The Uber tragedy earlier this month sent a shudder throughout the autonomous vehicle industry. Some of the companies working on the technology paused testing. Nvidia was one of them.
“We suspended because there is obviously a new data point as a result of the accident,” said Nvidia CEO Jensen Huang. “As good engineers, we should simply wait to see if we can learn something from that experience. We don’t know that we would do anything different, but we should give ourselves time to see if we can learn from that incident. It won’t take long.”
Huang made that remark during a press briefing at the 2018 GPU Technology Conference (GTC). In an earlier keynote address he discussed Nvidia’s role in developing autonomous driving technology—and a new product. His inner engineer shining through, Huang rattled off numbers to make his case.
“Civilization drives about 10 trillion miles a year,” he said. “In the U.S., 770 accidents happen per 1 billion miles. [Automotive] safety work in US has reduced the number of accidents, so now you have to drive a billion miles to produce 770 accidents.”
That means, he said, “You have to ask yourself how confident you are when your fleet of 20 test cars has driven one million miles.”
Autonomous cars need a lot more miles under their wheels than that, he indicated, to gain enough experience under variety of conditions so that their designers can fine tune—and then prove—their safety. The answer, Huang said, is doing the bulk of the testing in virtual reality.
“That’s where Nvidia can shine. We know how to build VR worlds. We could have thousands of these worlds with thousands of scenarios running, with our AI car driving itself through these worlds. If something happens, it could send us an email and we can jump in and figure out what was going on.”
“We could use VR to create extreme corner rare scenarios that are completely repeatable,” he indicated, “so engineers can debug these systems.”