I have worked with automated equipment, and I have worked with computers all my life. While nothing designed by humans is perfect, a lot of equipment can be made to remove the failures of human-factors, and we do end up with a lot fewer problems and better situations all-around.
Who among humans can say that he/she can do things more precisely and faster than a robot or a computer? If automated vehicles remove just 1% of accidents from our roads, our lives will have been improved tremendously. But, the promise is for removal of 100% or human failure from our roads; I’m willing to accept just 50% improvement in lowering in the number of accidents and deaths.
I have worked with automated equipment, and I have worked with computers all my life.
I have strong reason to doubt your assertion based on the naive confidence displayed in your statements.
The one nugget of accuracy that I found, is your acknowledgement that computers make mistakes faster than any human can react.
I design and build automated equipment for industrial applications, and have seen first-hand the absolute lack of subtlety with which a machine executes it's programming.
Have you ever had a headlight on your vehicle go out suddenly, and then later for no readily apparent cause, come back on? This same wiring, using these same connectors cannot ever be relied upon to autonomously operate a motor vehicle safely, and therein lies the absolutely fatal flaw in the autonomous transportation fantasy.
A sensor that fails in an automated assembly process, whether it is due to age, or neglect, or power transients, or simple, common dirt, can be made to fail in a safe state. That's the origin and meaning of "FAILSAFE"... That upon failure the sensor or switch output goes to a state, either ON or OFF that inhibits the machine from operating in the absence of meaningful input.
A motor vehicle, by it's very definition cannot be made to fail-safe. It is in motion, among others of its type and cannot safely
just stop operating.
Skipping over a few lines; as finicky as a simple optical sensor is, machine vision is orders of magnitude more complicated. To recognize a thing as a "thing" with no more resolution than presence or absence, requires that the item to be "seen" be in precise alignment with the sensors programmed expectations... Not too far away, not too close, certainly never rotated or canted at an angle to the sensor, and must be precisely illuminated as well. No machine ever invented, or ever to be invented will have the capability to identify an oncoming object and react appropriately in every circumstance. (or even the majority of circumstances) Solid or liquid, soft or hard, concave or convex... All of this has to be known IN ADVANCE in order to properly program the machine's reaction, which as you noted, will be irrevocably swift and final.
You say you've worked with computers "all your life" and I believe you. That also helps me understand your lack of judgement on this topic.
I'll tell you what: Come back when you can tell me all of the reasons why it hasn't been done already, and I might share your enthusiasm, but the truth is it's a horrible idea, and if people continue to insisit that "It's going to happen" regardless, then many people will die re-learning what some of us old farts already learned.