In addition to the topics described by the English term "deep learning", a quarterly reporting NVIDIA conference much attention was paid to the development of autopilot systems for cars as well as the role of NVIDIA products in this context. It was stated that "the car supercomputer" DRIVE PX2 already supplied more than eighty partners, who can use it to give their vehicles a different degree of autonomy in management.
DRIVE PX2 in the company called "real-time supercomputer" that uses artificial intelligence to control the car. As it turns out, in some embodiments DRIVE PX2 can work without two discrete GPUs GP106 generation Pascal, content with a single processor Tegra generation Parker. It can work with cameras and sensors to recognize objects and determine their relative location. At the same time the head of NVIDIA recognizes that the whole complex of problems on driving automation can not be solved with a single processor. For this reason DRIVE PX2 made scalable platform - to its most productive form two Tegra processor combined with two discrete GPUs GP106.
However, NVIDIA is considered the head of the Tegra processors are a key component of the autopilot system. If not for these processors, and then create an autopilot would be impossible, he said.
In general, according to NVIDIA CEO, creating autopilot - it is a problem to a greater degree program. And it is very difficult, because if it were not so, NVIDIA would long ago have done autopilot". For Tegra processors are already in millions of automotive entertainment systems, and production cars with the autopilot has no offers. In addition to the CPU, you need the software. By the way, the first cars with driver assistance systems that use components and NVIDIA technology, consumers will see this year, but the bulk of the announcements will take place in the next two years.