GM product chief says AI-defined vehicles aim to halve development time

Some elements were already known: GM had been developing hands-free, eyes-off driving automation and prototyping new battery chemistries, The Verge reported. The event sought to show how the pieces fit together. Under the “intelligence-defined vehicle” paradigm, GM is betting on AI, centralized computing and virtualized engineering to compress development timelines.

The overhaul comes as GM faces hyperefficient Chinese EV makers expanding globally and softening U.S. EV demand, forcing the automaker to balance multibillion-dollar investments in next-generation battery plants with software development costs. The Verge described GM’s bet on unproven software architectures and autonomous technology as high-stakes: if the AI systems falter or development is delayed, GM risks spending heavily while falling further behind software-first rivals such as Tesla.

In the interview, Anderson reduced the event’s message to one line: “The car’s getting smarter.” He described the vehicle as “a trusted, intelligent companion” and said he is especially focused on making the development process faster and better.

Asked whether speed comes at the cost of safety, Anderson said GM gains speed by virtualizing tests. He said GM historically could afford a limited number of prototypes for crash tests over six months; now it can run thousands of tests in the virtual world for faster feedback. “So I get speed, and I get safety, better coverage, and better understanding,” he said.

On AI models, Anderson said the choice depends on the task. Transformer-based foundation models enable generative design work, and GM has enough in-house data from decades of design files and analyses to train systems on its own data, he said. Co-training those models with data beyond GM tends to improve them, he added.

Anderson also discussed moving from deterministic code to probabilistic AI that learns from the structure of the world and from data. The published excerpt of the interview ended as he began answering a question about embedding those models in vehicles and how much processing runs locally versus in the cloud.

The Verge published the interview under the headline “Sterling Anderson doesn’t think humanoid robots should be building cars”; the excerpt focused on GM’s software strategy, testing and autonomy rather than that headline claim.