If you do actual ML research, the technologies are two sides of the same coin. Three days ago there was a paper posted here in HN on using the GPT-like transformer architecture for reinforcement learning problems (of which self driving cars is a partial subset of).