The AI-designed car is taking shape

The world of automotive design is full of advanced 3D visualization tools and virtual reality sculpting platforms, but your average new car still enters the world as a sketch.

These drawings traditionally see endless iteration and refinement from all angles before being transformed into 3D models by hand, some die-cast in the digital world, others sculpted in clay to better visualize lines and contours. This is just the beginning of the design and development process that often takes half a decade or more.

This means that many of the new cars arriving at dealerships this summer were first scheduled for 2020 or 2021. The initiatives began when alternative fuel incentives became widespread and electric vehicle chargers spread like wildfire. The days of internal combustion are numbered.

Today everything has changed. The Trump administration’s second law eliminated all types of electric vehicle incentives while eliminating tariffs and import and export restrictions. Automakers that once pledged to go all-electric by the end of the decade are now shoving engines into anything that moves, and factories are being hastily recommissioned to dodge the worst of the import restrictions.

Amidst all this, we have a boom in artificial intelligence, which a growing number of manufacturers are taking advantage of to take advantage of the 60-month design and development window for new cars. As with most aspects of AI, the potential is huge. So do some of the more worrying implications.

At GM, the new vehicle development process gets an injection of artificial intelligence at the design stage. GM Creative Designer Dan Shapiro walked me through the workflow, which always starts with human design. “That’s the point of these drawings, and AI helps us see them sooner,” he said.

By inputting hand-drawn graphics into a commercially available tool called Vizcom, Shapiro was able to create a 3D model and fully realized animation in hours, a process he previously said took “multiple teams months.”

Shapiro’s example was a concept car with aggressive lines that would have looked at home on the streets Night city. Writing prompts like: “Create a dynamic rendering motion shot of this Chevy concept car… empty high streets. Modern city,” he created a simple animation. Soon they were rolling along the kinds of perpetually wet roads that are essential in a cyberpunk future.

On some iterations, the vertical wheel covers were gone, but this was quickly fixed with some quick revisions and re-renderings.

For now, at least, this animation is only used internally as a rolling animation Mood boards To help GM teams see what works. Shapiro was adamant that it is always human designers who shape things, not artificial intelligence: “We’re still monks deciding what Buick, GMC, Cadillac, and in this case, Chevy looks like.”

But artificial intelligence is having an impact there, too.

Computational fluid dynamics (CFD) is the science of determining how a fluid flows around a given shape. CFDs help EVs move a little further versus charging, and larger trucks offer a little better wind resistance. Since 2018, a Swiss company called Neural Concept has been bringing the power of neural networks to the art of CFD. Tasks that previously took hours on supercomputers can be simulated in minutes on GPUs like those from Nvidia.

Neural Concept has applied its technology to everything from family sedans to Formula 1 racers (Williams Racing is one of its clients), and while most of its clients prefer to remain anonymous, keeping details of their design tools and processes confidential, Jaguar Land Rover (JLR) has recently been praising the technology. At this year’s Nvidia GTC, Chris Johnston, a senior technical specialist at JLR, said that aerial missions that previously took 4 hours are now completed in 1 minute.

General Motors is on the same path, developing what it calls an “AI-powered virtual wind tunnel.” Scott Parrish, a technical fellow and lab manager at GM’s R&D division, gave me a demo. “We have developed an artificial intelligence model to provide near real-time prediction of clouds,” he said. Designers and engineers can push and pull surfaces and get near-instant feedback.

It’s not just about reshaping cars. The GM process is also changing. Where previously designers would hand off models to CFD engineers, who would test them for several days or weeks before providing feedback, this is now more frequent. Since designers can produce 3D models quickly, CFD work can start early.

However, these automated procedures are not perfect. “We are building autonomous systems that design cars with strong human supervision,” said Pierre Paquet, CEO and co-founder of Neural Concept. “The value comes from combining the speed of AI with human judgment, not from removing the human from the equation.”

The way the car looks and its penetration through the air aren’t the only aspects contributing to the half-decade development roadmap. Programming has become an increasingly large task. The push toward software-defined vehicles has meant more complex integration efforts that have delayed launches It costs billions. Artificial intelligence is seen as a potential boon here as well.

At Nissan, the main focus is on automating some of the menial software development tasks, such as unit tests. These code generation tools “improve development speed as well as quality,” Takashi Yoshizawa, the Nissan executive responsible for software-defined vehicles, told me.

A common theme among companies diving into AI is that it will boost worker productivity by eliminating menial tasks, not cutting headcount. GM representatives were adamant on this point. “This impacts something that a number of people are concerned about, but the way we really benefit from it is by allowing people to do what they already came to GM to do,” Brian Stiles said. He is the Director of Design Innovation and Technology Operations at GM Global Design.

Pierre Paquet of Neural Concept said the same about his clients: “Our platform is designed to amplify engineering teams, not diminish them.”

Matteo Licata isn’t so sure. A former automotive designer, he is currently a professor at IAAD (Istituto di Arte Applicata e Design) in Turin. “Jobs in design studios may not disappear immediately, but in my view, only a fool would think that such a massive increase in productivity wouldn’t impact studio headcount one way or another,” he said.

This has some of the most troubling implications for Licata students. “Getting into car design was very difficult before artificial intelligence, and now it is even more difficult,” he said.

Whether AI is a blessing or a disaster depends largely on how wisely manufacturers deploy it. Some show better judgment than others. Dodge recently published some supposed “Old family photos“One of its most popular models in 20 years. In fact, images generated by artificial intelligence barely look like the real thing.

Marketing missteps aside, the goal now is speed. AI is already being used in GM’s design process in next-generation vehicles, but no one has commented on when it will hit the market. For its part, Nissan is working to achieve a 30-month goal for new vehicles as it works to regain momentum in the American market.

Is this fast enough? We’ll find out in 2029.