To keep it in the realm of the actual Daytona 500.
There are a huge number of rules regarding the mechanical attributes of the cars. This is to make it somewhat reasonable to compete, as well as safety. If you were to substitute the driver with mechanical, computational devices, while still having human drivers competing, I think there would be limitations imposed on many of the abilities of the device / devices. Human drivers can process an amazing amount of data / feel, almost subconsciously. But they can only see their current field of view. A device could have 360 degree view. Not fair. No degradation of physical abilities over time or environmental conditions? Not fair. The visual acquisition problem is fairly easy to match to human capability. But how to account for the wear and tear of the race on the human abilities? It is not just initial driving skill, but how the driver can keep the highest level through to the end of the race. The rules regarding the mechanics of the cars are also there to highlight the abilities of the drivers.
That sounds to me like in able to make the car competitive enough to even be permitted to participate in qualifying (which I doubt it would), it would be a killing machine.
And maybe I’m stupid, but I can’t wrap my head around a self-driving car drafting at 200mph, two inches from the rear bumper of the car in front. And how would it deal with the infamous NASCAR “rubbing” and “bumping”?
I feel I can safely say, “Not in my lifetime”.
Actually at Daytona? I suspect that there would be an “uncanny valley” effect, in that at least some of the drivers would be gleefully trying to nudge the robot driven car out of the race. If the car can make it more than ten-twenty laps (and the team doesn’t run out of tape), then it might have a good chance of winning.
It’s an impressive talent in a human. A robot could be programmed with a synthesis of a dozen hot drivers’ “intuition” - validated by science and engineering, and implemented with infallibly fast and precise servos.
Of course, what we’re basically talking about here is F1. 
I’d bet that short of a gasoline gang-rape, a robot car with adequate sensors could avoid most contact, or - if programming were allowed - “fight back” with infinitely better knowledge of the dynamics and immediate situation. It’d be bump-FLICK! and off goes the bump-er.
That’s one of the main reasons there’s such a push for self-driving cars, because they can do that. Even using the ultra-safe standards that would be required for consumer models. For a computer, that’s really easy. It’s just tough for us meatbags.
Another ‘cloud as god’ argument.
No, an AI machine cannot safely tailgate another car at 200 mph.
Your instantly-reacting servo still has lag, and the chunk-o-metal that just flew off a car 2 places forward will cause a ‘hiccup’ in the car being tailgated. Work out the time required to close 2" at 200 mph. How much slippage does the lead car encounter as it goes over the brand-new obstacle?
<insert quote about ‘there is no power to save us from ourselves; only we can do that’>
Just how fast do you think people can react?
And just how much “slippage” do you think is going to matter? Why wouldn’t a servo be able to react and calculate a trajectory as fast as a human? With better information of all the other items around it?
I’d believe that the biggest obstacle to winning would be the other drivers ganging up on the robot car to its detriment. And the robot being constrained by safety rules
That’s like saying that the only way a missile traveling at Mach 4 could possibly hit a plane taking evasive action would be if a human were guiding it.
And, we know that isn’t true.
I think a few concepts are getting confused here. “Self-driving car” doesn’t necessarily mean using the same software as the consumer models. Senegoid was just pointing out that such software has many parameters and can be tweaked to be more or less aggressive or whatever.
But I think if someone wanted AI to win Daytona, it would either be bespoke or a heavily-modified (likely very cut-down) version of the consumer software. It’s a very different, and indeed far more constrained, environment than the open road.
No - it’s a ‘machines can react faster than humans’ argument. There are already numerous examples of control systems that would be impossible for a human to operate due to the need for fine control and fast reaction times.
Modern fighter jets, for example - although still flown by a human pilot, may be inherently unstable and utterly impossible to fly without their dynamic stabilisation systems.
Why on Earth would we be closing that distance at 200 MPH? There’s no point in drafting behind a stopped car.
Let’s get them working on the cookie-cutter mile-and-a-half tracks first. I suspect restrictor-plate pack racing is considerably more complex than most of the intermediate tracks.
COULD a robot controlled NASCAR be built that could win? Absolutely yes, with unlimited budget and resources. The more you restrict budget/resources/time, then the ‘absolutely’ part starts to fade away. More restriction, yes turns into no.
Way to miss the point!
I am saying that your AI cannot do it EITHER!
I thought it obvious that a person cannot safely tailgate at 2".