Six months. That's how long Sterling Anderson says a round of crash tests used to take at General Motors, and that was with only as many prototypes as the budget allowed. Now, he says, his teams can start thousands of simulated tests and get answers much faster.
It's a small detail from a long interview, but it explains the rest. Anderson is GM's chief product officer, and last week (October 21) the company staged a big technology showcase in New York to argue that the car business is about to run on a different clock.
The pitch has a name: the "intelligence-defined vehicle." Whether it works is a question worth real money.
A roadmap with a lot of moving parts
GM's event pulled together AI-assisted vehicle design, new battery technology, more powerful in-car computers and driving automation. Little of it was a surprise on its own. The company had already talked about hands-free, eyes-off driving, and it was already prototyping new battery chemistries.
The news was the stitching. As The Verge reported, GM is betting that artificial intelligence, centralized computing and virtualized engineering can shrink a typical four-to-five-year development cycle to a little over two years.
Asked for the unifying theme, Anderson kept it short: "The car's getting smarter." Then he admitted he may be one of the few people who gets as worked up about the development process as he does.
He's probably right. Nobody has ever bought a truck because of how it was simulated. But the process is where the money is.
Why two years instead of five changes the math
The headline number is easy to nod along to without asking what it actually does, so let's slow down.
A four- or five-year development cycle is a long bet on a world that won't sit still. Think about what a carmaker has to guess at the start. What will batteries cost? Will buyers still want the thing? Which software features will feel dated by launch day? Every extra year between decision and showroom is another year for those guesses to go sour.
GM knows the feeling. American buyers have cooled on electric vehicles, and the company has had to rethink how fast it moves on electrification, all while still holding multibillion-dollar commitments to battery plants. A program locked in at the peak of EV enthusiasm lands in a very different market. Cut the cycle to just over two years and the window in which the world can change on you shrinks. You also get more attempts at a given model. That's a quieter advantage, and maybe the larger one.
Then there's the competition. Chinese EV makers have been expanding abroad with a reputation for efficiency, and Tesla was built as a software company that happens to sell cars. GM is a century-old manufacturer retrofitting that mindset onto an enormous organization. Speed is the most direct way to close the gap.
Back to the crash tests. Where a team once sent a limited number of physical prototypes into walls over about six months, Anderson says it can now run far more scenarios in software. In his words, "I kick off thousands of tests in the virtual world that give me feedback much quicker." He says that buys speed and better safety coverage at once, and he waved off the worry that going faster means cutting corners.
Believable? Partly. More simulated scenarios can surface problems earlier and from more angles than a handful of crashed prototypes ever could. But a simulation is only as honest as the model behind it, and nobody outside GM can audit that. Physical testing hasn't gone away, and regulators still care about what happens when metal meets metal. The claim to watch is whether the compressed schedule holds up on an actual vehicle program, with an actual launch date, rather than on a stage in Manhattan.
The cost side is just as unresolved. Software development isn't cheap, and GM is paying for it while it also pays for batteries. If the AI-heavy approach stumbles or slips, the company burns cash while rivals keep moving. The Verge called it a high-stakes gamble, and that's fair. You can't half-adopt a faster cycle. Either the engineers change how they work or they don't.
What an intelligence-defined vehicle means in practice
Strip away the branding and the idea is fairly simple. Instead of a car stuffed with dozens of small, separate computers, each doing one job, you put a lot of processing power in the center and let software do more of the work. That makes the vehicle easier to update and easier to design around. It also turns the car into a platform that improves after you buy it, which is how phones have worked for years.
AI shows up twice in that picture. Inside the company, it helps engineers explore designs. Anderson described transformer-based foundation models feeding the kind of generative design work he was talking about, drawing on a large pile of GM's own data built up over decades. Inside the car, it shows up as driving automation and the trusted-companion language Anderson used.
The catch for you, as a customer, is what happens to the data. The interview was framed partly around data and privacy, and the company has described data as a kind of currency in this strategy. A car that learns from its owners needs to collect something from them. How much, and on what terms, is a conversation GM will have to have in public sooner rather than later.
Robots, robotaxis and what Anderson is really arguing
The interview's headline takeaway is that Anderson doesn't see a role for humanoid robots on the assembly line. It's the kind of line that travels well. I can't tell you the full reasoning behind it, since the exchange isn't something I can reproduce here, but the framing hints at a broader philosophy: use AI where it fits the problem, not where it makes the best demo.
That's a reasonable stance in an industry currently in love with humanoid machines. A factory already has purpose-built arms that do one job very well. Replacing one with something shaped like a person is a showpiece until someone proves it's cheaper.
The same practical streak runs through the rest of the conversation. The Verge's summary says Anderson digs into the physics of multisensor vehicle safety and argues that personally owned autonomous vehicles beat robotaxis. Those are two separate claims, and each could carry its own article. The first is a technical argument for cars that cross-check what they see using more than one kind of sensor. The second is a business one: GM would rather sell you a car that drives itself than run a fleet that drives strangers around.
Robotaxi operators bear the cost of every vehicle and every mile. A carmaker selling self-driving capability to owners puts the hardware on someone else's balance sheet, which happens to be GM's whole business model. You can believe Anderson's case on its merits and still notice how nicely it lines up with what GM already knows how to do.
What to watch next
Three things will tell us whether this was a strategy or a slide deck.
First, a real vehicle program with a shortened schedule. GM can say two years all it likes. A model that reaches showrooms on that timeline, with clean safety results, is the proof.
Second, the eyes-off driving feature. Hands-free systems already exist. Taking the driver's attention out of the loop raises the stakes for regulators, insurers and anyone who has to explain a crash. How GM rolls it out, and to whom, will say a lot about how confident the company really is.
Third, the money. The EV slowdown is forcing choices on every automaker, and GM is trying to fund batteries, software and autonomy at the same moment. If any of those three has to give, we'll probably see it in the next round of investment decisions before we see it in a press release.