I don't understand how Musk thinks that 10,000 miles per intervention is something worth bragging about on an earnings call. That kind of rate would make scaling beyond a trivial number of cars totally unfeasible, and represent a significant public safety hazard. Was he trying to imply that they can improve that figure by orders of magnitude in a couple of months? I can't see how that would be possible. If such quick and easy improvements were attainable they'd have already been implemented long ago.
Good post Phil, what's tricky about Tesla and the AV industry in general is that there aren't binary measures of safety. It's all a bit subjective and anecdotal. I have people telling me that Waymo is not safe yet I have taken dozens of rides, including an 80 minute one the other day and feel completely safe. Obviously it's not a big sample size but I don't personally think Tesla is at Waymo's level of safety.
So come June, I won't be sitting in the back of a Tesla robotaxi but I know some/maybe many people will :)
This sounds like the ground version of air traffic control, which is of course is fraught with safety issues, especially on taxiways. If I see a Tesla robotaxi approaching, I’m going to pull over. (BTW, this post was informative as well as “entertain[ing]”!)
Phil, thank you for spelling out about a "remote-supervised" FSD scenario. Even if the practice of using a remote operator is not new, this opens a whole new can of worms!
I don't understand how Musk thinks that 10,000 miles per intervention is something worth bragging about on an earnings call. That kind of rate would make scaling beyond a trivial number of cars totally unfeasible, and represent a significant public safety hazard. Was he trying to imply that they can improve that figure by orders of magnitude in a couple of months? I can't see how that would be possible. If such quick and easy improvements were attainable they'd have already been implemented long ago.
Great post Phil, appreciate your work and articles about the AV industry!
Good post Phil, what's tricky about Tesla and the AV industry in general is that there aren't binary measures of safety. It's all a bit subjective and anecdotal. I have people telling me that Waymo is not safe yet I have taken dozens of rides, including an 80 minute one the other day and feel completely safe. Obviously it's not a big sample size but I don't personally think Tesla is at Waymo's level of safety.
So come June, I won't be sitting in the back of a Tesla robotaxi but I know some/maybe many people will :)
Thanks for the thoughts Harry.
The tricky part about safety is that it lives out at one-in-a-million, while everyday experience is one-in-a-dozen/hundred.
Yep totally fair. Which is why if I don't feel safe anecdotally in a Tesla, that's really bad :)
This sounds like the ground version of air traffic control, which is of course is fraught with safety issues, especially on taxiways. If I see a Tesla robotaxi approaching, I’m going to pull over. (BTW, this post was informative as well as “entertain[ing]”!)
Phil, thank you for spelling out about a "remote-supervised" FSD scenario. Even if the practice of using a remote operator is not new, this opens a whole new can of worms!