The FSD approach misunderstands the problem. Let's say hypothetically you could use cameras and ML to create an agent as good as a human driver. But is that goo

The FSD approach misunderstands the problem. Let's say hypothetically you could use cameras and ML to create an agent as good as a human driver. But is that good enough? No, no one will accept that standard. The expectation is that self driving is near perfect. To achieve that is going to take more than just cameras.