Maybe someone will point out that a computer program cannot possibly have the same ability to perceive and judge as a human. Yes, humans are fallible and cause accidents all the time but can distinguish between a human and a shadow. A computer program does not perceive. It calculates. Human perception is superior to that of a computer program with a camera lens.
Neural Networks are a significant part of these systems. We’re not just talking about lines of code anymore.
Actually, my take on this is that whatever issue you want to use for comparison, the software of the vehicle can and will be able to exceed the capability of the human. Even things like anticipation of potential problems can be handled (although I doubt that they are using these speculations in today’s versions — they can adapt and improve and someday will do just that.)
When we think of something that is extremely human, like intuition, a computer program can be designed so that it will appear to mimic the human mind and do something intuitively. Only it will be the result of massive calculations and Baysian statistics that comes out with the same decision.
The other thing that we know is that the vehicle will have software that will make decisions and sometimes these decisions will choose the “best” result for the vehicle. In other words, do you program the vehicle to hit the pedestrian or swerve into a lamp post and risk killing the occupants of the vehicle? This software will eventually be fine tuned and installed in these vehicles as well.