>Fisher engaged Autopilot while the car was in motion on the track, then set the speed dial (on the right spoke of the steering wheel) to 0, which brought the car to a complete stop. Fisher next placed a small, weighted chain on the steering wheel, to simulate the weight of a driver’s hand, and slid over into the front passenger seat without opening any of the vehicle’s doors, because that would disengage Autopilot. Using the same steering wheel dial, which controls multiple functions in addition to Autopilot’s speed, Fisher reached over and was able to accelerate the vehicle from a full stop. He stopped the vehicle by dialing the speed back down to zero.
If anyone is curious how to do it. From https://www.consumerreports.org/autonomous-driving/cr-engine.... Personally I wouldn't call it too easy to do; That summation makes me think it would be possible on a whim with no forethought (surely a wealth of the scenarios where someone would choose to do this), but looking at the picture in the article they needed a setup with a large weight (but not too large) tied around a length of rope (also something I don't have available on demand in my car and would have to go somewhere and buy).
I'm not saying a drunk idiot at home with some unusual materials couldn't do it but it's not as simple as something you can do anytime you feel like it.
On the topic of better systems to get the driver to look at the road: "By comparison, BMW, Ford, GM, Subaru, and others use camera-based systems that can track the movements of a driver’s eyes and/or head position to ensure that they’re looking at the road. Some vehicles—including those equipped with GM’s Super Cruise—can automatically slow to a stop if they detect that drivers have ignored repeated warnings to look at the road."
We shouldn't forget that there's a lot of prior art on this area of human-machine relations. One place where we've had automated driving for ages is in train/subway and getting their conductors to pay attention even when they don't have anything to do for hours on end is something that's been struggled with. I recall that they've tried things like adding a button that they have to press periodically, but it doesn't work (the brain still can't help falling asleep and it will automatically interact with the button). Same for other more advanced things like making them input some more complicated action to show they're awake. The brain's just too good at ignoring things it realizes aren't necessary. I think one of the best solutions, maybe the only solution that's been foolproof, is https://en.wikipedia.org/wiki/Pointing_and_calling. Not sure how we would utilize that idea for self driving cars though.