I am an expert in AI. Learned the symbolic kind before the AI winter, then worked on neural nets (stealthily) during the AI winter. Automonous driving cannot y

I am an expert in AI. Learned the symbolic kind before the AI winter, then worked on neural nets (stealthily) during the AI winter. Automonous driving cannot yet handle all the edge cases, and when it fails people die. What's worse is when it fails it cannot explain why it failed, so we just throw more training data at it and hope for the best. That's not engineering; that's ML cargo cult voodoo. Modern ML is great for finding family members in your photo collection, but nobody dies when it labels your dog your grandmother. Tesla makes great cars but I never use their autopilot. They should focus on what they do best and give up this obsession with autonomous driving.