GM's AI vision: Less sci-fi, more practical innovation

Driving the news: David Richardson, GM's senior vice president of software and service engineering, laid out the automaker's vision for AI in an exclusive interview with Axios.

• He joined GM from Apple in 2023 and is now assembling an ambitious AI organization led by Silicon Valley veterans like chief AI officer Barak Turovsky and John Anderson, head of AI research.

What he's saying: Richardson began the conversation by noting he's super-bullish on AI, before quickly poking a pin in the hype.

• "I do think there's a lot of hype in AI, and there's a lot of tendency to do AI for AI's sake," he said.
• "Right now, our vision is we want to use AI to actually drive real ROI for the business, as well as develop consumer-facing products that people actually need and they actually want."

The big picture: Disruption from artificial intelligence, particularly in office settings like tech and media, has fueled predictions of a white-collar bloodbath.

• So far, at least, AI is helping early adopters boost productivity without mass layoffs, according to MIT's State of AI in Business 2025 report.

Zoom in: AI's most-hyped sector is humanoids — dextrous, ambulatory robots that are trained to apply human reasoning.

• Morgan Stanley believes there's a $4.7 trillion market for humanoids over the next 25 years — most of them in industrial settings, but also as companions or housekeepers for the wealthy.
• Elon Musk envisions 1 million Tesla Optimus humanoids by the end of the decade.

GM had its own humanoid robot years ago. Co-developed with NASA, its Robonaut 2 flew to the International Space Station in 2011.

• But that doesn't mean it wants humanoids operating in its factories any time soon. As the Humanoid Robot Games in Beijing showed last weekend, they're still trying to master staying upright.
• "I don't want to say it's never going to happen," Richardson said of humanoid factory workers. "I think it's very early, and I do think there's a question as to what is the overall value you get from that?"

Instead, GM is training AI "cobots" — robots that work alongside humans to assist with jobs that are dull, dirty or dangerous — so humans can do less strenuous tasks and focus on craftsmanship, he said.

• "We've got so much proprietary data around how to do manufacturing," he said, "and that's like a treasure trove of data for us to train AI to do that."
• When building new production lines, for example, GM uses AI to create a digital twin of the manufacturing process to spot bottlenecks or safety issues in the virtual world before building the real thing.

• GM factories are also using AI vision systems to detect paint imperfections and welding defects, painstaking work that used to be done manually.
• AI also helps with predictive maintenance to avoid, say, a costly shutdown of a production line.

Beyond manufacturing, GM is deploying AI where it can have an immediate effect on productivity and profits.

• In vehicle design, for example, AI helps engineers quickly optimize things like vehicle aerodynamics, saving time and money.
• GM created an AI tool to analyze traffic patterns and EV usage data to help identify the best locations to install charging stations with partners Pilot Flying J and EVGo.
• Another AI tool enables GM to detect voltage anomalies in battery packs before they make their way into an EV.

The intrigue: More than 15% of the software code GM develops is now done using AI-assisted tools. AI has helped GM catch 10x more software bugs, far earlier in the development cycle.

• The company also created an AI tool to help GM dealers order vehicles with the right features and equipment for their particular customers.

🏎️ 1 fun thing: GM is even using AI tools in motorsports to try to win more NASCAR and IndyCar races — and starting next year, Formula 1 races, too.

The bottom line: GM's strategy is a reminder not to overlook the practical side of artificial intelligence.