NVIDIA self- driving AI chip - boom or bust for BRN?, page-14

Yes, but as the interview in the video points out, a lot of processing for driving decisions is made without going to the cloud (and by using 'AI').

Danny Shapiro (NVIDIA): "For the autonomy, that decision making has to happen on-board. Its too mission-critical. The amount of time it has to make a decision is too short to go to the cloud, do some computation, and come back. So, all that sensor data feeds straight into the car computer, and basically we have 1/30th of a second — one frame of video — to be able to identify everything in that scene and then make those driving decisions."

Interviewer: "How are you able to make those decisions so fast? Is it like, some specialised algorithm? What's going on at the software level to be able to do things so fast?"

Danny Shapiro (NVIDIA): "Another really good question. That's where artificial intelligence comes in. The world is too complex to be able to write code to say, 'If you see this, do this. If you see this other thing — there's too many things that could happen. There's too much in a scene."

Interviewer: "Right, so rules-based doesn't work."

Danny Shapiro (NVIDIA): "Correct. So, AI we can train to recognise all different types of objects and scenarios. In a fraction of a second, with a powerful super computer, we're able to understand that full environment."