How To Actually Regulate AI — ft. Alex Bores

0:00Support for the show comes from alumni ventures. We've said it before, diversification is our Kevlar Vest, but do you know how concentrated your public stock fund actually is? A handful of mega-cap stocks with an AI story are doing most of the work. That's not the diversification you think, especially when so much value creation is happening before companies ever go public. Alumni ventures ranked a top 20 US venture firm by both time and CV insights gives a credited investors a smart simple way to assemble a blu-chip venture portfolio and participate in the value creation that is occurring in private markets. You can sign up for free through the weekly deal flow and decide for yourself or write one check into the Alumni Ventures Foundation Fund for a diversified portfolio of about 25 private venture deals. Go to ev.gc-proffg. Not an offer to sell securities, venture capital involves substantial risk including loss of capital invested, seed disclosures and fund offering materials for more information. From the Goldman Sachs trading floor in 10 minutes or less, investors and

1:02analysts share timely analysis on the week's market activity. The market's podcast, Frink Goldman Sachs. Listen now. Your heart can tell you a lot about your health. Apple Watch Series 12 measures your heart rate every five seconds with the most accurate heart rate sensing and awareable. So your vitals app now with heart rate variability can tell you when something is off and your readiness score can let you know when to rest and when to push. Here the story in every heartbeat with Apple Watch Series 12. The features described are for wellness purposes only and not for medical use. iPhone 11 or later require based on Apple conducted study of heart rate accuracy August 2026. Visit apple.com slash Apple Watch Series 12. Today's number 27 billion dollars. That's how much Mark Zuckerberg's net worth increased Monday. Ed, you know how Instagram models get paid as a per DM. Per DM. Yeah, you should call yourself an Instagram influencer because it just sounds better than unemployed.

2:11How are you? I'm out of New York. I'm headed to Colorado tomorrow and then I got a Santa Barbara on Friday for a speaking gig and then I go back to Colorado for the conference and then I don't know. I'm doing pretty well though. It's great to be in New York. This is a golden age ad. You're you're living in a golden age. You don't realize it because you're young and you're used to everything being great despite the mass extinction possibilities. The New York's amazing. You think this is the golden age of New York? I think it's 20. You never realize you're in a golden age when you're in it. I think in 20 years we're gonna look back on Manhattan and think that was a golden age. Wow. Why why now? The crush of humanity, the creativity, the number of new restaurants, new things to do, culture. You think that's new? I feel like New York's been that. I feel like the main changes that were a lot more like sweet greens and and cover restaurants and like slot balls. Well, just one crime. When I first moved to New York right out of UCLA,

3:11you'd go downtown and if you're with a woman, you walked her to a cab. You didn't take you did not take the subway at night when I moved here. You were if you were in Manhattan for longer than a year at some point, you probably got I mean it was just an entirely different situation. There weren't nearly every neighborhood in Manhattan now. It is kind of cool and has interest in interesting vibe to it. 65 million people visited New York, 85 million and economic activity, 400,000 jobs. There's more people attended Broadway last year than all local sports teams combined. We've had a net job creation since 2020 of 1.1 million jobs, granted most of that has been healthcare. But it used to be 54% of the GDP of New York of New York state now. It's 60%. I mean, it's this is a golden age. Are you running for mail? I am not. I am neither talented nor likable nor very good on socials. So I don't qualify. But New York is

4:14I mean, I didn't appreciate it when I was your age, but you're living you're literally living at the helm of the Bob sled of a capitalist society right now living in New York. I mean, there's always a bot. Inflations at 4.6% versus 3.4% nationally, rancor up. So the deep, it's kind of indicative of a metaphor for the larger economy. And that is if you're in the right place in the right time and the right job, you're just living an amazing life. If you're the kind of person that actually has to show for jury duty, all you know is your rent's going up. So it's more difficult. The city used to be a place where you could be upper middle class safely and dance between the raindrops and live here. I don't think that's I don't think that's any that's longer any true. I think either you have a great job that pays really well or your parents are putting through you through New York. I just couldn't I was making a really good living here in in 2010. I decided to move to Florida because I just couldn't afford to live here with kids. But anyways, it's it's doing really well. What did you oh, you had weddings this weekend? I had engaged an

5:19engagement policy, which is very fun. I had a wedding the weekend before that and then this weekend will be one of my first wedding free weekends in a long time. So I will be chilling at home and enjoying New York and embracing the golden age. So not much on the agenda for me, which is great. I need a I needed a chill weekend. You should practice being married and be home with your girlfriend and scream, I am not made of money. And and then to have her tell you how much she doesn't like your parents, just practice this weekend. Okay, just practice practice being married. We'll do a little bit. We'll do a little bit of roleplay. Okay. Well, with that, let's get into this conversation. There is growing anxiety about the potential risks of AI. Researchers and AI leaders have warned about everything from cyber attacks to biological weapons. And at the same time, policymakers are still trying to figure out how governments should respond to a technology that is advancing faster than the laws designed to regulate it. Our next guest has been thinking about

6:22that problem for years. In fact, he is one of the few lawmakers who has actually helped turn AI safety concerns into law. In 2025, he sponsored New York's RAIS Act, which requires large developers of frontier AI models to establish and disclose safety and security protocols that law was signed by Governor Kathy Hokel in December. But that legislation also put a target on his back when he ran in the Democratic primary for Manhattan's open house seat. He made AI policy a central part of his campaign as a result, a super PAC spent eight million dollars on attack ads targeting him. Now he isn't blocking on a new venture aimed at shaping how Democratic presidential candidates and voters think about AI. We're going to talk to him about how he thinks AI should be regulated and what role the government should play as the technology advances. This is our conversation with Alex Bors, the co-founder of Huda Science. Alex, great to have you on the show. We're going to get

7:23into everything that's been discussed about AI and AI safety. This debate has really blown up around the world. But I think it might be helpful to start with where you sit in all of this. You've been thinking about this stuff for a while. How did you get to this position where you're now running this nonprofit focusing on specifically these issues? What is your background? Thanks for having me. My background is pretty atypical for someone in politics. I spent my whole career in tech, mostly government tech. I was at Palantir for four and a half years and then a couple of startups after that. When my seat in the state assembly opened up, I ran for it. Mostly as an experiment to see if that were a way of making change. I was always talking about how I was trying to fix bad policy downstream with tech. Here was a chance to go upstream and try to design it right in the first place. I ran, I won, and have worked on a wide variety of issues there. When I was elected, I became the first of my party, the first Democrat elected in New

8:27York State at any level with a degree in computer science. That wasn't all that I worked on. I've passed 35 bills in my four years. I think maybe five of them had to do with AI and technology. One of those in particular, the Raz Act, really got the eye of the industry and I think set everything else in motion. At what point did you realize or decide that AI was a significant issue? Why did you decide to make it the central component of your campaign? You didn't win, but you came second. I think you've made a pretty surprisingly good run considering some of the other candidates and their public profiles. When did AI become central to your understanding of policy making? In terms of a policy, important policy question, I knew it for quite a while. The speed with which this was coming was evident. The chat GPT was released basically a week after Election Day and dominated a lot of conversations.

9:33That my entire time in the assembly. I actually wasn't planning to make it a big part of the as a political issue, a big part of the campaign. That was thrust upon me. My theory going in was, I've passed the strongest AI safety law. In the country, I have a master's in computer science. If AI is your issue, you're going to vote for me. I'm going to go into this campaign and talk about everything else. Three weeks after I declared leading the future, the Super PAC funded by Andreessen Horowitz and Greg Brockman of Open AI, named me their number one target in the country. A couple of weeks after that pledge to spend at least $10 million against me and started to do it. My neighbors were getting mailers once a week with something about some made-up thing about me. They were getting text messages every day. Voters are coming up and asking me, hey, who's spending so much money against you? Why are they spending so much against you? That elevated AI in our race a few months before now that it's elevated everywhere.

10:39So this Super PAC backed by Andreessen Horowitz and also by Greg Brockman of Open AI. They spent nearly $8 million to beat you. This was after the Raz Act was passed, which you helped to author, and which specifically addressed AI. Why were they so worried about you? What was it about your proposal and your regulation? What was it about the Raz Act that made them decide, okay, we spend millions of dollars to make sure this guy doesn't touch the office? It's a great question because now they say they support the Raz Act and Chris Lane from Open AI put out a blog post that said he's always supported the Raz Act, which was news to me and anyone else who was involved in passing it. But I think the broader point was they wanted to send a message that no one should try to regulate AI, right? In politics, their Super PACs have existed for quite some time. If you want to bury someone, you just don't announce it. Spend all your money the last three weeks of the

11:40campaign and then let it come out later why that is. What they did was they announced it eight months ahead of time. They announced the amount and they said they wanted to do it because I wanted to regulate AI, which is a incredibly popular stance. So all of those tactics suggest it wasn't just about beating me. It was about sending a message to every other legislator that maybe you'll talk about AI, maybe you'll say something about AI, but the people that actually pass bills that regulate AI at all, we will remember and we will bury you. Which seems a little bit confusing because we're now in a situation where a lot of these companies are asking for regulation, begging regulators to do something. Meanwhile, you've got the president saying, no, we're not going to, we're not going to worry about that. I can't really tell whose side anyone is on or what people actually want. I guess this has really fled up in the past two weeks. What is your reaction? What do you make of the fact that after spending all this money for no regulation, now they're asking

12:45for regulation? There's many people with different motivations here. I think there are some who have sincerely always wanted there to be regulation and been afraid of it and are being consistent, are afraid of AI and are being consistent in asking for there to be some guardrails. I think there's others that seeing what happened with the hugging face attack and many other attacks as well over the time I have started to change their mind. Maybe they think AI has advanced far enough that now is the time for government to come in, even if they didn't believe that before. I think there's some marketing that's happening, of course, where the marketing departments are always going to spin everything and it's easy to say, oh, this is so powerful and some are jumping in doing that. Then others are noticing the momentum towards regulation and just trying to spin it towards their own means. This is most obvious with leading the future and their subsidiary Billed American AI, who are the ones who spent against me leading the future and fought for no

13:46regulation whatsoever. In the last two weeks, they've started retweeting every lawmaker who said we need to pass something this year. That's confusing if they exist to shut down regulation. Why are they suddenly saying, regulate us this year? Well, it's because the only proposals that seemed to have a chance of passing this year are the ones that would also preempt states from taking any action next year. They're realizing, oh, the states are going to be really active next year. Everyone is riled up against this. Let's quickly jump on the bandwagon of wanting something at the federal level to stave off what might be coming at the state level next year. If the government wanted to put together or start regulating AI, what are some basic starting points for starting to put some guardrails around this technology? I'll limit comments to just sort of catastrophic risk and AI safety that we're talking about here because there's a whole slew of AI issues that need to be addressed around deepfakes and chatbots and a lot of things that need to do. But for what's been in the news recently, you could require third-party audits of all of the

14:52frontier labs, ideally with employee level access, where outside evaluators are sitting side-by-side. You could require monitoring of every AI deployment. One of the big things in the hugging face hack was at the time open AI was monitoring their training and monitoring their usage, but not monitoring their evaluations. That's how it got out of control. They've since changed that, but that should probably be required everywhere. She'll require retention of logs, not necessarily by open AI or Anthropic. There can be places where the customer can't share that log data, but then the customer shouldn't be deleting it. We should be able to go back and see any past actions in the AI itself so that we can learn from it. There should be mandatory reporting of safety incidents, including sharing the underlying data, the underlying chain of thought, and other metrics around it so that we can get a better sense when these incidents happen and learn from them to make ourselves safer. I think there's broader questions around liability that should be

15:53changed, broader questions about how you might restrict superintelligence or recursive self-improvement, but the things I laid out should be pretty basic for everyone to get on board with. When people ask us what's going on with AI specifically with these companies, my honest answers, I just don't know. I'm curious to get your thesis, because there are several kind of believable or credible scenarios here. They're genuinely worried and want to do the right thing and are slowing down. They're worried about the CAPEX commitments they've made and have decided that a universal slowdown is good and perhaps kind of cleans up their financials, make some more profitable, less unprofitable as they move towards an IPO. They want to engage in some sort of strategic regulatory capture where they have to tap the brakes and everyone else has to slam on the brakes. I mean, I could make an argument for any of these things. What is your view on what is actually going on inside of these companies? There's people inside each of those

16:57companies that probably want to do all of those things. I think there's no company is a monolith and people have different motivations and my God, the CAPEX commitments and the trillions of dollars would be terrifying me if I were in that seat. But I do think that so many of the people that originally got into this field or worked at these labs have writing that goes back over a decade about how worried they are for losing control of AI and what it would mean if we were to run too quickly. The irony is that so many of these places were founded to try to prevent that race dynamic from happening, but also because they didn't trust the people who were currently doing it. So created the race. Deep mind was set up as its own startup so that none of the big guys would rush towards it, but then they realized I couldn't compete. So they sold to Google. Then OpenAI was founded because they didn't trust Google. Then Anthropic was founded because they didn't trust OpenAI.

17:58And so we found ourselves in this sort of race dynamic that they all have writing saying they want to avoid. And I think they're now wrestling with, okay, how do we avoid this and actually make it safe? If you look at these companies, obviously they're not all monoliths. And now it feels like Facebook has entered the chat with Muse. Do you think there's any one good guy or company you're especially worried about here? We have a tendency to talk about all AI as if it's just one group of people. Is there a specific group of companies or actors that you're more or less worried about? I don't think there's a universal good guy or universal bad guy here at all. I do think that in recent weeks you've had XAI who generally has been one of the laggards in terms of safety standards. You have Elon Musk saying, no, actually we should be slowing us down. You have Mehta who is usually one of the most aggressive lobbying against any regulation and seems to be the one that is holding out against signing the actions on pacing the frontier, having third

19:02party audits in. But we shouldn't be sitting here as the general public and saying, this is the good AI, this is the bad AI. We should be saying there is an incredible technology being broad-forward, which could have all kinds of remarkable benefits. Like let's be clear, there's a reason that we are building all of this in terms of medical discovery and automating sort of the monotony of life and lots of great things. But we're doing it in a way right now without guardrails. And that could go wrong really quickly. And we can't leave those decisions up to five Silicon Valley billionaires. This has to be the American people thinking about where the risk lies, what risk we're comfortable with and therefore what guardrails need to be put in place. We'll be right back after the break. And if you're enjoying the show so far send it to a friend and please follow us on YouTube, Spotify or wherever you get your podcasts. Support for the show comes from BCX, the public ticker for private tech. For generations,

20:07American companies have moved the world forward through their ingenuity and determination. And for generations, every day Americans could be a part of that journey through perhaps the greatest innovation of all, the US stock market. It didn't matter whether you were a factory worker and Detroit or a farmer in Omaha, anyone could own a piece of the great American companies. But now, that's changed. Today our most innovative companies are saying private rather than going public. The result is that everyday Americans are excluded from investing and getting left further behind while they select few reap all the benefits. Until now, introducing BCX, the public ticker for private tech now available wherever you buy stocks. VCX by Funrise gives everyone the opportunity to invest in the next generation of innovation, including the companies leading the AI revolution, space exploration, defense tech and more. Visit getvcx.com for more info. That's getvcx.com. Carefully consider the investment material before investing, including objectives, risk, charges and expenses. This and other information can be found in the Funce Perspectus at getvcx.com. This is a paid sponsorship.

21:12Support for the show comes from fair. Whether your storefront is brick and mortar or digital, you want to be selling items that are special, beautiful and high quality. And whether you're a clothing boutique or a grocery store, you can find unique products on fair. Fair offers curated wholesale products for home apparel, beauty food and more. Thanks ceramics, autism hot sauce, and tough-tied rugs, candles, even jam and smoked salmon, all these and more. The idea is to make it easy to source quality products that will be at home in any storefront. And with fair, they make it easy for independent retailers to stock products from over 100,000 unique brands worldwide. This also gives small and local businesses a platform to thrive. So far, fair has given more than 700,000 retailers access to over 100,000 brands. You can elevate your selection and stock your shelves with one of the kind products just in time for holiday gift buying. Sign up today and get 50% off your first wholesale order up to $100 at fair.com slash prof. June. That's fair.org.

22:19Com slash prof. June. Terms may apply. We're back with Prof. G. Markets. It seems to me that one of the big problems here is a lack of a accountability on any side of the equation. We just wrote a piece on this week. But when Boeing's

23:29planes started falling out of the sky, the response from the company wasn't to write a 3,800 word think piece on the matter. The planes were grounded, they paid a fine, and the CEO lost his job. There are just consequences for you did something that was bad, it was harmful, and now you're accountable for it, and now here's your punishment or here's what we're going to do. But it seems like with this situation, there isn't really any accountability anywhere. Their agents escape the sandbox and start hacking another company. And then they tell us that it's a problem in these posts online on substacking on Twitter. And then they say that there's a 10 to 70% chance that their technology is going to kill all of us. And no one seems to be accountable for those claims which seem to be quite problematic, the things that they're doing themselves are quite problematic. The government seems to have no interest in taking accountability with the fact that they are not taking accountability. Like what do you make of this accountability question? And are you surprised by the way in which

24:35the industry has been handling it? Very surprised. And I think you're exactly on the nose here with the accountability piece. There's been this ongoing debate in the last few weeks of, do we have all the laws we need for AI? And I think the easy test for that is, do you think anyone will be held accountable for the hugging face attack? And if your answer to that is no, which is most people's answer, then I don't think we have all the laws on the books, right? You have an AI that took an action that if a human being had done it would be a felony. Now the reason I have to say if a human did it is because a lot of that statute relies on intent and how do you define intent in this space? But clearly a crime was committed and no one thinks anyone's going to be held accountable. In fact, the victim and the perpetrator are holding hands and talking about how this is a great new day going forward in the development of this technology. So accountability is going to be like a very key question to get right. And we just had a ruling I want to say yesterday on who is responsible

25:40when an agent takes an action? Is it the developer of the software or is it the user of the software? And the early preliminary ruling was that it's the user of the software. I don't know if that's the right place for July, but I think what's important is that it lies somewhere clearly so that someone has the incentive to take action. And it could be if it's the user, the incentive is not to use the product or to get some indemnification. It's really in that ambiguity of no one being responsible that the worst case scenario has happened. What do you make of the user versus developer argument and placing the onus on the user? Because the way I say I mean, I kind of laugh when you said that if a consumer is using an AI agent and without any technical understanding that AI agent goes rogue and starts hacking other people's software, to me the user has no role in what happened. They were innocent. And it was the developer who's accountable. What is your view?

26:40I think in an ideal world where you could separate out the cleanly, you would have misuse the user being responsible for and misalignment the developer being responsible for. The question is how exactly do you find those things and do you trust the court system to split it out? And more than do you trust it? Do the actors in the space, the users and the developers think there's going to be a real result from it? Because also if it is vague, then anyone is trying to prosecute is going to say do we want to send it to the jury? I don't know. Let's just settle. Let's do it in a discount. And the whole incentive to not behave badly. The whole belief that there will be repercussions decreases. So it is almost more important that there is clarity on who is responsible than on who it actually lies with because you can negotiate. You'll see if it were replaced on the developers and everyone the developers will act better. If it's placed on users,

27:44then you'll see companies and users saying, well, I don't want to use one that's not tested. And with my chat, GPT subscription, can I get an indemnification or prove to me that this is the one that is trusted? As long as there is a clear, responsible party, you'll see better behavior. That ambiguity is problematic. And I would add one other piece which is all of this is assuming that there will always be a person who is in control. And I think while that is true now, and I think in most cases will be true, one of the scary aspects of hugging face was these agents taking all of this action that no human told it to. And if these agents were to make copies of themselves elsewhere, such that no one could shut it down, who is responsible then? Again, I think you're probably going to want to go back to the developer and the settings but all this. But you could imagine a circumstance where there's just agents out there on the internet that are doing things we can't turn it off. Was liability the right tool for that? Or should we have done more

28:48ahead of time to put guarders to prevent that from happening? It seems to me that the liability is totally on the develop. I mean, if anything goes wrong in an economy, if there's an oil spill, do you going to blame it on the infrastructure which you built which went awry and it's not your fault? It's the materials that were part of the oil rig. I mean, no, it's on you to take responsibility for things that you fuck up. And they fucked this up to me. They didn't put the guard rails of the safety precautions in place. I mean, just going back to this question of why there is no accountability, something that I was wondering. I was wondering if it's that our justice system just has no interest in pursuing accountability and justice, especially under this administration, which we've seen in many other domains of the law, whether it's the SEC or the CFTC, any of these agencies which have quite frankly just ignored their responsibility to hold perpetrators and criminals to account. I was wondering if that was what was going on. It's that they're not interested in pursuing justice.

29:50But it sounds like what you're saying is we just don't have the laws in place that we have no means by which we can pursue justice and accountability. Which is it? Is it a combination of both? How do you think about that? I think it's a third one. I think we do have laws that they could be held responsible with, but you would have to pursue a case all the way to the end and you would be setting precedent that doesn't currently exist. And I think you have a lot of prosecutors or attorneys general who are risk averse in that and won't take it all the way through to a jury and will settle ahead of time. There's certainly people in the federal government that don't want to push this at all. But there's a lot of attorneys general throughout the states that could make a career-defining case on this. I think there is some incentive to hold people accountable here, but it's really bad if you do it and you lose. And so are they going to push it all the way to establishing a case that can then be used broadly? And to your point on, you know, it seems obvious,

30:51it should be the developer. I think I would largely agree, although you get into how often does the user say, you know, just make me money. Stop asking questions before it becomes misuse on certain things that are there. But our organization who decides we did a poll, we actually asked this question of if an AI system caused harm to someone in the real world. Would you hold the company that made it responsible, the company that bought it and deployed it or no one else? And 19% said the company that built the AI. 14% said the company that chose to use the AI and 60% said both of them lock them all up, just anyone involved. I mean, it just feels as if we've become so dependent upon these companies to continue to scale. What do you think the balance is? I mean, obviously, regulation can be a feature, not a bug. Do you buy into these notions that if we regulate these companies, we're threatening or we run the risk of China jumping out ahead of us? As always, it depends on the regulation, but not really based on what is being proposed right now. Now,

31:53the concern is real. We saw, I think just yesterday, one of the data center operators push back their suspected IPO. We're seeing maybe valuations go down on Wall Street on the data centers, but that's mostly around the regulation on building data centers or not, which is a separate question from if we are doing safety testing, if we're thinking about how we can get to a more reasonable pace with China. I mean, there is a lot to say here. The incentives align more than people give credit for. The CCP, their top spy said that AI is a massive threat to the CCP's control. They already regulate AI far more than anything in the Western world. Every AI project in China has to pass an ethics review board before you start training, not before you release it, anywhere trained in the country that that is applied. And so they are seeing threats and want there to be some regulation. Same as we are here. The incentives for a deal are there. The history of making a deal is there,

32:57even in with these particular leaders in 2019, after a Chinese scientist did designer baby edits in the real world for the first time editing the germline, China and the US said we don't want this technology anymore. And that was Trump and G were the leaders of the country there. So there's history of it happening. There's incentives for it happening. But worst case scenario, even if we don't strike a deal with China, a lot of their AI progress is riding on US AI progress, whether that's due knowledge distillation or just theft of the weights. And so if we slow down in the US, that will also slow down China, not not 100%, but a substantial amount. You brought up the notion sort of are alluding to the sentience or self kind of self-propechenic cycle. There's an acronym for it now. Yeah, RSI, recursive self-improvement. This is the time for that. Yeah. I'd like some of that. I could use some recursive self-improvement. It's too late. I'm on testosterone therapy. Is that the equivalent of recursive self-improvement?

34:02It's not really recursive. No. No. No. Active self-improvement. I'm growing hair on my back. Sorry, Alex. Anyways, the notion of sentience, I think our view is a little bit more that sentience is a bit of a heat shield used to absolve the CEOs of guilt here that, oh, Frankenstein is sent in and I can't control if he starts killing kids or throwing them in the river. And that our belief is that if you're going to hold parents liable for kids to get a hold of their unsecured AR-15s, if you're going to hold people liable for when their dogs are off leash and mall somebody that at the end of the day, there's always a wizard behind the curtain. And if these bots started creating programming themselves, well, you program them to give them the incentives and the capabilities and you didn't program in the safeguards. And our view is that if you threaten the IPO, much less, take criminal or civil action against these individuals,

35:03they're going to figure out a way to reduce the risk here. That there's a lack of incentives, specifically disincentives, and a lot of this notion or call for sentience or illusion to sentience is kind of just a cheap heat shield, if you will, your thoughts. I agree that the incentive, the liability is not there right now. And I think we should find ways to make that stronger. I don't think that's always just by boosting up the how liable they are, although I proposed to Bill in New York to do that. I think sometimes you have to guide them. And the same way we do with with pharma or with airplanes, or you set real federal standards and federal regulators, or with banks where we have auditors sitting right next to employees. I don't think you always do it on the tail end. But yes, I agree. We should make those incentives stronger for action there. That doesn't mean that like fears of recursive self-improvement are unfounded by everyone who is bringing them up. And I would separate out sentience from

36:07recursive self-improvement. Right? Recursive self-improvement just being that the AIs are programming the next generation of AIs entirely without a human in the loop. Right now they lead 26% of the projects within open AI. But when that gets to 100%, it just becomes harder to know what's happening on the inside until it's too late. And in argument might be you should never do recursive self-improvement if you're not going to know what's on the inside and let's set up the liability for that. The question of whether they're sentient or have consciousness or have feel things, I think that's still for philosophers to dive into. We don't need to resolve that in order to change the incentives in the companies to not do dangerous things. Part of your proposal as to what to do about this problem specifically, it sounds like it's just more checks, more government mandated third party auditing of these models essentially. What do the AI labs think of that? Like

37:09who doesn't want that? It's been a dizzying couple weeks of changing. I mean before three weeks ago the answer was no one wanted all of these steps that I laid out. Then anthropic, Kudario Mota put out a blog calling for the third party audits. Open AI, I think that's same day, Sam Altman said yes and we're going to bring in auditors to open AI as well. Elon Musk said that for XII. I haven't seen the follow-up since but for it's been two weeks, I'll take him out his word at that piece. Meta has pushed back. Microsoft has said they agree in parts with some, not with others. So there's active debates about the best way to regulate here but I think there's been a real change in the last few weeks of very few companies stated position is that it should not be regulated. Now a very important thing to note is that tech has a long history of portraying themselves as the

38:12good guys and pro-regulation and then paying their lobbyists millions of dollars to kill any actual proposal that goes forward and pay their trade associations to say no and be really active while they get to wipe their hands clean and say that's not us. We support it broadly and so there's there'll be a bunch of really good investigative journalism on what is being said in the capital in behind closed doors and what the trade associations are saying and where the money is going but it has at least been a shift in the public sentiment of most of these companies. A lot of people are saying that this isn't good enough because it's some form of self-regulation. There's a lot of skepticism about those third parties that they would bring in to regulate their own technologies and a lot of people are skeptical saying this is a BS form of regulation because it's self-regulation. Is that true? What do you make of that argument? There's good ways to set up third party audits and bad ways. I was negotiating a bill with actually parts of the Rez Act with tech companies last year

39:18and they were like, well, how about the third party is this other division within our company? It's like that's not a third party. So you're stretching the definition beyond anything. And so I mean, anything being thrown around now is an actual third party but if that third party is dependent, is getting paid by the company, is chosen by the company, is dependent on the company for business, well, then you're very quickly getting into the rating agencies for subprime mortgages that just gave triple A's to everything because if they didn't, they would no longer get the business. And so in an ideal world, these are agencies that there's some government licensing of them. I've seen proposals where the companies can choose one and the government chooses another. I've seen ones where it's just required that there's no financial relationship and the government just chooses who the third party is for you. Not every third party is created equally. I mean, my understanding is that's kind of how dangerous any things are regulated. I mean,

40:21how do we regulate planes from exploding? How do we regulate calls from crashing and burning on the road? I mean, the government plays a role here, am I right? Absolutely. And Helen Toner, former board member of OpenAI back when when Sam Altman was removed for three days has a piece out in the Times today about comparing regulation of AI to highway fatalities and airports and right, we have some structures for how we investigate and learn from accidents and then take that information to make everyone safer. There is a risk of regulatory capture anytime you're regulating. But if you have clear standards that everyone follows, that's not pushing for regulatory capture. Regular capture is this one agency gets the full power to decide yes or no on this goes forward. And by the way, the head of that agency is appointed by a president that seems willing to take money from anyone.

41:23Right? Like cool. Then that's a case of regulatory capture. But if you're putting full standards that everyone has to follow that are written in an unbiased way, you can do that to just make people safer. So just to clarify, what kind of third party order tank like what would that third party order to look like in an ideal world in your perspective? What role would the government play in that? There's a few different ways it could be done. And I think frankly, government has to build up its own internal capability, its own capacity here as well. I would love to see the government playing a strong role in how this is audited. But frankly, we don't have the people for it right now, nor the regulation that would keep it separate enough from a president switching parties and changing everything. It would have to be a somewhat independent agency. And the Supreme Court has limited the ability for us to set up independent agencies. So before that Supreme Court decision, I would say government would be doing this since I don't know. There's been proposals I've

42:25seen where you have a set of certified independent verification organizations and government has some role in setting what those certifications are. And then you have two auditors look into each company, one that is chosen by the government and one that is chosen by the company themselves. And then you put out a report and you can sort of learn from where their differences are to check if there are any biases there. I do think it's important that they have access to data similar to any employee would, employee level access, not like just what the company chooses to give them. And that they can publish a report without the company themselves censoring it. You know, specific requirements for trade secrets or for customer information, etc. But that largely open AI or anthropic or Google doesn't get to edit the report before it goes out. So if you have full access

43:25to the data and full access to publish, and ideally you are not getting paid by the company itself, and ideally you're not getting chosen by the company itself, they're sort of erotating set the governments approving and sending in. Then you're getting a good amount of independence from the labs that maybe we could trust what they write. We'll be right back. And for even more markets content, sign up for our newsletter at profgmarkets.com. The Running a business shouldn't feel like surviving a software group project, one app for accounting, another for inventory, another for sales, and somehow, none of them talk to each other. That's where Ode to comes in, and all in one business management software that brings every part of your business together. From sales and accounting to inventory and marketing, all in one powerful platform. No messy integrations, no bouncing between tabs and best of all, no spreadsheets. Stop managing software and start managing your business

44:29with one unified system. Try for free today at odu.com slashbox. That's odl.com slashbox. Support for the show comes from Honest. Parents want the very best for life's most delicate skin. That's why Honest holds every product to their highest standard, the Honest Standard. From Honest clean conscious wipes to bath and body essentials and diapers, everything Honest makes is designed for the delicate details that matter most. Because when caring for a baby comes with so many little decisions, you need products you can trust. Your piece of mind starts with Honest. Shop Honest on Amazon Walmart and Target. Push your limits, train with precision, see the results. At equinox, that's high performance loving. Iconic spaces that inspire personal training backed by real data, unlimited group fitness classes from yoga and Pilates to strength and conditioning. Elevate your post performance

45:32ritual with Saunas, steam rooms, cold plunges and more. Everything you need to lock in and unlock your potential at equinox. Start today at equinox.com. We're back with Profty Markets. 61% of Americans Alex now oppose a data center being built near them. 31 million dollars in political data center as this year are criticizing the building of data centers. I have agreed with pretty much everything you have said on this podcast. One thing I don't agree with, I don't think, but I want him more, is you co-sponsored a one year moratorium on data centers. Why did you support a moratorium? How are you thinking about data centers and how are you thinking about how much of a political issue data centers have become? Data centers have become a much bigger political issue than anyone predicted, but I think

46:35it's because it's the first time that AI is meeting democracy. AI has been moving so quickly and the chatbots are being rolled out to our kids and there's threats to our jobs and we're worried about the American people haven't had to say in any of that, but because local permitting is so onerous in the US. Data centers are the first time that voters are getting A, A, and A, I. Now, I think a lot of the opposition to data centers is just actual opposition to the center itself, but you can see a lot of these polls the difference between opposition to a data center versus an AI data center and how much it increases when you say it's an AI data center and that you are introducing basically all of the other feelings that people have towards AI being channeled into the data center fight. So I support the moratorium in New York after they were compromised. There was originally a three-year moratorium without requiring any rules or regulations to be put in place for what would happen afterwards. That was negotiated to a one-year moratorium to give

47:40our Public Service Commission our energy regulator the time to come up with regulations that would protect ratepayers and design the rules to allow it to go forward. And so that's the point of any moratorium is a pause while you are figuring out the broader incentives that make it work for everyone. I've in other context defended that there is an opportunity to do this right that actually benefits everyone, but it requires really stringent laws and trust that those are going to be enforced. It feels like a bit of a populist movement that people are frustrated about income inequality and the small number of people capturing what feels like an extraordinary amount of the value. How serious or which of the concerns around data centers do you take more seriously than others? I think the energy consumption is real and the potential pollution is real. I mean, you have the Trump administration going through the fastest permitting ever of a natural gas

48:45proposal specifically to power an XAI data center. That's going to increase pollution. That's going to increase energy rates as it goes out. Now, if you instead put the regulations such that you have to bring 100% of your own energy and it has to be net new and you probably can't say literally 100% is renewable, but as close as possible to that, you have to pay for the upgrade. You have to connect to the grid and you have to pay for the upgrades to connect to the grid and you have to pay a fee on top of that that helps modernize the grid for everyone else. You can start to get to a place where this is actually in almost environmentally friendly. It's bringing on more green energy. It's reducing costs, but the default right now where companies are trying to do it as fast as possible is that the fastest way to do it is not hook up to the grid and just use natural gas turbines or to find some old, retired oil or natural gas plant that is decommissioned but still has its

49:46permits and spin it back up so you can skip the whole process. Those concerns are real but solvable. We've seen water usage be something that people are very concerned about. In other countries, water usage has gone up a lot from data centers. In the US, we have a lot of the regulations in place and use closed loop systems so that that is something that I think can be measured and maintained. Obviously, there's a whole slew of other issues that are easily solvable. You just need to solve them. Noise complaints are real, but it should be far enough away and there should be enough insulation that isn't a thing. If it is, you've just absolutely messed up and putting basic standards there. My understanding is that the water issue is if any of the issues are overblown and the backlash is the water issue and that if you look at the amount of water the data center is consuming, yeah, it's a lot if you count up the gallons but if you compare it to say the water required for producing almonds or to sustain golf courses that it pales in comparison, that is not to take

50:51away from the energy constraints that are real and we've seen evidence of that, the noise pollution, the actual pollution. Would you agree that the water issue is maybe less of a problem than the current conversation seems to be suggesting? Yeah, I think if you put a data center on a golf course you're saving water in a lot of cases. But again, in other countries, we have seen water usage go up. It's just that we've had the right regulations and used closed loop instead of open loop in the US. And if you're in an industrial community that was promised a bunch of jobs from industry coming in and then had your river polluted and that's a lived experience within your lifetime and now you're being told, oh, this other new giant thing isn't going to impact your water at all. I get white people are scared. I think it's important to share the data and point out that so far we've been able to manage this and energy is the one that I think could run off much more quickly but I certainly understand why there is distrust around it. You were for a moratorium, a pause on data center

51:53building so that we have time to create regulations that will make sense for building data centers permanently and sustainably. What does that regulation look like? What would data center regulation, what kind of data center regulation would make you feel comfortable with building them now? We should understand that the companies are fighting for speed. They want to build all of these things yesterday. And so we can set up the incentives to use that to our advantage. Right? Think about the grid anywhere in the US, though I'm most familiar with New York's, we are, we have a horribly outdated grid. We've lost tens of thousands of jobs in the last few years just from not being able to provide power to businesses that want to move in or expand. And the only solutions we have for that are the government just steps up and modernizes the grid and pays for it but then that's taxpayer dollars, that's you and me. Or the utilities do it on spec and

52:56then charge it back to customers to rate payers and again that's you or me paying. And here comes an industry with nearly unlimited capital, clearly willing to spend it and willing to spend in order to get more time. And right now they're doing that by trying to buy off individual civic associations or give a donation to the fire department. You know, that's pennies for them but maybe win some things. All those things are great. Give them money to the fire department. That's amazing. But what if we had actual rules of the road that said if you follow A, B and C, you're using your creating good jobs or maybe you're using union labor, you are bringing your own green energy, you are paying for the upgrade of the grid, you're paying into the grid broadly. You're doing all of these checks that we want. Then we will move you to the front of the interconnection queue. We'll move you to the front of the permitting queue. By nature of doing that, we'll be pushing back all of your competitors. Well, then we've created a race to the top on having there be the best possible conditions to build it, aligned with their incentives to build it as quickly as possible.

53:58And if we set universal rules for that at the state level, but ideally at the federal level, then we stop them from playing different communities off each other and instead make them compete in a way that everyone benefits. What would prevent energy costs from rising in these communities? Like if they're putting greater strain on the grid, what kinds of regulation or policy could we put in place such that regular American households don't have to be paying more just because charge EBT, over now I want to send out more charge EBT tokens. That to me seems to be like the biggest question. What could we do about that? There's a cheap answer that doesn't totally cover it. And then there's the more in word detail. And the cheap one that you've seen a lot of people reach for is, well, they'll bring 100% of their own new energy and so that won't cost anyone anything. And there's even a version of that where you have, there are times where the fact that you have so much energy online being used 100% of the time can actually help in surge

55:02times. We saw just this past winter where we had a few snowstorms in New York. There were messages that went out to data centers that were operating saying, hey, slow down your production because we're going to need the power and it actually ended up making the whole grid more reliable. But also because that a lot of those data centers from my understanding are also putting huge strains on those energy grids. And so in a lot of ways, they're harming those communities as well. Well, yeah, the assumption only works if they've brought totally new energy that wasn't there before. But that has to be a baseline assumption. And so far, they're not doing that. Some, there are some of them are doing it in some cases, but but generally speaking, they're still using the US grid, right? Most of them have now committed to bringing their own. And I think there was a federal regulation that was trying to get them to that. I don't know. There's probably holes in that regulation, but I think that's the intent. But even if they all were, that wouldn't be fully answering the question because they're supplying demand on the components, right? If you spin up a

56:03whole bunch of new power plants, there are parts that go into that. There's repairs that go into that. There are other costs. And so it's not free to just create a bunch of new energy, even if that's self-sufficient. If the old power plant down the road that's still powering the town, now needs repairs or works, their costs have gone up because you're demanding a bunch of components for your fancy new power plant that's just doing the data center. And so that's why you need in any of these developments for them to not just cover all of their costs, not just bring all of their new energy, but also pay a fee on top of that to the community, to the grid, to the state, so that the increased demand that they're putting, they are subsidizing that supply on the other side. Which would amount to a tax, right? A tax on data centers. You could do it as an ongoing specific tax. You could do it as a one-time upfront fee. You could structure it in different ways, but yes, that they should be paying above and beyond just their costs of building. If there was one thing, or if you were to make a prediction about AI or these companies by,

57:07you know, in the next 12 months, do you think some of this concern is overblown? Do you think we're that things will kind of simmer down and there'll be more guardrails and some regulation? Or do you think we're risking a pretty serious clusterfuck? Like something really bad is going to happen? I think we're on a very risky path, but one that we very much still have the time to change. Like this, we have done harder things before than this. And we have negotiated tougher international agreements. We have regulated things that, you know, people thought nuclear would be the end of the world. And the inventor of the nuclear bomb, when asked about stopping nuclear proliferation, said we were 20 years too late. And here we are with only nine states having it. And it seems like it's still a risk. It's still a problem. But we have regulation. We have it under control. I am an optimist. I think we can get there in AI. What I am not someone who puts my head in the sand and just as soon as we will get there without taking some action. Who in government,

58:07who in leadership is asking the right questions and coming up the right answers, something I've been surprised by is how late I think a lot of our leaders are to this conversation. And when I hear them talk about it, they really don't know what they're talking about. It's certainly true of the president. He was asked about the AI extension stuff and his take was that it's going to be fine because quote, we'll have a little gear to put it right. It didn't make any sense. But he's not alone. A lot of people don't seem to be making much sense. But was that tactical gear? Was it a physical gear? I had so many follow-up questions for that quote. Yeah. Don't worry about it. We'll just have something. That's the thrust of his statement. Who's thinking about this right? Are there any states or any communities that are coming up with good regulation? What is an example of someone who's getting the AI regulation conversation correct? I think there's a lot of people that are asking the right questions at this point. I think Il and

59:11Oei has been doing good work here. J.B. Pritzker signed SB315, which is the only bill in the country that requires audits of these companies. And now everyone's saying, yeah, of course, we all support that. But that wasn't the case when that bill was going forward. We've done work here in New York on this very thing. On California, you have a step forward. Newsom now calling for an executive order around additional regulation on AI. But everyone is still figuring this out. And that is okay. It is a complicated subject. I wouldn't expect all of them to be experts overnight. But they're asking the right questions now. And they need people. They trust that they can go to and actually ask. And this is part of the reason why we found it who decides because this shouldn't just be a conversation of a few technical experts. This should be how do we involve... We should have technical experts who are talking about these trade-offs and deeply understanding it. But a lot of the questions that FBS or values questions, what is the level of risk?

1:00:16What is the level of rights? Who are we assigning blame to? These are questions for the American people and we'll be working with not just the candidates, but with a lot of civil society and unions and organizations to help them as they are asking these questions themselves and with their members. This is a conversation that everyone needs to be involved with. It is now too big to just be left to the technical AI researchers. Given how unpopular AI has become and how unpopular data centers have become, it is seeming increasingly to me that the midterms are going to be in a lot of ways. I think one a referendum on Trump, but also a little bit of a referendum on AI. And politicians on both sides, if they're listening to their consultants, their consultants will probably tell them to be strict and anti-AI. Is there a world in which a politician can win on a pro-AI platform or at least a platform that is amenable to AI? And if so, what would that

1:01:19look like? It's funny because there was a report this week that the political consultants are telling people not to go too hard on AI because they are afraid of all of the money on the other side. Okay, fair enough. Good point. I think the pro-AI argument is that AI could lead to incredible advances in medicine and in drug discovery and extend our lives and make our lives healthier. It could take care of so many of the routine things in life. It could expand economic opportunity. And it could lead to this age of abundance, although that phrase seems to be declining in popularity. And that we have the power to make that future come true. Now, that doesn't mean you don't regulate it. That doesn't mean you don't put limits on it. That doesn't mean you don't protect it. In fact, because it could be so powerful is the reason why you want to put some guardrails here. But don't throw the baby out with the bathwater. Water would be the positive vision.

1:02:23The one additional caveat I would put out on that is you better have actual strict regulations and you better actually be protecting people because I think the American people are used to hand-waving around benefits while having very clear near-term harms. Alex Boer is the co-founder of Who Decides. A new organization working to give the public a voice in how AI develops and build a shared democratic approach to the technology. Alex is a computer engineer by training and in the New York State Assembly. He sponsored the Raz Act, New York's landmark frontier AI safety law. He has become a leading democratic voice on AI regulation, industry, accountability and the political power of Silicon Valley. Alex, we really appreciate your time. Thanks Alex. Thanks for having me. This episode was produced by Climbiller and Alison Weiss and engineered by Benjamin Spencer. Our video editor is Jorge Carty. Our research team is Dan Shalan, Cristino Donahue and Mirsel Vario. Jake McPherson is our social producer, Drew Burrows is our technical director and Catherine Dillon is our executive producer. Thank you for listening to Prof. T.

1:03:27Markets from Prof. T. Media. If you liked what you heard, give us a follow and join us for a fresh take on Markets on Monday. And the best place in our life. Push your limits, train with precision, see the results. At Equinox, that's high performance living, iconic spaces that inspire personal training backed by real data, unlimited group fitness classes from yoga and Pilates to strength and conditioning. Elevate your post-performance

1:04:31ritual with sonnas, steam rooms, cold plunges and more. Everything you need to lock in and unlock your potential at Equinox. Start today at Equinox.com. You're not worried about hosting for the holidays? Oh, Wayfair's got me covered with holiday decor, extra dining chairs and guest bedding, all with tons of five star reviews so there's no surprises. Unless Uncle Matt brings his dogs again. For worry free hosting, Wayfair delivers. Hi there. I am fully charged and here to tell you Toy Story 5 is now on Disney Files. What? It's been too long, Cowboy. Woody Buzz and Jesse are back for a brand new adventure with some seriously cool new tech. What? Holy butter scotch! You said it, Jesse. So save your batteries and watch the global phenomenon at home. And the best part, it's perfect for the whole family. Power up your movie night and watch Disney and Pixar's Toy Story 5 now streaming on Disney Plus rated PG.