If you have no interior camera, it's incredibly easy to get a strike. Just drive through about two miles of inactive construction zones with both hands on the wheel, and pay attention to the road like you're supposed to instead of noticing the flashing on your dashboard until it starts beeping at you, then do that two or three more times close enough together before leaving the construction zone. Boom. Instant strike.
No, not being sarcastic. The above sounds quite difficult to do. If you get the beeping, you just disengage and carry on driving, don't you? Seems like you then wouldn't get a strike. It would be evident right away that the car was being sensitized by the construction zone, so one would just abandon use of the driver assistance feature at that point. This doesn't seem instant at all. You have to repeatedly make the same error, which seems quite difficult!
Of course, it probably is a bit easier to get one without an interior camera, but it sounds like the flashing dashboard is super easy to see. Just have to be using peripheral vision.
Maybe I need to be more cognizant of & sympathetic to people who can't use peripheral vision, but for me it's just never been an issue seeing the flashing screen to my right, even if I miss the initial warning message. Obviously those unlucky enough to have retinal damage which results in peripheral vision loss may be worse off.
Anyway my point was that those who have a hard time avoiding strikes should definitely not have been in the early roll out group. There's no issue with excluding people from the initial rollout who have below average ability in this regard, regardless of whether we think it is "easy" or "hard." The fact is that there are many many people without strikes (ever!) who use FSD frequently. They'd be safer bets than those who have issues with providing full attention to the driving
& supervision task.
It's possible that Tesla gets better disengagement data from rolling to users who get strikes, though. They really believe in the system! This was a randomized rollout though it sounds like, so that was not their actual strategy.
100% preventable if they would have released it to the OG crew
Discussed
here and
here too. Kind of an asinine decision. Some of the people posting videos of v12...well...I guess it'll be released to everyone eventually, so maybe it doesn't really matter? It's really surprising to me there are not more accidents in general. I think it just goes to show how difficult it is to get in accidents if you're driving relatively "normally" and not drunk or far exceeding the speed of other vehicles.
Wonder if v12 is improved enough for a false sense of security and when it does something unpredictable the drivers are unprepared to react.
Yes, a very predictable consequence of NN approach, even to someone who knows nothing about how they work.
Looking forward to a new smoother v12 which will still be unusable with passengers in the vehicle, due to its propensity to do weird things unpredictably.
Hopefully no more accidents. Unfortunately the random rollout has substantially increased risk of an accident before the next release.
They should have followed the tried and tested same method of Influencers first and then Early Access. That would be a LOT safer and probably provided better feedback results.
Roll out to:
1) Users above certain number of miles (imperfect due to varying number of drivers, but ok)
2) No strikes
3) People who have submitted some bug reports, maybe.
This would result in a good random distribution geographically. And a fair amount fewer irresponsible driving demonstrations, perhaps.
Maybe it's just me, but I am far less likely to use V12 unless they keep a robust prediction visual on the screen. From a trust perspective I don't like the disconnect now that exists between what's in screen and what decision V12 makes
What? There’s literally no way to look at the screen while driving as a means of obtaining actionable info. It’s just not useful at all, except for figuring out what happened via in-cabin video recording afterwards.
It’s not like you drive along and make decisions about whether to take over based on what is on the screen! It would never work - display latency, plus human interpretive latency, plus mapping-visualization-to-actual-environment latency would be huge. Better to just keep eyeballs on what is happening outside the car. And in fact it is the only thing that works.
I am just hoping v12 slows down for red lights in a timely manner. v11: Literally failed to slow down for lead traffic at every red light on Mira Mesa this morning, when there is traffic things seem worse. About six interventions in seven miles. All reported to Tesla. Still had to use brakes sometimes, even after the intervention.
in the video had more ‘pay attention to the road’ alerts and had them more frequently than I had when I got the strike I mentioned above.
He had two pay attention warnings (did I miss one?). That is not going to yield a strike in this time period.
All of his wheel nags were satisfied before proceeding to great stridency.
You have to be more negligent than this (he was HIGHLY negligent: for some reason he did not have both his hands on the wheel at ~9 and ~3, which I just cannot wrap my head around, so he had many more torque nags than he should have, but they were all satisfied so they don’t count) to get a strike. (With current software - prior buggy versions, who knows - I never had problems but can’t exclude possibility of issues.)
- I think he was not paying as much attention as he should have, but was instead probably looking at James during the drive and not seeing the torque requests.
For both the nags he was caressing the screen to put in new waypoints or to check the route. Pretty irresponsible, but at least the speeds were low. If he has severe peripheral vision limitations as many on this forum apparently do, though, it could be quite dangerous.