[D] Why does Tesla not train a large transformer model? They have a lot of data.

I am not an expert of machine learning. Given what I know though, if I was head of AI at Tesla I would at least attempt the following approach to the self-driving problem:

The architecture will be a large transformer model, basically the GPT of self driving. Take your recorded footage and split it into clips of varying length (seconds to minutes). (Tesla can probably get up to hundreds of millions of such clips).

Tuple those camera recordings with the …

data machinelearning transformer