I'm with you up until the software being inherently unsafe bit. I've seen a lot of code I wouldn't trust with anything, let alone my life, but in your own examp

I'm with you up until the software being inherently unsafe bit. I've seen a lot of code I wouldn't trust with anything, let alone my life, but in your own examples: trains are partly safe due to automation, warning systems, etc. and cars have gotten safer at least partly due to things like ABS and traction control which rely heavily on algorithms. However, those are relatively simple systems compared to a self driving car and a lot more "bugs" will likely show up as more AI drives more miles.