It's a bit counterintuitive but the logic goes: We can't have full self driving unless vision-based perception is essentially 100% solved. If that happens, LiDAR would be only a very marginal improvement over vision and not worth the complexity.
youtube.com/watch?v=BFdWsJs6…