>The neural nets do not include any explicit “if X happens, then do Y” programming. Rather, they are trained to recognise and classify objects using examples of millions of videos and images from real-world driving conditions.
Real self-driving will need to include a robust expert system. This should obvious to anybody who thinks about how humans drive. Driving is not just pattern recognition but thousands of decisions and judgements about the real world every hour.
I don’t think we know enough about how the human brain works to make that statement.
There’s a lot of stuff that goes on outside of our conscious self.
The conscious part of our brain may just make up that rational, “I saw A, and because of B, I decided to do C” explanation after the fact.
Also, self-driving cars already have an expert system on top of their neural networks.
That’s what makes them start moving, decides where to take a right turn, etc. The neural networks only are there for deciding how to make that right turn.
Real self-driving will need to include a robust expert system. This should obvious to anybody who thinks about how humans drive. Driving is not just pattern recognition but thousands of decisions and judgements about the real world every hour.