The learning car.
- 9. January 2017
- Autonomous Driving
- Photos: Daimler
- Text: Steffan Heuer
Powerful supercomputers will fundamentally revolutionise the automotive. Artificial Intelligence as a megatrend at the CES 2017 in Las Vegas.
At this year’s CES trade show in Las Vegas, Mercedes-Benz and graphics chip maker NVIDIA offered glimpses into their collaboration to unveil a game-changing AI car in the coming year — a goal into which many tech and automotive companies around the world have devoted massive resources.
“It’s easy for humans to recognise things and distill these observations into a mental model of the world around them,” explained Jen-Hsun Huang, co-founder, president and CEO of NVIDIA in an Inspiration Talk at the Mercedes-Benz stall. Until very recently, though, that level of perception wasn’t possible for even the most powerful computers.
The challenge of autonomous driving
The breakthrough came with so-called deep learning, explained Huang, who had called AI “the solution to self-driving” in the show’s opening keynote. Deep learning involves neural networks building on layer upon layer of computation to teach themselves about the world. “We have dedicated ourselves to working on the most extreme computing platforms. Now we can say that AI is the future of computing and will revolutionise the future of automobiles.”
Sajjad Khan, Vice President Digital Vehicle and Mobility at Daimler AG, described how Mercedes-Benz and NVIDIA have been building a long-term partnership to bring AI to everyday drivers. “We need excellent software and great computational power to make things happen that come naturally to the user but require an immense technical input.”
Jen-Hsun Huang, Co-Founder and CEO of NVIDIA
As a glimpse of things to come, he pointed to upcoming features in the new C- and E-Class. Their software studies user behaviour to make personalised predictions, for instance automatically suggesting the morning’s routing or navigating to the nearest gas station to fill up, based on driving history, realtime sensor readings, destination and other factors. With an in-car office experience, incremental AI gives drivers time back in their busy days, said Khan.
Cars that understand their environment
Accomplishing true AI, both experts agreed, requires new thinking about how a car should be designed, starting with the computing architecture on which all systems will be built. It’s a path Mercedes-Benz and NVIDIA embarked on three years ago, but they wouldn’t go into details what’s in store for next year to keep it a “surprise”.
Once every vehicle houses a supercomputer and is connected to the cloud, a future Mercedes will evolve into a highly adaptive co-pilot that’s always looking out for its passengers’ comfort and safety inside and outside the cabin.
CES 2017 in Las Vegas – read more on specific subjects here: