NVIDIA (pronounced nah-vid-eya) has been around for 23 years, but you may not have heard of it.
That’s because it’s a company working on big projects behind the scenes. It’s first foray into the tech world was as a 3D graphics company, primarily designing the graphics used in video gaming. But that eventually led to a surprisingly natural partnership with automakers, with NVIDIA setting up virtual wind tunnels or crash tests to help advance the design and engineering of cars.
“It even looks a little bit like a video game, but it’s using real data so we can help automakers create safer vehicles by letting them very quickly design and test,” Danny Shapiro, NVIDIA’s senior director of automotive, told Business Insider of the virtual tests.
Over time, NVIDIA has progressively worked itself into vehicles you use or hear about everyday — like Tesla, which uses NVIDIA for the graphics on its giant 17-inch touchscreen. Shapiro said there are 10 million cars crossing 20 different brands on the road today using some form of NVIDIA tech inside.
But perhaps more importantly, NVIDIA is designing the brain running self-driving cars, like ones built by Audi, Ford, Mercedes-Benz, and Volvo. In total, there are more than 80 automakers, start-ups, and research institutes relying on NVIDIA’s brain to power their self-driving cars, Shapiro said.
“The ability for our systems to now be able to sense what’s going on around car, to interpret it, to understand it, and take action in a 30th of a second is what is enabling autonomous vehicles on the road today,” Shapiro said.
The brain running these cars is NVIDIA’s Drive PX 2 — an artificial intelligence computer that helps cars locate themselves and avoid obstacles. As Shapiro puts it, the Drive PX 2 gives cars the ability to “see,” so to speak.
Self-driving cars, like the ones built by Ford, use a suite of sensors to detect objects. They also use radar and lidar, which use radio waves and lidar waves, respectively, to detect obstacles.
Those various sensors are “more or less building a three-dimensional model in the brain of the car, of everything that’s going on around that vehicle,” Shapiro said. It’s then up to the car brain to see its surroundings and react appropriately and swiftly.
The performance of the Drive PX 2 is so strong that it’s equivalent to that of 150 MacBook pros, but it’s roughly the size of your forearm.
“[It’s capable of] 24 trillion deep learning operations a second,” Shapiro said. Just 10 years ago, that kind of operating power would have required a massive supercomputer the “size of a house” as well as “500,000 watts of power and a massive air conditioner,” he added.
“We now have this down to the size of a license plate,” Shapiro said.
It’s size is key, because fitting a massive supercomputer into a truck is less than ideal. Audi lines the Drive PX 2 along the sides of its trunk, and it’s capable of operating in harsh conditions without an AC unit ranging from negative 40 degrees Fahrenheit to 185 degrees Fahrenheit.
And the brain is only getting more sophisticated. For example, the Drive PX 2 can distinguish between an ambulance and a FedEx truck — that comes in handy so the car knows to pull to the right for one but not the other.
NVIDIA is bound to become a bigger player in the space as more car companies and tech giants rely on its AI.
And NVIDIA’s access to a massive pool of data is key. It recently announced a partnership with Baidu, often referred to as the Google of China, to create a cloud-to-car, self-driving system. That’s a big deal, because it can aggregate the data collected by all the autonomous car fleets who choose to use the system.
Having a bigger pool of data to pull from will help accelerate the advancement of the car brain, and it also opens the door for over-the-air updates. Considering Baidu is now testing in California in addition to China, NVIDIA is creating a treasure trove of information to improve its brain.
“We’re achieving levels of perception that are superhuman,” Shapiro said. “Very soon we are going to see our tech in production vehicles.”
Via Danielle Muoio, Business Insider