- Another 60m miles will be suggestive, but you need at least 2x-3x the ordinary fatality mile count to even have a 90% confidence that the results weren't a fluke.
- Tesla's FSD might operate in less-than-ideal conditions more often than competitors, but I'd posit that the typical human driver operates in much less ideal conditions than FSD. Humans are explicitly instructed to not enable it in sketchy scenarios and to be fully attentive when it's in control, and performance in that environment is at best an upper bound on what you might expect if deployed more widely.
- I skimmed the stats from https://www.tesladeaths.com and was hoping Cunningham's law would kick in if that was too far off base.
- Tesla's FSD might operate in less-than-ideal conditions more often than competitors, but I'd posit that the typical human driver operates in much less ideal conditions than FSD. Humans are explicitly instructed to not enable it in sketchy scenarios and to be fully attentive when it's in control, and performance in that environment is at best an upper bound on what you might expect if deployed more widely.
- I skimmed the stats from https://www.tesladeaths.com and was hoping Cunningham's law would kick in if that was too far off base.