If Tesla doesn’t have the right training data then no one does. Tesla has both simulations and vast amounts of real world data from cars all over the world. Extracting depth maps from 2d images is a solved problem, which negates the need for Lidar. Best I can tell Tesla is the closest out of everyone to solving self driving. I don’t think they’ll solve it by next year, but If I had to guess I’d say 2.5 to 3 years from now Tesla will have relatively safe self driving cars.