diplomat33
Average guy who loves autonomous vehicles
Actually you said “the Same FSD Beta software we have”. Last I checked nobody here has that software.
Correct. We have V11. Maybe I misspoke. I was trying to say that there is one system called FSD Beta. There are two versions of that system, V11 and V12. Tesla is testing both V11 and V12. Both V11 and V12 are supervised FSD but Tesla is hoping to remove driver supervision in the future. I was really responding to @willow_hiller 's question about testing L4. I disagree with the implication that Tesla is testing a secret version of FSD that is L4 or unsupervised FSD. I believe V12 is currently L2 and supervised FSD. I think when Elon talks about unsupervised FSD, he is not talking about some secret L4 version but rather, he is simply talking about the safety stats of V11 and V12 if driver supervision were removed. Tesla is able to look at metrics of how V11 or V12 would perform without supervision to see how close they might be to removing driver supervision. So when Elon says that "unsupervised FSD is trending well", I think he just means that the trend of the safety stats is looking good for supervision to be removed at some point in the future. For example, maybe he is seeing fewer safety critical interventions with V12 compared to V11 or maybe the simulation is showing that if left unsupervised, V12 would have fewer accidents than V11?
I just want to know more about v11 safety. It would be hugely important if it is safer as currently implemented and utilized by the public.
But it is so frustrating that Tesla has never provided anything to help show the world how much safer the roads are as a result of v11 use.
Just such a difficult analysis to do - this may be why it has not happened.
Yes, I would love to know more.
In theory, supervised V11 could very well be safer than humans based on how Tesla measures safety:
1) Tesla only measures crashes with airbag deployment or active restraint deployed. So Tesla is not counting many safety issues like minor accidents or near misses.
2) V11 can handle routine driving and lane keeping well. So V11 is unlikely to have one vehicle accidents that humans have when they are distracted or impaired. And any serious accidents that V11 could cause are hopefully prevented by the driver if they are attentive enough. Assuming most FSD beta users are attentive, they should be able to prevent most serious crashes. So we should expect the number of serious crashes to decrease.
We should also consider that FSD beta users may not be using it in more risky situations. This could be skewering Tesla's stats, making it look safer than it really is. We don't know if V11 is objectively safer than humans. But the keyword is "supervised". If the human driver is attentive, supervised V11 could very well be safer as Tesla defines it. Unsupervised V11 is a whole other matter.
Certainly, if you only measure serious crashes and FSD can avoid many human accidents and the attentive human driver can prevent many FSD accidents then, in theory, we would expect the combination of FSD and an attentive driver to be safer than a human alone. In fact, this is the very argument Mobileye uses for why a well designed "eyes on" driver assist system can be safer than a human driver.