Red Light FSD policy by-the-book?
Not that this specific edge case couldn’t be imagined. It’s more like, if this case were to be specifically imagined and trained on, there are nearly “infinite” other similarly obscure cases that could also be imagined and trained on. Infinite cases would preclude FSD from ever shipping. So instead, the basic general cases need to be trained. Then actually occurring cases need to be trained. Finally not before foreseen cases need to be checked for relevance and new training possibilities. Once FSD gets trained well enough to do the best possible steps given most unforeseen cases, the training will be pretty much done enough to start using. After that, there will be the further drive toward perfection that never really ends.video marker 14:20
Give Tesla break on FSD performance, nobody could imagine this edge case. Same person who drew the CT sketch drew the temporary graphics on the intersection.
