There seem to be all hese suggestions to develop technology to monitor driver using self-driving cars, for example to check if they are alertness by tracking eye movents, or checking if the head is tilted weirdly(dozed off). I wonder if anybody considered resorting to old-tech.

Why don’t we use the legacy human verification system known as CAPTCHA to confirm that driver is awake and not drunk or otherwise temporarily disabled? Since the car is self-driving, the driver could take a few seconds to solve a puzzle. The puzzles should be chosen to require the perceptive and cognitive abilities relevant to driving the car. In an extreme degenerate case, the test could be to ask the human river to take a drive in an simulated environment. This would detect a driver failure ex machina. If he’s drunk, tired, or on drugs, then he fails and the auto-pilot can take some actions:

1) turn on the flashers. And I think self-driving cars should have special signals beyond the turn, break, parking and headlights to indicate to other drivers that the AI is failing and human is not in control.

2) shine bright lights, blast loud sound or cold air, blast oxygen enriched air, spray very cold water, inject adrenalin, etc. to attempt to revive driver.

3) pull over (AI should slowdown or speed up to try to reach a safe lane to temporarily stop and wait for help)

4) call the cops

5) “phone a friend” by dialing up a remote-ops center to have a trained (human) professional guide the vehicle to safety.

6) call the owner’s insurance company.

7) do all of them at the above at the same time.

I’ll perhaps hate this when implemented, but for now it seems much simpler and more effective than AI passively ascertaining proper driver condition.

P.s. the two birds being AI and human driver.