IBM's AI Rollercoaster, Demis Calls for AI Watchdog, NY Pauses AI Data Centers | Diet TBPN

0:00John Coogan:IBM is absolutely nuking. The stock is down 25%. Boom! IBM. Well, that is a crazy chart. What is that? That's the one week chart? It looks better on the five year because the stock is actually way up in the AI era since the launch of ChatGPT. IBM has done really, really well. The stock has basically doubled since the introduction of ChatGPT. During the AI era, you know, are you going to be a winner or a loser? Are you going to get steamrolled, slopped, something like that? But it's been doing well up until today when the company reset the narrative around their server business specifically. So the high-level reason that IBM is not well-positioned in the token path to use the Brad Gerstner and Gavin Baker parlance is that AI spending is currently flowing into GPUs, memory, networking, hyperscale cloud computing, and frontier model inference.

0:53John Coogan:IBM is not a major winner in those categories. So just to refresh on IBM, because it's an interesting business with a great name, International Business Machines. The first business machines they made were punch card systems. They made clocks. It's like you're running a business. You need a machine. You're going to need a great clock. You're going to need a clock. No, that really was part of the business. Not just any clock. Like a clock that works really well. That's right. Professional clock. Clock Pro Max. That keeps you on time. Exactly. Clock Pro Max. It's lighter. It's the lightest, thinnest, best looking, fastest.

1:22Not fastest.

1:23John Coogan:You don't want a fast clock. But tabulating machines, basically a bunch of different ways to process information mechanically. And that foundational insight was pretty simple. It was businesses will continually pay forever to automate record keeping. And at a high level, that's sort of been working forever. And they're continuously doing it. Do you know if they ever tried to sell a clock as a SaaS product, time as a service? Hmm. If you really, really squint Red Hat, Kubernetes, it's keeping time between distributed systems. Maybe there's something there. But when you're running a database across a bunch of different servers, there's some timekeeping aspect that's important.

2:05John Coogan:But no, I don't think they ever did. The IBM that people know, the mainframe business, that started in 1964. System 360, it was a compatible family of devices, which is interesting. It's not just one, people think one mainframe, but it was actually a whole bunch of different systems that you can upgrade piecemeal without redesigning the entire workflow. So you need a little bit more storage, you upgrade that. You need a little bit more compute, you upgrade that. And this turned IBM into the dominant supplier of corporate computing banks, insurers, airlines, manufacturers, governments. They all used IBM as the central system for their hardware and software.

2:42John Coogan:This was the mainframe era. And the whole reason that IBM in particular became dominant in mainframes was they focused on high reliability, long customer relationships, expensive switching costs. It's very difficult once you're in the IBM ecosystem to weed your way out. Proprietary software tied to the hardware, certain software would only run on IBM hardware, so you couldn't replatform. You had to rip everything out, which is very difficult for a large bank or a large airline in the 60s and 70s. Good for business. And they also had huge, huge support and consulting contracts associated with all the software and the hardware that they were delivering.

3:16John Coogan:Sort of a precursor to the forward deployed engineer, if you squint a little bit. But the PC era was the real turning point. So the IBM PC launched in 1981. This legitimized the personal computing market and set up two new companies, Intel and Microsoft, to capture immense value during the next computing boom. So the IBM PC ran Windows and used an Intel chip. At the time, IBM was doing$30 billion in revenue, Intel was doing less than$1 billion, and Microsoft was only doing$17 million in sales. And so I think Microsoft had like 120 employees. And all of a sudden, those two companies became ultimately way, way, way bigger, like 10 times as big.

3:54John Coogan:So the market eventually fractured and proposals to break IBM into separate companies started to pop up. the market fractured because once you had an Intel chipset and Windows operating system, you could run Windows on a different chipset and you could have a different chipset with different operating system. And the value capture piece, there were just other PC manufacturers that came in. And then obviously Apple with their anti-IBM, like challenge the man campaign. So the market was fracturing and there was a bunch of proposals during the 80s and 90s to break up the company into separate units.

4:24John Coogan:Lou Gerstner, who became CEO in 1993, rejected that idea. And he said, quote, we do not necessarily need to manufacture every piece of technology. We need to be the company that makes all of it work together. So we have to work together. We're going to be the integrator, the systems integrator. His strategy ultimately produced three things, IBM global services, large outsourcing contracts, and a vast consulting organization. And that's a lot of what we know about IBM today. So services businesses do have limitations though, lower margins, higher headcount, slower organic growth, price competition, etc.

4:56John Coogan:In 2019, IBM acquired Red Hat for$34 billion and spun off its traditional managed infrastructure outsourcing business in 2021. So today, you can think of IBM as sort of three key businesses. They have software, which is 44 % of the business. That's at 80 % gross margins. Great business. 31 % of their business is consulting. That's under 30 % gross margins, though. And then 23 % of the business infrastructure, which is just shy of 60 % gross margin. And so for the last three years, the stock's been doing really well, up 77 % before dividends, and the Red Hat acquisition started paying off. And the Z17 mainframe cycle was surprisingly solid.

5:34John Coogan:But the problem is that they just called out a shift away from mainframe spending with customers shifting capital spending towards the physical AI buildout. Demand for AI and associated hardware is strong, but IBM is losing share of their customers' technology budget. IBM still does have a strong asset for the AI era, Red Hat OpenShift, which is their enterprise Kubernetes platform for orchestrating workloads across multiple computers. But there are so many other companies offering AI capabilities up and down the stack that they're getting a little hammered today with the biggest share drop in its 115-year history.

6:10John Coogan:Rough day for IBM, but an interesting story nonetheless. the last. Demis Hassabis from DeepMind, the DeepMind chief, he has called for a U.S.-led body to test frontier AI models. He says society has a precious window to prepare for technology advancing at historic speed. He's a Nobel laureate. The Financial Times has the story, and there's an article that he posted on X that will sort of click through and give you the takeaways from the Financial Times. Google DeepMind chief executive Demis Hassabis has called for the creation of a U.S.-led standards body to test new frontier-class AI models for national security threats, arguing that urgent action from international regulators is needed to address the risks posed by rapidly advancing technology.

6:57John Coogan:I'm surprised. Has he never proposed this before? This feels like something that has been proposed many, many times, but maybe I'm just misremembering the AI 2027 people and the AI 2040 people and the OpenAI white paper and what Anthropic said. It feels like we've seen this before, of like we need to have a regulatory body of some sort. All the way going back to when the All In podcast was talking about an FDA for AI back like two or three years ago. But it's now here, and it's coming from a DeepMind executive, which hits a little harder. The warning from Hassabis, a Nobel laureate who leads Google's AI efforts, follows the White House's abrupt export ban on Anthropics' most advanced models last month, alongside a fresh wave of warnings about the potential for AI to disrupt the global economy and financial system.

7:44John Coogan:We talked a little bit about The Economist. They got together with a much more moderate proposal, I think, because it wasn't actually calling for any sort of change to the development of AI whatsoever or the rollout. And they were just saying AI could get better in the next 10 years. Yes. Which is a very sort of... Could get a lot better. That's what they said. They didn't just say better. They said a lot better. But the actual pitch from the economists was we need to have economists and government officials think about responses. If there is job displacement from AI, what is the impact? What will the reaction be to sort of like prep the legislation so you can be more ready when things start to happen, whether that's retraining or stimulus or jobs programs or all sorts of different things.

8:35John Coogan:So this intervention from Demis is the most detailed proposal yet for AI regulation from Google, which is vying for AI leadership with Anthropic and OpenAI. Quote, we've already seen the challenges frontier models pose for cybersecurity. Good point. And other threats, including nuclear and bio risks, may soon emerge as capabilities continue to advance. The rapid progress we're seeing in AI requires a new approach to testing frontier AI model capabilities that is dynamic, adaptable, and rigorous. The US is well positioned, given its economic and technical standing, to take the first step in developing such a framework.

9:08John Coogan:My big question is, it seems pretty easy to go to the leading labs and say, hey, you have to go through this process. But do we have a good framework in the United States for reviewing code that China just sort of throws over here, open sourced, because I mean, as we've seen with the Kimi K2 and GLM, like if you tie someone up in an FDA like review for even six months, let alone a year, let alone what the FDA timelines are for drug development, five years, 10 years, sometimes you are going to have open source models that are way, way more advanced. So new frontier model. Yeah. It's going like a one-month review and then there needs to be with ai it could be a two-minute review yeah i would hope but let's say it's like a three-month yeah a three-month delay or six-month delay then when it eventually does get released it gets distilled likely within even less time than that is publicly available yeah i would like to see so we keep seeing these like letters and proposals And they always come, one, with a request for urgent action, but they rarely come with super concrete scenarios, like near-term scenarios.

10:27I want, here's what's going to happen in six months. Here's what's going to happen in 12 months.

10:31John Coogan:Or even just like a trigger. Like, it would be interesting if somebody said, if the unemployment rate goes above 10%, I would recommend a stimulus check of$1 ,000 be sent to everyone and means test it so it only goes to the middle class and lower class. Like, that is a very reasonable thing. That's basically what happened during COVID, right? Like, the unemployment rate went to 15%, and then, boom, there were checks in the mail. And that's a very concrete proposal that you could say, if this happens, then this happens. Yeah, I want someone like Demis. Basically, the world of LessWrong and AI 2027 and 2040, they're willing to lay out super concrete scenarios.

11:10And they can at times come across as very sci-fi. But there's always, at least so far, been some element of reality in them. I hear what you're saying. I want somebody who's generally more kind of moderate to come in and just say, here's a few potential scenarios and this is what I think. because I don't believe it's, you know, Demas could suggest what he thinks that the government should do. The U.S. government, in this case, he's, you know, encouraging like a U.S. watchdog. He's in London though. So I think it's going to be on our lawmakers and our government to understand in these different scenarios, at least start thinking through in these different scenarios, how would we approach them?

11:54Yeah.

11:55John Coogan:I just, I always have a problem with the timelines and predictions because those can get so nitpicked and they're so hard. I'd be more interested in less of like... No, but don't you think that'd be helpful if... I don't think it's helpful. No, no, no. I actually don't. I think it's much more helpful to say if the unemployment rate goes to 10%, create a new government body that hires people to do something. Like create the next TSA or send out stimulus checks or lower interest rates, right? If you tell Washington DC, AI models are very good at hacking computer systems and they're going to get better at hacking computer systems, there's not really much for them to do with that because hacking computer systems are already, it's already illegal.

12:38And the solution there is for companies to beef up their own cybersecurity, make sure they're using the most advanced models. And so if you play out more concrete scenarios where like here's a timeline for the trucking industry and potential job displacement within trucking or any of these other categories. I just think it allows people in Washington, actual lawmakers to start thinking about it.

13:01John Coogan:I just think that's always wrong. Like they're always wrong about those predictions. It's so much better to just say, look, if the trucking industry goes through mass job displacement, then here's what I actually propose. Here's the solution. As opposed to just saying like, there might be a problem and I think that there's a problem coming down the pipe. I don't know. Like it's like, what are you actually advocating for? Other than just being like, the sky might fall. I have a P-doom of this number and it's your job to go figure it out. It's like, you're smart. What do you suggest? UBI? Higher taxes?

13:29I just don't think predictions are always wrong. There's been so many examples over the last decade where people have gotten predictions dead on.

13:37John Coogan:Yeah. Situational awareness. Tyler, what do you think about this? Yeah, I mean, I'm probably in the camp of productive regulation. Usually it has some bad consequences. It doesn't really work out. But also, I was just going to say, what he's describing is basically just Casey, the Center for AI Sanderson Innovation, which is under the Commerce Department. Yep. And it's like a slightly beefed up version because right now Casey is like very much you opt in. Yeah. But it seems like he should just have said like you should specifically beef up Casey, add these policies, do these like specific things.

14:08Yeah. And I think that would be much more palatable or well-received or like something like, like what do we actually do with this letter? It's kind of very, you know, like I'm not sure what we actually do with this.

14:18John Coogan:Yeah. It feels like if, I mean, to go back to cybersecurity, it's like if it's a national issue, like the NSA works on this stuff, increase their budget, maybe raise taxes to increase their budget, issue more debt to increase their budget. If it can be solved by the private market, it's like go support CrowdStrike or start a new company that can help with cybersecurity. I don't know. The actual concrete recommendation boiled down from what Demis wrote is something along these lines. create a U.S. frontier AI standards body that's like Casey, but probably more beefed up. He's also advocating that it's overseen federally, but funded by AI companies.

14:55John Coogan:Define and regularly update benchmarks to determine which models and labs qualify as frontier. So that's something that doesn't exist yet. Require frontier labs to submit models for testing up to 30 days before release. That's sort of nice because that would allow someone who's just going and building and recommender system on Netflix that's actually, it is using AI technically, but it's not a frontier model because it doesn't qualify for that. So then they can just go and ship the latest recommendation algorithm on Netflix, no big deal. Test models for cyber, biological, nuclear, deception, autonomy, and guardrail passing capabilities require strong cybersecurity, personal vetting, model cards, watermarking, and substantial safety research.

15:37John Coogan:Use national labs, federal agencies, and independent third-party auditors to conduct evaluations, develop independent confidential tests so labs cannot train specifically against the benchmarks, require labs to fix serious vulnerabilities discovered after release, apply the rules to all frontier models deployed in the United States, including foreign and open source models while exempting smaller models. Okay, so he wants to apply it to foreign open source models. That feels very tricky, but I guess you could get like sort of DMCA notices to Hugging Face and GitHub so that it doesn't proliferate across the web.

16:11Yeah, I mean, there's probably some power lot to where people are actually downloading

16:14John Coogan:and inferencing the model. And I guess if you go to all the NeoClouds and all the open source folks, and you're like, okay, this model is actually a bad model, you've got to stick to GLM 5.2. Yeah, it's much easier to regulate the compute. This is going to be very controversial to the open source fans. Yeah, this is kind of like the George Haas nightmare take. You're tagging every single GPU. For sure, for sure. Coordinate a slowdown among Frontier Labs if testing reveals sufficiently serious risks and turn the US framework into an international system of shared frontier AI standards. Well, I like the general direction.

16:46John Coogan:I like the idea that he's just sharing his viewpoint more broadly. I think all of that is good. I'm not sure that there's enough to dig into here exactly how this would, like where the rubber meets the road, how this would be implemented or what effect this would actually have on the industry. Like this could be really good for open source because it could just slow down the frontier, closed source labs. It could also be really bad for open source if it's much more cumbersome because an open source project might not have a regulatory budget to actually massage a model through the approval process.

17:15John Coogan:There's a reason why small biotech companies get acquired by big pharma before they launch their drugs. It's because the big pharma companies have offices in Washington, D.C. and can walk the legislators through the whole process. So in general, I'm sympathetic to the view where people say, Oh, regulation benefits the biggest companies in the world because historically that's how it's played out. Maybe that's different this time. Who knows? It could just slow down the frontier. But, you know, he does work for a leading lab. So let's move on. What's going on in New York? I will tell you what's going on in New York.

17:48John Coogan:Today, New York Governor Kathy Hochul signed an executive order placing a one-year pause on new AI data centers in the state. This is the wah-wah for everyone except 70 % of Americans. People that are probably excited about it. But the order establishes a moratorium while New York develops a regulatory framework and conducts environmental impact assessments examining data centers' energy demand, water use, water quality, air quality, and effects on the electric grid. You would think there's a decent amount of oversight around those things generally already, like air quality. Like whether you start a new barbecue restaurant or a coal plant, you would imagine that there's just a general rule about not polluting the atmosphere that would apply to data centers by default.

18:41John Coogan:But it seems like there's a little bit of a special case here. And so they're working on this in particular. The move immediately drew criticism from the tech industry, which argues that restricting data center construction will cost local communities jobs and weaken America's position in the global AI race. Earlier this year, Maine considered a similar moratorium, but Democratic Governor Janet Mills vetoed the proposal after concerns it would block a major data center planned for a town still struggling after the closure of a local paper mill. Hochul's Republican challenger, Bruce Blakeman, also opposes the moratorium, arguing that local governments, not the state, should decide whether to approve projects that promise significant economic benefits.

19:22John Coogan:If it stands, the order would make New York the first state to impose a broad moratorium on large-scale AI data centers. There hasn't been that much of a data center boom in New York State that I'm aware of. It'll be interesting to see how they define AI data center. Or will they do it on energy or what type of GPUs you're racking or what's going on there? But more to dig in. Ken Griffin was on Goldman Sachs' podcast, the Exchange's podcast. And it was circulating this week, even though it was, I think, recorded last month. And he was talking about how, yeah, in his view, what an error this would be to the data centers are going to get built.

19:58And if they're not built here, that means hundreds of billions of dollars of revenue basically flowing through other countries.

20:06John Coogan:other countries probably go to other states first well he was talking about if New York does it there's going to be a lot of other states the meme is like China wins in this scenario US Senator John we did this with nuclear we did this with manufacturing we're mistakes there's also an article in the Wall Street Journal can a prettier data center curb the community backlash people have been batting this idea around for a while, but let's pull up this image and you tell me, would you be okay with this going into Malibu, the Malibu compute company? Would you be cool with this? If it was puking out diesel fumes 24 seven, if it was clean and it didn't drive up energy, didn't use any water, it was all closed loop and it looked like this, tolerable, right?

20:55I wouldn't just be okay with it being in my town. Oh, demand it in my backyard.

21:00John Coogan:Yes. True Yimby over here. That's right. Uh, in an effort to soothe local opposition architects plan data centers that resemble tech campuses or art museums rather than bland boxes. You have to imagine that the money that they're spending on the data center for a facade like this has got to be very, very cheap by comparison. It looks like a slot wall. Half a percent. Less. And all of a sudden, just every time it's screenshotted, like there was that hot Google presentation where they were in front of those crazy tanks and they put the logo on there and it made it look like they were taking like a brewing facility and turning it into a data center.