AI for landlords, Grok-2 unleashed, and the latest attempt at AI regulation

This week on Equity Kirsten Korosec and Mary Ann Azevedo were joined by one of TechCrunch’s resident AI experts, Devin Coldewey. They broke down their deals of the week: WeRide, a Chinese autonomous vehicle startup, seeking an initial public offering in the United States at a $5 billion valuation, EliseAI, the company offering AI solutions to landlords, raised $75 million and became a unicorn, and Grok-2 is now available on X but it will cost you. Then the team got into a couple recent startup shutdowns: Tally and Score. Tally was a nine-year-old fintech that helped consumers manage and pay off their credit card debt. Score was a dating app for people with good to excellent credit that was only around for a few months before it got sunsetted. And last but not least, we did a deeper dive into a California bill known as SB 1047 that is aimed at stopping real-world disasters caused by AI systems before they happen. Equity is TechCrunch’s flagship podcast, produced by Theresa Loconsolo, and posts every Monday, Wednesday and Friday. Subscribe to us on Apple Podcasts, Overcast, Spotify and all the casts. You also can follow Equity on X and Threads, at @EquityPod. For the full episode transcript, for those who prefer reading over listening, check out our full archive of episodes over at Simplecast. Credits: Equity is produced by Theresa Loconsolo with editing by Kell. Bryce Durbin is our Illustrator. We'd also like to thank the audience development team and Henry Pickavet, who manages TechCrunch audio products. Learn more about your ad choices. Visit megaphone.fm/adchoices

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:11Speaker 00

  Hello and welcome back to Equity, TechCrunch's flagship podcast about the business of startups. Today is August 16, 2024. I'm Maryann Azevedo and joining me today is TechCrunch Senior Reporter and Editor Kirsten Korosec. Kirsten, what's going on in your world this week?

• 0:27Kirsten Korosec?

  I mean, the only thing really that I'm interested in this world right now is the great Waymo Honk-a-thon that's been happening in San Francisco. I don't know if you've heard about this. To be clear, it has been resolved now, but this software engineer set up a livestream from her apartment, looking down at this parking depot that Waymo robotaxis are using. And every morning around 4:00 a.m., when they would come home to like, sleep, I guess, as they were parking, they are programmed to honk, like, you know, if a car is reversing into them. But when you have like dozens of these robotaxis all parking at once, it was like a chorus of honking every morning at 4:00 a.m. So she set up this livestream with this like lo-fi beats in the background now

• 1:12Kirsten Korosec?

  so you can observe them. And now the honking has been resolved, but that's been the chatter around the San Francisco streets and on my beat.

• 1:21Speaker 00

  Oh, my goodness.

• 1:22Speaker 02

  Lo-fi chill beats to honk in the park, too. Beautiful.

• 1:26Kirsten Korosec?

  Yes.

• 1:27Speaker 00

  Yeah, beautiful indeed. And hey, that voice you're hearing, that's Devon Coldaway. He's a Senior Editor at TechCrunch and one of our resident AI experts. We're thrilled to have him on the show today. Devon, how are you?

• 1:38Speaker 02

  Hey, I'm great. Yeah, AI is nothing but great news and all kinds of fun stuff all day long. So we're having a wonderful time covering AI these days.

• 1:48Speaker 00

  Yeah. Devon, how does it feel to be like the resident AI expert? Pressure?

• 1:52Speaker 02

  Well, it's funny because I I started covering AI because it was just like a weird scientific topic years and years ago. It was just like one of the many things that you'd cover is like, oh, there's a paper came out. They can now like use machine learning to do this and that. And that was in like 2016. And I just covered it as that thing. And then all of a sudden, AI becomes like this industry transforming technology. And everybody's like, Devin, you know about AI, right? And I'm like, no, no one does. I guess I'm the closest thing we've got. But like, literally, no one has any idea about this technology. But it's been interesting to learn along with everybody else. But it's nice to have the context of 10 years of writing about and learning about the technology in its infancy.

• 2:32Speaker 00

  Yes. So you were writing about it, learning about it before it was even, I guess.

• 2:38Speaker 02

  Cool. Yes.

• 2:40Speaker 00

  Very happy to have you on the show today. And so I'm going to run down our deals of the week. We've got some diverse but interesting deals. We're looking at WeRide, Elise AI, and Grok. Then we'll be taking a look at a couple of startup shutdowns that took place this week. And then lastly, looking at California's new AI bill. So Kirsten, China's autonomous vehicle startup WeRide is seeking a US IPO at a $5 billion valuation. That's a lot of money. Tell us more.

• 3:10Kirsten Korosec?

  Yeah, so you basically just gave the TLDR. They're hoping to debut coming up here soon on the Nasdaq. It's an interesting deal in that they're expecting to raise about 96 million. Although, if underwriters, you know, go for all the options, that's going to be about just shy of 120. Then on top of that, there are some investors, specifically the venture arm of Renault, Nissan Mitsubishi Alliance, that have agreed to buy shares worth about 320 million in a concurrent private placement. So that's the money part. The interesting part about the WeRide thing is, this is a Chinese autonomous vehicle company. They're considered one of the leading

• 3:51Kirsten Korosec?

  ones in China. They actually do testing here in the United States, in California specifically. The real story here to me is like the geopolitical intrigue because there are a ton of risks for this company tariffs, the geopolitical tension between China and the United States, increasing US government controls on exports that could potentially limit access to things like semiconductor chips, the data privacy, security concerns by the US government of, let's say, a Chinese autonomous vehicle company testing in the United States. So all of these risks, and yet We Ride is still willing to go to the US markets, which to me is the most interesting part about this.

• 4:35Speaker 00

  Pretty bold, huh?

• 4:36Kirsten Korosec?

  Yeah, I mean, I think it shows that it is difficult to raise money in China. I mean, there's certainly plenty of state government-supported incentives and subsidies, but in terms of raising money, The New York Times highlighted a survey by Deloitte that showed the number of IPOs in mainland China fell something like 75% in the first half of 2024. And this is a result of like Chinese market regulators tightening listing requirements and things like that. And so in spite of all these risks, We Ride's like, okay, we want to scale, we want money, let's try the US marketplace.

• 5:15Speaker 02

  That's always really interesting to me. I've seen this with some other Chinese companies that have come over to the US, and I wonder whether they really just have to budget it out as like, look, this could be a total crater, but the potential upside is so high and the potential of getting our foot in the door is so valuable that it's worth just burning that $50 million or $100 million, whatever it might be. It's just interesting to see the different businesses like enter this differently regulated environment and seeing what they do.

• 5:43Kirsten Korosec?

  Yeah, totally. And on that note, this same week, We Ride received approval to California is a little bit wonky when it comes to regulating AVs, and I'm not going to get into it here, but basically they've received another permit that's going to allow them to Test driverless with passengers in California. This is through the CPUC or California Public Utilities Commission. So you see that they're like, they're scaling here in the United States. Granted, still testing and stuff, but it used to be always viewed as like, oh, this talent play or, you know, grabbing technology. And I think it's changed. It's different now. I mean, I think talent's a big part of it, but it's also just playing in two totally

• 6:27Kirsten Korosec?

  different markets and accessing cash.

• 6:29Speaker 00

  Yeah. I mean, I don't know about everyone else, but I'm not super excited about more robo taxis anywhere, but who cares what I think?

• 6:37Kirsten Korosec?

  Well, you happen to live in a place with a lot of robo taxi testing. So I think that like, you are naturally, let's say cautious about them in your hometown.

• 6:49Speaker 00

  Yeah, I guess you could say that. Well, next up on my deal of the week is a startup called Elise AI. And Elise AI is this company that says it's developing a suite or has developed a suite of AI powered property management tools for landlords. They raised a $75 million Series D, now valued at $1 billion, which these days is not very common. I was intrigued by this. I know that on the team, we have some differing thoughts on this sort of tool, but what this company claims to do is help landlords, I guess, communicate better with prospective tenants and existing tenants. So if you're interested in apartment tour or learning about a certain complex, you

• 7:31Speaker 00

  can communicate with this chatbot and set up a tour. Or if you are a tenant, you can send in a request of, oh, I need something fixed and this chatbot will respond to you. So it's had some reviews that the chatbot is not always successful in terms of handing off requests to an actual human. So, you know, there are always limitations when it comes to bots. We have to be clear. But I was kind of surprised at how many customers this company has. It's been around already for about seven years. They have more than 350 customers, including 70% of the top 50 rental housing operators in the US.

• 8:07Kirsten Korosec?

  Oh, Devin has opinions on this one. I can tell.

• 8:10Speaker 02

  Well, I mean, I'll tell you what the biggest red flag, the whole thing is that 70% of the top 50 rental operators are customers. I just simply think that when it comes to landlords and property owners, any tools they use are designed to save them money, not renters. Rent is, keeps going up, and yet somehow the renters, the people who own these properties are using these, like constantly, there's tools that are cutting costs, they're making things more efficient, and yet the rent keeps going up. Isn't that a little weird? We've already seen automation, AI being put to the purpose of price fixing and collaboration between rental operators using platforms like this. They don't call each other and

• 8:52Speaker 02

  say, hey, we should raise rents in this area, but they all talk to a platform that automatically says, hey, everybody could probably raise rents a little bit in this area. And it's basically a form of collusion with a third party, which is a very common form of shifting the responsibility for collusion. Like antitrust authorities are very familiar with this kind of thing. And here you have something that is like, you know, a chatbot, and that's totally different. But I just think that if all these people are using it, the only explanation to me is it saves time and money for the property owners.

• 9:23Speaker 00

  Right. Yeah. Well, to your point, Devin, the company says that on average, the chatbots have boosted lease tour bookings by 125% and decreased overdue payments by 50%. Ding ding, that's what you were talking about. Yes, they want their money. They want their money on time. So that probably is a big driver here. Of course, they're saying, yes, we can help get tenants' requests sooner and help them with maintenance stuff, too. But I see your point. There is obviously this component of, we want to get paid on time. We want more people to be able to view our properties as well.

• 9:57Kirsten Korosec?

  So I was, Devin, you'll be happy to know that I was a villainous property owner, you know, landlord at one point in my life.

• 10:05Speaker 02

  Some of my best friends are landlords. It's okay. Seriously.

• 10:08Kirsten Korosec?

  It's okay. We're all, we're all human. And this was like a single home. So to be clear, this is not like some giant operation. So I would never use a product like this. I see a benefit of using an AI chatbot in terms of like scheduling tours or handling payment questions. I think where it becomes clearly a downside potentially for the renter is when it's about services that need to be provided to the renter, like a plumbing issue, or a leaky roof, or things like that. Now, if it takes away that friction and delay, great. But if the AI chatbot is simply just replying and then nothing is happening and

• 10:48Kirsten Korosec?

  you can't contact a human, that becomes really frustrating experience. And I could see, to Devin's point, why it would be incredibly popular with, you know, large scale property owners. To me, what actually was the most interesting thing about this deal is that they are expanding into healthcare, which, which I was like, whoa, okay. So I guess the founder thinks that the company's tech stack can be adapted to serve health clinics' administrative needs. There are a lot of AI chatbots out there already in healthcare. So, Mary Ann, I had some questions about that. Do you know how this would be applied? It seems like a competitive area.

• 11:28Speaker 00

  Well, I didn't report this out, but basically, yeah, like you said, that they want to try to help clinics do things like schedule appointments, with billing, with payments. And they actually, it was last year, I think, they launched a product called Health AI. So it's already out there. I find it kind of interesting because you're right, that's definitely a more crowded space. I think a lot of doctors' offices or clinics use some sort of technology like this. I don't really know more about it other than that, but definitely the CEO said that, exactly what you said, that they think that much of the company's tech stack can be adapted to serve these clinics' needs. And they said several providers are already using it. There are competitors, like one called Hirow,

• 12:10Speaker 00

  which also uses AI to handle text and voice conversations between healthcare organizations and their patients. I will say, I've been dealing a lot with healthcare related stuff, and I have seen an increase of like text messages, mostly for reminding us about payments. Appointments, of course, but there is something to be said, though, when I get a text about a payment as opposed to getting paper in the mail, I'm a lot more likely to click the link on my phone, set up the payment than I am to get that paper and then have to deal with, oh, okay, I'm gonna pay my bills. So anyway, I don't really know what Elise AI's chances are in this space. I think it's crowded, but I guess they have enough confidence in what they've built to try it out.

• 12:55Speaker 02

  Yeah, I think that there's certainly merit in meeting patients and renters and other folks in in those positions where they are. Whereas before, like you said, it's often, hey, call the office, send in the mail, mail the check to this or that or whatever. It's good to have these better tools. I'm just concerned that the tools are in the hands of the people who already, you know, screwed it up. You know, they already made it hard enough and they're like, oh, here's some tools to make it easier for you.

• 13:21Speaker 00

  Yeah, well, no, I think we're right to be skeptical. Speaking of tools, Devin, your deal of the week is XAI released Grok two and added image, added image generation on X. I don't know why that was so hard to say. Okay, I am clueless. I'll be honest. Please fill us in. What the heck does this mean?

• 13:40Speaker 02

  Well, you know, it's not like a deal deal so much, but like, you know, the AI world is so weird because it doesn't really make money. It's not really clear how anything works or who built what with whose money. And Grok is the most like the weirdest of all because it's Grok is the large language model, you know, like ChatGPT and Claude. It's a large language model, but it's built by XAI, which is Elon Musk's AI company that he spun up basically out of spite after breaking up with OpenAI and Sam Altman and all that. And it's made this chatbot, which, you know, by all accounts is fine. You know, it's very hard to tell how good these chatbots really are because there's no good measurements of

• 14:22Speaker 02

  like, oh, you know, they they test them on these benchmarks, but they're not the benchmarks are just like, you can game them. So he's putting it out there. He's making it available to everybody on Twitter. And there's a new version, version two that just came out. I think it's only available to paying. People right now. And so as such, it's being used to like, you know, generate poems about how beautiful Elon Musk is and stuff like that. But they also added, well, the thing is, it's like, there's a lot of chatbots being used in lots of different places. So when somebody releases a big new one, there's immediately a flurry of activity to find out like, is this significantly better? Is it worse? Has this been a huge waste of millions of dollars in computes in training it? So everybody's sort of scrambling to evaluate, and that's the phase we're in right now.

• 15:06Speaker 02

  But the more sort of immediately interesting part is that they tapped a startup called Black Forest Labs to provide the image generation capability. They're a German company and they are funded by Andreessen and a bunch of people. It's and they they have a totally uncensored image model or not uncensored completely. You can't do like porn on it, but you can, you know, like show and like everybody is showing pictures of Trump and Musk and everybody else. You know, like crashing planes into the World Trade Centers and stuff like that. It's like, oh, my God, you're seeing the imagery that's going around on Twitter right now. And it's clear that this is just a huge bull in a china shop situation.

• 15:51Speaker 02

  Then that is sometimes good because it shakes up other parts of the industry. So we may see Grok and X attract a lot of legal attention, almost certainly, and a lot of regulatory attention that might otherwise have had to be tested out by a smaller company or a more mainstream company. So it's just interesting to see someone just barge into the room and knock everything down. But nobody really knows what's going to go on.

• 16:21Kirsten Korosec?

  Well, Elon is definitely the person who barges in and knocks

• 16:24Speaker 02

  Yeah.

• 16:24Kirsten Korosec?

  Stuff down. Of course, the big flag here for me is, and it's nothing new, but heading into an election, the large amounts of misinformation that's already out there that AI is being used to propagate. And this is another tool that does that. And it just seems like a fire hose of more information that we as humans can like possibly disseminate and filter and digest. And so here is another one that has zero guardrails on it. So I'll be curious to see on the regulatory front, leading up to the election here in the US, if there is pressure. For Grok to to put some limits to its use. My guess is that there will be a big fight and that there won't be, but we'll see.

• 17:09Speaker 02

  I guess the problem is simply that most of these companies have voluntary limits. There's very few legal limits on what you can do, even when it comes to generating images of politicians and stuff like that. Like, if you're careful, you can tell Grok to generate a news release from the Biden-Harris campaign and an image of Harris and Walz doing something they never did and produce this whole fake narrative and spread it on X. And there's no guardrails against it because it's private speech on a private platform. And there's no FCC or FTC or FEC. There's no limits at all on this kind of private speech. But if it becomes amplified, as it does, it could have a huge effect. But there really is no statutory.

• 17:54Speaker 02

  You know, the senators can send angry letters, but they can't say, hey, you're under arrest for election interference, you know, because the rules are behind. And that's why we're trying to set up the, you know, new regulatory systems.

• 18:07Speaker 00

  It'll be super interesting to see how all that plays out. We've got a head to break, but after the break, we'll get into some sad news of startup shutdowns. So stay with us.

• 18:22Speaker 00

  So this week, we had not one, but two startup shutdowns, and both happened to be in the fintech space. Ouch. You know, contrary to what our readers might think, when we get scoops about shutdowns or layoffs, it's not something we enjoy reporting on. You know, yes, of course, we like to get scoops, but these are not fun things for us to write about. It's a bummer when companies have to shut down and lay off people. So many lives are impacted. So just want to say that upfront. So on Monday, actually, I covered the shutdown of a fintech called Tally. This one got a lot of attention, I think, because it raised so much money and was around for about nine years. It raised $172 million in funding.

• 19:05Speaker 00

  It had some pretty high-profile backers, including Andreessen Horowitz, Kleiner Perkins. And it was valued, I think, close to $855 million at the last time it raised, which was 2022. So you know, here's a company that had a lot of things going for it in terms of capital behind it, you know, high profile backers. And yet it shut down and it shut down because after exploring all options, quote, "it was unable to secure the necessary funding to continue its operations." So Tally was a fintech that helped consumers manage and pay off their credit card debt. They were what they kind of were dubbed as a debt robo advisor. And supposedly they were trying to help people consolidate debt into like lower interest rate loans.

• 19:47Speaker 00

  And I think, you know, just this current macroeconomic environment with the high interest rates and a lot of delinquencies, I don't know for sure, but I think all of that contributed to the company's demise.

• 19:59Kirsten Korosec?

  Yeah, I mean, here's the thing, there was also another shutdown this week, which was Score, and that was the dating app for people with good to excellent credit. And I want to talk about that in a minute. But in this one, there's been a number of shutdowns in your area, in my area of expertise in transportation. And I don't know if this is the case with this one, so I'd be curious what your opinion is, but I saw a ton of companies during COVID scrambling and raising a bunch of money, over hiring, over indexing. And then in 2023, and now sort of the lingering effects 2024, blaming it on economic conditions, but also they had completely spent more money than really their runway would allow them to exist.

• 20:42Kirsten Korosec?

  And so at least in my area, I'm seeing that happen. Companies shutting down because they just ran out of runway and they had over hired in 2021, 2022. Is that the case with this or is it really more like you said, an interest rates issue?

• 20:58Speaker 00

  Yeah, you know, I really, I really don't know. I did see a chart floating out there on social media of how the company's employee or their headcount had dropped pretty dramatically over the past year or two. Earlier this year, they actually were shifting. They pivoted from B to C. They were saying they were going to actually sunset their consumer app and shift to B to B, which was a huge red flag that something was not going well. If it had been doing this for like nine years and all of a sudden was shifting to B to B is not a good sign. So yeah, I really don't know for sure. I wish I knew more about what exactly went on behind the scenes. I will say though, that when I posted about this, a lot of people seemed genuinely like kind of bummed. Like this was one of those companies that actually had, was perceived to have more of

• 21:42Speaker 00

  a noble mission and not so much just out there trying to make a ton of money. But apparently really had this noble mission of wanting to help people who had found themselves in a lot of debt and wanted to dig themselves out of it. But yeah, I mean, when we're dealing with really high interest rates, we're dealing with more people having trouble making payments. I'm sure all of that contributed, but I cannot tell you for sure exactly what went on behind the scenes.

• 22:05Kirsten Korosec?

  So this.

• 22:06Speaker 00

  Oh, in the case of Score.

• 22:07Kirsten Korosec?

  Yeah.

• 22:07Speaker 00

  The case of Score was a little different.

• 22:10Kirsten Korosec?

  Yeah, so here's the thing, like, is this a fintech story or just another dating app that is having problems? Because there's a bunch of dating apps out there right now that are like, it seems like their time right now is over for a bit and they're just all struggling. So is this a fintech story or a dating app story?

• 22:27Speaker 00

  I mean, it's a good question. I think it's a little bit of both. So Dom, Dominique Madori Davis wrote about Score. Gosh, I guess it was a few months ago, February. Wow. Yeah. And so it was a dating app for people with, quote, "good to excellent credit." And of course, that attracted also a lot of attention because people were like, what? So you can, you get on and you're talking about your credit scores. I mean, huh? And is this classist? There was accusations of it being classist towards people who didn't have high credit scores. But the company maintains that it was only supposed to be a pop-up app and it was only supposed to be around for 90 days. But it received so much user interest that the parent company, which is called Neon Money Club, decided to keep it open for six months. It racked up about 18,000 users, made 8,000 matches, and then decided to shut down.

• 23:12Speaker 00

  So yeah, I don't know. I think it's probably a little bit of both. I mean, I've never used a dating app. I realize there are a bunch of them out there, but it feels like this was a very niche category or target market.

• 23:24Kirsten Korosec?

  Yeah, I want to also say like, what a weird idea to find someone with a high credit score. But I don't know. Maybe I'm an oddball here.

• 23:32Speaker 00

  Well, you know, I think the CEO, he said, or he told Dom that it was, it was really not just about that. He said, if anyone called this app classes, that they didn't read the mission behind it, the app connected a lot of good people who are looking to prioritize financial health. And it made people more aware of their credit health and it connected them to educational resources to help them with their credit journey. So, I mean, I don't know. You know, I was surprised when I first read about it and I'm, I'm not shocked about it shutting down, but I, I think I'm a little surprised about how quickly. But, you know, they're saying this was sort of the plan to begin with. And they're also kind of putting a call out of, hey, we're happy to let someone acquire what we've built or learned so far. If you're interested, call me.

• 24:13Kirsten Korosec?

  Hmm, interesting.

• 24:14Speaker 00

  Not me. That's what the CEO said.

• 24:15Speaker 02

  I think it's classic. Classist, let it die. I'm, I'm happy with letting this become a footnote in 2024's weird history.

• 24:25Speaker 00

  Yeah, I mean, I have to, I have to kind of agree that it, I don't love the concept.

• 24:30Kirsten Korosec?

  Yeah. So also, before we move on, I know we have one more thing to talk about, but like, are we going to start seeing a trend of like, when a startup dies, that they're like, oh, actually, JK, it was just meant to be a pop up all along.

• 24:41Speaker 02

  Just a pop up.

• 24:41Kirsten Korosec?

  Yeah. And not really something that was successful and supposed to make money, just a fun thing. You know, like, I don't know if I buy that, but.

• 24:49Speaker 02

  We're pivoting, pivoting to failure.

• 24:51Speaker 00

  Yeah, I would say, yeah, there, these were two very different types of shutdowns. One where the company had been around for a long time, raised a lot of capital, another that really seemed very much more experimental. But as mentioned, I think anytime there are shutdowns, we do, we do find it kind of to be sad, because again, there are actual humans being impacted by these things. So, moving on, we are lastly going to talk about this new AI bill in California, SB 1047. We touched upon this last week. It's supposedly aimed at preventing AI disasters, but a lot of people in Silicon Valley are warning that it will actually cause disasters. So, Devin. Of course, would love to hear what you think about this.

• 25:33Speaker 02

  Well, it's tough to say because AI regulation, everything is necessarily kind of ad hoc because no one knows how AI is developing or what it's really capable of. I'm not talking about like AGI and it's going to turn into Skynet or anything, but really like, is it good enough to take over this category of jobs? Is it good enough that we can rely on it for this government loan program or whatever? It's not really clear and a lot of people think that having any regulation right now is premature. But then everybody else is like, well, we have to have something because these guys are just running around making AIs that can do all kinds of stuff with no regulation whatsoever. Personally, I like the idea of sort of erring on the side of over regulating at the start and you can always pull that back. You can make little easements in the law.

• 26:16Speaker 02

  And I think that California has done that generally with emerging categories. They have been the one out front saying, hey, we're going to regulate with an eye to safety and this other stuff. And if you don't like it, deal with it.

• 26:31Speaker 00

  You know what? I neglected to actually describe what this bill is specifically aiming to do. So apologies, it's aiming to regulate AI at the model level and mandates that frontier models that cost at least a hundred million dollars and use something that I is a language that I don't understand, 10, something 26 flops during training, that they must conduct safety testing.

• 26:55Speaker 02

  No, as you mentioned, it's, this is the key thing is that it's in the law that it's a hundred million dollar models. Like if you make an AI model that costs you a hundred million dollars to train, or I think it's, you know, something like 10 terra flops. It's some enormous amount of compute to create, which at present it's not clear. Like maybe GPT four falls under this category. The next version of Llama might fall under this category. These are huge, huge, huge, expensive models made by giant companies. And the opposition to it is largely along the lines of basically, oh, this is going to chill the startup ecosystem. And it's like, look, if you've got a hundred million dollars, I would not call you a startup.

• 27:38Speaker 02

  And if this isn't even a hundred million dollars in funding, it's a hundred million dollars that you're spending on a single foundation model. Like that kind of outlay requires some kind of oversight, in my opinion. And so like when they say, oh, it's going to be a problem, it's going to harm our budding AI ecosystem. That's according to Fei-Fei Li, which she is an incredible scholar and entrepreneur. But she's, I think her, her company is also on the larger side and she's going to be doing foundation models. And all these companies would rather be unregulated. That's as clear as you can make it clear.

• 28:13Kirsten Korosec?

  Yeah. I mean, Devin, I think the question here is like, why do you hate innovation so much?

• 28:17Speaker 02

  Yeah, exactly. You know, and I, I hate it personally, but you know, it doesn't, we don't need to hate it structurally.

• 28:22Kirsten Korosec?

  But I did have an actual thing. We were talking about this earlier because I wanted to sort of like, understand exactly what was going on. And you said something that was interesting, which was like, hey, if you have the ability to spend a hundred million on building a foundation model, you probably have the million dollars to hire the lawyers required to comply with the regulations. And that this is really not hitting that startup ecosystem. My question for you is, is there any downside to regulation? For instance, startups at a very low level, maybe aspirationally will hit this a hundred million dollar or be so large that they would even qualify.

• 29:05Kirsten Korosec?

  Either setting up shop in a different country or state, or is, is there any downside to this bill in California?

• 29:12Speaker 02

  There's always the risk of flight, but I think that the risk is being deliberately overstated by the opposition. But it's paperwork, you know, it's not like they're saying, oh, you need to include a government backdoor. Like we sometimes hear from the intelligence community with privacy bills and stuff like that. This is not a substantive thing. Like you're filing paperwork and you have to have the ability to shut down the model on short notice. You know, if it's producing catastrophic harm, which I would hope you would want to have that anyway. So it's, I really don't think that this is too onerous of a requirement. And it's not even going to cost you a million bucks. Plus you'll probably be able to pay some startup that will appear to do this kind of compliance work for you.

• 29:54Speaker 02

  I just don't think that it's that big of a deal. And the fact that it targets such large models, which can only be created. By a handful of companies and people with this amount of capital. It's like when you have here in Washington, you know, there was an attempt to have a head tax or like a seat tax, basically, but only for companies that had, you know, more than a thousand employees or something like that. They're like, oh, this is going to be, this is going to stifle innovation. It's going to make people move out of state. And it's like, oh, what innovation like there's a thousand employees. This is only going to hit huge companies that can afford the extra like 1% or something like that. It's not even going to touch the small companies that are in a fragile state and, you know, need every penny to keep the lights on while they're researching their new frontier model.

• 30:39Speaker 02

  I just don't buy the argument against it. Although I do think the companies involved like OpenAI, Anthropic and venture capitalists have been lobbying against it and suggesting some amendments that kind of undercut some of the bill a bit. And I do believe we are going to see some of those amendments adopted. I don't know exactly which ones or when, but my guess is the bill will pass with some significant kneecapping. But it'll also, it'll serve as a useful ground floor to test out what the actual cost of regulation like this is. Cause in a year or two, when it comes into effect and all of a sudden everybody finds out, oh, it costs like $5,000

• 31:22Speaker 02

  and two hours of a lawyer's time to comply with this law. Then everybody who opposed it saying, oh God, it's going to shut down, you know, Meta is going to look like a fool and a total cry wolf kind of crybaby.

• 31:35Speaker 00

  Yeah, I have to say though, I kind of, well, I do agree with you. Let's err on the side of caution here and the state Senator Scott Weiner, who authored the bill. He said, we have a history with technology of waiting for harms to happen and then wringing our hands. Let's not wait for something bad to happen. Let's just get out ahead of it. You know, that kind of makes sense to me, but yeah, we'll see what happens. We'll see if there's amendments. We'll see if this passes. Let's stay tuned. Anyway, we're going to have to leave it there, but never fear equity will be back next week. Until then you can find us under the handle at Equity Pod on X and Threads. Have a wonderful weekend. Bye bye.

• 32:11Speaker 02

  Bye.

• 32:12Speaker 00

  Bye.

• 32:15Speaker 00

  Equity is produced by Teresa Loconsolo with editing by Kel. Bryce Durbin is our illustrator and we'd like to give a big thanks to our audience development team and Henry Pickavet, who manages TechCrunch Audio Products. Thanks so much for listening and we'll talk to you next time.