Why the AI Bubble Debate is Useless

**Nathaniel Whittemore** (0:00)
Today on The AI Daily Brief, why the AI bubble debate is, for most of us, pretty useless. Before that in the headlines, that White House AI executive order gets paused. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI.
All right, friends, quick announcements before we dive in. First of all, thank you to today's sponsors, KPMG, Blitzi, Rovo, and Robots and Pencils. To get an ad-free version of the show, go to patreon.com/aidailybrief, or you can subscribe on Apple Podcasts. And if you're interested in sponsoring the show, send us a note at sponsors at aidailybrief.ai. Lastly, today is the last day that you will hear me yapping about the AI ROI benchmarking study, at least until we have the results. I'm gonna close this survey up at the end of the day on Tuesday so we can dig in to all the numbers. So last chance to get in your use cases and get the full readout, you can find it at roisurvey.ai. Welcome back to the AI Daily Brief Headlines Edition, all the daily AI news you need in around five minutes. Today was supposed to be the day that we were getting a big White House executive order on AI. However, the plans for that executive order, including its approach to preempting state AI laws, have been scuttled as the White House faces pushback from Republican lawmakers. Last Wednesday, President Trump posted on Truth Social, Investment in AI is helping make the US economy the hottest in the world, but overregulation by the states is threatening to undermine this growth engine. We must have one federal standard instead of a patchwork of 50 state regulatory regimes. A draft of the executive order was later leaked to the press and disclosed some pretty heavy-handed tactics. The draft order, for example, instructed the Justice Department to set up an entire task force for suing individual states over the constitutional validity of their AI laws. There was also talk of withholding national broadband funding to states that had passed their own regulations. On Friday, Reuters reported that the executive order had been put on hold, citing sources but not having many details. Washington trade paper Punchbowl News had the scoop later in the day. They wrote that lawmakers were looking to negotiate a legislative compromise rather than have the White House address the situation by executive order. One solution being pushed by House Majority Leader Steve Scalise is to insert a provision into the must pass National Defense Authorization Act. Some lawmakers went on the record to criticize the administration's approach to the issue. California Republican Jay Obernolte said, I don't think the executive branch has the authority to enforce preemption on the states. If they found some legal angle, I haven't heard about it. Tom Tillis, who cast the deciding vote against preemption over the summer, said he'd quote, rather do it through the law rather than executive order, saying that an executive order would not quote, give us long-term certainty. Now, this is a little bit deeper than we normally get into the political jostling in DC, but the chain of events does seem to suggest a shift in power and perception around AI. Until now, the White House has enjoyed a lot of latitude to dictate policy direction on AI. However, now we're seeing multiple Republicans breaking ranks and doing so publicly. And it's very clear that Republicans don't like the idea of their White House being involved in multiple lawsuits against the states to block AI protections during an election year. Politico's Friday newsletter asserted that voters are ready to turn against the GOP on this issue. Tim Wu, a Columbia professor and Biden-era tech policy advisor, argued that the politically savvy move coming into the midterms is to come out in favor of AI regulations. He said, I don't think the public is too excited about losing their jobs to an army of robots. It is very clear to me that the political winds are shifting on this issue. And I sort of think the midterms are going to be a bit gruesome. Moving on, corporate insurers are seeking to exclude AI risk from their policies. The Financial Times reports that AIG, Great American, WR Berkeley have asked regulators for permission to offer policies that exclude AI risk. So what is actually going on here? There are a lot of nuanced issues. When it comes to underwriting the risks of using major LLMs, Dennis Bertram, the head of cyber insurance for Europe at Specialty Insurer Mosaic, commented, It's too much of a black box. Rajiv Dattani, the co-founder of an insurance startup called the Artificial Intelligence Underwriting Company, says no one knows who's liable if things go wrong. A handful of corporate liability claims have been brought so far and suggest that insurers could see a lot of exposure. A solar company called Wolf River Electric recently sued Google for $110 million claiming defamation. Google's AI overview feature had falsely claimed the company was under investigation by the Minnesota Attorney General. Last year, a tribunal ordered Air Canada to make good on a chatbot's offer of full refunds for travel that had actually taken place. Now, the damages were only a few hundred dollars, but the case highlighted the massive potential exposure if chatbots go haywire. For insurers, the issue isn't so much about idiosyncratic problems and one-off claims. Aon's head of cyber, Kevin Kalanick, said, Insurers can afford to pay claims in the hundreds of millions for isolated losses. But what they can't afford, he continued, is if an AI provider makes a mistake that ends up as a thousand or 10,000 losses, a systematic correlated aggregated risk. Moving over into the wild world of compute, Google needs to rapidly expand their AI infrastructure to keep up with demand. At an all-hands meeting at the beginning of the month, Ameen Vadat, a VP at Google Cloud, addressed the company on the topic. He remarked, The competition in AI infrastructure is the most critical and also the most expensive part of the AI race. A slide deck viewed by CNBC included a slide titled AI Compute Demand, which read, We must double every six months. The next 1000x in four to five years. The meeting came on the same week that Google reported earnings, forecasting a jump in 2025 capex for the second time this year. Now, unlike Meta and Microsoft, the market responded positively to Google's earnings. Many analysts concluded that Google's revenue growth was strong enough to support the more ambitious target. The reveal here was that behind closed doors, this year's infrastructure buildout is a tiny sliver of what's expected over the coming years. Vadat told employees that Google's job is of course to build this infrastructure, but it's not to outspend the competition necessarily. We're going to spend a lot, he added, but the real goal is to provide infrastructure that's more reliable, more performant and more scalable than what's available anywhere else. During the meeting, CEO Sundar Pichai told staff that 2026 would be intense, arguing that they're in a good place to deal with a boomer bust. He said, We're better positioned to withstand misses than other companies. It's a very competitive moment, so you can't rest on your laurels. We have a lot of hard work ahead, but again, I think we are well positioned through this moment. OpenAI and Johnny Ive continued to build out their consumer device team at the expense of Apple. In his Apple-focused newsletter, Bloomberg's Mark Gurman reported on the latest big news out of Cupertino. He said that rumors of CEO Tim Cook's imminent resignation were blown out of proportion. Last week, many linked the rumors to a failure to execute on an AI strategy, but in Gurman's opinion, Cook still has years left in the role. Succession planning is underway at Cupertino, but Gurman believes the reports of a resignation by the middle of next year were, in his words, simply false. However, some real news is that OpenAI continues to poach from Apple's hardware engineering team. Multiple Apple hardware executives had already moved across to Ive's new company, however, Gurman reports that Apple is facing a full-on exodus of talent. Over the past month, Gurman wrote, OpenAI has hired more than 40 people for its devices group, with many of the new hires coming over from Apple. He added, From what I've heard, Apple is none too pleased about OpenAI's poaching and some consider it a problem. The hires include key directors, a fairly senior designation, as well as managers and engineers. And they hail from a wide range of areas. Camera engineering, iPhone hardware, Mac hardware, silicon, device testing and reliability, industrial design, manufacturing, audio, smartwatches, vision pro development, software and human factors. In other words, OpenAI is picking up people from nearly every relevant Apple department. It's remarkable. Lastly today, Brett Taylor's Sierra is the latest AI startup to reach $100 million in ARR. Taylor, who also serves as chairman of OpenAI's board, founded the company in February of last year with former Google Labs executive Clay Baver. Sierra provides AI customer service and sales agents to enterprise clients. In a blog post, the co-founders remarked on reaching the milestone within seven quarters, writing, That's a heck of a lot quicker than we expected and makes Sierra one of the fastest growing enterprise software companies in history. In addition to their long list of tech forward customers like Discord, Ramp, Rivian and SoFi, they also have brought on customers with much older businesses like Vans, SiriusXM and Rocket Mortgage. Indeed, the co-founder said they were a little surprised that so many older companies were comfortable with integrating AI into their customer service workflows. I could do a whole episode honestly on this. What Sierra got right out of the gate was that doing AI for enterprises was not going to be creating a flashy app, agent or demo and hoping they figured it out on their own. It involves a lot of messy work of going in wiring systems together, making sure the company has the right type of dev support and enough of it. I think that this milestone is validation of an approach that many enterprise AI companies would do well to copy. For now, that's going to do it for the headlines. Next up, the main episode.