I don't know if you're in the market for a new car, of course, but chances are that soon, possibly the next time you buy a vehicle, it will have so much processing power and artificial intelligence that you won't won't be able to keep up. And that's a very good thing.
Because the smarter cars get, the safer we become. It's estimated that we could reduce traffic fatalities by 90% -- or 30,000 lives every year -- by 2050, once cars start driving themselves.
To get there, tech companies are creating hardware and software that make semi-autonomous and fully autonomous cars a reality. NVIDIA (NVDA +1.80%) and Alphabet's (GOOG +0.05%) (GOOGL 0.04%) Google are two leaders in the car tech space -- and they're just getting started.
Advanced hardware
NVIDIA released two huge steps forward in automotive technology recently: its Drive PX 2 system and the DGX-1 supercomputer.
Drive PX 2 is the next iteration of NVIDIA's Drive PX system, which already helps power some of world's most advanced autonomous cars. Drive PX helped an Audi A7 drive nearly all by itself from San Francisco to Las Vegas and is used by other automakers, as well as Google.
But its new Drive PX 2 packs an even bigger punch than its predecessor. The new system can process 24 trillion deep-learning operations every second, and is paired with on-board cameras, sensors and other tech to give autonomous cars situational awareness. This gives driverless cars the ability to distinguish the difference between an ambulance and delivery truck and how it may have to respond to each one differently.
And then there's the the company's recently unveiled $129,000 DGX-1 supercomputer. The DGX-1 will act like a cloud-based server, helping to process 170 teraflops of data, in addition to the Drive PX 2's processing prowess.
The DGX-1 has eight of NVIDIA's new Tesla P100 graphics processors, two Intel Xenon processors, and has the same processing power of about 250 servers. It can also track 15,000 data points per second, from each of a vehicle's on-board cameras.
The supercomputer was designed, in part, to conduct research for universities, and NVIDIA says it could help discover "better cancer drugs." Now imagine that it's synced up with your car, making your driving decisions. It would be nearly impossible for it not to drive better than you.
But semi-autonomous and autonomous cars need more than just hardware to succeed. They need artificial intelligence systems, too.
Thinking like a human, only safer
To make the most of the on-board and cloud-based processing power, Google is implementing its own deep-learning system to help cars drive more like people, only better. Deep learning is a subset of artificial intelligence (AI) and allows computers to build on previous information that they've learned.
Google has learned that driverless cars need to have some similar attributes to human drivers so that passengers feel more comfortable driving in them and so that other drivers on the road understand what a driverless car is doing.
To do that, Google's used deep learning to make its driverless cars drive in ways that we understand. A Wall Street Journal article mentioned last year that Google's cars used to be too cautious, which lead it to tap the brakes when it didn't need to, or never pass a parked car if it meant slightly crossing over a double yellow line.
Now though, Google's cars are applying learned techniques that make them bend the rules a bit, while still driving safer than humans. That means they inch forward a bit after stopping at a stop sign so drivers know they intend to go, they cut into corners a bit more instead of making unnaturally wide turns, and they know how to navigate an intersection if the stop lights have gone out.
Google's use of deep learning for driverless autos is key to making cars that not only react faster but also make better driving decisions than their human counterparts.
Foolish final thoughts
If you pair supercomputers together with artificial intelligence, you get massively intelligent automobiles that will only get smarter the more they drive.
Imagine that your next car will not only be better than you (at driving, of course) as it rolls off of the assembly line, but over-the-air updates (just as our phones receive operating system updates) will make them even smarter. Tesla Motors already does with its vehicles and others are following suit.
The result is cars that are vastly more capable at driving than we are, and constantly improving their skills. That's the automotive future that's quickly approaching, and NVIDIA and Google are leading the way.
