Trump & AI Policy + Overrated/Underrated on the Cerebral Valley Podcast

[00:00:04] Hey, it's Eric Newcomer. Welcome to the Cerebral Valley Podcast. I'm here with my good friends and Cerebral Valley co-host Max Child.

[00:00:12] Hello, hello.

[00:00:14] And James Wilsterman.

[00:00:15] What's up, Eric? Glad to be here.

[00:00:18] And we're going to talk about the miserable, happy, terrifying topic, the thing on everybody's mind, the election. We're going to try and focus a little on AI policy, but also just Silicon Valley's strong influence on the incoming Trump administration.

[00:00:34] James, how are you feeling right now?

[00:00:37] Well, I do feel like a bit of deja vu from 2016. And I would say I don't feel like shocked by the outcome in the same way. You know, I do feel like we're gonna, we're just going to go back on that roller coaster of Trump being so-

[00:00:52] We're resistance libs now? We're all sort of pink hats on or-

[00:00:57] I don't think we'll be, I don't think the world will look like that, at least not in the immediate days of the Trump administration.

[00:01:03] But who knows what could happen?

[00:01:05] Right.

[00:01:06] Right.

[00:01:07] Yeah, I mean, I think I had the same conversation like eight times yesterday, which is everyone's like, I'm pretty upset. I'm also not surprised. I'm way less surprised than I was eight years ago.

[00:01:20] And to the resistance lib point, I think everyone's kind of like, you know, I'm tired of this crap and we're going to have to fight and slog our way through the next four years. But I also don't think it's going to go quite as, you know, cuckoo pig hat direction as you were referencing.

[00:01:36] Well, I think there are a couple of things, you know, and I wrote a big piece, but, you know, one, that Trump actually won the majority makes a little bit more and that the majority voted for him really knowing what he's about. So it's like, man, they, this is what they wanted, you know?

[00:01:52] So I do think, I mean, none of us, you know, I think, yeah, I mean, I didn't vote for Trump. None of us voted for Trump. This is a, you know, coastal leak crowd right here. But yeah, I mean, this is what America wanted.

[00:02:04] I mean, the point I would make is just that, again, I mean, Kamala's going to probably run about four, five points behind Biden or something. That means one in 50 people switch sides, right?

[00:02:15] So if you imagine you're in a restaurant with 50 people, just one of them that voted for Biden last time switched over to Trump, right?

[00:02:22] Except in every single county.

[00:02:24] Except in every single state and every single district. So you could say, oh, the uniformity of that one in 50 switching is surprising to me.

[00:02:29] But, but if you look around a restaurant and you're like, yeah, I can see one to two people in here switching size, you know, it's not that shocking.

[00:02:36] And I mean, I guess it's maybe surprising that it happened everywhere.

[00:02:39] But after they saw like a train wreck, that's what's surprising. It's like, get that, put that conductor back in there.

[00:02:44] But the data, the data shows that people think the country is a train wreck right now, right?

[00:02:49] We, we all are like January 6th was a legendary train wreck, but the data shows more people think the country is a train wreck right now than like ever before.

[00:02:57] And whether or not you think that's right or wrong, that's how people are feeling.

[00:03:00] So one in 50 or two in 50 switching is just, it's just not that surprising.

[00:03:04] And I think everyone I talked to yesterday had that.

[00:03:06] And the prediction markets were right. I mean, I, I believe in the prediction markets, but I feel like the whale really made me skeptical.

[00:03:13] You were skeptical that that that whale was driving a lot of that shift towards Trump.

[00:03:17] No, I believe. Yeah, yeah, yeah. I kept going on the results. Now I'm like, oh, he's smart.

[00:03:21] Well, it turned out there was a story that he had paid a pollster to run polls.

[00:03:26] You know, there's this strategy where you say, what is your neighbor believe, you know?

[00:03:32] And, and that sort of reveals things that people otherwise might try to hide and that his sort of neighbor polling showed people even more supportive of Trump.

[00:03:42] And so it wasn't, you know, this guy was putting, you know, many millions of dollars to work on this bet.

[00:03:47] It makes sense that he, he did his research. Yeah.

[00:03:50] I mean, I'm not clear to me that his 30 million or whatever was actually driving the prediction markets on polymarket.

[00:03:57] Like the liquidity is pretty high. I think a lot of people were betting on Trump.

[00:04:01] So I do wonder how much it was intelligent, you know, analysis versus like the crypto industry leans sort of towards Republicans.

[00:04:10] Anyway, there are endless, endless takes to be out on Trump.

[00:04:13] I'm sure you're going to, you know, the pod bros or, I don't know, Matt Iglesias coming to us for, for our tech.

[00:04:21] So we will, we will now, now dig into what this means for AI policy.

[00:04:27] And tech policy in general.

[00:04:28] And tech policy.

[00:04:29] Yeah. I mean, I wanted to like frame this up.

[00:04:31] I think, you know, we talked a little bit earlier about where we think the Trump administration can have a big impact on tech and AI specifically.

[00:04:37] And kind of laid out like five key areas that I think would be good to walk through.

[00:04:41] And obviously if you guys feel like I'm missing anything, we'll throw it in there.

[00:04:44] But I think the big ones are energy policy.

[00:04:48] So, you know, nuclear, solar, how are we going to, you know, generate enough energy to power all these advancements in foundational models and so on.

[00:04:56] I think chips is a huge one.

[00:04:59] CPUs, GPUs, NVIDIA, Taiwan, you know, TSMC in Arizona, all that kind of stuff.

[00:05:04] Um, regulation, uh, as we discussed recently, the, uh, California state bill that was meant to regulate really large AI models was vetoed by Governor Newsom.

[00:05:14] But maybe there'll be a national regulation.

[00:05:16] Maybe there won't.

[00:05:17] Where do we think Trump leans there?

[00:05:19] Uh, markets and M&A is a huge one.

[00:05:21] Um, the mergers and acquisitions, the roll-ups of small tech companies have been severely depressed under the Biden administration.

[00:05:29] Lena Kahn, his Federal Trade Commission chair, is sort of openly anti-tech and anti-acquisition.

[00:05:35] Will that change under Trump?

[00:05:37] And then finally, I think you just can't leave out the old chaos, black swan madness of the Trump administration.

[00:05:43] Wild shit seemed to happen on a weekly basis.

[00:05:46] January 6th being the real cherry on top of that Sunday.

[00:05:49] But, uh, what, what kind of crazy stuff could happen?

[00:05:52] And how does that affect, you know, tech and AI?

[00:05:55] Obviously, hopefully not crazy enough to lead to, you know, bigger negative consequences.

[00:05:59] But I think-

[00:06:00] When you say nuclear.

[00:06:01] Uh, uh, well, nuclear or just Cold War, you know, Cold War II, you know, maybe get all that.

[00:06:07] Like a pandemic or something?

[00:06:08] Uh, yeah, like a crazy global pandemic.

[00:06:10] Yeah, I think that's, that was a good, uh, black swan.

[00:06:12] Not Trump's fault, though.

[00:06:13] I think we can, we can agree, maybe.

[00:06:14] Let's talk energy.

[00:06:15] How do we make enough energy to power GPT-8 or whatever's going to come in the next four years?

[00:06:21] Um, again, I think to frame this up, we're already seeing a loosening in nuclear policy and regulation.

[00:06:26] Uh, Texas is building more solar panels than the rest of the country because it's easier to build there.

[00:06:30] How are we going to make enough energy?

[00:06:32] And is Trump good for energy or not?

[00:06:34] Well, I will say that, uh, the bid to use nuclear to power that Amazon data center was rejected by the Federal Energy Regulatory Commission.

[00:06:45] Under Biden.

[00:06:46] So I think we'll see what happens, but I could definitely see Trump administration being pro-nuclear and reversing some of these decisions to reject efforts by big tech to use nuclear power, restart these nuclear reactors.

[00:07:00] Hopefully, you know, I, I think more likely than not he's friendlier to nuclear, but I don't think we, we know for sure.

[00:07:07] I mean, I think nuclear is a tough one because everyone's afraid of meltdowns, you know, Chernobyl, yada, yada, yada.

[00:07:13] Uh, even though I think you can objectively say nuclear hasn't killed that many people, nuclear energy for the record, not nuclear weapons.

[00:07:19] Right.

[00:07:19] It's just very scary to people.

[00:07:21] It's just got this fundamentally scary element to it.

[00:07:23] I think like populist, uh, yeah, popular.

[00:07:26] Yeah.

[00:07:26] It's not a populist energy source.

[00:07:27] And Trump is nothing if not very sensitive to populism.

[00:07:30] I think solar is a much more interesting one.

[00:07:32] I, I just keep seeing more and more data about how much cheaper solar has gotten in the last couple of decades.

[00:07:37] You know, it's now basically cheaper than any new source of energy.

[00:07:40] And Texas is rocking.

[00:07:41] Right.

[00:07:42] I mean, that is like just deregulate.

[00:07:45] Don't tell people what to do with their houses and you'll get a lot of solar.

[00:07:47] Yeah.

[00:07:48] And so that's Texas solar is really the argument that, you know, a deregulatory environment is perhaps better for the climate than people who vibe more with climate change, climate change being an issue.

[00:08:01] Like Texas is probably the most impactful state in the union in terms of progressing our clean energy future, because you can just throw up solar panels in like a couple of months there.

[00:08:09] And you can't do that anywhere else in America, certainly not California.

[00:08:13] And so arguably Texas is having a more positive impact on America's transition into clean energy than any other, you know, major state government or real federal government.

[00:08:23] Certainly.

[00:08:23] I hate that the first one is pro Trump.

[00:08:25] That's our bias.

[00:08:26] I showed it right there.

[00:08:27] I mean, one thing I mean, you know, yeah, I do think it's not surprising that on the measurement of unrestrained energy production without much regard for environmental impact, who's going to win any Republican generic Republican, I think is sort of a traditional answer there.

[00:08:47] I mean, I think to your point, we can caveat all this with Trump changes his mind all the time.

[00:08:52] He did Operation Warp Speed and now he basically wants to ban vaccines.

[00:08:55] So like, you know, no guarantees that their stated situation is where we'll end up with solar panels.

[00:09:01] But I think trying to be a little optimistic in the midst of sort of darkness of this experience that maybe this will be really good for clean energy in the United States.

[00:09:10] And obviously there's also just the risk of like favoritism where it's like Amazon maybe never gets their nuclear contract approved, but Elon gets three of them.

[00:09:20] So that that'll be another issue.

[00:09:22] Totally in play.

[00:09:24] Chips.

[00:09:25] This should be a winning issue for Biden.

[00:09:28] He passed legislation called the Chips Act.

[00:09:31] Could not be more overtly aggressive on this issue, trying to make sure that, you know, America is not dependent on China really to produce chips that power, you know, all our devices.

[00:09:43] Of course, today, you know, Taiwan, which at the end of the Biden administration is not China, is, you know, through TSMC, a major producer of silicon for the United States.

[00:09:56] I don't know what.

[00:09:57] Yeah.

[00:09:57] What do you think is Trump a gain or loss on the chips front, James?

[00:10:02] Trump in the last few months of the campaign was saying on the trail that the Chips Act was so bad that the only way to get these companies or the best way to get these companies back to manufacturing in America is to put a bunch of tariffs on their imports.

[00:10:18] So the they're forced back to America.

[00:10:21] So I'm a little skeptical that he's going to be sticking with the Biden strategy here, even though it seems to be succeeding in the U.S.

[00:10:33] But I could see him changing his mind to once kind of the clarity of the issue is in front of him.

[00:10:41] Yeah, I think like solar where or sorry, energy in general, where we were sort of like, yeah, based on what they're saying, this policy could be pretty good for energy production.

[00:10:50] I think we have to go based on what they're saying to James's point.

[00:10:54] And it's not good.

[00:10:55] Like, you know, they hate the Chips Act.

[00:10:58] Mike Johnson, the Speaker of the House, said they're going to try to repeal it, even though his like coasts.

[00:11:03] Speaker at an event was like, actually, that was really good for my district.

[00:11:06] And I'm a Republican.

[00:11:07] So can we not do that?

[00:11:08] So bizarre.

[00:11:09] It's industrial policy.

[00:11:10] Like, I don't understand.

[00:11:12] Tariffs are a terrible way to deal with this problem.

[00:11:15] I mean, I fundamentally will get this in the black swan section, but I think that Trump makes it much more likely that China will try to take over Taiwan in the next four years, which would be the worst possible thing for chips in the global on a global scale.

[00:11:27] So obviously the reason we need to bring semiconductor production here is because of the fear that we don't have access to Taiwan.

[00:11:37] As long as we do, we're in a pretty good spot.

[00:11:40] It's obviously up for debate, but I pretty strongly believe that Trump is enabling China and Russia.

[00:11:46] But let's just focus on China right now.

[00:11:48] And there's sort of leaders to act in a more dictatorial fashion, dictatorial fashion.

[00:11:51] And clearly the number one thing Xi Jinping wants to do is retake Taiwan.

[00:11:58] They have like a date set, right?

[00:12:00] Having said like 2026 or something.

[00:12:02] They're doing like practice blockades in the ocean around the island.

[00:12:06] They're like, oh, we're just going to throw like nine battleships in a little circle around Taiwan just to just to feel out the waters.

[00:12:13] So it's not it's not looking good, in my opinion.

[00:12:16] I don't know.

[00:12:17] I'm pretty I'm pretty pessimistic about the future of the Taiwan China conflict.

[00:12:22] All right, we're moving on to regulation.

[00:12:24] I think it's pretty obvious that the VCs and Elons of the world who helped get Trump elected believe that Trump just won't regulate.

[00:12:35] You know, that is Doge Department of Government Efficiency.

[00:12:41] We're going to cut everything.

[00:12:42] Why would we add any regulation to get in the way of AI?

[00:12:45] Let it rip, baby.

[00:12:46] We should talk about what is Elon trying to achieve from a regulatory perspective because he was supportive of SB 1047, the California bill that Governor Newsom vetoed this year.

[00:13:02] So I am a little confused.

[00:13:04] Maybe you guys have a better sense of what Elon prefers from a regulatory perspective.

[00:13:09] He prefers he be the one to do it, you know, and everyone else be regulated is my checks out.

[00:13:16] Didn't he didn't he suggest that SpaceX executives take be part of the Pentagon?

[00:13:20] Yeah, I think that was a story that came out today.

[00:13:23] It's it seems like it's what's good for Elon is apparently good for America.

[00:13:27] But yeah, I mean, Elon is extremely worried about sort of, you know, the big existential risks of AI.

[00:13:35] That's why he helped create open AI in the first place.

[00:13:38] Now open AI is another source of worry for him.

[00:13:42] So he's been supportive of rules to try and make sure we don't create some super intelligence while at the same time now building his own foundation model company with XAI.

[00:13:54] Yeah, the interesting thing does does Elon look for regulation when he's behind in an industry?

[00:14:00] And then, you know, where he's ahead, he looks for deregulation.

[00:14:04] I think if you look at even even things like SpaceX or Tesla, you know, he's advocating for deregulation because and potentially he would benefit from, you know, electric car subsidies being revoked because he's so far ahead that other companies are more reliant on those subsidies.

[00:14:22] Similar for SpaceX, you know, he's so far ahead that anything that allows him to launch more rockets faster is good for Elon.

[00:14:29] AI, not not as clear, because if he, you know, if open AI and Anthropic and others just kind of go at the pace they're going, you know, he's he's a little bit behind.

[00:14:39] Other AI related regulation, though, I guess would be something maybe we should discuss around wokeness of the AI or free speech.

[00:14:47] Right. Clearly, that's sort of racial bias and stuff is not going to get regulated.

[00:14:53] I also think, you know, Hollywood has a potential, you know, strong copyright interest in AI and they seem sort of out of the fold in a Trump administration.

[00:15:02] So you could have imagined a Harris administration sort of trying to do something to help protect copyright holders.

[00:15:09] And maybe that doesn't happen.

[00:15:11] I also think it's interesting to your point on the wokeness of AI issue, like, you know, J.D. Vance kept bringing up this like censorship of social media as like a terrible crime of a democratic administration.

[00:15:23] Right. And obviously, I think we all know that the current top of the line AI models, GPT and Claude from from Anthropic are, you know, there's a lot of quote unquote guardrails around what you can ask it.

[00:15:38] Right. You can ask it if you ask it to build like a pipe bomb or whatever.

[00:15:41] It's not going to tell you. Right. And then there's a lot of other sort of.

[00:15:45] I don't I don't know if you categorize them as woke, but you'd say safety related sort of actions taken by those companies.

[00:15:54] And my question is, like, is someone like J.D. Vance in the VP slot going to go crazy on these companies that are, you know, regulating the speech of AI, quote unquote, and try to open up these models and make them a little wackier and maybe a little less less safe when you talk to them.

[00:16:12] Yeah. The weird part about all this is there is no regulation on this right now.

[00:16:15] Now it's all sort of internally regulated by the companies.

[00:16:19] Yeah. OpenAI has a set of things they don't want ChatGPT saying to you.

[00:16:23] And it's true of XAI, of Grok.

[00:16:28] I believe when they launched their image model, you could ask for literally any copyrighted characters.

[00:16:34] You could ask for like Mickey Mouse holding assault rifles or, you know.

[00:16:39] Oh, yeah. It was a fun moment.

[00:16:41] Celebrities. Yeah.

[00:16:41] And I think Elon himself, you know, also rolled that back, whether that's, you know, due to existing copyright law or, you know, their own internal desire to not have that be what users can make in their in their tool.

[00:16:55] It wasn't really some unique new AI regulation that had him do that.

[00:17:00] Importantly, the Biden administration introduced an executive order that, in the words of the Andreessen Horowitz crew, regulated math or made, you know, try to make math illegal.

[00:17:12] It's like if you're doing a lot of math, if it's big enough, you need to register with the government and we're going to keep track of it.

[00:17:18] And so that that was mostly monitoring, but it certainly was a sign that, OK, government is going to be much faster this time around than when with social media to sort of police the development of AI.

[00:17:31] So there there's a big question of does the Trump administration pick up the ball and say, oh, that was a good idea.

[00:17:38] We should keep keep doing stuff there.

[00:17:40] Or do they say, oh, no, like our tech people, they don't really like that.

[00:17:44] And I think part of what we're sort of dancing around with Elon is like some of the tech people, you know, Mark Andreessen, Ben Horowitz, they endorse Trump.

[00:17:52] They don't want AI regulation.

[00:17:54] But then you've got Elon who seems to be very worried about it.

[00:17:58] So it's not clear that there's actually a Silicon Valley Trump supporter consensus.

[00:18:03] And so therefore, we don't know exactly what is going to what Trump is going to believe in terms of how much the, you know, either executive orders or pushing legislation needs to come come along in AI.

[00:18:16] I just really I think that I think this issue is going to come up around like job losses and job replacement at some point, like like all like like half of American politics.

[00:18:26] It's all about jobs in the end.

[00:18:27] And I think that if and when we get to the point where there's meaningful job losses among a certain sector of workers, some party is going to step in and be the anti-AI party for that aggrieved group of workers.

[00:18:39] If we get self-driving trucks, you know, pretty soon and the 1.2 million truck drivers are losing their jobs, like that's probably a Republican anti-AI group, right?

[00:18:50] Like, you know, F the AI that is driving our cars, like we need a regulation that mandates like there's always a human behind the wheel or something.

[00:18:57] And then on the other side, like for, you know, Democrats, if it turns out it's a bunch of like college educated, like middle management, white collar workers, you know, who, you know, legal associates or something who are all getting replaced.

[00:19:10] And there's like massive job losses in those markets.

[00:19:12] I feel like the Democrats might step up and start being the anti-AI party or the, you know, the white collar job party to protect against AI.

[00:19:21] So I just my prediction is that at some point in the next four years, there's going to be a meaningful anti-AI backlash built around some sort of job losses.

[00:19:29] And I think it's really about what job losses come quicker or what group of, you know, employees starts being more and more nervous about getting replaced by AI.

[00:19:39] Because if you look at the public polling, people are pretty nervous about AI already.

[00:19:43] And I think that's just going to keep escalating on a, you know, monthly, quarterly, yearly basis.

[00:19:48] Couldn't be more opposed to that.

[00:19:51] But yeah, I agree with you that both parties seem likely to take up the mantle.

[00:19:56] I mean, I guess it's going to be a popular viewpoint.

[00:19:59] Like they took our jobs.

[00:20:02] You know, America's version of like a safety net is like you should be able to get a job.

[00:20:07] And so if we're not going to give, you know, much stronger social benefits to people, then we're going to have to sort of take a heavy handed approach to making sure that they can get jobs.

[00:20:19] I would much prefer to say, you know, we attack jobs and find efficiencies, give people more health care coverage and safety net benefits than these sort of backwards looking job protections.

[00:20:32] I guess maybe a counter take to that would be that one of the main lessons of the last election was that people seem to hate inflation a lot more than they hate unemployment.

[00:20:44] So we'll maybe if AI is somewhat deflationary, you know, in terms of...

[00:20:50] Eggs, your eggs can be cheaper, all right?

[00:20:52] Yeah, productivity going up.

[00:20:54] There's robots.

[00:20:55] Robots farmers now.

[00:20:57] Picking those eggs.

[00:20:58] Yeah, I don't know.

[00:20:59] Maybe then it'll be hard to justify some of those regulations if, you know, you're seeing the benefits of a bit of the cooling of inflation.

[00:21:09] And I just don't...

[00:21:10] I really disagree with the idea that people hate inflation more than they hate unemployment.

[00:21:14] I think it's all about the order of magnitude of both of these things, right?

[00:21:17] Again, like, yeah, people dislike 9% on, you know, inflation much more than they dislike like 4% to 5% unemployment.

[00:21:24] But if you go back to the great financial crisis in 08, like, unemployment's the reason George Bush and John McCain got completely incinerated in that election, right?

[00:21:32] It wasn't inflation.

[00:21:33] It was, like, job losses and the economy crashing and everything.

[00:21:35] So I just don't...

[00:21:37] I think that if 5% of American workers lose their jobs due to AI in the next four years or something or 10, like, it's...

[00:21:44] People are going to hate that.

[00:21:45] That would be my prediction.

[00:21:47] If there were a prediction market for that, I would throw some dogecoin on it.

[00:21:52] I think one of the key reasons we all oppose Trump, among many, is just, like, he's a chaos agent.

[00:21:58] Who knows what's going to happen?

[00:22:00] Like, America is actually great right now.

[00:22:02] Like, why blow up the status quo in some ways?

[00:22:04] Obviously, we could make continued incremental improvements.

[00:22:07] But just, like, yeah, chaos is not necessarily a good thing.

[00:22:10] But, yeah, so I think predicting what's going to happen in the Trump administration without considering sort of insane things going to happen makes him seem sort of more normal than he is.

[00:22:21] We've already touched on Taiwan.

[00:22:23] And, of course, if China invades Taiwan, that hurts semiconductors.

[00:22:27] Are there other scenarios you can imagine where having Trump as president and this continued improvement of AI causes some weird problem or benefit?

[00:22:39] Yeah, I mean, I think we should talk about what the world would look like under Trump if we achieve AGI in the next couple of years.

[00:22:47] Maybe that's an early time frame, but that seems like what Sam Altman is often.

[00:22:52] Well, you know, some of these people think AGI will solve democracy.

[00:22:57] So maybe we can get rid of Trump early.

[00:22:59] We plug it in the computer.

[00:23:00] It's like, why do we need democracy?

[00:23:03] Or just, you know, it's like solves all these optimizations.

[00:23:06] I'm joking.

[00:23:07] I honestly don't believe that.

[00:23:08] And Dario at Anthropic wrote like a review of how he thinks like AGI could move us forward.

[00:23:14] And I think he was also skeptical that democracy would be one.

[00:23:18] But yeah.

[00:23:18] Sam Altman recently said on an interview that he expects us to kind of, you know, just blow by the AGI sort of goalpost and then society not to change that much.

[00:23:32] Similar to maybe how we don't feel like society changed that much by passing the Turing test.

[00:23:38] So maybe it just is irrelevant.

[00:23:40] But it could be pretty interesting if that's not the case.

[00:23:44] And AGI totally upends society in some ways.

[00:23:49] Max already mentioned the kind of job loss scenarios.

[00:23:52] Is there other anywhere else that we might see AGI like completely change American society?

[00:23:57] What about AGI, you know, what about AI mayors and AI governors and sort of AI actually taking positions of power in our government?

[00:24:05] Right.

[00:24:05] I mean, like I think that there's some candidate in I forget in Texas or somewhere else who said he's going to like put every decision that he has to do as mayor through chat GPT.

[00:24:15] Like, are we going to see that become a more meaningful part of the actual functioning of American government?

[00:24:21] That like decisions are being filtered through AGI or AI agents are carrying out administrative roles?

[00:24:26] Like, I think that's possible.

[00:24:28] Possible.

[00:24:29] One thing I think I would flag, though, is if Trump, whoever happens to be president, if there is this big inflection point in AI, while we, you know, fantasize about downsides could actually be beloved.

[00:24:42] You know, it's like, oh, all of a sudden we're making way more progress on health care without spending government money.

[00:24:49] All of a sudden we have, you know, these improvements.

[00:24:51] Things are going great.

[00:24:53] There's a world which like, especially if his message is like, I'm going to let capitalism run wild and capitalism delivers AGI and AGI actually is good for people.

[00:25:03] You know, the president, you know, sort of gets to benefit from whatever whatever is happening.

[00:25:08] And, you know, if there are huge improvements in the next couple of years, it could just sort of boost people's impression of a Trump presidency.

[00:25:17] I do.

[00:25:18] I do think that's possible.

[00:25:20] I could see maybe, Max, what you're saying.

[00:25:24] You know, there have been people who have tried to, I think, use chat GPT as their defense attorneys.

[00:25:30] At some point, like, does chat GPT just become better than your average government provided attorney?

[00:25:37] Maybe then it starts to be an interesting legal question of whether you're allowed to do that or not.

[00:25:42] It's hard to say whether Trump would be supportive of any of this.

[00:25:46] All right.

[00:25:47] So a quick look back.

[00:25:48] I think energy, we think, you know, Trump could be good for creating the energy production we need for AI, though, of course, could have any sort of terrible impact on climate change.

[00:26:00] Chip production, I think the risk that he escalates tension with China and therefore hurts our ability to get chips from TSMC in Taiwan is probably a negative.

[00:26:09] He's also somewhat a terrorist.

[00:26:12] He's somewhat opposed to the Chips Act, which is actually going well.

[00:26:15] And it seems like some of these plants are actually working.

[00:26:20] Regulation seems like he has competing points of view from his supporters.

[00:26:25] So it's unclear who he's going to listen to or how that plays out or whether they're sophisticated enough to do it.

[00:26:32] We didn't really dig into markets, but I think it's clear that right now the stock market is bullish about Trump.

[00:26:38] If there's not a lot of M&A review, there could be, you know, if antitrust is loosened up, there could be a lot of acquisitions.

[00:26:45] And then it's very good for M&A.

[00:27:15] Yeah.

[00:27:16] It could be quite bad for everything across the world, not just AI.

[00:27:20] But I think that that's always the black swan you have to throw in there with Trump because he's clearly shown an intolerance for being removed from the White House.

[00:27:29] Great.

[00:27:31] Well, I mean, I agree with you 100 percent, like to be clear.

[00:27:35] And I think generally the point is that you have to be alive and a functioning society to benefit from AI.

[00:27:42] And that if you are not, you won't.

[00:27:44] And so that is why I think many of us thought that was the number one issue of the markets and whether anybody can acquire anybody or not.

[00:27:54] All right. So now we're going to move on to our next segment here.

[00:27:58] A little game we like to call overrated, underrated or properly rated.

[00:28:04] The goal of this podcast is to sort of assess where we are today in AI, how various companies are doing.

[00:28:10] And I think what's interesting and what James and I have discovered as startup founders is there's often quite a lot of data available to anyone who signs up for the right websites

[00:28:21] about how companies are actually doing.

[00:28:24] And for these companies, a lot of which we've talked about in the AI startup space,

[00:28:29] the key indicators of success tend to be either revenue or users and usage.

[00:28:36] And you can find this kind of data for both categories.

[00:28:41] You can find it for companies that sell to other businesses, which are more revenue focused and consumer companies, which are more about usage and uptake.

[00:28:49] And so I thought it would be fun to kind of walk through about 10 top AI startups, talk through their actual data that we were able to collect from various sources,

[00:29:00] and then walk through if we think given this data, if the company is overrated, underrated or properly rated based on fundraising and hype and so on.

[00:29:10] And just to lay this out up front, we pulled the revenue data from the information, which we believe to be a fairly reputable source with respect to company revenue.

[00:29:18] Very reputable.

[00:29:19] Very reputable.

[00:29:19] But it's clearly a chart, right?

[00:29:22] So it's some of its directional.

[00:29:24] Very reputable.

[00:29:26] And then the usage data is pulled from two sources.

[00:29:29] One is called SimilarWeb, which I would consider to be the best in class tracker of website usage for more website-driven AI companies.

[00:29:39] And then data.ai is probably the best for app data.

[00:29:45] So iOS and Android app usage.

[00:29:48] Again, none of this stuff is like perfect within, you know, five or 10%, but it tends to be very good about order of magnitude.

[00:29:55] Is this a company with millions of users or tens of millions of users or thousands of users?

[00:30:00] And it also tends to be good directionally.

[00:30:02] Is this company growing, flat, or shrinking?

[00:30:05] Which I think is very valuable information.

[00:30:07] Especially in AI world where companies will have raised at billion-dollar valuations a year ago, and now people just sort of move on from them.

[00:30:16] And the only reason you might know why is if you look at their growth charts and they're reversing.

[00:30:21] Yeah, absolutely.

[00:30:24] So I'm going to pick 10.

[00:30:25] We're going to go through.

[00:30:27] Debate overrated, underrated, properly rated.

[00:30:30] I will frame it up first with the big kahuna OpenAI, i.e. ChatGPT.

[00:30:36] According to the information, OpenAI is currently making over $4 billion, with a B, in annualized revenue.

[00:30:44] And I was able to pull the data both on app usage and website usage for ChatGPT and OpenAI.

[00:30:51] OpenAI has over 25 million monthly users on its app in the United States and has over 284 million monthly users worldwide on its website.

[00:31:02] I think you could pretty safely say it probably has near a half billion monthly active users if you were to add in the global app data.

[00:31:10] US iOS, 25 million.

[00:31:12] Yeah.

[00:31:14] And I think most interesting is the growth curve.

[00:31:17] According to basically all third-party sources, OpenAI has grown almost 50% just in the last three months, meaning it's like 1.5x bigger than it was three months ago.

[00:31:26] That could be the back-to-school boom, as we've discussed in previous episodes.

[00:31:31] But it has also grown over 2x year-on-year in terms of usage, as in it has twice as many people using it as it did a year ago.

[00:31:37] Crazy.

[00:31:38] Huge growth.

[00:31:39] Yeah.

[00:31:39] So, overrated, underrated, properly rated.

[00:31:42] This is obviously the highest rated startup in the AI space, but I'm interested in your guys' take.

[00:31:47] James, interested in your take.

[00:31:49] Overrated, underrated, properly rated.

[00:31:52] Okay.

[00:31:53] So, $157 billion valuation.

[00:31:57] There's maybe some question marks around that valuation in terms of how investors get paid back and things like that.

[00:32:06] I would say probably properly rated.

[00:32:09] It does seem to be the technical leader in the foundation model space.

[00:32:15] The one thing that I found really interesting is their CFO said recently 75% of revenue comes from consumers, so mostly chat GPT subscriptions.

[00:32:24] The reason I would say properly rated is it's a huge amount of revenue from consumer subscriptions, and very few companies have that level of consumer subscription model, business model.

[00:32:36] But I do think that what's interesting is that the core of the company doesn't seem to be much of a consumer company.

[00:32:43] They did hire Kevin Weil, so maybe there's some efforts internally to bring in kind of consumer entrepreneurs and operators.

[00:32:52] But I think, yeah, I'm a little nervous about whether they can actually capture the value of having this massively successful consumer product sort of accidentally inserted into a research lab.

[00:33:03] I'm raring up and ready to go.

[00:33:05] Go, baby.

[00:33:07] I think it's overrated.

[00:33:09] I mean, I've written a whole column about how it is overvalued.

[00:33:13] You know, I think certainly the revenue growth, which is what, Max, you're focusing on, is astounding.

[00:33:19] And if you look at that growth and compare it to Google and Facebook, you're like, this valuation makes tons of sense.

[00:33:25] But, you know, it's burning like $5 billion this year.

[00:33:28] It's going to burn probably more next year.

[00:33:32] And all that money is being spent creating models that are getting commoditized.

[00:33:37] So I think that's a big challenge.

[00:33:38] But the thing I'll add, besides the piece you can go and read, is that I think it's also overrated in that people in Silicon Valley are excited about perplexity and anthropic and everything else.

[00:33:49] And there are lots of other sort of competitors that people don't know.

[00:33:53] And so the public massively rates OpenAI, ChatGPT.

[00:33:57] That's what they know.

[00:33:58] That's what they're using.

[00:33:59] And maybe that means that OpenAI runs away with it.

[00:34:01] But if you were just talking to the consumer, they absolutely overrate the quality of OpenAI relative to what's available on the market.

[00:34:10] Yeah, 100%.

[00:34:12] I'm not going to wait in too long, but the growth is freaking nuts.

[00:34:15] I think given that the models have not improved that much in the last year, we can all say 2x growth year on year and 50% quarter on quarter growth is absolutely monstrous.

[00:34:25] So to me, this kind of re-rated me to be in the more properly rated zone if it's multiple at 35x revenue.

[00:34:31] Like, that's just crazy growth given the models haven't even gotten that much better.

[00:34:35] If we actually get a better model and you continue to see even faster growth than this, the valuation is completely justified.

[00:34:42] So I'm sort of fascinated to see if this growth continues.

[00:34:46] All right, moving on to our next one.

[00:34:48] We'll go with the classic pairing with OpenAI, Anthropic, makers of Claude.ai.

[00:34:55] Anthropic is making about $100 million in revenue, according to the information.

[00:34:58] It has almost 10 million monthly users on its website.

[00:35:04] Not super meaningful app usage, maybe a half a million to a million.

[00:35:09] And it grew 20% last quarter in contrast to OpenAI's 50% growth in the last quarter.

[00:35:17] So less than half as fast growth of OpenAI.

[00:35:20] So this time I'll kick it off with you, Eric.

[00:35:23] Overrated, underrated, properly rated.

[00:35:27] I'm going to say underrated.

[00:35:31] Interesting.

[00:35:32] One thing that I think is interesting is just if Trump really loosens the M&A environment and Amazon is in a sort of dire position in the AI race, does Amazon acquire Anthropic?

[00:35:48] I think it becomes much more likely all of a sudden.

[00:35:51] And so I think it's the value of Anthropic is, you know, the amazing technical staff.

[00:35:58] They keep recruiting defectors from OpenAI.

[00:36:02] Obviously, I believe in the power of revenue and money coming in and OpenAI is dominating it.

[00:36:07] But in many ways, Anthropic has sort of the same thing.

[00:36:11] And if this is a great category, there's going to be more than one player and they'll figure out how to get customers.

[00:36:18] James?

[00:36:20] James?

[00:36:21] I'll agree with Eric. Underrated, I think they have a diversified strategy. Most of their revenue is coming from not from consumers, right? So they have like every reason to go after enterprise adoption and starting with what everyone basically agrees upon in terms of the comparison to OpenAI, that their coding models are better.

[00:36:45] Maybe that will change and who knows what OpenAI has internally, but they seem to be outperforming OpenAI in that very key market of developer adoption and coding capabilities. So I think underrated for that reason.

[00:37:06] Max?

[00:37:06] Yeah, I will quickly say based on this data, I think super overrated.

[00:37:12] Yeah, based on the data. Yeah, we're just ignoring the data.

[00:37:15] I mean, you guys can say whatever you want. The vibes for the record before I pulled this data, I would have been where you are, which is like Anthropic.

[00:37:22] The model's dope. Everybody loves it. Developers. It's like Silicon Valley darling. It's the cool, cool, artsy version of OpenAI.

[00:37:29] OK, guys, OpenAI is making 40 times as much money and has 25 times as many users as Anthropic and is growing more than twice as fast right now.

[00:37:39] And they're currently valued at like a third as much. It's absolutely cuckoo for Cocoa Puffs. It makes no fucking sense other than it's an M&A play.

[00:37:47] So I think Anthropic is great. I love the model. But based on this data, Anthropic's valuation is absolutely ludicrous in comparison to OpenAI.

[00:37:56] So I'll say super overrated. Let's keep going. Midjourney, the darling of the image generation universe.

[00:38:04] Now, I will caveat this with Midjourney is the one consumer company where data is kind of hard to come by because they kicked off on a discord, meaning a huge quantity of their usage is on a discord server.

[00:38:14] As far as we know, they haven't taken outside investment, right?

[00:38:16] As far as we know, they haven't taken outside investment. They I will give three pieces of data here.

[00:38:22] One is the annual revenue of Midjourney, according to the information, is two hundred million dollars or more and growing.

[00:38:32] I was able to get data on their web usage of Midjourney dot com, which is four million monthly active users and growth looks pretty flat on the website as in not growing, not shrinking, not much either way.

[00:38:45] Flat. So, James, Midjourney, overrated, underrated, properly rated?

[00:38:53] I think overrated at this point. I think they were clearly pioneers in this image generation space and they had a nice advantage.

[00:39:04] You know, people probably a lot of this revenue is recurring revenue from subscriptions that they probably acquired a while back.

[00:39:12] Like you said, maybe not so much growth recently.

[00:39:15] I think that the image generation model space has gotten more commoditized.

[00:39:20] There's really good other models out there.

[00:39:22] There are flux image generation being one of them that gets higher, higher, higher ratings in terms of realism and accuracy.

[00:39:30] So just a commoditized space.

[00:39:33] And I'm not sure that the consumer revenue side is going to keep growing from for Midjourney.

[00:39:38] Eric, I agree with everything James said, but I'm going to say underrated for the very reason that they just didn't take venture capital.

[00:39:45] It's all like house money. They're making money and they're pocketing it.

[00:39:50] And then so like even if they like aren't going to be a sort of grow, grow, grow story where they're going to be the next, you know, public $100 billion company.

[00:39:58] And still they made a killing along the way.

[00:40:02] So I say kudos to them.

[00:40:04] Yeah. Underrated.

[00:40:06] Yeah. I think I see both sides of the debate.

[00:40:08] So I mean, I support both of your takes.

[00:40:11] Well, then you're properly rated.

[00:40:13] You're properly rated.

[00:40:14] Yeah, I think properly rated.

[00:40:15] I think the lack of growth is spiriting.

[00:40:17] But to Eric's point, it's an absolute ATM machine for the founders.

[00:40:21] So who cares, I guess.

[00:40:22] All right.

[00:40:22] Moving on.

[00:40:23] Jasper.

[00:40:25] Frequently ridden a hyper roller coaster.

[00:40:27] Been high, been low.

[00:40:29] You guys can weigh in on where it is in the middle right now.

[00:40:31] According to information, making around $60 million in annualized revenue.

[00:40:35] It has 600,000 worldwide web views in the last month, according to similar web.

[00:40:41] And it's basically flat on growth.

[00:40:44] 600,000, not 600 million.

[00:40:46] Yeah.

[00:40:47] Overrated.

[00:40:49] Overrated.

[00:40:50] All right.

[00:40:50] Over.

[00:40:51] Over.

[00:40:53] Overrated.

[00:40:53] All right.

[00:40:54] James.

[00:40:54] No, no.

[00:40:54] I'm saying like, does the company even matter anymore?

[00:40:56] It was like, you know, we were fighting to get them at our March Cerebral Valley last

[00:41:02] year.

[00:41:02] And now I feel like I didn't hear anything about.

[00:41:04] Okay.

[00:41:05] James.

[00:41:06] I guess I, maybe I'm calibrating to what I thought was a company that was already to

[00:41:14] Eric's point, sort of like, sort of it's, it's time had passed.

[00:41:20] But I guess based on where I was coming into this podcast and, and the, the 60 million revenue

[00:41:25] figure, like, I think that's a, maybe they're a little underrated, uh, you know, given nobody,

[00:41:30] nobody's really talking about them, but they, they have a pretty substantial, uh, revenue

[00:41:34] base.

[00:41:35] If it's almost like more than a half of Anthropic.

[00:41:37] I don't know.

[00:41:38] That's pretty interesting.

[00:41:39] Um, maybe that's outdated data or something from the information.

[00:41:43] I'm not sure how recently that, uh, 60 million was reported, but, um, if it's, if it's kind

[00:41:48] of in that, uh, 60% of Anthropic, uh, I would, I would be a little bit longer Jasper

[00:41:54] than I was coming into this.

[00:41:56] All right.

[00:41:56] Yeah.

[00:41:57] I think flat growth is tough.

[00:41:58] I, I, I would only say properly rated because I think, uh, folks like Eric have already

[00:42:02] danced on their graves at some time.

[00:42:04] So, uh, properly rated as fairly negatively.

[00:42:07] Um, okay.

[00:42:09] Let's do Suno, the music generation darling, uh, that lets you create cool, uh, online songs

[00:42:16] using generative AI.

[00:42:17] They are at 40 million in ARR.

[00:42:20] They are at around 12 million in worldwide web views, or web users.

[00:42:25] I'm sorry.

[00:42:26] Uniques, uh, over the last month.

[00:42:27] And I think most interestingly, uh, according to similar web, they appear to be shrinking

[00:42:33] a little bit over the last three months.

[00:42:35] Um, again, caveat that the status from a third party source, but I think we can safely say

[00:42:39] they're not growing and they might be shrinking in the last three months.

[00:42:43] Um, so yeah, Eric or sorry, James.

[00:42:46] Yeah.

[00:42:46] I had some questions on this.

[00:42:47] So did you, you mentioned their, uh, mobile app adoption that, that they launched that?

[00:42:52] Like, I'm wondering if some of the web traffic is kind of, uh, is cannibalized by the mobile

[00:42:56] app.

[00:42:56] Yeah.

[00:42:57] The mobile app is quite small in comparison, the website right now.

[00:43:00] Um, so I, I don't think it's enough to offset the shrinkage, uh, which looked like 15% quarter

[00:43:06] on quarter, uh, on similar web.

[00:43:08] But, um, yeah, there's probably some cannibalization of the web app by the mobile app.

[00:43:12] But again, I just don't think it's enough to offset.

[00:43:14] The mobile app isn't growing super fast or something.

[00:43:17] It's growing fast, but it launched like two months ago.

[00:43:19] I mean, so basically doubled month on month from like a hundred K to 200 K, you know?

[00:43:23] So it's not like life changing numbers here.

[00:43:26] Yeah.

[00:43:27] What is it?

[00:43:27] So, uh, James overrated, underrated, properly rated Suno.

[00:43:32] I know you love Suno.

[00:43:34] Yeah, I love Suno.

[00:43:35] I love the product.

[00:43:35] They use it a lot.

[00:43:36] Um, I think they have definitely the best music generation model that I've seen.

[00:43:42] And they're also like really, uh, shipping pretty fast.

[00:43:45] You know, they shipped this pretty cool feature that, uh, allows you to.

[00:43:49] Take pictures and generate a soundtrack, uh, based on the picture.

[00:43:53] Uh, they have these cover features that are allowing you to kind of, um, you know, sing

[00:43:59] into however you sing, sing a version of the lyrics that you want.

[00:44:03] And it'll kind of put a style, um, onto that music.

[00:44:06] And you can actually save styles now that you're generating from music.

[00:44:09] So I just think they're very good, um, product company.

[00:44:13] And, um, I would say the 40 million is very impressive, uh, ARR.

[00:44:18] So I'm going to say underrated.

[00:44:19] I think, I mean, I think everyone knows there's going to be a bear case here of like the legal,

[00:44:24] um, uh, status of the company, but I could also see that being, you know, taking 10 years

[00:44:30] to sort out or something or leading to some sort of like Spotify acquisition.

[00:44:34] So, um, I don't know.

[00:44:36] I'm, I'm pretty bullish on them.

[00:44:37] Um, and, uh, I would say on, on underrated.

[00:44:41] Eric.

[00:44:42] I'd say properly rated.

[00:44:44] I mean, they were extremely buzzy at, uh, cerebral Valley, New York.

[00:44:49] Their CEO spoke.

[00:44:50] Um, so definitely I am saturated with the hype.

[00:44:56] Uh, and I have churned, you know, I didn't, I didn't find sort of continued use to sort

[00:45:03] of a fun, it's fun to like whip it up and like show somebody, Hey, I made a song about

[00:45:07] this thing.

[00:45:08] Um, but I am impressed.

[00:45:11] I mean, they're, they're like second on our list in terms of engagement, right?

[00:45:15] I mean, they're, they're ahead of everything, but yeah, I mean, they are people using it

[00:45:20] and using it.

[00:45:21] Yeah.

[00:45:22] It's like a clear new use case.

[00:45:23] Like so much of, you know, uh, AI is like dreaming up use cases or over extrapolating

[00:45:30] from chat GPT.

[00:45:31] And here is an actual generative product that people are using.

[00:45:36] So properly rated.

[00:45:37] Yeah.

[00:45:37] And I think we've talked a lot about like, you know, how far, how long it's going to

[00:45:41] take to make a Hollywood level movies with AI or even like novels, right?

[00:45:47] None of that's happened, but like music were basically there.

[00:45:50] Like the quality of Suno is, you know, I think 90% or 95% of, you know, the average,

[00:45:56] uh, pop song or something.

[00:45:58] So I mean, it's pretty high.

[00:45:59] Yeah.

[00:46:00] Maybe that's a little high, but at least like, I definitely think you could, you

[00:46:03] could just like take, take, you know, you'd have to like work at it and make some,

[00:46:07] some songs that, you know, you know, uh, Hollywood is going to come for

[00:46:11] them.

[00:46:11] I don't know.

[00:46:12] These things are all plagiarism machines, but, um, I think, I think, I

[00:46:15] think if you played songs like from someone who just woke up from a coma of

[00:46:19] the last pop songs of the last year and Suno songs, like you might, you might

[00:46:22] see some, uh, interesting data now.

[00:46:25] I think it's like self-driving cars.

[00:46:26] It's like, you need to get to a hundred percent, like 90% is like an

[00:46:29] uncanny Valley where you're like, uh, it sounds like a good song.

[00:46:37] I'm, I'm, I'm going to go like overrated as a consumer product.

[00:46:40] And maybe underrated as like a B2B product or, you know, some sort of like

[00:46:44] API.

[00:46:45] Cause I think that shrinking at this stage as a consumer company is, is really,

[00:46:49] really bad.

[00:46:50] I think that like, just, you know, whatever, like whatever spin you want to put

[00:46:53] on it and how cool the thing is.

[00:46:55] I just think shrinking is quite bad or being flat at 11 million, you know, even

[00:46:58] given the level of engagement error, but maybe, but they are, they are a

[00:47:01] subscription business.

[00:47:03] Right.

[00:47:03] So I don't think their revenue is shrinking.

[00:47:05] Yeah, no, I don't think it is, but I just mean the usage.

[00:47:07] Like, um, if we think it's a consumer product, like tons and tons of people

[00:47:11] are going to create songs like for fun.

[00:47:13] And that's going to be, I think that's how they see themselves.

[00:47:15] I don't think they see themselves as an API company.

[00:47:17] Yeah, I agree.

[00:47:19] I don't think that's how they see themselves, but I think like the idea of

[00:47:22] creating music is very valuable to a million different creative

[00:47:25] professions.

[00:47:26] Um, and I think that like, you know, gaming that we work in, I think, you

[00:47:30] know, having generated music is incredibly valuable and maybe it's an input to

[00:47:33] actual pop songs and so on and so forth.

[00:47:35] So I just think that, um, the product, the model is insane.

[00:47:39] The, the, I think the revenue is good to your point.

[00:47:41] I'm, I'm pretty worried about the shrinking consumer size.

[00:47:44] Well, I love it as an artist tool where it's like pay for performance and

[00:47:47] you know, gets all its money for like, we're one of three authors on this

[00:47:50] song and we get the royalties as a publisher.

[00:47:54] Yeah.

[00:47:54] Yeah.

[00:47:55] Okay.

[00:47:55] Let's, let's move on to one of my favorites.

[00:47:57] Uh, perplexity, uh, the, the search engine that's going to destroy

[00:48:02] Google according to Eric, um, perplexity 20 million ARR.

[00:48:07] According to the information, I thought there was recent reporting.

[00:48:09] They put it more in the thirties.

[00:48:10] Uh, so I, it might be more in the thirties now.

[00:48:14] Um, MAU, uh, of their iOS app in the U S.

[00:48:18] Um, I can pull the global MAU, but I, I find that iOS us users to be the

[00:48:22] most valuable for this kind of product.

[00:48:23] Um, is 900,000, uh, app users, uh, under a million, uh, and growth quite

[00:48:31] good to X in the last eight months.

[00:48:33] Um, I am also able to pull downloads, uh, data from, uh, data.ai and

[00:48:39] downloads are roughly flat over the last eight to nine months.

[00:48:42] Um, so kind of an interesting mix of growth narratives there.

[00:48:46] So clarification there, uh, the wall street journal reported recently that

[00:48:50] their revenue went to 50 million.

[00:48:52] So, okay.

[00:48:52] So that's much, much higher than the information reporting.

[00:48:55] So yeah, up to 50 million in revenue.

[00:48:57] So if that changes your perception, that's good, good upfront information.

[00:49:00] So James perplexity overrated, underrated, properly rated.

[00:49:05] I'm going to go, even despite the, uh, fast growth and revenue here, I'm

[00:49:09] going to go overrated, um, given, um, that they're reportedly raising, trying to raise

[00:49:16] that, uh, I think an $8 billion valuation.

[00:49:18] I think it's like one of the hypiest companies in Silicon Valley right now.

[00:49:23] Um, and what's interesting is that, um, they have a ton of competition, including

[00:49:31] open AI, which just released this, uh, search GPT product.

[00:49:34] I mean, who knows if any of these can compete with Google on search?

[00:49:38] Um, but yeah, it seems like, it seems like a Google is going to compete and be like open

[00:49:48] AI is going to compete aggressively in this space.

[00:49:51] Um, and you know, perplexity doesn't seem to have like a ton of sustainable technical

[00:49:56] advantages or like a super large subscription base currently of consumers or, you know, and

[00:50:02] also has a lot of, uh, potential risks on the copyright and a legal side as well.

[00:50:08] So I'm, I'm, I'm saying overrated.

[00:50:10] Eric?

[00:50:11] Eric.

[00:50:12] It's overrated.

[00:50:13] I mean, it's just so hyped.

[00:50:15] It's so hyped.

[00:50:16] And like, I do think, you know, Google pays a bunch of money to be in search bars.

[00:50:20] And like, I do think a lot of the search competition is just like being where people go to search

[00:50:25] and there's going to be a heavy lift, uh, winning on quality, but maybe they strike good

[00:50:32] partnerships or something.

[00:50:34] Yeah.

[00:50:35] I think overrated as well.

[00:50:36] I just think that the $8 billion valuation that they're supposedly raising at, I mean,

[00:50:40] like for context, like I think their total MAU globally across iOS and Android is maybe

[00:50:45] like eight, 8 million or something like that.

[00:50:47] And they're worth 8 billion.

[00:50:48] Right.

[00:50:49] Uh, Reddit is currently worth about $20 billion.

[00:50:51] So we'll say like two and a half X as much right now on the open public markets.

[00:50:55] And Reddit has about 1.2 billion active users.

[00:50:59] So they're currently being valued at about.

[00:51:01] And I'm sure they're much more engaged too.

[00:51:04] Yeah.

[00:51:04] About 40% as much as Reddit, uh, with like one, 1,000th, the number of users, or sorry,

[00:51:11] one, 100th, the number of users, like a hundred, uh, one, 100th or 150th of the users.

[00:51:16] Um, I think the only justification is the 10 to $50 million ARR growth between March and today.

[00:51:23] Basically.

[00:51:23] Yeah.

[00:51:24] I think the ARR growth is killer, but I just don't think the usage is that killer.

[00:51:28] Okay.

[00:51:28] I think that's enough companies.

[00:51:29] A lot of good takes in there.

[00:51:30] Uh, we'll save some more for the future.

[00:51:32] I think it would be great to get a wrap up and given your guys perception of these companies

[00:51:36] and the data we've seen so far, what do we think this tells us about the AI industry?

[00:51:41] What's valuable?

[00:51:42] Is it revenue?

[00:51:43] Is it usage?

[00:51:43] Is it growth?

[00:51:44] Is it a certain type of company?

[00:51:45] Do you have a more nuanced viewpoint?

[00:51:47] I'm fascinated to hear.

[00:51:48] So Eric, you want to kick it off?

[00:51:51] Uh, I mean, I think it's notable that, and not surprising, but notable that, you know,

[00:51:57] companies, some companies have peaked already and that there are a lot of mini hype cycles.

[00:52:01] Um, I would say, you know, like these companies are making money when they're hypey because they're

[00:52:06] subscription businesses in some cases.

[00:52:08] And so I think that is good for them.

[00:52:11] But yeah, the fact that there are sort of mini hype cycles is, is notable.

[00:52:15] I think, I think what's really interesting is that the majority of the revenue we discussed

[00:52:20] coming from consumer applications, right?

[00:52:22] And that includes open AI, um, includes mid journey, includes, um, perplexity.

[00:52:28] I mean, for the record, there are a lot of B2B companies making plenty of money.

[00:52:32] So this is just where we have more interesting engagement data, but yes.

[00:52:35] But I do think that a lot of it is with consumer.

[00:52:38] Yeah, that's, that's my take as well.

[00:52:40] And I think it, it's sort of interesting if you think back to, um, the internet era, like

[00:52:46] a lot of the companies that kind of emerged first and became the biggest companies were

[00:52:51] those consumer companies.

[00:52:52] Um, and then you had this later wave of like SAS companies, right?

[00:52:58] Like kind of copying a lot of like the consumer internet mechanics and, and, um, graphical

[00:53:04] interfaces and sort of, uh, adoption patterns.

[00:53:07] But I think, um, you know, maybe we're seeing something similar that, that consumer kind of

[00:53:12] leads the way and then, and then you get, uh, you know, later, later companies kind of

[00:53:17] adopting those, that technology and B2B.

[00:53:21] Yeah.

[00:53:21] I think I, I have a similar perspective to you guys, which is that this seems like the

[00:53:25] greatest consumer revenue, like boom time I've ever seen.

[00:53:29] I mean, I just never, I've never seen people sell stuff for money to consumers and in the

[00:53:34] internet post internet era.

[00:53:36] Right.

[00:53:36] I mean, the whole idea of the internet was like, everything's free.

[00:53:38] Give it away for free.

[00:53:39] Just get as many users as possible.

[00:53:40] Like sell ads later, you know, Google and Facebook and Snapchat and et cetera, et cetera.

[00:53:44] But like every single one of these companies is monetizing, like very effectively, uh, well,

[00:53:50] they're making revenue.

[00:53:51] The one thing we haven't been talking about is they're giving people a service that might

[00:53:55] caught, you know, that it might be more expensive to produce than it is.

[00:53:59] I think unit economics might be the issue here.

[00:54:01] Yeah.

[00:54:01] Upside out.

[00:54:01] Yeah.

[00:54:02] Yeah.

[00:54:02] I think that's fair, but I, I sort of, I don't know.

[00:54:06] I doubt that these things are deeply out underwater on unit economics.

[00:54:09] I think open AI can afford to sort of incinerate money and probably anthropic, but to your point,

[00:54:14] I think mid journey, our understanding is a super profitable, you know, my bet would be

[00:54:18] that a number of these other, you know, companies are, are unit economically doing quite well,

[00:54:23] whether it's like Suno or perplexity or, um, you know, any of these other ones.

[00:54:27] So I think that, yeah, it just feels like monetizing consumer products is like the narrative of

[00:54:34] the last year.

[00:54:35] Um, and it'll be interesting to see if that continues.

[00:54:38] All right.

[00:54:38] That's our show.

[00:54:39] Thanks for listening.

[00:54:40] Thanks Max, uh, for putting all that data together.

[00:54:43] Uh, cerebral Valley is November 20th.

[00:54:46] Uh, we will have talks from Dario at Anthropic.

[00:54:50] Yeah.

[00:54:50] Yeah.

[00:54:51] You'll get to hear, uh, on our YouTube channel and in this podcast feed from some of the people

[00:54:55] we're talking about here.

[00:54:56] And thanks so much for listening.

[00:54:58] Thank you.

[00:54:58] Thanks.