That may be true. However, self driving cars are an entirely different matter. While they are really cool, do you really want to be in one hurling down the highway at 85MPH (I'm in Utah) and trusting that the automated systems are going to know the difference between a coyote or a tumbleweed?
We can prove if software is reliable. But humans are very unreliable. You already have people going 85MPH in Utah while drunk as a skunk. Your feeling of safety is already unjustified, and doesn't agree with the real risk you face.
There is a lot that can be done with image recognition and sets of rules for the correct behavior in every driving situation. And when I say every, I mean every single one. That's because we are able to categorize every event, and work out the right reaction for those categories.
Most of the reactions are really simple.
Tornado? stop the car. UFO landing in the middle of the highway? stop the car. Car in the turned over in the middle of the road? stop the car. Huge grass fire covering the road? stop, turn around, inform the local authorities of the situation. Person in a wheel chari crossing the street? yea, stop for them, then proceed when safe to do so. Jack ass passed you through a 4-way stop (that happened to me)? proceed slowly until safely out of the flow of traffic. As a human I barely knew how to handle that one at the time, it wasn't something in driver's ed. but hindsight it seems obvious now. But please don't criticize my examples too harshly, they're only examples not based on any real car firmware implementations. It's presented rhetorically to demonstrate that we can think through the problems and work out rules on how to handle situations, even ambiguous ones.
The rest of the every day situations are basic rules of the road, which hopefully we already are familiar with. Getting enough data to reliably detect the situations is the hard part, making the decision is not nearly so hard for the SW. Lots of sensors and image processing is done, and we're only scratching the surface right now. Eventually the car's awareness of the road conditions will vastly exceed a human's awareness, and in some ways it already does exceed average drivers.
As more devices become connected, and the processing power increases, we'll see some sophisticated capabilities for every common things like cars. The barrier to autonomous driving won't be technology, but rather a social resistance to the change. Something you're already demonstrating. I'm not saying we shouldn't be cautious about radical changes to our lives, especially where safety is concerned. But there is a level of rigor in the engineering for autonomous driving that isn't present in the driving test that Americans are taking today, so the assumption that humans are safe or more safe than a computer is on a shaky foundation.
