Most car companies have not yet gotten the hang of software. Use the central stack in a modern Volkswagen, Honda, Audi or Mercedes and you’ll find it’s a far cry from smartphone-grade tech.
But these companies have learned that software leadership will be crucial in the next era of automotive features. So like every company terrified of being left behind, they’re all throwing money at the Next Big Thing: artificial intelligence. AI was everywhere at CES a few weeks ago, and a huge focus of the many companies at the show operating in the automotive space. They all made clear that one day, AI will be in your car.
There’s just one issue: Almost no one can explain why you’d want it.
(Welcome to another installment of Power Moves, a column on the winners and losers of the EV transition. I’ll break down what’s happening, why you should care and who’s going to come out on top.)
AI Everywhere, Doing Nothing
Sure, they all have an idea of what it might do. Ask BMW, or Honda, Volkswagen or Sony-Honda Mobility what their AI features will do, and you’ll get the same basic answer. It’ll be conversational. It’ll be personalized. It’ll make recommendations about places to eat or charge or visit. Will any of this be worth your money? They’re still working on that.
You can now use ChatGPT in many VW models, though I’m still not sure what to use it for.
Mercedes and Volkswagen, for their part, have already put ChatGPT in their cars. ChatGPT is an undeniably useful tool, helpful for coding, for text summarization and as a jumping off point for learning about new topics. But its key feature is its ability to generate and synthesize large volumes of text, something that’s hard to enjoy from behind the wheel of a car, via voice. You can’t really draft emails or edit stats code while driving. It’ll do this via voice, enhancing the car’s speech recommendation. But the automakers themselves struggle to provide great examples of how this will help drivers.
“AI can provide information on tourist attractions, report on past football tournaments or help solve maths problems,” the Volkswagen press release read.
It’s surely fun for general trivia, but I can’t see the utility of working out complicated math problems via voice. If you do want to do that, for whatever reason, Google Assistant or Siri works perfectly fine. My colleague Tim Levin asked Mercedes CEO Ola Källenius what kind of queries company’s new Google Gemini-based system could handle that the previous-generation system couldn’t. His answer was telling.
“I don’t even need to think about it, because whatever you want can be done,” Källenius said during the round-table interview. “Whatever the large language model provides can be done.”
He went on to explain how the company’s new Google -powered “conversational navigation” can find charging stations with specific criteria. I genuinely believe Mercedes’ Gemini integration is the best example yet of in-car voice-based AI, really the first good one. But I’ll get to that. For the others, the focus on “personalization” betrays a lack of better ideas.
Photo by: Honda UK
Sony-Honda says AI will be key to selling the Afeela, but the demo they gave me wasn’t impressive. Plus, the UI shows that even the basic software needs some work.
Sony-Honda Mobility’s Afeela demo showed how the AI could be used to—I swear I’m not joking—“make the theme more Japanese.” That verbal command caused the system to change its wallpaper and display theme to one based on the PlayStation samurai game Ghost of Tsushima, a hilarious choice for depicting Japan broadly. But the demo also showed how it was conversational. The Sony representative said he was at CES, and the AI assistant asked how he was enjoying it.
I can’t fathom a feature less useful than a robot interrupting my drive to feign curiosity about my day. In an interview, Sony Honda Mobility President Izumi Kawanishi offered a couple examples of what the in-car AI could do, neither of which sound compelling.
Sony-Honda Mobiity CEO Yasuhide Mizuno introduces the Afeela’s “Personal Agent.”
Photo by: Sony Honda Mobility
“Number one, [the AI agent] can help change the internal appearance of the car,” SHM President Izumi Kawanishi told me through a translator. “We showed some examples where you change the lightning, and the general environment itself. Second, the car—specifically linked to the voice agent—can recommend specific items to the driver. For example, songs, based on what the AI agent knows of the driver’s preferences[…] So those are some specific examples of how they get customized to the user.”
I’ve never needed a robot to tell me what color I want my ambient lighting to be. And Kawanashi admitted that the car would recommend songs to play on Amazon Music or Spotify, its two supported streaming services. Both have their own robust recommendation algorithms, with far more information about your music preferences, but I suppose there’s some utility here.
BMW, for its part, says its next iDrive will use AI to personalize the experience. The company also showed a demo that was, well, mostly just customizing lighting and the mode of the car. These automakers say you’ll be able to personalize the driving experience, but there really are only so many settings for a damper or a throttle pedal. You’re not going to find much beyond what Normal, Sport and Eco already offer, in my view, and many cars have had “Adaptive” or “Auto” driving modes for far longer than their makers have been marketing them as “AI-powered.”
Photo by: BMW
BMW’s Panoramic iDrive offers plenty of customizable screen real estate. But when the demo display features an Air Quality Index readout, it starts to look like they ran out of useful information to display.
It’s no surprise, then, to see German companies diving headfirst into a new technology before figuring out how to make it genuinely useful for the average driver. BMW did that with gesture control, Mercedes does it constantly and Audi’s never far behind. But even Honda is getting in on the AI hype. Its new Asimo OS for the 0 Series EVs is built around many AI functions.
“By constantly updating the in-vehicle software based on the ASIMO OS through over-the-air (OTA) updates, even after the purchase of the vehicle, functions and services will be continuously advanced in accordance with the preferences and needs of each individual user,” the press release said. “These OTA updates to both the digital UX and integrated dynamics controls will allow Honda to deliver a personalized ownership experience that will enhance the joy of driving.”
Photo by: Honda
That’s right: Customization. Preferences. A digital UX. Sounds familiar.
More and more, I get the impression that automakers are using “AI” language to show a general orientation toward technology, rather than focusing on any meaningful value add. I sat down with executives at Faraday Future, a company that says it is building the world’s first “AI EV” and “Range Extended Artificial Intelligence EV,” or “RE-AIEV for short. But when they showed me the RE-AIEV, they didn’t mention any AI-powered capabilities. I pressed a representative to explain what the AI features were.
“Well, it’s a general term,” he said.
The Killer App
Don’t think that AI is irrelevant to the car market. While it’s quite fun to tease these companies for their confused, questionably useful ideas, AI proficiency will be crucial for automakers for one key application in particular: Autonomy.
Autonomous and driver-assistance systems are the most natural place for AI to provide real value.
Automakers are increasingly relying on AI to power their driver-assistance features, including “Level 2 plus” systems. That’s according to Ed Kim, President and Chief Analyst of automotive research firm AutoPacific. Systems like these are not autonomous—they still require constant driver supervision—but are the first steps toward autonomy. They take some of the burden out of highway driving, and better AI systems can make them smoother, safer and usable in a more diverse array of conditions. Tesla’s misleadingly named “Full Self-Driving” suite was likely the first time most people heard of AI powering a key feature, and these systems are the biggest area where AI is driving sales directly, per Kim.
“People want active safety features. They may not necessarily know that these are powered by AI, but increasingly they are,” Kim told me. “So it’s not that [consumers] are necessarily asking for AI, but they’re asking for technology that has a direct benefit in their daily driving.”
A Bright, But Uncertain Road Ahead
ChatGPT integration probably doesn’t fit that criteria, at least not yet. Kim said that only 18% of consumers show interest in an AI-powered voice assistant, for instance.
But he’s right that this technology can drive real, marketable advantages. AI is already transforming parts of the auto industry, especially when it comes to car design and manufacturing. Scout Motors’ Chief Technology Officer Burkhard Huhnke told me that AI-powered aerodynamic simulation has drastically decreased the company’s iteration time, allowing them to adjust and re-analyze vehicle aerodynamics an order of magnitude quicker than in days past.
AI can improve aerodynamic and crash-test simulations, allowing for rapid iterating.
It’ll no doubt drive manufacturing efficiencies, too, while helping companies design cheaper-to-produce vehicles that still pass crash tests. AI is a transformative advancement for most industries, and the car world will be no different.
Yet technological advancements require us to learn new paradigms. Asking AI to do the same voice assistant work you’ve already built is like asking ChatGPT to do arithmetic. It can do it, but it’s not going to drive meaningful improvements. Leveraging this technology will require companies to solve existing consumer problems, rather than inventing situations in which AI could theoretically help.
I don’t need to hear an AI summarize the Bolshevik Revolution while I drive to Chipotle. I need an AI that can find me a charging station with a good restaurant within walking distance, preferably with options that fit my passenger’s dietary restrictions.
Mercedes’ “Conversational Navigation” feature in the upcoming CLA could be a great AI-powered feature, assuming that the underlying model doesn’t hallucinate or provide incorrect information.
Photo by: Mercedes-Benz
That’s where Mercedes’ Gemini tie-up comes in. The brand’s “Conversational Navigation” feature promises to solve those kinds of complex queries. You can ask it for Italian restaurants near a charging station with gluten-free options, and Mercedes says it’ll find you one. It’s a perfect use of AI: Gathering varied information from publicly available sources, analyzing it and using it to find a solution to a specific set of criteria. It’s a task that no existing navigation system can accomplish, and that any human would find annoying.
It is a perfect example of how this technology can be transformative. Its beauty is that it’s solving a real problem that cannot be reasonably solved without AI. You don’t have to care about AI to use it, you don’t have to buy into the marketing hype. You have a problem, and Mercedes will sell you a solution.
That is a recipe for success. But if all your AI can do is make the infotainment system look “more Japanese,” don’t expect customers to care—or pay up.
Contact the author: Mack.Hogan@insideevs.com.
