What Freedom Means In the Age of AI [feat. Venice.ai Founder Erik Voorhees]

Today on the AI Daily Brief, a conversation with Venice founder Eric Voorhees. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. To join the conversation, follow the Discord link in our show. Notes. Welcome back to the AI Daily Brief. Today we're doing something a little bit different. I have a conversation with Venice co founder Eric Voorhees. The idea for this conversation had been building for a while. One specific catalyst was the arrest of Telegram founder Pavel Durov, which as you'll see, I think has much bigger implications for the state of the relationship that we have with Internet platforms in general. But some of the conversations surrounding that also were reflected, I think, in the debate we've been having around California's SB 1047, which is of course AI regulation. Now, Eric and Venice are sponsors of the AI Daily Brief right now, but this episode is not sponsored in the sense of it being a precondition of the sponsorship that they to be on the show. Their sponsorship was totally standard. It's just the mid roll ads that you hear every day. I wanted to have Eric on to have this conversation because he's been interested in and fighting around these types of issues for a very long time. It precedes, in other words, his relationship with artificial intelligence. So in this show we talk about the arrest of Pavel Durov, the implications of SB 1047, and how Eric thinks about the fight for the future of AI and freedom. I think that even if you find yourself disagreeing with Eric, you'll come away better for having heard how he thinks about these particular issues. Tomorrow we will be back to a normal episode, but for now, enjoy this conversation with Eric Voorhees, the co founder of Venice AI. Eric, welcome back to the Daily Brief. How are you doing?

I'm doing great. Thanks for having me on the show again.

Yeah, no, this one was. It feels particularly opportune to have you here to talk about this. And you know, it's one that I think is really important. And so broadly speaking, what we're going to be talking about today is, you know, there's a set of sort of catalytic or catalysts that make this conversation matter right now, the arrest of Telegram's founder. Also the debate that has been happening around AI regulation, specifically California's SB 1047. But the broader context is sort of a renegotiation of our relationship with the Internet and our expectations around platforms. And you are someone who both is engaged in this conversation as a citizen, as issues that you think about all the Time and have for. For a long time. But also as a founder of the type of platform that's going to be implicated by how we come to conclusions about this. So I guess as a, as a way to start, what were your, you know, when you heard the news about Pavel Durov being arrested, what were your sort of first thoughts?

Like what?

You know, was this a feeling of inevitable. Was it a surprise? You give me, give me a sense of just sort of like the immediate, instantaneous, personal reaction.

Well, I think my general sentiment is that the Internet has carved out a lot of space where people can interact with each other without being seen at distance. And it basically allows people to just have their own space without anyone watching, if you're using it correctly or in certain ways. And this, I think, is really troublesome to governments that believe it's their right and their responsibility to observe everything that their citizens are engaged in. And so, you know, something like Telegram or Signal, these chat apps, it's not, you know, they're closed source, so it's not clear really how secure they are. But the belief among people using it is that they are secure, meaning they're private, they're encrypted, no one can watch the conversations. So hearing of Pavel's arrest, like, very sad. I mostly just had a lot of questions, right, like, what are the allegations? What are they upset with? And over the subsequent days, it sounded like they're upset that people used his platform for bad things. And that, of course, is a very, very dangerous, very dangerous foundation on which to build policy, because if you must be responsible for the uses of your platform, it turns everyone in the world basically into a spying agency for their own government. To me, that seems so much worse than any particular criminal activity which might be going on on a specific platform. So it's like the one of those cases where the cure is so much worse than the disease. So I'm glad that Pavel has been released, and I guess this will probably be a long, drawn out legal situation that we're all watching very intently.

So one of the things that has been really interesting to watch over the last sort of five or more years, I guess, um, you know, first in the context of crypto, and now I think it's really coming to the fore with AI is I think that, you know, at least me, and I want to speak for other people, I sort of took it for granted. Like, I took the way that the Internet had grown up more or less for granted, rather than understanding it as a byproduct of some amount of specific decisions that were made, particularly, you know, section 230, which explicitly deals with this particular issue of whether platform owners are culpable. And my impression, the first time I remember sort of getting this impression was when there were Libra hearings being held when Meta wanted to, you know, create its own stablecoin or its own version of a stablecoin. And it was so clear from those hearings that the average person in Congress on both sides of the aisle was wished that they had a time machine that they could go back and not have that regulation in place basically to make, make organizations responsible, platforms responsible for, for their users. It feels to me now like we are heading into a sort of crescendo around this particular issue. And it feels like AI regulation is, is going to be the place that this, that this comes to a fore. Is that something that you thought about as you were founding Venice? Is that something you think actively about now? I mean, how much does this sort of risk that you have to consider as you're, as you're building this company?

Well, I'd be lying if I, if I didn't say that like a huge impetus for the creation of Venice in the first place was the, the assumption or the presumption that AI services would become entrapped, ensnared in all sorts of government regulation and would end up spying on their users entirely in a very Orwellian manner. And not only spying, but censoring and controlling and curating the conversations that people had with machine intelligence. So I think that's, it's not that that's a risk that that could happen. That's what will happen. These, these large AI companies that are centralized will be and are being captured by the state. And because they're seeking profit and because they do not hold strongly to any moral or ethical principles of privacy, they will allow themselves to be captured so that they have the permission to continue operating. I think that leads to a very scary place. And Venice really was started as a permissionless alternative, open source, permissionless, uncensored, not spying on people. So that as the world wakes up to what's going on with the centralized providers, they have an alternative. And just as I've seen, like Bitcoin is a lifeboat for the financial system as it falls apart and collapses in the future. So too hopefully, can Venice be an alternative to the sort of mass scale surveillance and control of machine intelligence?

Let me ask you a question which is maybe routing a little bit deeper into Venice, but it's something that I've seen some people ask and want to know about. There's this sort of world that you're describing that is clearly the one that sort of we're in slash heading into where there's just ever more pressure around surveillance and what data. You know, basically the more data that is available about you, the more data that regulators and authorities want. Right. It's a very clear pattern. We've seen it in finance and you're seeing it with, you know, technology more broadly. The logic is about the bad stuff that doesn't happen in a, in a world where they can control everything. Right. You know, you have less of certain types of behavior that, that the society finds objectionable. Right. This is the sort of the, it's always the logic. It's like you give up freedom in exchange for, for, for safety. The counter logic that you've articulated really well on this show and in other places is about what we give up to, to sort of get that, that safety and why that's sort of problematic for, for society. But I guess a thing that comes up sometimes is a question of how do, how does society deal with those undesirable behaviors in sort of the world where it chooses not to surveil, where it chooses that that trade off isn't worth it? You know, what basically it's like, what is the art, what is the vision of a future where, you know, what are the trade offs really? I guess, you know, how much, how much is that? Is it sort of, you know, you give up safety in acceptance of freedom? Or is that a sort of a false dialectic that's basically just presented by those who are going for the safety?

The first issue is that people believe that if you give up the freedom, you get safety. And this is sort of just assumed that like, if the government has more tools and more power, it can reduce the harms of society. I have not, in my experience as a human on this earth, seen that to be true much at all. It has not seemed to me, if you look at history or if you look at your own world and experience around you, that the larger and more powerful the government gets, the less problems there are in society. It seems quite the opposite. And as you approach the extremes, as you approach a totalitarian government, those have unequivocally been the darkest periods of human history. So I don't buy this notion that like, if you just give up your freedoms to this coercive, monopolistic, controlling organization, that society is safer. If it were, then we could have that discussion about like, what are the right trade offs here But I just don't think that's, I don't think that's true. And I think the continued grasping of these governments to get more and more control betrays something which people should pay more attention to, and that is that they are not effective in their stated purpose. If governments were good at policing society, they wouldn't need continuing powers over and over and over. We're in this phenomenon where like government clearly fails to stop much bad activity. So they ask for larger budget, they don't ask, they force larger budgets and more control. Their results don't get any better and they use that then as more justification for more budget and more control. And this process simply continues because everyone wants to believe that there's someone in charge above them, you know, like raining down justice on society. And I just don't think that that's true.

Makes sense. So basically the core argument is that the fundamental premise, you can't really get to discussing the trade offs because the premise of the trade off is just not real.

Right, right. It's not a discussion of freedom versus safety. Yeah. Our only option is how much freedom do we give up. And our safety generally tends to be the same or worse. So it's a pretty bad deal.

Sort of building off of this and just for context, I think the reason that I'm asking some of these questions is sort of the foundation upon which later questions of how we should regulate, how we should think about AI regulation and things like that will be built. What do you think as you sort of survey the landscape of problems with AI, negative externalities of AI, the types of things that people cite in their sort of, you know, attempt to say that the need for regulation, which do you find sort of more or less compelling as issues that society actually has to find some way to deal with, whether it's through regulation or not.

I think the, the existential question of like the, the scary AI that takes over and destroys humanity is a super interesting topic, however, and I think that it is totally plausible that an AI could potentially do that someday. Like that's a, that's very sci fi, but it is plausible. At the same time, I don't think we're anywhere close to that right now. And there's no evidence of any such thing actually happening. So it remains in the theoretical and sci fi world. And I think the fear of that outcome causes people to not react very rationally to the discussion of AI as it actually exists in the world. And I think the bigger risks and the bigger danger tend to be not that AI will take over the world and cause all sorts of problems. But that AI is an extremely powerful tool under centralized control invites a totalitarian world that is very scary. Not because the AI was bad, but because the people that were controlling it were bad. And those people always believing themselves to be good. Right. Like all, all tyrants in history believe themselves to be good. So the world I'm, I'm worried about is one in which, and this isn't very far in the future where, where large powerful AIs are watching everyone. And that means watching all of your conversations in all of your chats, all of your conversations, in your emails. It is analyzing your videos and your photos that are all being stored in the cloud, and it's a complete digital Panopticon. And then people who want to control society for all the good purposes that they believe they are involved in, use that information to control people and shape them and design the lives of citizens. And it is a power which no person should have. No one is smart or wise or good enough to have that kind of power. And the potential for abuse and bad outcomes I think is huge and severe. So that's really, I think the, if you were to graph like likelihood and, and damage caused, that's the one that I think is the worst, because it's very likely. It's sort of like the inevitable path and the damage is, is so bad that that's really what we should be paying attention to. And the, and the only antidote to that isn't that you create some regulation to prevent the bad actors. It's to keep. To permit people to have open alternatives, to permit markets to function, to permit freedom, to permit privacy and prevent this kind of like grotesque Panopticon from forming in the first place.

You know, it's funny, if you define the sort of the chief problem of AI as such, that means Venice is actually an AI safety startup.

We are an AI safety startup. Yes, yeah, absolutely. Startup.

Yeah. So, okay, so wonderful, like background context for the next discussion that I want to get into, which is the debate that's been happening around SB 1047, I guess. First, how closely have you been following it versus sort of it's a thing going on in the background and you're keeping an eye. But you know, just that that'll help my framing.

Yeah, I'm following it. I would say medium closely, and it doesn't surprise me that it's. That it has emerged. Cool.

So I guess a little bit of background for listeners who, who haven't been following along as closely this is a bill in California that's primarily focused on big existential risk type issues of AI, right? So this was something that came out of AI safety circles, not as we justify in them, but as sort of normally people talk about them. The ease and you know, the sort of open philanthropy kind of aligned part of the world and came through the California State assembly in the form of Assemblyman Scott Weiner and was in the progress for a while, but really only sort of picked up steam over the last few months as a going concern. And in the lead up to, to, to this process, there have been a lot of different sort of debates around it. The two big issues that people cited, I think in their opposition to this bill were one, its focus on issues of X risk and sort of the things that are, you know, often pejoratively labeled sci fi over any sort of misinformation, deepfake kind of contemporary problems. That was where you saw a lot of the critique from like for example, members of the California congressional delegation who were opposed to the bill. And the second piece was a concern that it would really put a huge damper on open source development and open development, particularly around some of these issues of, of culpability for, you know, for, for model creators, you know, that, that, that were being built into this. Now this part of why this picked up steam is that we don't have any sort of regulation that's being debated really on a, on a national level. So the whole conversation focused in on sort of this Cal, California piece. But it was also amplified because you had sort of, you know, an increasingly louder set of people joining in the conversation. Notably Elon Musk at one point came out in favor of, of SB 1047. When you were looking at the bill, Eric, or you kind of, you know, observing it media medium closely, what stood out to you as, as particularly concerning about the way that it was sort of set up?

So one, the premise that you could regulate AI by having a California bill regulate AI is very silly because California is but a small place in a very large world. And even though the preponderance of commercial AI is built there currently, that could change without too much difficulty. So even if you believed that the terms within that bill were good, it'd be like trying to hold water in your, in your fingers. It's completely ineffectual. But I think most concerning about it is that it basically centers around the regulation of model training based on how much math is going into sets these thresholds, that if a model is trained with a certain amount of computation or parameters or it costs over a certain amount, then it is subject to X, Y and Z different regulations. That is such an easy thing to get around. If you can't train 10 to the 26th parameters, you train 10 to the 26 minus one parameter. And if you can't train $100 million model, then you train a $99 million model again. It's like, what's the point, really? More importantly, from an ethical perspective, how can you say that it is ethical to train a model that's like 10 to the 25 parameters, but not 10 to the 26 parameters? What is the ethical difference between those two things? Does anyone in the world believe that the politicians in California are wise enough in the framing of AI to decide where these kind of thresholds should be? The bill is obviously reactionary. It is obviously a reaction to people's fear about this new thing called AI and their hopes that government will come in and protect them from this fear when there's not actually a crime that's been committed. So, yeah, I don't like the whole thing. I'm sad that Elon came out in favor of it. I'm his biggest fan, but sometimes he just misses the mark completely.

Yeah, I mean, whole separate conversation around the standards that we hold, people that we admire, to having to. To, you know, be perfect and everything, but. So I think that I want to tie this actually to something that you were mentioning before as you were talking about where you sort of sit with. With some of the X risk kind of questions. You said basically that it is plausible to you that there's a future scenario in which some of those fears become manifest. However, it's very clear that that's not where we are now. And this tends to be the place that I find myself. I think that the. Where I kind of diverg of the conversations around AI safety is that they seem to be predicated upon the idea that we're somehow going to sleepwalk into these terrible scenarios, that there won't be some moment of recognition. And, you know, the, The AI safety folks, I think, would say there have been all these moments where we've reset the goalpost and, you know, there's fast takeoff and it'll move faster by the. By the time we recognize it, it'll be too late kind of a thing. But I just don't know. It feels to me like, you know, there's so much focus on this. There's, you know, the. And no one thinks that these models are problematic now. So sort of where Is this line and, and ultimately then regulation that's trying to prevent those bad outcomes in the future is sort of, you know, arbitrarily drawing this line at some, you know, future state that, that, you know, is sort of unproven. So I don't know. I think, I think that, that the arbitrariness of that line is, is sort of one of the biggest problems for me as well.

Yeah, I think that's, I think that's fair. Although if I'm to play devil's advocate here, if and when a nefarious AI intelligence emerges, the amount of tricks it can play on people to deceive them into not understanding what's going on I think is pretty severe. But I take a couple different approaches. First of all, if you're worried about AI safety, then open source to me seems like the obvious right answer. If you're worried about X risk. Open source is really important. The ability for all of humanity to be watching what's going on with models, to contribute, to learn, to deal with it is really key. Open source tends to be far safer than closed source because you simply have more eyes on it. It's harder for things to hide when, when there's sunlight, right? So that's one and two is like just a very, almost like a Zen acceptance of the evolution of man and technological progress. There is no way to stop the development of AI. You cannot regulate it. You cannot put it, you cannot put the genie back in the bottle. You can't pause it. You cannot pause it. It cannot be paused. It's silly to have a discussion about whether to pause it or not, because we can't deposit. And we should just accept that and realize that this is an unfolding which no one controls but all of us can influence in certain ways and we can help shape it. And so to not be like, have this sort of savior complex about being able to stop something that is unstoppable.

One of the big points of discussion that comes up in and around all of this is the sort of the specter of China, right? And this is used to argue all sorts of different sides. The, the specter of China is used to argue why we have to move faster and we can't slow down. It's also used to argue by, by some folks that why we shouldn't have open source AI because then they'll get it right. And so it's, it's sort of a, not a, a clear cut what, what the specter of China means, but between what we're seeing in Europe now, where you have Meta not releasing Lama multimodal, you have Apple not releasing Apple intelligence, you have them arresting the, you know, the CEO of Telegram. Between that, you know, the sort of, you know, the Chinese model of, of AI development in the U.S. do you think that sort of Balkanization of AI and Balkanization of the Internet is inevitable? And how, how does that play out and impact, you know, how we think about AI's development?

I think that people have too much of a presumption that these things can actually be controlled. Like they can be controlled on the surface. The market for illegal drugs is a good example here. People believe that illegal drugs are bad and so they make all these laws trying to restrict them. It does not appear empirically that that actually works. Drugs are pretty much available in every city in the world, even the ones with the very most harshest consequences. And so you get this security theater where all you've done are prevent the most law abiding people from making their own choices. But the people that you're actually worried about are going to behave pretty much the same anyway. Things on the Internet tend to be accessible from anywhere, right? Like China cannot block people from accessing open source AI or anything available in the west to anyone with a modicum of like, technical sophistication. So it just, it cannot, it cannot be stopped. It cannot be blocked. And I think that's ultimately a good thing. I think it's good that politicians do not have the power to curtail these things or to prevent their citizens from free and open exploration of essentially what is mathematics.

So I actually want to take this in a slightly different direction then, kind of relating to SB 1047. So SB 1047 is happening, or the debate around it is happening, and OpenAI comes out opposed to 1047, but then a couple days later announces with, alongside Anthropic that they've reached a voluntary agreement for advanced model previews with, with the US Government with the nist. How do you think about arrangements like that that are sort of nominally voluntary? Like, is that a better path forward? You know, to the extent that there's going to be a relationship between private industry and government, or do you see problems with that model as well?

Well, first it's never voluntary. All this is happening under duress, softer or harder forms of duress. But I also think, like, I'm probably the outlier in believing that working with government agencies is bad. I think that the average person believes it's good for these large companies to have oversight because they have this belief that the government is some kind of benevolent force made up of a different clay than the rest of mankind. I just don't believe that. But I'm the extremist. I'm the outlier here. I think it's very damning that OpenAI has a ex NSA board member. That's so weird. That's so weird. And it's weird that Sam would come out and sort of like highlight it as something he's proud of. And he may be being authentic there. He may believe that the NSA is like a good agency that actually keeps the world safe and is important, but the whole thing just feels so bad to me. Again, the antidote here is open source. I think the big AI firms, they have a lot of challenges and they don't want governments to shut them down. They don't have a strong fundamental belief in some of these principles of privacy, of lack of censorship. I think they hold those views weakly, if at all, and so they don't mind sacrificing those things in order to keep operating, in order to not be shut down or various other consequences. So I think it's sad, but it's inevitable. And if Sam wasn't in charge at OpenAI, another CEO there would have done the same thing.

Yeah, it's interesting. I think, that you are. I mean, I don't know Sam, so I don't want to put words in his mouth or thoughts in his mind. I do think that from some of the comments that he's had, he has. It feels to me like he has a romantic sort of moonshot, 1960s notion of. Of what this could have been, that the US Government should have been leading the charge for sort of beneficial AI. I mean, he talked about how they went to the government right at the beginning of the product and said, you should invest in this, you know, and so I think that you might be right, that, you know, you can have a total difference of opinion, that he may actually think that that is. Is the best way forward.

Totally, yeah. He. He doesn't see the US Government as the primary threat, whereas I do. Right. So that's a fundamental difference, which I.

I think, you know, the part of why I love doing the show is that I think fundamental differences make us much better. It's also why. Why I still like America in spite of everything.

And, and to the degree that's true, fundamental differences making us better is, is. Is the entire reason why we should not allow artificial intelligence to be captured by one or a few large tech companies that are controlled by the government because they will shape people's thoughts into one monolithic. You know, like it'll destroy the ability to have diverse and different opinions. And if that kind of diversity of thought is important to humanity, which it obviously is, we can't, we can't accept that future.

So perfect segue because this is what I was going to ask you about next. So my read is that part of why again, if you sort of step back and look at the trend lines, big tech technology as an industry has been getting more and more powerful basically since especially the dawn of sort of social media, when all of a sudden platforms went from, you know, hundreds of millions of sort of engaged users to dominating big portions of our time. Right? Billions and billions of users, you know, algorithms shaping perspectives. I mean, I think, you know, TikTok, the culmination of this pre. I was sort of the debate that we've been having around TikTok and what that algorithm. But there's the, the point is that big tech, relative to the previous relationship between, you know, sort of the private sector and the government, there's been a, it's thrown it off balance compared to where it was. Big tech is obviously much more powerful than, than previous generations of companies. It seems to me that the, the government is trying to. There's an inevitable, that there will be an inevitable push for the government to grab more power as a bulwark against big tech. And so you bas two sets of actors who are, you know, getting more powerful at the same time, nominally in sort of in an attempt to constrain the other one. But the sort of the outcome is just these much more powerful centralized entities all around. Right. Like the net, sort of like average power of a centralized entity that you deal with has gone up. You speak eloquently about kind of how open source and decentralized solutions, you know, potentially provide a bulwark to that, especially with AI. How do you play out the scenario where that actually can happen? Like what, what is, what is the sort of the fundamental determinant of whether that will help and hold aside like regulation, whether it's allowed to. Is it consumer choice? Is it consumer education? Is it people? You know, what, what actually kind of constrains the power of big tech, for example?

Yeah, such a good question. I think like the English language needs two different terms here for power because the nature of power from a government and the nature of power from a company of any size are fundamentally different. When we say power of a government, a government has power to literally murder people at scale and does so frequently. Google doesn't murder people at scale or ever. A government has the power to imprison people at scale for their own personal choices, to take away their entire life and freedom. And I don't really care how big Google gets because it doesn't ever have that kind of power. It's a different, A coercive power is a different thing than market influence. Why am I not scared of TikTok? Because I don't have to open the app. And I feel like culturally we have removed so much personal responsibility from people that we've, we've forgotten or like lost the plot that someone can just not use the app. And people say, oh yeah, but society, you know, is like pressures us to do these things. Well, you know, take some, take some responsibility. Like if you don't like an app, don't use it. If you don't want your children influenced by TikTok, don't let them use TikTok. The idea that we need the government to come in and ban something because we simply can't make the choice ourselves really betrays that we are becoming sheep under a shepherd. And I don't think we should try to construct society in that way. So what constrains the power of big tech as you say we do? I constrain Google by not using it. I constrain TikTok by not using it. They can never force me to use it and I'm just one little guy and they have billions and billions of dollars and yet I don't have to use it. That's so fundamentally different from a government where they can do whatever they want to me, including kill me. That's fundamentally different. I'm more worried about how we constrain that power because I can't just not use the government. I can't opt out of that. So I think that's the crux of this conversation is that those two types of power are very different. And if you constrain tech or corporate power by giving more to the coercive one that can murder you, don't be surprised when you end up with a society that is much darker.

This might be the sort of optimist in me, but I do think that sometimes we are living through genuine big patterns of history changes I think right now. And in the midst of that, when you're sort of in the eye of that storm, things that historically will be seen as very short periods of time feel very long. And we're calibrating right now what we want our relationship to be with these uber powerful algorithms right and it feels like it's been a long time, like we've been watching this sort of happen. You know, Eli Perizur wrote the Filter bubble in like 2009. Right. The guy who founded moveon.org and you know, that sort of problem has gotten bigger and bigger and bigger. It's like it's 15 years later. 15 years is a blink of an eye in the context of history. Right. And if you take that sort of larger view that the way that this has to get addressed is just by people making different decisions, you know, in aggregate. There are some early indicators to me that there's that starting to happen and it's being reinforced by, by markets. And so here's the example that I'll give. Schools by and large haven't banned smartphones yet. But it's such an obvious thing to do. It's like, would you, you know, I think that the analogy that I've often seen is like, would you have let people bring a TV and a VCR and their full DVD collection, you know, to school back in the day? Is it like. Of course not. You know, that's like schools have that choice and they're just starting to do this now. But the other thing is hold aside just the schools parents are starting to find the line that's not, you can't be in touch but you also can't have free access. And it turns out it's like non smartphones. The Nokia brick is coming back. Nokia is actually having a resurgence with this like nostalgia driven, you know, it's got, it literally has snake on it and this is becoming the popular phone and like I have a three and a five year old and so we're a little ways from having to fully think about this, but it kind of feels like an obvious in between to have a thing where it's like yes, you can text your friends and stay, you know, touch and feel connected, but no, you can't, you know, just wander onto the darkest course of the Internet with, with unlimited, you know, freedom. And that's like a, that's. You see parents making those decisions that are different. You see schools starting to make those decisions. And again, like I said, it's, it's very nascent. I don't want to overstate the type of trend, but I think it's reflective of what you're talking about that like there are actual solutions that don't involve, you know, banning TikTok. It's sort of, you know, people are figuring out how to limit access to it.

Yeah, like should the federal government dictate how and when children can use cell phones. Obviously, cell phone use for children or adults can be and is damaging and harmful if done in certain ways. Right? It's. It's apparent that damage can, can happen if used poorly. Does that mean that the federal government should be telling people how to use phones? No, like, that's up to the individual people. And that's part of, like, what makes a diverse, free society is when you decentralize those kind of decisions. And even though you say it's obvious that schools should ban phones, that might be obvious to you and is probably obvious to many other people, but others might have a different opinion. And I think that's okay. Like, I think schools should choose for themselves what's happening, what their policies are in the school. I think parents should choose for their children what happens with their children. This decentralization of property rights and control is how you prevent bad decisions from becoming systemic. Because as you said, 15 years is a very short amount of time. Historically, we don't really know what the right way for humans to use cell phones is. And cell phones are changing. In five years, they might be quite different than they are today. None of us really know what is best for every other person on Earth. And we should have that humility. While we believe we know what's best for ourselves or our children or maybe our neighborhood or something, we should have the tolerance to permit different views and different beliefs. And that's why we should not outsource and delegate all control of everything in our lives to one set of politicians thousands of miles away. They believe they know what's better for everyone. And sometimes they might be right, but often, often they're wrong. You know, the food pyramid in the United States is a great example of this, right? Like that food pyramid the FDA put out. And that food pyramid has been gospel in the United States for a few generations. I grew up understanding that food pyramid and believing that that was like pure science, right? This is how you're supposed to structure your diet, and if you do that, you'll be healthy. And over the last 10 years, a lot of people are like, actually, it's like, maybe inverted. Not just that it was marginally wrong, but the whole thing might be inverted. We should not trust anyone to control the lives of every other person on Earth. That seems obvious to me, but I don't know why so many people struggle with that concept.

So let me ask, this is another perfect segue. It's like, we planned this. There will be a conversation on A national level about AI regulation.

Right.

Whatever happens. In fact, I think that if Governor Newsom decides to, I mean, he might veto SB 1047 if he decides to advance it. It almost mandates that we have a national conversation faster because the California dragnet then scoops up every company that does business with residents of California in ways that become problematic. Right. This was a central pillar of the opposition for many of these companies is that, you know, it's not just making a choice for Californians, it's making a choice for everyone. Right. In the same way that, you know, money laws, you know, that are state by state, create, create their own problems as well.

Yeah.

As we head into that sort of national conversation around AI regulation, you obviously are sort of totally admitted and clear about, you know, being on a, on a sort of fully to one side of sort of a set of philosophical and ethical principles in this. What would, what would you want to contribute to, to sort of the regulatory discourse, or at least the underpinnings of the regulatory discourse? If you could, if you could sort of enshrine one way of seeing or thinking as part of the conversation that started the sort of regulatory discourse, what would it be.

That the biggest danger is? Some kind of curtailing or banning of, of open source development in this field? I would obviously love there to not be any national regulation on AI, period. But it will happen, unfortunately, because people believe the government should regulate everything. But I think the most important bulwark against the bad outcomes is simply that open source be permitted. As long as open source is permitted and isn't hamstrung in certain ways by certain laws that drastically favor closed source models or something. If it's just permitted, then I think the world has the tools that it will need to navigate any of the challenges brought by AI. The really dark outcome is if open source gets curtailed in some way legally. That's really what I'm fearful of.

I have another question, but I think we shouldn't do it now. It's a preview for a future conversation. One of the things that I'd love to discuss with you in another show.

Is.

We'Re sort of in like this first paradigm of generative AI where the modality is chatbots.

Right.

And assistants. And we are quickly heading into potentially another paradigm where we're outsourcing even more of the actual decision making and reasoning to, you know, agentic systems. And I'm really interested in your take on how that amplifies or, or changes some of these issues that, that we're discussing here.

It's hard to know because we haven't really seen it occurring yet. It's sort of like always on the immediate horizon, and it's obviously inevitable, but it's not quite here. So it's hard to see how that manifests and comes out. I think it's going to draw a lot of attention to how do you protect yourself from a world of different agents, some benevolent, some evil, and many in between. And it might actually bring a lot of attention back to public key cryptography. And I think public key cryptography is really the one thing that we know AI can't break or can't do right? Like everything else is like, AI will do that, it'll do that, it'll do that probably better than all of us, but it can't break strong encryption. And so that that confinement really becomes like a foundation on which humans can protect themselves and understand what's going on in their own world. I often think of things through the cryptocurrency lens, and when you have agents running around the world, but you can protect your own wealth and control your own wealth through public key cryptography, then you really don't need to care too much about what those agents are doing. Similarly, while it's hard to know what if something is fake, you can prove that you created something through public key cryptography, right? You can create a key, publish that key, and then when you create something, sign with that key. And that proves mathematically to the world that that thing originated with that identity of the original key and those primitives, you can really build like a lot of very honest, honest parts of our modern society that have a great deal of integrity. And it's about like trusting that math again instead of trusting people or laws or regulations.

So one thing I think is clear is that we're likely to see a lot of realignment and readjustment. Right. I think that right now we're in a moment where the introduction of sort of these wildly new technologies and different systems. It's not clear how any one type of actor or when any one type of person is going to react to this. I think a great example of this is Zucker, Zuckerberg and Meta becoming the sort of, you know, the loudest champion for, for open source AI. I don't think that's something that many people would have had on their, on their bingo cards, you know, a little while ago. Zuckerberg recently also wrote a letter that deals with a lot of these issues that you've been mentioning on the show. What did you think when you saw that letter maybe give just a little bit of context for those who hadn't seen it.

Yeah, I was surprised that this wasn't a bigger story. Zuckerberg basically wrote a letter to Congress, like, three or four weeks ago. It was both a mea culpa, but also he was basically revealing to everyone what was sort of thought of as a conspiracy theory, which was that the government was pressuring Facebook, and obviously they were doing this with others to censor certain information during COVID or to not report on one thing or to not say one thing, or to control the information that was getting out. And I think a lot of people felt like that was probably happening, but it was always one of these, like, debated things. But he basically came out and said, yes, the government was telling us to not talk about certain things or to censor other things. And what was great and inspiring, and he deserves a lot of respect for this. Not only did he say that, but he said that he wishes he had pushed back more strongly on that, that he wouldn't have been as compliant or as willing to make those kind of censorships spec them. So that took a lot of courage for him to do that. And it's been really impressive to see over the last couple years. There's something obviously different about Zuckerberg. He's changed his behavior in some interesting ways, but this is obviously relevant for the AI space, because people don't know what truth is often, and especially not in the moment. And if we have a world where AI is controlled and licensed by government, then when there is something big and consequential and controversial, the government's truth and facts and science will be what everyone is supposed to abide by. And when we all, you know, we all lived through Covid, and it was scary and confusing and difficult, but I don't think that we would want situations like that to have their narratives controlled by central parties in the moment. Imagine a near future where the US Is in some kind of big military conflict with China or something. How would you feel if artificial intelligence was only permitted to talk about certain things in that conflict, couldn't answer certain questions, or had to answer certain questions in certain ways. Would that serve the American people? I don't think so. And really, what makes, I think, America special is that we do not have some kind of totalitarian government telling us what is right. We can believe for ourselves what is right, and we can debate what is right. And truth with humans often needs to emerge through vigorous discussion and skepticism. So that letter from Zuckerberg Was, I think, a big eye opener or should have been for many. And I wish it had gotten a lot more publicity. Like, why aren't people outraged that that was happening? It seems like maybe the media was, just because it was happening with all of them and they were involved in that kind of censorship and so they were a little embarrassed to even talk about it much.

Could be. You know, I think it's. Zuckerberg is fascinating. I mean, having watched his whole career, I think the one consistent thing is that he does pretty constantly reinvent himself. Not in like a flip floppy way, in a. Takes in new inputs, learns new things and comes to different conclusions sometimes, you know, and he's, he's one of the few people who has. One of the byproducts, I believe, of the total control that he's had over Meta for its entire life is that he has the privilege of actually just being who he wants to be and making the decisions that he wants to make without being sort of threatened. So he, he's kind of been able to do things a little bit differently. And I think this latest manifestation of that has been a positive one that I think a lot of people have been excited about. But going back to the broader point, because obviously Zuckerberg's sort of much more secondary than the, than the point of the letter, you know, or taking it into a slightly different place. You know, it's inescapable that right now the context that everything is being seen through, at least in the US is, is the presidential election. Have you been surprised by the, like, the, the level of, you know, or, or lack there of, of misinformation, explicit sort of, you know, use events of AI to, to influence conversations. Like, you know, I think the, one of the biggest fears heading into this election cycle was that AI would be running rampant, you know, tricking voters and, and trying to push them in different directions. You know, what do you think, you know, is. Is what we've seen sort of what you've expected, or, or, you know, what, what do you think about that whole side of things?

I think the fear that society understandably has had over the last few years was that like, deep fakes were going to be a really serious problem for humanity. And actually I think it isn't because people quite quickly realize that deep fakes can look completely authentic and sound completely authentic, and very quickly everyone just built a skepticism. So if you see some video come out that is just unbelievably bad in some way, many people's first reactions are going to be like, oh, that's probably not real. There's a healthy skepticism, I think, that limit the damage that deepfakes can do, and that's going to be a learning process. Society will have to learn to be skeptical of what they see and hear, which is probably a good skill anyway. Two is, no matter how nefarious AI bots are, I'm not sure that they can ever lie more than, than the actual politicians themselves. Like, I watched the debate a couple nights ago between Trump and Harris because I feel like it's important to watch these things, and I'm just embarrassed for the country. Like, both candidates are not respectable people. The amount of lies and deception, just in terms of, like, how they frame the debate, what they claim, what they don't acknowledge. You could go to any high school debate group and see people with more rigorous intellects talking. And these are supposed to be like the cream of the crop that democracy has brought up. For us to consider that we have Harris and Trump as the cream of the crop of American democracy is pretty damning. And I think far more, far more of a problem than, like, what the AI bots are doing. Like, the candidates themselves seem like the problem there is.

Joe Rogan had a conversation, I don't know, maybe a year ago now, maybe a little bit less than that. Everything is faster with, with Lex Friedman, I think. And they were talking about whether Rogan would vote for a bot for president at some point, whether, you know, just let. Let AI take in all the information, make different decisions. And he was sort of, you know, for it, maybe a little bit tongue in cheek because he was, he wasn't actually being forced to consider it. But I joked recently at a presentation that when that came out, a lot of people were super skeptical. And like, the farther into the election season we've gotten, the more people I see being a little bit more open to that possibility in the future.

Yeah, well, I mean, the bar is pretty low. I'd like to just be able to vote for, like, an inanimate object. Like, if I could vote for, like, a FICA or an orange to sit in the Oval Office, that would be great. They would do less harm, they would lie less frequently, they would be less deceptive, and I would really know what I was getting. So I would prefer that as, like, an option on the ballot.

So I think based on that answer, I have a fairly strong guess of what you're going to say. But how consequential do you think this election is for future development of AI in sort of the immediate term in the US.

Yeah, I don't. I don't know. I don't know. I feel like it's not consequential on the bigger issues. And the bigger. The biggest issue, I think, is the federal debt, which, amazingly, wasn't even asked about during the debate. Yeah. I think AI is so new in these candidates, it's not clear which candidate is actually better for AI or not. Frankly, I don't trust either of them. I'm not going to vote for either of them. Neither of them are respectful people, and I would never endorse either of them for anything.

I think it's a reasonable take. I tend to think that we don't really know yet where political alignment is going to lie as it relates to AI. It has not calcified and hardened into clear partisan lines in the same way that every other issue ultimately kind of becomes, you know, calcified and hardened into partisan lines now that there aren't indications of where it might go. But, you know, I think Chuck Schumer, when he came out with sort of a roadmap recently, or not recently now, six months ago now, it was. It was much more rah, rah innovation and we got to compete than I think people anticipated. I think people kind of had just assumed that Democrats would be pro safety and pro, you know, pro control. And. And it was. It was sort of differently positioned than that. Although, again, it wasn't policy. It was just suggestions for. For where they go next. So I. I tend to think that it would be overstating the case that it's, you know, that either. Either choice is going to lead to clear directions that the other one wouldn't. When it comes to AI policy.

Yeah. It's hard to know what they will do because they don't base their policies on any kind of consistent moral framework or principle. So I don't even know what the Democratic Party is today or the Republican Party. Right. Like, they both just grow the government. One of them acknowledges that they want more government, the other pretends they don't, but still keep voting for it. I don't even know what these parties are. Like, Dick Cheney is now endorsing Kamala. Like, the whole thing has just become such a circus. And unfortunately, like, in a circus, you can just leave. But here we're stuck in the tent.

So let's flip it around. Let's come to the other side of this and talk positively. Not for the sake of beautiful symmetry, but because, look, you're. As someone, I believe that it's impossible to be A builder in the way that you've been a builder without being on some level, base level, an optimist, or at least sort of have belief that things can change.

Change.

What's most exciting to you right now in AI, which could be technological, could be the way that people are responding to it, or changing, could be discourse around it. You know what gets you hopeful, jazzed, excited.

Time savings. Time savings. I think whenever, whenever civilization creates tools which save time for a given unit of work output, society gets substantially wealthier. True when the shovel was invented. True when we had a tractor instead of a horse on the field. True when we had electricity. All these things allowed humans to build and create what they want with less input cost. And AI does that in a powerful way across many industries. And perhaps the most powerful being, like coding and engineering itself, a competent coder now with AI can produce just an incredible amount. And that's a huge unlock for civilization. So, yeah, I'm definitely an optimist on the world. And it's the savings of the input costs that make AI so powerful. And I don't think people appreciate or respect that enough. Often people are, like, fearful of it. They think that now that AI can do all these things, people won't have jobs. It's like, yeah, 99% of people used to be farmers. And you know what? We lost pretty much all those jobs. Do we want to go back to that world where everyone was on the farm and had a job on the farm? I don't think so. We should not look at work as something that we want. We should look at the output as something we want. And if we find a way to get the output more easily, that's virtuous.

Yeah. Listen, I am in complete agreement whenever I have this conversation as it relates to likely impact on jobs and economies and stuff, the thing that always stands out to me is that the fear of, of job loss, at least when you move to the medium or short term or in the long term. Right, the short term, of course, there's going to be lots of chaos and transition and, you know, lots of pain and that will have to be dealt with. But when you look at the long term, it's like in a world where, you know, you can make the same amount of code with one tenth of the effort, are we all going to say, cool, the stuff that we have now, the stuff that we built, that's enough stuff, that's enough creation, we're just going to just do that one 10? It's like, no, we're going to do 10 times as much stuff, but it's actually going to be 100 or a thousand times more stuff. And I think the entire pattern of human history shows that we have basically unlimited appetite to create more and consume more. And that's never going to stop.

Yes, our base needs are somewhat limited, but our desires are unlimited. And many desires that we have today weren't even conceived of long ago. So the human mind is an incredible creative spirit. And when we have more powerful tools, I think net that's just good for us. Absolutely. It'll cause disruption. And I think while I use that farming example, the difference with AI is that I think these changes are going to happen much faster. So I don't want to be glib about it. There will be pain and a lot of confusion and struggle, but we should recognize those as consequences of civilization moving ahead rather than things that we should reverse.

Couldn't agree more. Well, Eric, awesome conversation. Let's have you back again and we'll keep talking about these themes. But always appreciate your time.

Cheers. It's been great. Thank you.