The real problem, which I think the article does a poor job of making clear: Tesla sold millions of cars before the current generation Hardware 4 vehicles with $10,000 full self driving packages which never really materialized convincingly ‘full’ self-driving capability. There’s fair arguments for the HW4 vehicles not having FSD either, maybe because it needs to be supervised or isn’t perfect or whatever. But the HW4…
Seems like HW3 has been pretty good since FSD v12.3+ came out. If we describe HW4 as 99% of the way to true FSD, then HW3 is probably 95% of the way. Though approaching 100% (maybe 2x the human standard) is going to be exponentially harder to get to.
If we describe current LLMs 99% of the way to AGI and full sentience then https://en.wikipedia.org/wiki/ELIZA was probably 90% of the way.