Comment by stephc_int13

Companies are going to die on that hill. I would not call it a waste of money.

It might work in the end, and even if the current tech is not sufficient to attain the goal of fully autonomous driving, the car is a very good guinea pig for robotic research.

We're collectively learning a lot in this process.

I am not surprised that it turns out to be much more difficult than anticipated.

I think it is 95% vaporware, but I liked the ideas behind the latest Nvidia presentation.

Machine learning is all about the training quality, useful robots (such as autonomous vehicles) have to operate in the real world.

We can't have robots that kills by accident.

It is difficult/risky/costly to train them in the real world. Good virtual twins are needed to achieve safe training with infinite repetitions in extreme situations.

Not sure if that is sufficient, but it is worth trying.