Two things to note about the Uber car killing

Since this is the first fatality I think one thing that is interesting to highlight is that this sad accident is not of the type autonomous vehicle people love to bring up: how will the car decide between the baby and two adults... blah blah. Just like almost all accidents aren't. Thus, those corner cases are simply not important. That should really be remembered. What is important is to avoid regular accident due to imperfect sensor data and poor detection algorithms. Like this one. This is same as the accidents when humans are driving.

Further, there are calculations now claiming autonomous cars are about 25 times more likely to hit a pedestrian than a human driver. This was interpreted as that the is not safe and not ready and so on. However, I would like to claim the very opposite. Looking at technology development, we know Moore's law, indicating a doubling of processor performance in every 18 months. If this type of scaling i applied to the performance of autonomous vehicles, it gives that they will perform on term with a human driver in five years. In five more, the autonomous vehicles will outperform humans by a factor 32. This is the kind of scaling we can expect. Since we are only a factor 32 away from human performance today, we should be celebrating, we are almost there. Making cars much safer than they used to be. This is great news!