$12K for FSD is insane

[AndreP]

I don't think that would explain why the system has gotten safer over time, unless you assume new owners have magically gotten MUCH better at figuring this out.

A more likely explanation is the system improved, since its accident rate declined even compared to its own #s from earlier times.


How so? You're saying people who are inherently bad drivers use it less? Or use it more? Or what?

it's also not accurate.

Hilariously we JUST in the posts above yours saw yet another example of it.

People very commonly use it outside of highways despite the manual telling you not to.

Then complain anytime it does something wrong there to double down on the hilarity.

So the idea "AP miles are all highway miles" simply is not so

This would be very fair criticism especially driver age if you're only comparing "All tesla owners" to "all drivers".... but the stats also compare the SAME Tesla drivers with and without AP...and STILL find AP much safer.

FWIW I agree Tesla could offer a lot more useful breakdown of the stats.

And I wouldn't be at all surprised to find there's a ton of interesting caveats in the degree to which AP improves safety.

I'd be VERY surprised if it showed AP was ever less safe over any significant # of miles compared to non-AP in the same situation.

The system becoming "safer" over time could suggest that people have, over time, been able to determine where Autopilot functions well and adjust their usage accordingly thus driving down the accident rate.

It would be interesting to put the AP accident rate trend against uptake of new people purchasing Teslas and dipping their toes into AP for the first time.

The regulators etc don't just look at single numbers, accident stats are broken down with all types of criteria. And there surely are areas where AP excels. Like if we could build in a system that detects intoxicated drivers and only permits driving by an autonomous system, that would be an obvious win.

I'm sure there are great use cases, and I don't think the published stat tells us anything of real value.

 

[COS Blue]

Cruise control from 1989 would just slam into slower traffic in front of you.

No, because I would have hit the brakes. But cruise control from 1989 definitely would have hit that bridge shadow full-on at 80mph. Which is kinda what I would want.

I will say one thing, you definitely picked the right avatar for your personality .

 

[2101Guy]

its hilarious to me, that the CEO CONTINUES to state "any human intervention is error".

Yet, there are CURRENT youtube videos where FSD crosses over a DOUBLE YELLOW LINE and into the oncoming traffic lane. And it took (repeatedly) human intervention, to get the tesla BACK into the proper lane.

If it werent for human intervention, I can only imagine the number of FSD accidents that would have occurred

 

[Knightshade]

No, because I would have hit the brakes.

Well...one hopes you would have.

And yet the data shows us sometimes people don't brake manually when they should- and cars with FCW and/or automatic emergency braking are measurably safer than those without it

Automatic Emergency Braking (AEB) System Impact on Fatality and Injury Reduction in China

The automatic emergency braking (AEB) system is an effective intelligent vehicle active safety system for avoiding certain types of collisions. This study develops a national-level safety impact evaluation model for this intelligent vehicle function, ...

www.ncbi.nlm.nih.gov

AEB reduces rear-end collision rates by 43% [3]. AEB has been shown to reduce front impact crash rates by 27% [22]. The effectiveness at preventing pedestrian fatalities and injuries due to impact by the front of a car was 44% and 33%, respectively

If you prefer a US data study those show similar numbers.

Effectiveness of forward collision warning and autonomous emergency braking systems in reducing front-to-rear crash rates

'Effectiveness of forward collision warning and autonomous emergency braking systems in reducing front-to-rear crash rates' by Cicchino, Jessica B.

www.iihs.org

FCW alone, low-speed AEB, and FCW with AEB reduced rear-end striking crash involvement rates by 27%, 43%, and 50%, respectively. Rates of rear-end striking crash involvements with injuries were reduced by 20%, 45%, and 56%, respectively

I'll take "reduces accidents 30-50% but once in a while annoys a passenger with a false positive" over "keep driving a 1989 deathtrap"

 

[Knightshade]

its hilarious to me, that the CEO CONTINUES to state "any human intervention is error".

Can you cite where he said that?

Only recent such comment I'm aware of is this:

https://twitter.com/x/status/1481724767363338246

Which is about input, not intervention.... and about the infotainment/climate system, not driving.

 

[ucmndd]

its hilarious to me, that the CEO CONTINUES to state "any human intervention is error".

Yet, there are CURRENT youtube videos where FSD crosses over a DOUBLE YELLOW LINE and into the oncoming traffic lane. And it took (repeatedly) human intervention, to get the tesla BACK into the proper lane.

If it werent for human intervention, I can only imagine the number of FSD accidents that would have occurred

To be fair, I believe the “almost all input is error” is shorthand for what he actually means to imply - “almost all (human) input is in response to (system) error”.

So while I think that general premise is stupid, particularly in regards to the sorts of human preference inputs they needlessly buried in the v11 UI, I don’t think he really means what you’re implying in the case of the autopilot override example.

 

[2101Guy]

Any Required Human User Input is an Error for the Tesla Self-Driving Autopilot | NextBigFuture.com

Elon Musk views any human user intervention is an error situation for the Tesla Autopilot. Elon means that whenever a human has to take control from the Tesla

www.nextbigfuture.com

 

[2101Guy]

So the question is, if Human Input = AP/FSD error. And based on the trend of INCREASING need for human input (more phantom braking), whats the new date for the promised Robotaxi service? And why does Elon refuse to provide any sort of update for Robotaxi?

 

[Daniel in SD]

So the question is, if Human Input = AP/FSD error. And based on the trend of INCREASING need for human input (more phantom braking), whats the new date for the promised Robotaxi service? And why does Elon refuse to provide any sort of update for Robotaxi?

Haha. He provides updates all the time. He just did an interview with Lex Fridman and said robotaxi reliability is coming this year.
I should note that if you look at California autonomous vehicle collision reports robotaxis seem to get rear ended all the time so I wouldn't count on it to fix phantom braking.

 

[2101Guy]

Haha. He provides updates all the time. He just did an interview with Lex Fridman and said robotaxi reliability is coming this year.
I should note that if you look at California autonomous vehicle collision reports robotaxis seem to get rear ended all the time so I wouldn't count on it to fix phantom braking.

I mean, Im supposed to have been making $100k a year via Robotaxi service since December 2019 I think. Im missing out on renting out my Tesla and letting it drive itself to pick up people and take them to their destinations!

 

[Knightshade]

So the question is, if Human Input = AP/FSD error.

Yeah- that's...kinda different from your original claim.

He's stated- correctly- that if a human needs to intervene in a self-driving system that's an error by the self driving system

Which is...pretty obvious, and not especially funny.

And based on the trend of INCREASING need for human input (more phantom braking)

Can you provide some data demonstrating phantom braking is increasing overall.... or are you just generalizing from your own 1-car personal experience?

Because in my own experience I virtually never get such braking on normal public-release AP on highways. Like couple times a year maybe.

I DO get it on local roads using FSDBeta... but even there it's far better in the current version than it was in earlier ones....slowdowns now tend to be brief 2-3mph regen affairs instead of 10+ actual braking ones of earlier beta releases.

 

[2101Guy]

No, you are correct. FSD has gotten so much better and phantom braking is not a thing anymore. No one is complaining about it. And all of the numerous industry reports on how poorly FSD performs are all fake news. 100%.

 

[Knightshade]

No, you are correct. FSD has gotten so much better

I mean, it objectively has if you're referring to the beta folks have been testing since late 2020 now.

Just comparing videos from then to now shows his pretty clearly- as does talking to those of us actually doing the testing.

Still has a ways to go of course before wide release.

and phantom braking is not a thing anymore. No one is complaining about it.

I see you're doubling down on misquoting people and misrepresenting what they actually said.

Bold strategy, let's see if it pays off!

 

[JustJohn]

On the contrary- Teslas system continues to offer excellent performance compared to all others.

I'm sorry you're triggered by my posting a bunch of sources showing your clams "nobody" else has phantom braking are factually untrue- and in fact not only do all the other brands have it, most have it far worse.

A good solution would be stop posting untrue things.



Anyway- AP is literally the thing that sold me the car, since at purchase time 95% of my driving was highway. I tested other L2 highway systems from Lexus/Toyota, Nissan/Infiniti, BMW, Hyundai/Genesis, Honda/Acura, etc... all of them were terrible in comparison.

I've got over 30k miles on it now (and it'd be at least 2-3x that if not for Covid moving most work to home), still 90%+ on highway using AP and it remains awesome.

I continue to maintain that it's an embarrassment to use with other people in the car. Thank you though for all of the unsolicited homework assignments you've turned in. Maybe someone will read them.

Autopilot is not useable.

 

[Bouba]

I mean, Im supposed to have been making $100k a year via Robotaxi service since December 2019 I think. Im missing out on renting out my Tesla and letting it drive itself to pick up people and take them to their destinations!

Why wait for robotaxis? If a taxi driver can really earn 100k per year then we should start picking up fares right now

 

[ucmndd]

Why wait for robotaxis? If a taxi driver can really earn 100k per year then we should start picking up fares right now

Robotaxi will obviously be solved by using TeslaBots to drive the cars. It's the future!

 

[Knightshade]

Autopilot is not useable.

Neither are your dishonest arguments

billions of miles get driven on AP.

Who's using it exactly if it's "unusable"?

MEANWHILE, BACK IN REALITY....

Consumer reports rated 18 different L2 driver assist systems.

Including from Cadillac, Ford, Mercedes, Audi, Subaru, BMW, Porsche, Volvo, Lexus, Nissan, Mazda, and more.... They even included 3rd party system Comma AI.

Guess who ranked higher than anyone else for capability and performance?

Tesla. The only brand to score a 9 out of 10.

(a couple others managed an 8... Subaru and BMW a 7... most of the rest score in the 3-5 out of 10 range).

Where CR ended up dinging Tesla was essentially 2 things:


Tesla lets you activate the system in places you shouldn't (exactly as discussed earlier in the thread regarding a road where you live!) and Tesla wasn't monitoring the driver to insure he was paying attention (ie not detecting bad drivers rather than an issue with the system itself....and something Tesla has since fixed with an OTA update to use the interior camera for this).

But in actual use it's the best thing on the market.

Why wait for robotaxis? If a taxi driver can really earn 100k per year then we should start picking up fares right now

If the taxi driver can operate 24/7 (minus refueling the vehicle time) and not take any salary at all 100k/yr should be pretty easy right now.

Otherwise you're gonna need to wait for robotaxis.

 

[2101Guy]

Why wait for robotaxis? If a taxi driver can really earn 100k per year then we should start picking up fares right now

My bad. It's the actual value of FSD that will EXCEED $100k. As soon as regulatory approval happens. Because we all know, THATS the only hold up with allowing Teslas to drive themselves on the road with zero human input/zero human intervention. The federal regulators.



Elon Musk
@elonmusk

The FSD price will continue to rise as the software gets closer to full self-driving capability with regulatory approval. It that point, the value of FSD is probably somewhere in excess of $100,000.

 

[2101Guy]

Whenever Elon says "probably", thats the equivalent of a certain former president when he would say "LOTS of people are telling me..." (which translated to..NO ONE had told him anything. )

 

[Knightshade]

My bad. It's the actual value of FSD that will EXCEED $100k.

That seems pretty accurate to me.

How much does the salary of a taxi driver cost?

Now multiple that by 3 since it can do 3 shifts a day.

Now multiple that by the # of years you can operate the software.

That's almost certainly more than 100k by quite a bit.

As soon as regulatory approval happens. Because we all know, THATS the only hold up

Weird you claim this, then immediately contradict yourself below...

The FSD price will continue to rise as the software gets closer to full self-driving capability with regulatory approval. It that point, the value of FSD is probably somewhere in excess of $100,000.

See the bit in bold?

They'll need THAT and then need regulatory approval in all the places it'll be required.

Whenever Elon says "probably",

We covered this a while back... Elon has already told us, years ago, not to believe him when he gives a target date for a thing he has never done before (and in this case NOBODY has ever done before).

People should not ascribe to malice that which can easily be explained by stupidity." (LAUGHTER) So-- so it's, like, just because I'm, like, dumb at-- at predicting dates does not mean I am untruthful. I don't know-- I-- we've-- I never made a mass-produced car. How am I supposed to know with precision when it's gonna get done?

Sub "never made a mass-produced car" with "never made a self driving car" or "never made a reusable rocket" or "never landed on mars" and the warning remains the same.

If it's something that hasn't been done then nobody knows how long it'll take-and thus you probably can't rely on someone telling you how long it'll take before it's done.