Why GM Will Give You Gemini — But Not CarPlay  ## Metadata - Author: [[Decoder with Nilay Patel]] - Full Title: Why GM Will Give You Gemini — But Not CarPlay - Category: #podcasts - URL: https://share.snipd.com/episode/8fded384-c25e-45c2-9aff-3b7d6af27427 ## Document Notes ## Highlights - **GM's Multiyear Bet On Converging Technologies** - GM is pursuing EVs, autonomy, AI, robotics, and home energy as interconnected long-term bets. - Mary Barra calls EVs the company's North Star while remaining flexible to meet customer choice. Transcript: Nilay Patel You have the introduction of a Google Gemini-powered assistant in the cars in 2026, a next generation hardware and software platform that will come first on the Escalade IQ in 2028. That will enable the next generation of Super Cruise, which will let you have eyes off driving, additional hands off driving in the Escalade IQ. There's a new robotics division, which is a big bet, and then new batteries and new home energy solutions. All of that is coming in the context of what I would call huge changes in the car market, end of the EV tax credits, tariffs that are upending the global supply chain, what appears to be A brewing trade war with China over rare earth metals, which are essential to all of these plans. Did I miss anything? Mary Barra Well, we've been pretty busy. So I think you captured most of it. ([Time 0:05:11](https://share.snipd.com/snip/1adf82e2-3ea4-4156-a80a-ae88d134ca98)) - **EV Growth Will Slow But Continue** - GM expects slower EV adoption after the loss of the $7,500 tax credit and adjusted plans accordingly. - The $1.6 billion charge reflects a near-term slowdown, not abandonment of the EV strategy. Transcript: Mary Barra I would say, first of all, it's being very customer focused and making sure we have great products with the right design, the right quality, the right features, performance, the right Software and services that we can put on top of that that customers want to buy. So we start there, and that's where we want to head. We still believe in EVs, and I consider them to be our North Star. But clearly, when you have a dramatic shift that we're seeing in the regulatory environment in the United States and potentially other places, as well as a very substantial change In the consumer tax credit, that's going to change buyer behavior. When you think about vehicles, they're either the most important or the second most important purchase that a customer makes. I mean, it's very significant, and we get to be a part of that with our consumer. So as we look at these changes, we need to make adjustments. But we're going to meet the customer where they are, whether they want a great ICE vehicle or they want an EV. The adjustment that we made today from the charge was really reflecting what we believe now will be a slower EV adoption, but we still have a great portfolio of award-winning EVs that We're going to be offering. ([Time 0:07:17](https://share.snipd.com/snip/e9035289-6ceb-4514-8de6-5b99d0f32507)) - **Pivot Factory Plans Fast** - Adjust manufacturing plans quickly when market signals change to avoid capacity mismatches. - Use flexible factory plans to shift production where demand is stronger, like expanding Lake Orion capacity. Transcript: Mary Barra One of the things that we made an adjustment already was the fact that we were planning to have one of our factories in Lake Orion, Michigan, be an electric plant. And when we looked at the shift and what's happening with tariffs, we decided that we were going to expand the capacity because right now in some, for instance, full-size utilities, We can't build enough. The demand is so strong. And so we're going to use that plan. So once you make a change like that, that's something that already has happened. I don't think we're going to know for sure what true EV demand is until next year, because like you and many others, there was definitely a pull ahead for people who wanted to be able to Have access to the $7,500 tax credit. So we'll see potentially a little bit lower in the fourth quarter. We don't know yet. And as we get into next year, we're going to see that. But again, when you talk about $7,500 on a vehicle, that's pretty substantial. We do expect, and I think the industry expects and the external forecasters believe that we're going to see slower EV growth. ([Time 0:09:21](https://share.snipd.com/snip/b88faade-ecef-435c-bf45-609cdfb91321)) - **Ownership Drives Repeat EV Demand** - Owners who try EVs tend to buy EVs again because of performance and lower maintenance. - Broader charging infrastructure expansion remains key to sustained EV adoption. Transcript: Mary Barra Well, first I would say, I don't think we played games with residuals. I mean, General Motors has been very disciplined for more than a decade about understanding what pricing does and when you make dramatic changes in pricing, what that does to the residuals. And then it takes almost a new generation of that vehicle to get the residuals to be where they're at. So we manage that very, very carefully because we want to make sure the people who buy our vehicles keep that value. Now, when someone leases, that then becomes the leaser or in our case, many of our vehicles are bought through GM Financial. But we look at that as well because it's still something very important. But again, I think it's too soon to tell. I mean, generally what we see, there's a new car buyer, and then there's people who buy used vehicles. And so we're just going to have to see what the equation is. But I think one of the things you're getting at is the consumer is very rational. So I think they're going to be looking across the board. Sterling Anderson If I can add one thought there. And Eli, you're now in a Vistic, so getting a little more of this experience with some of these cars. For many, getting over the hump of that kind of initial activation energy of being comfortable with an EV, being comfortable with the range, getting comfortable with setting up level Two charging in their home and these other things. I think there is incremental value to, as you say, kind of the penetration that we get in the secondhand market for people that get comfortable with these vehicles. They are better cars. They're better on a number of dimensions. And as that comfort grows, I do hope that it's value accretive to kind of future purchases. Mary Barra Well, and to your point, Sterling, we have already seen with our consumer feedback or customer feedback that once they own an EV because of those benefits, they very much are likely To buy another EV. ([Time 0:11:01](https://share.snipd.com/snip/10a9953c-5214-4342-9d38-8cd940cced97)) - **Functional Scale With Brand Expression** - GM organizes functionally with centralized product development and brand-specific execution. - Sterling oversees global product to get scale while brands keep distinct customer promises. Transcript: Nilay Patel Let me ask you the decoder questions. GM, I think people understand it as nameplates, right? There's GM and GMC and Chevy and Cadillac and Buick and all the rest. How is GM structured as a company, as an operating company? How is it structured? Is it just the divisions and the brands, or are you more functional than that? Mary Barra I think we're more functionally – and Sterling is responsible for our products globally as the chief product officer, because we leverage that to get the scale. But then we also understand for the different brands, what we need to do to make sure we're delivering on that brand promise, whether it's true luxury with a Cadillac value, but the unique American heritage that Chevrolet has. Buick is premium and has been doing very well in growing. And then GMC, which is a very premium truck. So each of our four brands are very specific for what they mean to the customer. So we make sure when we do a portfolio of vehicles or when we're doing, for instance, full-size trucks, that we understand what it means to be a Chevy, to be a GMC or a Cadillac. We look at all of that. And so it's really a cross-functional team working together to make sure we deliver on that. And then we do have regional sales and marketing team that really understand the customer in each of our markets. So at broad brush, at the most senior level, that's, I would say, the way the company is organized. ([Time 0:27:14](https://share.snipd.com/snip/5a282e73-eae9-481a-a95b-987737acb356)) - **China Competitors Need Context** - GM benchmarks Chinese EVs but notes subsidies and regulatory differences make direct comparisons misleading. - Safety and local connectivity rules require market-specific technology choices. Transcript: Sterling Anderson Yeah, a couple of quick things. As Mary said, as we benchmark across these vehicles, if we strip away the subsidies and the other support that the Chinese government is providing to many of these players, we compare Favorably on most dimensions with our EVs. In particular, we've got some work that we're doing around a next generation electric vehicle and electric vehicle architecture that compares very favorably to what we see. So I'm impressed by what they've done. I'll start with that. I think we don't want to be resting on our laurels that way. I think there's more that we can be doing, but we're on it. We recognize very acutely kind of the speed at which an industry that is so heavily subsidized tends to move, and we don't intend to wait for it. Mary Barra And I think Sterling makes a really great point. So we are regularly looking at how do we make every part of our business more efficient. We don't control what the government policies are. It's evident by what's changed this year. So we just need to have the best product that we have as efficiently as possible. And that's what we're focused on every day. Nilay Patel We just had Ford CEO Jim Farley on the show. My friend, Joanna Stern, interviewed him. I was on leave, so she filled in for me. That was very nice of her to do. He is very clear that BYD makes a better car than Ford. He drove around in BYD for a while and publicly announced, like, this is better than our car. On our show, he said to Joanna, ([Time 0:33:01](https://share.snipd.com/snip/7cb30350-d009-4c41-84a3-fb098ec6c87f)) - **Scale + R&D Over Isolated Skunk Works** - GM avoided isolated skunk-works and pursued steady EV architecture improvements across the company. - Scale, manufacturing fixes, and battery R&D (LMR chemistry) are positioned to reduce EV costs by 2028. Transcript: Sterling Anderson A couple of quick comments. So I, I like Doug a lot. We worked together. I worked with Alan Clark, who's, who's been leading the Skunk's Work program for Ford, both great people. Um, so, so, uh, nothing about that. What I will say is I've seen a fairly oscillatory behavior from some OEMs when it comes to EVs, right? In one moment, it's a, we'll force an electric powertrain into an ICE platform and we'll end up with a fairly compromised product as a result that turns people off to certain segments, Right? Electric battery, electric trucks. There's some things that just weren't great about the way that that started in some of this space. And then I see a pendulum swing all the way to a, it's got to be Skunk Works project. It's going to be done fully independently from a clean sheet. And then we're going to somehow try to ingest it into the broader organization. That ingestion is where the risk lies. I think what I've seen that General Motors has been doing since before, I only got here four months ago, so there's only so much that you can attribute at all to me at this point. But what I'll say is General Motors has taken a much more steady approach to this development. Electric vehicles designed and developed and manufactured on electric vehicle architectures, which are not nearly as compromised. A focus on the cost basis for these vehicles, on what are the value-added components that we can put in them, and what are the things that just don't add that value that we can strip out And kind of refine. Whether it's reduction in menial stuff like spot welds, right? So some of our next generation platforms, dramatic reduction in those. You'll never see those. You'll never hear about them. But they drive an enormous cost. I think at one point the estimate was $40,000 per spot weld if you advertise kind of all the costs of the robots and everything else across it. But to the battery chemistry and some of the battery innovations, I think you may be aware that we've been pioneering, we are today the largest manufacturer of cells in North America. This is larger than Tesla today. We have been pioneering work in lithium manganese rich or LMR battery chemistry that has something closer to the energy density of high nickel batteries that you're familiar with At closer to the cost of lithium iron phosphate or LFP. And those will come to market first in GM vehicles, in some of our larger vehicles in 2028. So what we've done is oriented the company and some of this kind of the functional strength that Mary references in common design, common engineering, to innovate on many of the foundational Enablers of compelling battery electric platforms done through the streamlined kind of process that we have for our global product development across the company. So my expectation is when this hits, as it hits, as LMR hits, as our architectures hit, as our manufacturing improvements hit, as some of our robotics work that we can talk about hits, What you'll see is those immediately kick in. And what's done initially in one model, in one plant, in one brand, rapidly scales across a massive portfolio of vehicles. ([Time 0:35:05](https://share.snipd.com/snip/797e1b6e-0832-415b-b065-311039a900a8)) - **Phone Projection Out For A Unified HMI** - GM removed Apple CarPlay/Android Auto from EVs to push users onto a unified, safer in-vehicle HMI. - Mary and Sterling argue an integrated system reduces distraction and unlocks vehicle-specific services. Transcript: Nilay Patel We're back with GM CEO Mary Barra and head of product Sterling Anderson. So far, we've spent a lot of time asking Mary about the car industry's biggest problems and how she's reorienting the company, trying to solve them. But if you go look at the day-to conversations about GM EVs on the forums and subreddits, you'll see that they're dominated by one very important decision. The choice to ditch Apple CarPlay and Android Auto. GM has been very adamant about the decision to move on from phone projection in its EVs, and now as chief product officer, that's Sterling's decision to justify. So I wanted to hear it from him. Does he still feel good about moving on from CarPlay and Android Auto? And what's he building in the future that will make this entire debate irrelevant? The big announcement that I'm really curious about is Gemini powered assistant in the car. That's coming obviously in 2026. That's, I would say, an expansion of the idea that GM should have a software platform in its cars, right? We're running Android automotive in the cars. There's an app platform. There's a data platform. There's Google Assistant in my car. My daughter was asking it to play Taylor Swift. That's basically all we use that platform for today because she won't let us do anything else. But it works. It's there. The big decision there, and I know you know this question is coming, is that you bet against putting smartphone projection in your EVs. There's no Apple CarPlay or Android Auto in the EVs. But the gas cars still have it. How did you make that decision? Mary Barra Well, I think it's really a question of timing as we look at that because really, and I want to make sure that we get Sterling's input on this as well, but as we looked at it and as we made that Decision, we were getting a lot of feedback from customers that it was very clunky moving back. It wasn't seamless. And frankly, in some cases, could be distracting to move back and forth between if you were doing something that you could do on a phone projection type of system versus if you needed To do something in the vehicle. And we also know that's only going to increase when you look at some of the things we're going to talk about of how we can just make your life better and assist you as we move forward. I think we're at the very, very early stages of services we can have on a vehicle to improve the overall customer experience and make the journey smoother. And so we looked at that, and we decided that we needed to have a great system in the vehicle that allowed people to have one system, and we're going to continue to make that better and add New features. And Sterling... ([Time 0:47:50](https://share.snipd.com/snip/d962de9f-14b9-41f1-a255-0174c7cadd2a)) - **S-Curve Shift: Remove Old Interfaces** - Sterling frames the CarPlay removal as a necessary S-curve jump to a superior vehicle HMI environment. - He compares it to removing disk drives: short-term pain for a better long-term experience. Transcript: Sterling Anderson I agree with that. Yeah. Nilay, one thing that I'd say here is what we're talking about is the inevitable performance degradation when jumping between S-curves. And what I'll describe as those S-curves is, you know, the first one was, you know, for some time, I think you and others got attached to phone projection applications largely because The in-vehicle HMI was pretty bad. Your opportunity for doing some of these things was better when you were using that. We've now gotten to a point where, I mean, you're driving a Vistic, I understand. I mean, you've got Dolby surround audio. You've got giant screens. You've got giant displays. The analog I would use here is we're on this new S-curve. There is inevitably a jump that has to happen for you to get over to it. That's uncomfortable for many, right? But frankly, it's a very Jobsian kind of approach to things, right? And the removal of the disk drive, nobody liked that. Everybody on the forums and Facebook was complaining about it. But it was really a, look, guys, flash storage really is the future. Get on board. You'll see that. That's kind of what we're saying here. That's exactly what we're saying. In fact, and the analog that I use for folks is, look, you've got, if you're talking to me about CarPlay, it's very likely you've certainly got an iPhone. You've probably got a MacBook. You have the opportunity to use phone projection on your MacBook, right, a phone mirroring application. How many of you are accessing online services like email, social media, and otherwise through the phone projection app in your laptop? Almost none of them do. Why? Because you've got a much larger screen on your laptop. You've got a much more convenient HMI via the keyboard. You've got better speakers. Now take that same analog to the car and ask the same question. Is it in a car that has not only just laptop speakers, not only a laptop screen, but something better and that can move you and that can integrate with, you know, charging infrastructure With super cruise availability on your maps, all of these other things. You're in a much more immersive environment that can do so many more things. Why would you use the equivalent of a phone mirroring application on a laptop in your car? And so for us, it's a look, we're taking out the disk drive, guys. Get on board with flash storage. That's where the future is. It's kind of the analog. ([Time 0:50:17](https://share.snipd.com/snip/cb63ae6a-9630-4abd-b73d-24e1e856df0a)) - **Eliminate Login Friction For In-Car Apps** - Reduce login friction by exploring federated IDs and smoother profile handling in cars. - Prioritize ease of use to get users to adopt native car apps over phone mirroring. Transcript: Sterling Anderson And do you have to log into each of these different services and applications in your car? Because if you do, you get some breakage. Some people just will never do that, right? It's a pain. We're looking at that as well, right? What can we do about federated IDs? What can we do to eliminate that friction of you engaging with your car? I'm not sure I quite follow the whole, like, it's illegal to use your phone when you're driving and not when you're on your laptop. I think that cuts against your argument a little bit, because if you're sitting with your laptop and your phone, you don't pick up your phone. Nilay Patel If the cops see you looking at your phone while you're driving, you get pulled over. If the cops see you looking at your phone while you're using a laptop, I don't think they're going to do that. Sterling Anderson My point being, if you've got your phone on your laptop or you've got your phone in your car, if you're sitting next to your laptop, you don't typically pick up your phone to answer an email. You'll answer on your laptop. If you're sitting in your car, you wouldn't pick up your phone to use a map application if it's sitting there on a bigger screen in front of you in the car, right? And if you have a convenient HMI by which to dictate— Wait, hold on. Nilay Patel Not to get too weedsy, but I know we have a lot of readers who have Windows laptops, for example. Okay. And they are constantly picking up their phone to use iMessage on their iPhones because that is not an experience you can project to Windows in a meaningful way. Sterling Anderson And that kind of gets to Mary's point, right? Which is don't count out the expected. Don't expect that we won't have a growing list. I mean, today we use Google Automotive Services in these cars. That's got access to the Play Store. Quite a few apps available through that that will only grow in time, as will availability of apps from other places. Mary Barra The consumer a great experience in the car that's not clunky, that allows them to getting to this point of safety that one of your viewers talked about, that you can be the least distracted And pull even more information from the vehicle to improve the way your whole experience overall. So, you know, we're at the, I say, the early phase of making that shift and now continuing to add more to it, continuing to make it more intuitive. How do we make, you know, access to more apps and more things you want to leverage? ([Time 0:55:15](https://share.snipd.com/snip/ebc45709-f9e1-4b88-b5d9-e1c94e5d30a7)) - **Gemini + Centralized Compute Roadmap** - GM will add a Gemini-powered conversational assistant in 2026 and centralize vehicle compute in future architectures. - The 2028 Escalade IQ's centralized compute and Ethernet network will enable faster updates and advanced features. Transcript: Nilay Patel We obviously use Google Assistant in our car today. AI has often been described to me as a platform shift. And in particular, like natural language interface has been described as a major kind of platform shift. And then maybe the way we write applications will change with MCP and these other technologies. But the platform shift is you're going to talk to the computer. It's going to understand you. It's going to take some action for you. Are you feeling that way about the car, that AI and particularly these assistants will let you have a platform shift inside the car? Yeah. Sterling Anderson Yeah, we are. So we're thinking about AI across the business, right? Not just in kind of what it does to the product experience, but what it does to our development of the product, what it does to kind of what we can do to bring to bear the massive data that This company has in our production systems, in our CAD and development systems, to leverage in development of AI for production for the product. As it relates to the product, though, what I'll say is this is one of the really important enablers of speed in a company with a portfolio this broad is development of a common platform Or undergirding infrastructure on which we can deploy, including over-the software updates. As Mary mentioned, there are some legacy platforms that will kind of work their way out of the portfolio. But one of the things we're working on in the software-defined vehicle space is a new electrical architecture that centralizes all compute in the vehicle. It doesn't just move the compute to zonal aggregators. It centralizes all of it. In the upper trims, it leads to about a 35 times increase in the computational power of these vehicles. It allows for... Nilay Patel This is the 2028 Escalade IQ, right? Sterling Anderson It will arrive first on the 2028 IQ. That's right. What it does is, from a networking perspective, it moves to Ethernet-based networking, which allows us to move to sub-millisecond response times. Take magnetic suspension systems or dampers as your example. From accelerometer through controller back to actuator, we're talking sub-one millisecond. You're talking 1,000 hertz. That's a 10x improvement over previous electrical architectures. It's a massive, massive opportunity, not only for the dynamics, which I found General Motors is extraordinarily good at, but for the software that we can deploy on this. Because now we've got a centralized architecture. Now we've got a central compute on which we can run a variety of applications to which we can deploy updates those applications at a much higher pace than what we do today. ([Time 0:58:31](https://share.snipd.com/snip/947ef3c1-b139-4682-9f85-47b816ed0f5d)) - **Keep Safety-Critical AI Local And Tailored** - GM will run safety-critical, time-sensitive controls locally and tailor AI models for vehicle functions. - Generic LLMs will be augmented with specialized, fine-tuned models that interface with controlled vehicle functions. Transcript: Sterling Anderson There are a number of things that we're going to infer directly on the vehicle. And we've got quite a bit of, as I mentioned, the computational power is there for us to do a lot of on-vehicle inference. What I'll say, though, is step one, what you're referencing is in the Gemini-based conversational AI work that we're going to release next year, that's really about kind of vocal interaction, Voice-based interaction with your car. It's about asking about your destination. It's about asking about a number of other things where it can give you kind of contextual responses. I think you're hitting at the heart of, well, but many elements of the car, if you're going to directly control them, are safety critical. And how can you deal with the kind of accuracy that you get out of an untailored LLM if you're giving it direct access to those things? And the short answer is we're not. We are developing, in addition to the contextual AI that comes out next year, a more powerful, specifically tailored AI that will learn from not just your preferences and not just the Data that you have available via a variety of other mechanisms, but it will be specifically tailored, customized, and learn from the vehicle itself, its functions. It will have specific access to specific functions. It will have other controls in place that a generic LLM doesn't have. So I think that the value of these large models oftentimes can be realized in much greater effect through tailoring and fine-tuning for specific application areas, and that's what We're going into post the Gemini conversational AI work. ([Time 1:02:21](https://share.snipd.com/snip/3009fddc-fad1-4365-bd89-e8904a65b323)) - **Safety-First Builds Trust Over Speed** - GM emphasizes conservative, safety-first deployment for autonomy and in-car AI. - Mary positions safety as a core customer promise that underpins slower, trust-building progress. Transcript: Mary Barra The only thing I would add, at General Motors, we always prioritize safety. We were the first to have a driver assistance technology out with Super Cruise, but we very much prioritized the safety of that and enabled features as we saw it demonstrate that it could Meet the safety standards that we have. I think that's what Sterling's talking about now is as we do this, we won't just go wide open, but we'll make sure that the integrity of the safety systems are there. Sterling Anderson If it's controlling a temporally sensitive or safety relevant function, it will happen locally. Nilay Patel One of the reasons I'm asking about that very specifically is I'm watching your partner, Google. I'm watching Amazon. I'm watching Apple. They control a lot of that stack, right? They control the LLMs. They control the inference. They control the smart home platforms. And they have not been able to connect those assistants in the way that you're describing, right? Alexa Plus exists, and it has to fall back on some complicated deterministic system to flip lights up and down, and they have to do some orchestration. Google Home, they're doing the same thing over there. Gemini is rolling out, and it has to fall back. And it's working however it works. Apple has not been able to ship this product, which they have promised people. I say, hey, Google, turn off the lights. By the way, my Google is going to turn off the lights, so I shouldn't say that. But if I ask my assistant here to turn off the lights and it doesn't happen, maybe the worst outcome is my wife is once again annoyed at me for trying to voice activate our home. If I get that wrong in the car, many other things can go wrong. That's right. Sterling Anderson That's right, 100%. This is something that we are very keyed on, right? Whether it's the development of our self-driving system, where we are acutely familiar with and aware of the challenges that arise when you have a kind of monolithic end-to system About which you can't make any strong guarantees, to other controls across the car that are safety relevant. So, as I said, we are – I mean, that's probably a much longer conversation than we have time for here. But this becomes architecturally relevant in the design of our self-driving system. It's architecturally relevant in the design of the rest of our AI and how it interfaces with vehicle controls. And we are taking that very carefully. Mary Barra And we have a safety group within the company that really looks at that and is looking at all those decisions are made because they can think through second order, third order, you know, Down the road of really thinking through to make sure that we've interrogated it to make sure we're making good decisions and we're not going to put the driver or the passengers into A situation that no one wants to be in. ([Time 1:04:01](https://share.snipd.com/snip/f05e7ea3-6caa-4746-ace5-5b3730e2fa51)) - **Train Robots With Rich Factory Data** - Use factory data, CAD, and work instructions to train embodied AI for robotics and cobots. - Deploy autonomous mobile robots and collaborative systems to improve safety and throughput before external commercialization. Transcript: Sterling Anderson Yeah. The only thing I'd add is, we are focused on growth areas that are both tangential to and kind of synergistic with our core business. When you say batteries and energy storage, we've recently announced a partnership with Redwood Materials, for instance, where our Second Life batteries are currently powering The largest second-use energy storage system in the world, the largest microgrid in North America in Sparks and Nevada, it's 63 megawatt hours, I think, I remember that being. So there's tremendous opportunity there. When we talk about kind of grid rebalancing, particularly as we move into LMR, which has much higher cycle life, we get into a world where by plugging your car in when it's at home or your Place of work, your car can be an active participant in the rebalancing of the grid, pulling power from in lulls and pushing power to at peak times. This is an enormous opportunity for the industry. It's an enormous opportunity for kind of, I mean, energy is going to be a very constrained resource. If the last several months of discussion in the AI space has taught us nothing, I mean, we're talking about a triple kind of the capacity needed online. We can play a pretty big role in that. As the largest cell manufacturer in the United States, there's a lot we can be doing there. In robotics, I'll just kind of paint the picture for a moment, particularly in a world of learned models, right, which is effectively where AI is taking us, data is really important, Right? And the quantity of data, the cleanliness of that data, the availability of that data in training systems that can take on more tasks than what they've ever done before. Look at a business like General Motors, and you see a product, each of whose parts is known in detail. It's sitting in CAD. We know the mass, we know the materials, we know the response to manipulation. We know the flows of that material through a production system. We know what happens at each step of the production system. Every single step is meticulously outlined in work instructions. That is an enormous kind of database from which to develop really competent physical AI systems, right? Embodied AI systems that can be powerful enablers, both of our own production and the safety of our employees. So today we've got, I think, 30,000 robots operating alongside on the order of 97,000 production associates in 11 facilities. ([Time 1:10:09](https://share.snipd.com/snip/1691293d-2513-4923-a20b-e8caf43a909c)) - **Seat Memory Doesn’t Swap User Profiles** - Nilay described a practical frustration: seat memory buttons don't switch infotainment profiles in his Cadillac Vistic. - Sterling promised to investigate, calling the fix "can't be hard." Transcript: Nilay Patel Have one very, very small feature request, but it's very important to me. All right. The buttons that control the seat memory on the door in my Cadillac Vistic do not also switch the user profiles on the infotainment system. So when I get in the car, I need to push the button to move my seat to my seat memory and switch the profile from my wife's profile to my profile. And I have no idea why I have to do both. I just want to be able to push the seat memory button and have that switch the profile as well. Can you fix that for me? Sterling Anderson I'm going to take a quick look. That can't be hard. Let me take a quick look. All right. Nilay Patel If you read the forums, I will tell you that is the basic shape of the response to this problem is that can't be hard. Sterling Anderson I'm going to take that one back. Let me take a look. These are different systems. Mary Barra And also what it would take to do it. We'll look at both. We are open to customer feedback. And we love the fact that you're driving a Vista. ([Time 1:20:06](https://share.snipd.com/snip/66e550ae-6b0f-48dd-8895-c4d67c865ce2))