EVNow
Well-Known Member
Even a baseline probably won’t mean much to us.
These are internal stats. Marginal improvements, what I expect at this point. We are definitely seeing a “local maximum” hit in performance and need 11 for a step improvement.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
FSD Beta accidents
- Thread starter Terminator857
- Start date
These are internal stats. Marginal improvements, what I expect at this point. We are definitely seeing a “local maximum” hit in performance and need 11 for a step improvement.
Elon pretty much suggested in the recent Lex Friedman interview that 11 will be a step backwards at first until things are trained up again, so I wonder if we'll get -xx% numbers for these items
That would be both hilarious, but also great from an honesty perspective.
But, probably won't happen. :-(
EVNow
Well-Known Member
That would precipitate a NHTSA recall.
The world can’t handle that much of truth
ps: Even when we had such regressions for internal tools in companies I have worked for, bosses would never let us put that in release notes.
I mean really
My work involves a lot of progress and cost reporting and transparency is key for any type of realistic measurement. In a case where you over-progress something and back it out, or if you change the baseline and that results in a negative % complete for a period measurement, you'd show that. Tesla doesn't need to disclose any of this stuff, but the precedent is there with reporting numbers in the first place...
Honesty and transparency are how you inspire confidence in your customer
I would love to see that as I do agree that it would come across as being honest, and transparent.
The reason for my skepticism is that I think the release notes are being used as marketing.
EVNow
Well-Known Member
Doesn’t look like marketing … more like back-patting by individual devs.
If you ask your devs to write one line on what they achieved in a sprint, this is what you get.
DanCar
Active Member
Los Angeles Times: Musk said no self-driving Tesla had ever crashed. Regulators had counted 8.
www.latimes.com
Musk said not one self-driving Tesla had ever crashed. By then, regulators already knew of 8
Tesla's CEO has claimed a perfect record for the safety of his company's robot-driving system. A public crash-reporting database contradicts him.
www.latimes.com
Funny, when I had FSD beta, I probably drove more miles combined than all of the folk in the list of “troubled customers”. Never had an accident. Guess I’m doing it wrong. (33,xxx miles/year, various situations everyday.). I’ll give credit where it’s due, the phantom braking could/is be an issue, never was for me. Luck I guess?
Daniel in SD
(supervised)
What was your accident rate before FSD Beta?
DanCar
Active Member
The accidents are for people who need an excuse for not paying attention.
AlanSubie4Life
Efficiency Obsessed Member
Los Angeles Times: Musk said no self-driving Tesla had ever crashed. Regulators had counted 8.
Musk said not one self-driving Tesla had ever crashed. By then, regulators already knew of 8
Tesla's CEO has claimed a perfect record for the safety of his company's robot-driving system. A public crash-reporting database contradicts him.
- Troy, Mo.: A Tesla was turning through a curve when “suddenly, about 40% of the way through the turn, the Model Y straightened the wheel and crossed over the center line into the direct path of the oncoming vehicle. When I attempted to pull the vehicle back into my lane, I lost control and skidded off into a ditch and through the woods, causing significant damage to the vehicle.”
This must be the one we had a video of here a while back? Matches the OP's description. Definitely seemed confirmed to be FSD Beta.
Some of the other descriptions also match FSD Beta, since it is steering on city streets.
This one is more surprising - Getting FSD Beta in 11 days after car purchase is crazy - perhaps Tesla should put some limitations on that!
- Jackson, Mo.: A Model 3 “jerked right toward a semi truck, then jerked left toward the posts in the median as it was accelerating and FSD would not turn off.… We owned this car for 11 days when our wreck happened.”
They claim the data covers from January 1st, 2021, which is pretty early for the wide release of FSD Beta. But I guess these must have been FSD Beta since the LA Times says so. (To be fair, it's not clear specifically which data they are referring to in the article when they give the timeframe, and whether the INCIDENTS spanned that timeframe.)
There's no question that accidents have occurred on FSD Beta and it's ridiculous that Ross Gerber and Elon would make these pronouncements since there were some pretty notable fairly well confirmed incidents by that point in time.
That being said, I'm not convinced all of these were with FSD Beta, and of course some appear to have occurred on the freeway, which would not have been FSD Beta even if they were running FSD Beta.
Given the amount of time/energy that one needed to put into just getting the Beta means the customers likely knew they were testing FSD Beta. Yet, none of the reports even mentioned it.
I only see a couple of them where they were even really plausible.
The reporters from the LA times don't seem to understand there is a difference between FSD and FSD Beta.
Tam
Well-Known Member
The article seems to make no difference between FSD and FSD beta.
First, it said: "Full Self-Driving “beta” program".
Then " incidents involving Full Self-Driving", but it failed to write down the words "FSD beta".
Again: "One driver reported FSD automatically “jerked right toward a semi truck” ". That is missing the specific term FSD beta.
Thus, when Elon Musk said FSD beta, the article might confuse it with Full Self-Driving in general.
It's an easy mistake because Tesla Public Relations closed down and failed to explain the difference to the press.
AlanSubie4Life
Efficiency Obsessed Member
Yes, which is annoying, because a couple of them were definitely on FSD Beta, and it detracts from the reporting.
Yes, a couple at least were very likely on FSD Beta. I'm assuming the one that I posed links to really was on FSD Beta, though I guess we only have the driver's (@tr6990 from Missouri) representation of that (seemed extremely likely from details given in the initial post). The ones in turn lanes must have been too, though.
Tam
Well-Known Member
Their data from 1/1/2021 does not make sense because it was only for a very few and they all were under NDA at that time. That meant, there were no free public postings of accidents on the website.
By 1/27/2021, there were about 1,000 people on it but mostly were Telsa employees with NDA.
By 9/17/2021, there were about 2,000 people on it but again mostly were Telsa employees with NDA.
That's why NHTSA wanted to open up an investigation on the NDA.
And that's why we got Safety Score instead of NDA.
And that's why in mid-October, 2021, only 100/100 passing scores were admitted to the FSD beta program without an NDA.
So, it would make sense that their data should start from mid-October 2021.
The same as it was while using the beta.
Yeah, this. I’ve had it disengage with keeping my hands on the wheel a little too tight when it was making a turn.
Daniel in SD
(supervised)
Note he says "as it was accelerating." Maybe he didn't turn off TACC.
Similar threads
