I agree with you. ImageNet is so far from really understanding the world, that based on this article I wouldn't put any money in this company. Tesla had a lot

I agree with you. ImageNet is so far from really understanding the world, that based on this article I wouldn't put any money in this company. Tesla had a lot of presentations about the complexities of detecting the right light to use in a crossing with about 20-30 traffic lights for example. Depth detection is also crucial, and as you wrote there are lots of strange objects/animals/people to handle as edge cases.