Question: When will a robot car win a car race against human drivers?

One of the classic questions in Artificial Intelligence is when will a computer beat a master human in some game (Chess, Jeopardy). Well, the next test is not a simple board game, but something involving control of a physical device at high speeds. So the next test is when will a self-driving autonomous vehicle without a human in the drivers seat (at all or via remote control) beat the humans in a NASCAR, Indy, or Formula One car race?

My best guess is 2025, though I am fairly confident before 2029 (the bicentennial of the Rainhill Trials and a year before the bicentennial of the Tom Thumb on the B&O), but this is without a lot of basis. We don’t have enough data points of cars vs. humans in a race. The automakers are not going to be too keen on this, since it sends the wrong message perhaps, and their is a great deal of risk, so it will be after robo-cars are already somewhat widely deployed.

Obviously if you took the safeties off, and didn’t worry about crashing, a robo-car could drive faster than humans now on an empty track. But with other drivers, this would affect the computer, how would/should it respond?