[stilist]

Is it actually better than a bad human driver? (And what’s your metric for ‘bad’ — tired, drunk, distracted, a teenager who just got their license but didn’t actually practice enough?) Because I’ve watched enough videos of people having to quickly override Full Self-Driving in ordinary situations that I’m really skeptical that it’s better than most drivers. I’d be willing to say that in practice FSD usage is small enough to not be a serious threat to public safety, but I haven’t seen evidence that it’s better than a human.

[jjoonathan]

Every time one of these threads come around I search youtube for a FSD video and qualitatively evaluate what comes up.

If you google FSD fail compilation, you'll get a FSD fail compilation. You can do the same for humans.

[ClumsyPilot]

> If you google FSD fail compilation, you'll get a FSD fail compilation. You can do the same for humans.

Please continue this line of thought -

When I see a human fail video, where that person is an idiot, or drunk I conclude they shouldn't be driving. And usually their legal right to drive is taken away.

When I see FSD driving like a drunk idiot, I also conclude FSD shouldn't be driving. And their right to drive shouls also be taken away.

Lets treat them the same.

[stilist]

I’m sure that’s true, but the premise of giving the responsibility to a machine is that it will consistently outperform humans. There should be less evidence available for FSD critics.

[advael]

And if there's a video of FSD that isn't a "fail compilation", it was probably filmed and/or uploaded because it showed off the FSD doing well, and may even be subject to, effectively, the editorial review of google (whether via algorithms or moderators), which could easily depend on the size of google's advertising relationship with tesla on a given day (or how heavily they want to push the safety of self-driving cars, a business they are also in)

There's so much bias at every step of the process of looking up youtube videos that it lends essentially no credibility to any claim in any direction on this matter

[jjoonathan]

> if there's a video of FSD that isn't a "fail compilation"

If? FSD drives a million miles a day. There are a few.

[advael]

I think that was neither unclear from the sentence nor a particularly interesting detail to fixate on

[jdminhbg]

> google's advertising relationship with tesla

Tesla spends $0 on advertising with Google.

[runarberg]

I think your parent’s question is completely justified by having anecdotes. They aren’t stating a fact, they are raising doubts of a claim. One does not need scientific evidence to raise doubts about a claim which it self has limited—and sometimes no—evidence behind it. Repeated real world examples are sufficient for such doubts.

Now if FSD proponents want to stop people from having doubts, they need to run several experiments in very diverse settings (as diverse as real world driving). In the absence of sufficient evidence, a skeptic is completely justified.