I’m gonna say it. I think the FSD project hit one of those roadblocks Neroden warned about (endlessly).
This sudden recruitment drive does not gel with “clear path to self driving” as discussed on Autonomy day. Further evidence of a delay was in the third row podcast part one where they talk about a rewrite to hand more of the primary decision making to the neural net.
It’s reassuring that ARKs targets are pretty good without FSD. I also like the talk of launching a human driven Tesla network without waiting for robo-taxis.
I see this differently. If Tesla FSD was in such a crisis mode, Elon Musk wouldn’t be casually chatting about the initial release being within the next few months on the third row podcast. He also said they were in the midst of a significant foundational rewrite of the code. What he described sounded like a stronger approach made possible by Tesla’s FSD team and custom high-power processing chips.
Elon also mentioned that the initial rollout of FSD would give cars “a chance” of navigating a city without intervention. Tesla is not claiming FSD will enable full autonomy in all situations - yet.
It seems to me that Tesla has been working to establish a strong foundation for Their FSD architecture. But undoubtedly there is a huge amount of work left to handle edge cases. It may be that the FSD rollout will be used to identify and classify the most important edge cases to tackle. Perhaps that’s something new AI recruits will work on.