Hacker News Reader

Comment by shmatt | Hacker News Reader

If Tesla's competitors were able to get government FSD permits without all that data. Maybe that data isn't really important to achieve the target

View on HN

if we're talking about overall effectiveness, then I think it is. Maybe passing a minimum threshold and having a safety effective of say 95% vs 99% or something. It's obviously going to be a very high percentage threshold that needs to be crossed to allow the public to use it.

So, right now it's like a trial experiment with FSD, but I definitely feel like a data advantage would play out into a higher safety rating before it's in GA. Maybe eventually others will catch up, that could be quicker than longer, but it's a valid potential advantage for Tesla.

just to nitpick, but with airline travel about five or six nines (99.9999%) safe, that is, one fatality every million miles of travel or so, 95% vs 99% is nowhere near good enough.

https://www.flightglobal.com/airlines/how-safe-were-commerci...

One million flights not one million miles of travel. There'd be a lot of dead frequent fliers if it was a death per million miles.

Yes it will be likely needing a 99.99%+ accuracy. I just made a vague suggestion, as I didn't have time to look it up. Elon mentioned it too, it doesn't just need to better than a human, but massively better. I don't know the exact accident stats for human drivers. But self-driving cars will need to be much higher.

Reply on news.ycombinator.com