I looked over that paper, and honestly it is nothing that impressive. Basically it just described applying off-the-shelf algorithms to a particular kind of data. It probably would be published and is interesting to look over, so I don't mean to deride the paper itself, but still it is just not that novel. Nothing about how their end-to-end self driving works or how it's better than the many competitors. So to me this was hugely underwhelming, really.