LIDAR, paired with a very high definition map, can be very good. It's quite unlike the vision sensing that Tesla is trying to do. LIDAR, in general, won't hit things because LIDAR is very good at sensing obstacles (particularly in dry clear weather). So if you are a risk averse self-driving company then LIDAR makes a lot of sense because, above all, you don't want to hit anything.

Tesla's vision system still tries to hit things, a lot. It is still confused by what is actually in front of it, a lot. If you set a Tesla loose in Phoenix without a driver it would probably hit something within an hour. Also, if I'm not mistaken Waymo is also using vision to supplement the LIDAR. I think Waymo used the kitchen sink approach throwing virtually every kind of sensor at the problem of not hitting things.