Audi and NVIDIA give an AI a crash course in driving
Rolling around in a deep-learning SUV that only has four days of experience behind the wheel.
Many of the self-driving demonstrations at CES involved systems required months or even years of training. NVIDIA and Audi decided to see what they could do in four days.
The automaker and chip company gave the AI installed on an Audi Q7 images of what it should perceive as a road including, white lines, orange cones, rows of rocks and a dirt road and that's it. It didn't program a path or even add any additional sensors beyond the single forward-facing camera.
While the results were impressive, it's important to point out that this is only a very small part of an autonomous driving system. But it does show how powerful AIs have become and how quickly they can make sense of the real world.
It's also helpful to know that if your path to work is under construction and lines have been redrawn, cars will be still be able to determine where they need to go. While it's important to have highly detailed maps for path planning, this research is important when those navigation tools haven't had a chance to take note of a roadway change.
Click here to catch up on the latest news from CES 2017.
