That's because the system isn't trusted, yet.
By the manufacturer.
Great.
Gotcha. Another point of view would be that Tesla doesn't care if you watch the road that well. They hardly have a DMS worth its name. The camera placement and lens sucks for DMS and it lacks IR. Goes hand in hand with the rest of their tone of voice and markering
Mine has IR, but the monitoring has gotten a lot better and more strict in some areas while loosening in others. In City Streets, especially with people and heavy traffic, it knows exactly where I'm looking and for how long and lets me know. In general the highway is just less complex, so there's less scrutiny.Great.
Gotcha. Another point of view would be that Tesla doesn't care if you watch the road that well. They hardly have a DMS worth its name. The camera placement and lens sucks for DMS and it lacks IR. Goes hand in hand with the rest of their tone of voice and markering
I think we mostly agree though.
I think it will work such that liability will slowly transfer from owner to insurance, and insurance will adjust according to whether the owner is driving manually or by using the autonomy features. If you're letting autonomy drive and there's an incident, then there's no strike against you on your policy. If you're driving, then the incident will affect your policy. I assume that this is one reason that Tesla got into the insurance game. "Legacy insurance" just won't quickly embrace this kind of thinking.
I'd be in agreement with that if Tesla had some skin in the game. As it is it's a pig in a poke with no criteria to say FSDj is meeting min performance goals. So far it's crowd sourced vaporware. Caveat emptor to those thinking about buying it outwrite.
FSD + Human has better accident rate than just human. Looks like a good perf goal that is being met.
That's a nice thought, but not a fact. Right now FSDb is too "interesting" (ie bad). If it improved 10x it would still be 100x less safe than human on its own, and the human monitoring would be over-trusting it. Dozing off while on L2 is not "safer", it's automation complacency.
That link didn't work for me, but here's YouTube's copy of the presentation. Impressive stuff. It's a bit of a surprise that Waymo isn't farther along.
Does Tesla's eye tracking even work at night or through sunglasses? I recall seeing a video that made it seem like it didn't.
Answer for sunglasses is definitely no. I can look at the screen all I want with my polarized sunglasses on and AutoPilot doesn’t say boo. But pull my sunglasses off for a moment to look at some details (while on AutoPilot of course) and I almost immediately get the “Pay Attention” nag.
The one delusional here is obviously you.
Now, this is a two year old video, so Cruise has improved quite a bit since this was shot, but they still get stuck ALOT. But getting stuck is more impressive than needing a driver to intervene no prevent an incident, and requires a whole different safety architecture tbh. Tesla and other L2:s requires a human to stop it from doing the wrong thing. L4:s sometimes require a human to help it navigate situations the system knows it cannot safely operate in.
