Jay Gierak at Ghost, based in Mountain View, California, was impressed by Wayve’s performances and agreed with the company’s overall stance. “The robotic approach is not the right way to do this,” says Gierak.
But he’s not sold on Wayve’s utter commitment to deep learning. Instead of a single large model, Ghost trains hundreds of smaller models, each with a specialism. It then hand-writes simple rules to tell the self-driving system which vehicle to use in what situations. (Ghost’s approach is similar to that of another AV2.0 company, Autobrains, based in Israel. But Autobrains uses another layer of neural networks to learn the rules.)
According to Volkmar Uhlig, co-founder and CTO of Ghost, breaking the AI into smaller pieces, each with specific functions, makes it easier to determine that an autonomous vehicle is safe. “At some point, something is going to happen,” he said. “And a referee will ask you to point to a piece of code that says: ‘If there’s a person in front of you, you have to brake.’ That code needs to exist”. Code can still be learned, but in a large model like Wayve, it would be difficult to find, Uhlig said.
However, the two companies are pursuing complementary goals: Ghost wants to create consumer vehicles that can drive themselves on highways; Wayve wants to be the first company to bring driverless cars into 100 cities. Wayve is currently working with UK grocery giants Asda and Ocado, collecting data from their urban delivery vehicles.
However, by many measures, both companies are far behind the market leaders. Cruise and Waymo have spent hundreds of hours driving without people in their vehicles and have made robotaxi services available to the public in a few small locations.
“I don’t want to downsize the challenge at hand,” says Hawke. “The AV industry teaches you humility.”
