For a deep learning machine with LIDAR , there's only small difference between animals, indian traffic, bad drivers, and narrow/unlit/poor roads. It's just a q

For a deep learning machine with LIDAR , there's only small difference between animals, indian traffic, bad drivers, and narrow/unlit/poor roads. It's just a question of more training(altough a recent NHTSA? regulation says the planning stage cannot use black box techniques like DNNs). And once self-driving is a business, i suspect we'll find ways to get said training data for cheap enough and fast enough. For example Tesla/Uber drivers as trainers .