I will never trust the law enforcement that this wasn't Uber's fault. Even if the law is perverted to reach this end before possible revolution....
http://www.latimes.com/business/autos/la-fi-hy-uber-self-driving-20180319-story.html
I will never trust an AI without incredible restrictions with these as the example of the SMALLEST of precautions, set to automatically go off, with several over redundant remote activators if it so much as considers anything I MAY not like. Also I demand an activator set as a deadman switch:
https://en.wikipedia.org/wiki/Explosively_pumped_flux_compression_generator
I find these statements baffling. Humans already hit other people with cars far,
far more often than AIs do; it's completely provable that giving humans a manual override can only increase the amount of damage. I don't trust
you with the power to override the AI's decisions with your faulty human brain, to be frank.
And I'm not sure why you think an EMP will save you. EMPs don't really work the same way they do on television, and, besides, basic radiation hardening is going to be an essential part of safety design for any of these systems - because an EMP caused by lightning or intentional sabotage could otherwise cause fatalities. So, again, your expectations are unacceptable in a real-world situation.
Look, Stephen Hawking was great at physics, but he wasn't an AI researcher and had no idea what he was talking about. Self-driving cars cannot "go skynet". And if you're worried about that then let me just say that you really wouldn't like some of the things I've been involved in developing, heh heh.