While at a recent industry event, Toyota offered a thought-provoking announcement that its upcoming line of self-driving cars would be powered by a Nvidia platform. While many may think that Nvidia is just a tool to drive video game graphics, there's a lot more than that going on under Nvidia's hood. Toyota is putting Nvidia's Drive PX supercomputer to work as part of that self-driving car line.
The self-driving cars are set to become available in just a few years, with some projecting that by 2020, the first ones will be on hand, able to drive where drivers—or perhaps now, occupants—tell said cars to go. This is also part of a larger development from Nvidia, which is actively trying to bring artificial intelligence (AI) to more products.
However, some parts of Toyota's projections are vague at best, as neither it nor Nvidia has defined which cars get the systems. There are also mixed reports about some cars that are completely autonomous for the elderly and the disabled, but other cars come with only partial autonomy; the “guardian angel” system driving such cars gives humans control, but takes over only to prevent an accident.
What's more, Nvidia's system hasn't exactly had a lot of fire-testing; the Drive PX system has only been around since 2015, reports note, with the improved Drive PX 2 version coming out just a year later. Though the processor driving it all—the Nvidia Xavier—has more than enough power for such operations at 30 trillion deep learning operations per second, it too is a comparatively recent invention, only coming out in September.
This is the inherent problem the system faces; Nvidia and Toyota are basically set to take these cars to market in 2020, by some reports, but to do so on the strength of systems that have really only existed since 2015. How would Nvidia and Toyota overcome the trust gap that almost certainly would arise from saying “Hey folks, get in this self-driving car! Its technology has existed for five whole years, and that's like an eternity in Internet time!”? If it can focus instead on concurrent testing and development, it can instead focus on man-hours spent on development.
Nvidia and Toyota will be asking a lot of people with these early-generation self-driving cars. In order to get around that issue, it's going to have to demonstrate safety unequivocally, especially if its primary market is the elderly and disabled. Still, the rest of us are eager for these cars to arrive as well; city traffic alone is reason enough to let much more reliable machines do the driving.