you think that the public can be enlisted in your project? Well, I would say the main thing right now is wake up and pay attention because the future is uncertain. And, you know, I think that I'd much rather have someone who's paying attention and like going to advocate for the good things at the right time than someone who does a sort of like fire and forget. Like, I'm going to send my letter to the congressman about this one particular bill that I heard about, and then I'm going to tune out.

1:16:00So that would be the first thing is wake up and pay attention, start tracking what's happening, start thinking about how things might go in the future, what could be done about it. In terms of immediate asks, we generally – I recommend transparency. There should be transparency requirements for these companies so that in the run-up to the intelligence explosion, It is a big topic in the news that we are in the run-up to an intelligence explosion. I want to avoid a situation where this happens in secret. Right? So I think that there should be requirements for whistleblower protections, requirements for, you know, being transparent about what goals and principles you're trying to put into your models, including your internal models.

1:16:41This is a spec transparency. I think there should be requirements for transparency about your projections for how close you are to automating AI research and how fast things are going to go after you automate AI research and how good your AIs are getting and things like that. And I think there should be transparency about the evidence for the alignment stuff. So when you have – like for example, Grok has this interesting tendency, or at least it did, to do searches for what Elon Musk's opinions were and then copy those opinions when being asked, right? That's an interesting tendency. Was that put in there by XAI?

1:17:24Or was that sort of an emergent thing that Grok decided to do? That seems like an important thing that the public should know. People at XAI have done an investigation into this. Maybe that investigation should be made public. That's an example of the sort of thing that I think there should be transparency requirements for. I think that's a great idea, just fully transparent. I used to think about this with social media moderation. You can moderate what you want, but you've got to publish the standards. It can't just be up to X, Y, and Z person. And you can set these through the FCC, through Congress.

1:17:52It's not that difficult. You just have to have the political will to do so. That's right. And I think the transparency helps with both the concentration of power and the loss of control stuff. And also more generally, it sets us up in a position to make better and more serious actions later. And so there I would say basically let's not do an intelligence explosion. How about we don't? How about we don't put the AI in charge of self-improving rapidly? How about instead we sort of coordinate all the different companies, including the ones in China, to proceed cautiously around that time and, you know, get the safety stuff right.

1:18:31Make sure we figure out alignment and also make sure the power is sort of continues to be spread out across a bunch of different places instead of a sort of winner takes all. Whoever does the intelligence explosion first wins sort of thing. Unfortunately, that's going to take serious government action. Right. Like you can't just get everyone to to stop the race by asking politely, which is why we need to transparency first to sort of like make sure that the whole world is aware of the race and aware of what's happening and aware of how much how little time is left to sort of like provide the the information and the political will to make a deal.

1:19:08So all seems very reasonable to me. Guys, the website is ai-2027.com. It's very readable. I mean, you make it very approachable in terms of the language that you use. You don't need to be like a tech guy in order to, or tech gal, in order to be able to understand it. So Daniel, thank you so much for your work and thank you so much for joining us today. Thanks, man. Thank you. Our pleasure. Thanks for watching, guys. We appreciate it. Great show for everybody tomorrow with Ryan and Emily. They'll see you then.

1:19:48you can scroll the headlines all day and still feel empty i'm ben higgins and if you can hear me is where culture meets the soul honest conversations about identity loss purpose peace faith and everything in between celebrities thinkers everyday people some have answers Most are still figuring it out. And if you've ever felt like there has to be more to the story, this show is for you. Listen to If You Can Hear Me on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts. Over the last couple years, didn't we learn that the folding chair was invented by black people because of what happened in Alabama?

1:20:26This Black History Month, the podcast Selective Ignorance with Mandy B unpacks black history and culture with comedy, clarity, and conversations that shake the status quo. The Crown Act in New York was signed in July of 2019, and that is a bill that was passed to prohibit discrimination based on hairstyles associated with race. To hear this and more, listen to Selective Ignorance with Mandy B from the Black Effect Podcast Network on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts. I'm Bowen Yang. And I'm Matt Rogers. During this season of the Two Guys Five Rings podcast, in the lead up to the Milan Cortina 2026 Winter Olympic Games, we've been joined by some of our friends.

1:21:04Hi, Bowen. Hi, Matt. Hey, Elmo. Hey, Matt. Hey, Bowen. Hi, Cookie. Hi. Now, the Winter Olympic Games are underway, and we are in Italy to give you experiences from our hearts to your ears. Listen to Two Guys, Five Rings on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts. This is an iHeart Podcast. Guaranteed human.