Hilarious that he gives us a deadline of "10-12 revisions" and now we start seeing revisions like 8.1
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Hilarious that he gives us a deadline of "10-12 revisions" and now we start seeing revisions like 8.1
Dan Detweiler
Active Member
Not to be rude, but who cares how the revisions are numbered? I'm sure the change makes sense to the company.
Dan
boonedocks
MS LR Blk/Blk 19”
It does for sure. @elonmusk on twitter said FSD would start rolling out general release in ~2 release points and with his new numbering it is now back by 2 so...the OP does have a point....not to be rude.
Dan Detweiler
Active Member
Sorry, but no he doesn't. It's a renaming...that's it. If you want to know how many revisions there have been then just go back and count them on the countless videos that have been released on YouTube.
Dan
diplomat33
Average guy who loves autonomous vehicles
No, because we did not actually go back 2 versions. We are still on the same version. Tesla is just calling it by a different name. So if Elon's prediction still holds, we should still see wide release in 2 more releases.
Soooooo maybe late 2024...ish...give or take a year...?
No, he said it'll start going to the Early Access Program then. We're many months away from "general release"
But he never gave a deadline. He said: "A few more revisions needed. Probably Beta 10 or 11."
There is no commitment or deadline there. Just a prediction about when it might expand to more of the early access participants.
No, he did not.
@diplomat33 gave a goal of about 1 intervention per 150k miles as a target rate for fsd "safety related" interventions, based on accident data. Regardless of what kind of interventions we choose to compare to this number, I think that it is overly strict. Every dangerous maneuver we make doesn't lead to an accident, in large part because other drivers will usually react to prevent the accident. It usually takes errors from both parties involved . So if we assume the combined probability can be equally attributed to both drivers, the probability of 0.000007 accidents per mile is the product of the individual probabilities of 0.00265 when they both make a mistake at the same time. That's an error rate of 1 in every 378 miles per individual. So, shouldn't fsd be held to this rate or better?
diplomat33
Average guy who loves autonomous vehicles
I don't think this works. We cannot assume the probability is equally distributed to both drivers. We don't know the percentage of errors that are fatal. We don't know the other driver will prevent the accident. Also, "safety related" interventions mean interventions that prevented a real accident. So it already factors in that the other driver did not prevent the accident. Plus, there are accidents caused by other drivers that you want your AV to avoid. It's called defensive driving.
I think your rate of 1 error over 378 miles is way too low.
Also, you don't want your AV to do any dangerous maneuvers, period. It is not ok for the AV to do a dangerous maneuver just because other drivers might avoid an accident. You want your AV to drive as safely as possible all the time.
I admit interventions might be a poor metric. There are too many unknowns.
So let's look at Mobileye's CES 2020 presentation that gives us some hard numbers:
The probability of human injury is 10^-4 injuries per hour of driving or 10,000 hours of driving per injury.
The probability of human fatality is 10^-6 fatalities per hour of driving or 1,000,000 hours of driving per fatality
So to match humans, FSD needs to do 10,000 hours of driving per safety error. Factoring in some margins, Mobileye multiplies it by 10 and concludes that you need 10M hours of autonomous driving without a safety critical error to be "good enough".
Source: Mobileye CES 2020
Whew, glad we got that worked out.
This is how it works. On any given drive we don't know whether we're the one to change lanes into a car in our blind spot, or the one who fails to slow down or swerve because we're distracted.
By definition, on average, each of us is making boneheaded mistakes that would have caused an accident if another car was there at a rate of 1 every 378 miles. That's why two cars come together every 150K miles.
diplomat33
Average guy who loves autonomous vehicles
Anecdotally, most of the ones I see involve more than one car. But, Ok, 1 in every 500 miles.
diplomat33
Average guy who loves autonomous vehicles
Sorry but I still don't see how 1 error per 500 miles can equal 1 accident per 150,000 miles. 1 error per 500 miles means 300 errors in 150,000 miles. That would mean only 1 safety error per 300 errors. I think humans make safety errors more often than that.
Put differently, you really think that an autonomous car that makes 300 errors in 150,000 miles would only get into 1 accident? I doubt it.
Doubt away, but that's how probabilities work. If you have two independent probabilities, the probability of both of them happening is the product of their probabilities. Just look around you as you drive, people are making mistakes all the time without causing accidents. Fortunately, every time somebody turns from the wrong lane there isn't someone else there. But if there had been, an accident could have occurred. Running stop signs and traffic lights doesn't always lead to an accident, but happens frequently.
diplomat33
Average guy who loves autonomous vehicles
You cannot just multiply both probabilities. The probability of an accident is not the probability of both drivers making a mistake at the same time. You can have accidents that don't involve both drivers making a mistake at the same time.
Ha. Like Karpathy said, one would think it would be easy to train a neural network to identify stop signs…
(Both stop signs aren't detected/visualized.)
(Both stop signs aren't detected/visualized.)
diplomat33
Average guy who loves autonomous vehicles
The car does still stop for the stop sign though. I'm guessing it used the map data to know the stop sign was there.
When you multiply them you are accounting for the fact that some of the accidents will be caused by the other guy. The product just means the cars will come together at this rate regardless of who was at fault. It is fundamental that the probability of two independent probabilities is their product.
diplomat33
Average guy who loves autonomous vehicles
They are not independent probabilities. The probability of driver A making a safety error is not independent to the probability of driver B making a safety error since drivers can see and react to what other drivers are doing. So, the probability of driver A making a safety error multiplied by the probability of driver B making a safety error is not the probability of an accident happening.
Similar threads
