advertisement
advertisement

Uber teams up with Nvidia to power the AI in its self-driving cars

Uber teams up with Nvidia to power the AI in its self-driving cars
[Photo: courtesy of Nvidia]

Since 2015, taxi service Uber and AI chipmaker Nvidia have been like Sam and Diane from Cheers, or Mika and Joe from MSNBC. They work together every day, and everyone thinks of them as a couple. But it took a long time to make it official, which finally happened today. The two companies announced that they are in a self-driving car relationship for the long haul. Nvidia, which recently squeezed a trunkful of computers onto a circuit board the size of a license plate, will supply the artificial intelligence hardware and software to replace Uber drivers, as well as freight truck drivers, in the future.

The self-driving relationship will actually be a threesome, as Uber confirmed that the Nvidia chips will continue going into its growing fleet of Volvo XC90 SUV taxis. (In November, Uber announced its intention to buy up to 24,000 of the cars.) The Volvo-Nvidia robo-taxis have already carried passengers on 50,000 trips, covering over 2 million miles, in pilot projects that began in Pittsburgh in 2016 and Phoenix in 2017. Uber hasn’t settled on a supplier of trucks for its future autonomous freight-hauling business, however. And none of the companies are saying when autonomous taxis will go national and global.

Nor do they rule out this three-way taxi tie-up becoming an open marriage. “We don’t have any exclusive relationships,” says Danny Shapiro, Nvidia’s senior director of automotive. Uber says the same in response to an email I sent them. Intel, for instance, is racing to catch up to Nvidia with its own in-car computer. Googling “Intel” brings up a lot of bad news of late, as its processors are subject to serious security flaws, called Meltdown and Spectre, and the software fix could hit performance–possibly critical for an autonomous car that has to make split-millisecond decisions. However, one or both of the bugs could also affect some cell-phone-style ARM processors, as well; and Nvidia’s AI chip for autonomous cars, called Xavier, includes an ARM processor. “We’ve analyzed to determine which [chips] are affected, and we’re preparing appropriate mitigations,” says Shapiro. “But we’re not aware of anything that’s been an issue.”

advertisement