Over half a century in failed AI research is also a very long time.
Yes many projects fail. It they all did not fail then I would argue the research goals were to easy and to conservative. You should want a certain percent of failure in advanced reattach.
But there is also huge success. Look at Siri. OK it is not perfect, not even close but continuos speech recognition was at one time an unrealized team of the AI community and now it is common place.
One other thing, don't fault AI because it does not have better than human capability. For example human driven cars killed 1,250,000 people last year alone. That is like exploding an atomic bomb over a small city every year. Self driving cars don't need to be perfect, they only need to kill less then a million people for be an improvement.
Yes, over a million killed world wide, 33,000 last year in the US alone. 260,000 in China
Humans are not really so good at what they do. We misunderstand each other and ask the other person to repeat, we run cars into trees we invest money in scams and kids fail math class. If AI were only as good at playing chess as the average human we'd say the AI sucked big time. If AI were only as good as humans at driving we'd bad it from the roads.
Think of it this way, if AI, self driving cars killed 500,000 people in one year it would mean it saved the lives of 750,000 people. It does not need to be perfect to be better then human because humans are really, really bad at driving. We are on average really bad at math. Only a few people are good at some things, AI only has to be better than the average human not better than the best human
[doublepost=1461004760][/doublepost]
I think there will be a gradual reversal of the tendency to defer to non-automated cars. Certain lanes and even roadways will be designated as Autonomous-only, like HOV lanes are today. Since the automated cars can cooperate actively and passively, traffic will flow more smoothly in these lanes.
It will creep in gradually. Already we have an announcement that in a couple years EVERY car (in the US) will have automated breaks. Next steering controls that prevent you from turning into some object like another car. These safety features will be added one by one until finally you can take your hands off the wheel and turn around to talk to the person in the back seat without hitting anything. Once the basic car is safe enough that even a drunk can't hit anything, then you can turn the driving over to the AI. The safety features will prevent the AI from crashing the car the same way they prevent the drunk from crashing the car. The layers are independent so the whole thing does not fail at once
The safety systems are layered. Will still have dual cylinder brakes and the next layer is anti-skid sensors on the heels and then we have the collision sensor to automatically apply the brakes, they will add more and more layers until finally you don't need a driver, so it will be gradual, I think over 20 years or more.
