I have to disagree with the article.
"Tesla itself have been anticipating the first fatal accident"
Not anticipating it very well, since they could have reduced the Autopilot feature to make it much less likely for this kind of accident to happen. For example Mercedes Benz anticipated it much more strongly by their system requiring a hand on the wheel at all times.
And this is just my opinion, they should have known any image processing program (robot) is guaranteed not to be able to cope with the complexity of the real world therefore should not have released the system at all.
I used to work in the robotics industry 16 years ago, although only very briefly (8 months). When I was there one of the projects was to replace human mushroom pickers with image processing gantry robots. The main challenge of course as with all robotics was to program the robot to interact with the real world, in this case via image processing. The programmers showed us the problems they were facing, the huge difficulty trying to program a robot to recognise ripe mushrooms, to understand 3D that some mushrooms are below others and not to try to pick shorter mushrooms first, etc. So many factors make it is impossible to code for every eventuality, slight changes in lighting conditions, change in hue due to mushroom growing at angles, etc.
If I remember correctly they gave up in the end, mushrooms are still picked by hand.
What caused auto manufacturers to think they can code image recognition for the even more varied and complex world of roads and traffic?
I say it is impossible and should not even be attempted. There is a good reason why rails exist, and even then accidents happen.
At the end of the day, there is no beating the human brain, its ready for anything.
http://neurosciencenews.com/neurosci...primates-4439/