Tesla’s Elon Musk sees self-driving cars in 15 years — and they won’t be evil

Elon Musk, the chief executive of electric car maker Tesla, is worried about unbridled artificial intelligence taking over Earth in the future. But he’s OK with the A.I. built into self-driving cars, which he expects to be dominant in about 15 years.

Musk spoke about the topic of self-driving cars with Nvidia chief executive Jen-Hsun Huang at the GPUTech conference in San Jose, California today. Musk said that the A.I. in self-driving cars is relatively narrow in scope, even though it seems like a huge computing problem today.

“Tesla is the leader in electric cars, and we’ll also be the leader in autonomous cars,” Musk said. “If anyone is interested in working on autonomous cars, we’d love to have you at Tesla. It’s going to be the default thing and save a lot of lives.”

Nvidia revealed more details about its car supercomputer, the Nvidia Drive PX, during a keynote speech by Huang. That supercomputer is designed to be the brains of a self-driving car, but the supercomputer alone still costs about $10,000.

Musk said that self-driving cars can take advantage of current sensors like ultrasonic sensors. But over time, cars will add new sensors that give them much better 360-degree detection of threats and safety issues. And the car will need much more computing power to process data from those sensors, Musk said.

As far as safety goes, Musk said, “The evidence is already overwhelming” that self-driving cars can be better at things like alerting us to brake lights ahead than human reflexes are.

Convincing regulators that self-driving cars are safe won’t be easy. But Musk said that one great way to gather evidence is to have car computers operate in “shadow mode,” showing what a real driver does while driving and what the car computer would have done in the same situation.

“We’ll take autonomous cars for granted in short time,” Musk said.