The brains behind self-driving vehicles are rapidly evolving to meet the ambitions of automakers and tech companies that intend to launch truly driverless cars over the next few years.
Silicon Valley chip maker Nvidia, one of the industry?s most prominent suppliers of artificial-intelligence systems, unveiled a supercomputer tailored for use in autonomous taxis that can reach Level 5 automation?that is, operation without human oversight or input?and handle every conceivable road situation. The company?s system, dubbed the Drive PX Pegasus, can process more than 320 trillion operations per second.
That?s about 13 times faster than the company?s current Drive PX 2, which crunches 24 trillion operations per second. As more than two dozen companies set their sights on deploying fleets of driverless taxis, industry analysts say that?s the kind of processing power necessary to fuse raw data provided by lidar, radar, and cameras, allowing self-driving vehicles to detect obstacles and determine where they are on the road. ?Today, dozens of companies are racing to develop robo-taxis, but they are still gated by the massive computation needs of a truly driverless car,? said Luca De Ambroggi, senior principal automotive analyst at IHS Markit. Nvidia?s latest, he said, ?shows the path to production for the automakers, startups and automotive ecosystem working to deliver this.?

?
?We need more computational horsepower to train

cars on how bad drivers drive.?
? Danny Shapiro, N...