Oh look, yet ANOTHER FSD Beta (wheel cover curbing) Collision

[ShelJW]

correction. I implied earlier that the rate of persons opting to get FSD was 10%.

I was wrong, and I apologize.

It's 7%.

Tesla's Low FSD Take Rate Offers Growth Opportunity

Most Tesla owners don’t buy Full Self Driving. Increasing FSD orders could present a huge opportunity for Tesla.

www.torquenews.com

Yeah. HA! ONLY 7%!!!

So, Tesla sold 100,000 Model 3's and Model Y's in the 1st quarter of 2022.

100,000 x 7% = 7,000
7,000 x $12,000 = $84,000,000. That's 84 MILLION DOLLARS.

Ha! What a joke! Tesla ONLY made $84 million for selling SOFTWARE that, after development, doesn't really cost them anything.

Figuring their sales will continue at the same rate (it'll most definitely increase), Tesla will ONLY bring in $336,000,000 in 2022 for U.S. sales of FSD alone! A third of a billion.

Not too shabby...

In fairness... I am sure that Tesla spends a large sum on development, coding, lawsuits, etc. It isn't free money, and I happen to feel that they're doing a service if they can, indeed, develop full self driving which ends up being safer then us stupid humans, but hey... that's one persons opinion.

 

[EVNow]

I was wrong, and I apologize.

LOL. You have no idea how wrong you are.

The basis of all that is a self-selected order entry form on TMC (Teslike is a terrific guy, but the data he gets is from self-selected people).

BTW, a good way to figure out how representative the data there is - is by comparing ASP from that sheet to the one we get from Tesla. Good luck with your homework

 

[S4WRXTTCS]

correction. I implied earlier that the rate of persons opting to get FSD was 10%.

I was wrong, and I apologize.

It's 7%.

Tesla's Low FSD Take Rate Offers Growth Opportunity

Most Tesla owners don’t buy Full Self Driving. Increasing FSD orders could present a huge opportunity for Tesla.

www.torquenews.com

I'm not going to tell you you're wrong as I don't know what the current take rate is at the $12K price point.

I will say however that once Tesla launched the subscription pricing that it didn't matter anymore.

Having it as a subscription allows reoccurring revenue where they can adjust the price to best fit the market demand.

In fact I look forwards to the day when FSD is no longer sold with the vehicle.

 

[S4WRXTTCS]

If the car attempts to run the stop sign, then it's Tesla's fault. If it succeeds in running the stop sign, then it absolutely is the drivers fault.

If the car is really determined to run that stop sign it will time its acceleration just right to catch the human off guard.

 

[2101Guy]

If the car attempts to run the stop sign, then it's Tesla's fault. If it succeeds in running the stop sign, then it absolutely is the drivers fault.

“All human input is error”

 

[EVNow]

“All human input is error”

Not while testing.

 

[mswlogo]

Maybe it was due to the stupid yoke he has installed! Can we blame that?

I don’t understand people who let the car steer itself when using FSD! The driver should always be steering! It is so weird what people do…

Literally...you're driving...so why wouldn't you be steering? Bizarre.

I perfectly understand how it could happen. That’s why I stopped using FSD Beta. It can screw up faster than you can react.

I was driving along no cars, well marked road, at night, going maybe 30 mph and suddenly it just wanted to slam into a guard rail that was maybe 5 ft away. It was way before an up coming turn. I still don’t know how I caught it. Or what confused it. My confidence in FSD Beta never went lower than that.

Clipping a corner is tricky because you might be assuming the car took it wide enough since it did fine the last 100 corners but you find it didn’t when it’s too late. And sometimes when it does it’s harmless.

OP is lucky that’s all that happened.

 

[AlanSubie4Life]

It can screw up faster than you can react.

I definitely agree that it can screw up, dangerously fast, but with both hands on the wheel, always torquing away from the worst case scenario, it’s very hard to imagine it being too fast to react to.

But maybe it will surprise me one of these days! I’m used to it trying to veer off the road and towards oncoming traffic at this point, though, so not sure such an event would even cause a ripple in my FSD Beta zen state.

I still don’t know how I caught it. Or what confused it. My confidence in FSD Beta never went lower than that.

Basically have to run a GoPro on infinite loop to have a chance to figure this out, I think. Sounds pretty alarming though!

My biggest regret has been not running FSD (and GoPro) when I’ve had to make a collision-avoiding stop for someone blatantly running a red light as I was making a protected left. Really wanted to see if it would catch it! It’s on my turn into work, though, and people routinely (and mysteriously!) don’t see that red light (it’s now happened twice, just to me, and once to someone I know), so a pretty good chance it will happen again. Bonus is that I will be ready.

Could probably get some good clicks for a video showing FSD avoiding an accident! Not sure it can see the traffic behind other vehicles that I am tracking, though. Pretty sure it can’t, so a best-case scenario would be a sudden slamming of the brakes.

 

[mswlogo]

I definitely agree that it can screw up, dangerously fast, but with both hands on the wheel, always torquing away from the worst case scenario, it’s very hard to imagine it being too fast to react to.

And what happens when a bad case is on both sides? Or it suddenly veers towards where your are torquing. Now you are in even worse shape.

With on coming traffic on my left and a guard rail on my right. I’m supposed to torque left and right with every passing car?

I was ready at any intersection or bump. Not at straight, simple and narrow. There is little purpose to run it if takes 5x the work of normal driving. I don’t think of driving as a Video game. Tesla has plenty of feedback to get it right.

Which is why I stopped running it. If that’s your thing go for it.

I’m not surprise OP got fooled. He hit off cheap. I’ve seen it cut corners unexpectedly close but I was lucky.

It is an accident waiting to happen. OP got off cheap.

 

[AlanSubie4Life]

And what happens when a bad case is on both sides?

Death grip, with or without torque. Disengagements are fine!

There is little purpose to run it if takes 5x the work of normal driving.

Agreed. It’s certainly not supposed to make driving easier, and it does not!

Which is why I stopped running it. If that’s your thing go for it.

Yeah, not a hardcore user but use it fairly frequently when I feel in the mood and can put up with its nonsense.

It is an accident waiting to happen.

Certainly (which is why they have happened on a few occasions now). I believe running it increases my risk of an accident slightly, but it’s the cost of entertainment/curiosity.

 

[mswlogo]

Certainly (which is why they have happened on a few occasions now). I believe running it increases my risk of an accident slightly, but it’s the cost of entertainment/curiosity.

That’s what I was doing. Trying it in each release out of curiosity. It would faithfully let me down within 5 minutes every time and then I’d wait for the next release.

 

[Twiglett]

That’s what I was doing. Trying it in each release out of curiosity. It would faithfully let me down within 5 minutes every time and then I’d wait for the next release.

I run it a bit longer than that for sure.
I still use my torture test route, which is less than five miles but is measured in disengagements per mile.
Its getting better, but still somewhere in the 8-10 in that five miles so nothing like I'd let someone else use.
But on my other, much longer route of 50 miles each way, it fares much better. There are 30 minute long sections of twisty turny where it performs flawlessly, including one tight 270 degree loop from one road to the next which it handles really well.
Most of the time I know where I think it's going to fail and just disengage it early and manually drive - talk about first world problems.
The most common disengagement I have on this long route is just the nav choosing a route that maybe shorter, but its all traffic lights and busy traffic. My preferred route is longer and with almost no traffic and significantly more chill.
For me FSD was never about letting it drive the whole thing, just be better on more of the journey and slowly but surely it is doing just that.

 

[KLam]

I am beginning to realize that all those Youtube clips of zero disengagement are false indications of the progress of FSD beta. There you see all the testers with their hands off the steering wheel. That's no good, as we are required to put our hands on the steering wheel to sense what the car is trying to do and to intervene when we sense the car is going to make an error. All of us beta testers knew that on turns the steering wheel oscillate and jerk wildly and we tend to take our hands off the steering wheel to stop from disengaging. But that's dangerous as we would not know where the steering is when the warning to take over happens.

So the real progress of FSD beta should be judged by how accurately the Steering wheel reflect the actual direction of the car being steered so that we can put our hands on the Steering wheel to give the necessary torque to not disengage,

 

[Phlier]

It can screw up faster than you can react.

Man, I wish more people understood that.

I've had it screw up faster than I can react enough times now to have me well and truly gun shy. I just can't stand to see videos of people using FSD beta without their hand ghost steering the entire time.

Even though I ghost steer non-stop while using FSD, there are still times when it gets scary.

IMO, if you're ghost steering to the point of maximum safety, you should literally *never* see a nag to move the steering wheel.

Yeah, I know this is coming across as virtue signalling and holier-than-tho, and I'm sorry for that. But I do think it's so important that it's worth coming across as being a bit of a tool and/or overly safety conscious ninny.

The point of all this: DRIVE THE DAMN CAR. AND DRIVE THE DAMN CAR EVEN MORE WHEN USING FSD.

 

[father_of_6]

Ditto. It's 4,000 lbs of metal moving tens of miles per hour. It's a real people killer if you let it be. Don't be that Uber driver.

 

[DrChaos]

Not my domain but how would you set up the optimization? What would be the output variable and what would be the loss function?

All this conflict about defining "collision" reminds me of Bill Clinton trying to get a definition of "xxxual relations."

Regardless of your definition of what happened in the video, the fact remains that FSD shouldn't curb your wheels. Yet at the same time, the driver of said vehicle shouldn't have allowed FSD to curb his wheels, either.

This ain't release level software. It's beta. And if you don't treat it with the respect it deserves, curbed wheels will be the very least of your problems.

(btw Clinton asked if sexual relations included oral sex, and the prosecutor said 'no'. After which Clinton said his famous line, which was reasonable and truthful.)

Yes, this looks like a FSD fault, causing a minor accident on a fairly easy and avoidable situation. What's interesting to me is why did this happen?

I noticed something: if you look at the right side camera looking down at the curb, there is significant light bleed from the right side turn signal. At the exact moment of one of the turn signal turning on, the camera images washes out in red and the alert triggers. I think newer light & camera assemblies don't have this problem.

To me this means that the FSD system lost track of the right curb location, recognized it lost it, and couldn't find it again in time. This looks like a fault of inadequate sensor hardware.

I also think that the front cameras should be stereoscopic which would better distinguish the raised curb vs a texture on the road, which the single image only system now has to do.