FSD Beta Videos (and questions for FSD Beta drivers)

[rxlawdude]

I think there could be some very useful discussions along those lines (may be not in this thread though?).

Difficult trying to take in international view points, but seeing a discussion around US perspective might have broader relevance.

What are you buying when you pay for FSD?

Is it reasonable to sell a product that could effectively remain in very long term beta?

Has Tesla changed its FSD offering over the years, are any terms / undertakings enforceable especially if Beta?

Can you construct a (meaningful / enforceable) contract of sale around a Beta product in public consumer markets?

I'm sure lawyers will find far better questions, but sooner or later I think the answers will be of interest to more people who actually bought FSD.

Your first question is the salient one. What are you buying?

This isn't the thread to discuss the debated answer to that, but chime in on the ones better suited for this topic.

 

[Battpower]

The real danger is once FSD beta gets to be really good its going to be hard for the human to oversee it. So we'll start to have accidents when people stop paying attention.

Complacency is almost unavoidable at some point (assuming development keeps progressing).

It seems to me like a very non-linear performance curve, where you need to achieve consistent safe operation given changing use expectations.

 

[rxlawdude]

I agree as there will most likely not be any specific answer in this case scenario since Tesla constantly change their contracts and marketing info. Also everyone's purchase situation are not the same.

It does bring me back to the days of sitting at the LA convention center for the bar though.

The ones who've had fun were taking the Bar when power failed; or the time there was an earthquake. They get an asterisk next to their Bar card number.

 

[PDX-Y]

Chuck's videos show it fully entering the first lane from a left turn and just stopping there.

I don't think that's correct, at least not in the video under discussion, where's it's attempting to enter the turn lane to the left (which legally might have been occupied by a car in the left lane, but wasn't in this case). There was no collision you can reasonably predict with any cars but the one in the far travel lane, which it did see and flag. Are you maybe talking about something else?

 

[S4WRXTTCS]

I'm not sure what the take rate of FSD is from back then. But I would assume there are several factors... arbitration clause, most people are not litigious, general sense of just accepting FSD as it is and moving on, etc. It is also hard to quantify what your damages are because you did get certain aspects of the FSD so it's not like the full FSD isn't there.

As for myself, I got my car in Oct of last year and made my decision based on what FSD could do at that time. I passed on FSD but would have gotten enhanced auto pilot but waited too long. I am using the subscription version but mainly for the ability to lane change without disengaging. I suspect a lot of people who are upset at their FSD purchase are the direct result of Tesla's advertising at that time. They were told full FSD would be coming soon. Even last year when Elon raised the price to 10k he intimated the increase was because full FSD was imminent. I believe the verbiage in the older purchases over promised and perhaps did not have the same disclaimer as what is in the current contracts. People bought into Elon's hype and their marketing either didn't or didn't do enough to inform buyers of the time frame or capability.

There is also the possibility those who were persistent and got their refund never posted to the forums. They simply got their refund or credit and moved on.

I'm no lawyer, but even when I bought FSD (2018) I knew Tesla was committing fraud. I simply ignored the whole fraud thing, and bought it anyways.

A lot of us like myself who paid for FSD knew the way it was being marketed, and the timeframe given for it was completely fraudulent. That it would never be capable of doing what Elon said it would be capable of let alone on the time frame given by Musk.

I often avoided having a conversation about FSD to friends/family interested in a Tesla because it's an awkward conversation to have. It would have been much easier if Elon promised something more tangible.

Now days it's a lot easier to have a conversation about because FSD is available as a subscription. I fail to see how Auto-Lane change is worth $200/month to someone, but they're not taking your money while promising you anything more.

My advice for someone who doesn't want to commit fraud:

Don't sell people things you don't have control over.

You can't sell a person a planet for that same reason. You have no authority over the ownership of a planet.
You can't sell a person a self-driving car because you have no authority over whether your self-driving car will even be allowed to be self-driving.
Even pre-selling a L2 semi-autonomous car is risky because regulatory bodies might not allow the full feature set of what you promise.

You should probably complete the feature first, and once you have regulatory approval then you can sell it.

 

[Chazman92]

FSDBeta 9.0 - 2021.4.18.12 - 5 way stop signs with Drone view

 

[S4WRXTTCS]

Complacency is almost unavoidable at some point (assuming development keeps progressing).

It seems to me like a very non-linear performance curve, where you need to achieve consistent safe operation given changing use expectations.

To me that's the biggest danger in trying to morph an L2 system into an L4 system.

Insurance companies are going to be keeping a close eye because it's on them to cover the claims if the driver becomes complacent.
The NHTSA is keeping a close eye by requiring car companies to report L2 and above crashes.
The media is going to continue their close scrutiny of Tesla's approach.

Tesla's entire approach comes down to being allowed to release FSD Beta to the general public. In some ways all the attention to the FSD beta right now is a good thing so everyone will get bored when the real danger starts to emerge when it significantly improves.

 

[Daniel in SD]

I don't think that's correct, at least not in the video under discussion, where's it's attempting to enter the turn lane to the left (which legally might have been occupied by a car in the left lane, but wasn't in this case). There was no collision you can reasonably predict with any cars but the one in the far travel lane, which it did see and flag. Are you maybe talking about something else?

Now I'm confused. You're not supposed to enter the intersection and wait to turn left if the left turn arrow is red. If the protected left turn were at the beginning of the cycle you might not have an opportunity to complete the turn. While it's likely there would just be a lot of honking and cursing and no collision it doesn't sound safe to me.

 

[Battpower]

Insurance companies are going to be keeping a close eye because it's on them to cover the claims if the driver becomes complacent.

Did Tesla start offering their own insurance in some countries? I thought I saw some time ago that they were making moves into motor insurance which seems to me the only way they will be able to make the L2 to L4 transition until there is established risk data in each market territory.

 

[AndreP]

I'd say most of the L2 ADAS features are safety features provided a few IF's are met. The most important IF is if they work well enough for people to use them, and IF the people don't abuse them.

Adaptive Cruise Control -> I strongly feel as if there is a very real tangible safety benefit in that it's a second set of eyes to brake, and it leads to more appropriate following distances. Probably the biggest overall safety benefit is its more relaxing, and people don't stress out so much. For me it also has a strong convenience factor because I live in an area with a lot of stop and go traffic. I'm not particularly content with Tesla's adaptive cruise control implementation because it doesn't respond to cut-in's/out's fast enough. Needs to be a tad bit smoother as well.

Lane Steering (AP) -> It's actually been demonstrated that Lane Keeping systems, and lane steering systems do help safety in that they keep the car in the lane. It's pretty self-evident if you look around, and you realize 50% of the your fellow drivers are texting. So either we need driver monitoring systems to cut down on phone usage or better lane keeping/steering systems to keep these idiots in their lane. For myself I don't particularly need lane steering, but its nice on occasion when you want to rest your arms. I don't particularly like Tesla's implementation as I feel like it constantly doing microcorrections and its not smooth enough. I'd prefer a well tuned hands free system like Supercruise, and hopefully Blue Cruise will work well. For my next car I definitely want a really solid hands free Lane steering system.

The problem I have with Adaptive Cruise is that, in my SUV anyways, it doesn't notify you when it slows down to match traffic in front of you. The braking is so gradual that you don't even notice the vehicle has slowed to match someone who is doing 10 under the limit when you actually want to be going 10 over the limit, and that leads to me passing them and getting a little frustrated.

Only time I'll turn Lane Keeping on is when I plan to be distracted, like if I want to text while having that little safeguard in place. And I feel that's likely a dangerous path to go down, that it will likely make you more inclined to let yourself be distracted and reliant on a system that can just as easily fail. I suppose it's probably better than nothing for people who will text anyways, but I also know I'm surely not unique in feeling like it enables the behavior.

When distracted is also the only time Lane Keeping is tolerable because then you aren't watching the road as closely and aren't as readily noticing the strange things it does. If I have Lane Keeping on and am watching the road closely + steering the wheel, it's disconcerting to have the wheel pull in a direction you don't think it should.

 

[PDX-Y]

I'm no lawyer, but even when I bought FSD (2018) I knew Tesla was committing fraud. I simply ignored the whole fraud thing, and bought it anyways.

If you knew what you were buying, it's not "fraud". That's a definitional thing. Fraud requires deception. If you weren't deceived, then in fact your existence (and mine, FWIW) is an affirmative defense against a fraud claim by someone else, because reasonable people exist who weren't deceived.

 

[PDX-Y]

Now I'm confused.

I think the problem here is that you're splitting; you want the argument to be between "FSD was wrong and unsafe", and "FSD was fine". My contention is that "FSD was wrong, but the basic safety features were working and the degradation was clean". Not every failure is a "safety" problem of the same magnitude and it's important not to make arguments of the form "FSD should never have done that!" because you're setting an impossible bar.

At the end of the day all autonomy systems are going to fall down in traffic like this, because the truth is that traffic signage isn't an objectively solvable problem even for people.

Should it have failed here? Probably not. They should fix the median detection. Most of the arguments about "It never should have turned at all!" though are bunk.

 

[Dan D.]

Its too bad there is no way to see these things play out... I'm curious as to how it would have proceeded had there been no intervention or vehicles in the opposing side.

We already know from other videos. Some of Frenchie's on 8.2 it fully goes into oncoming lanes from a right turn. Chuck's videos show it fully entering the first lane from a left turn and just stopping there.

If there is no intervention from the FSD driver then the result would be whatever reaction the other driver does to avoid a collision. FSD sets up the collision potential, it does not always get the car in a safe condition, only the intervention or the other road users saved the collision.

I don't think that's correct, at least not in the video under discussion, where's it's attempting to enter the turn lane to the left (which legally might have been occupied by a car in the left lane, but wasn't in this case). There was no collision you can reasonably predict with any cars but the one in the far travel lane, which it did see and flag. Are you maybe talking about something else?

rijc99 asked a general question about how "these things play out". FSD Beta has been improving but there still are general behaviours that we've seen, so I'm referencing Chuck & Frenchie's videos in general. v9 is doing silly stuff just as v8.2 did, and the ones before it. Of course we can't know exactly what would have happened if the driver had not intervened, but we can infer the possibilities based on all the videos for the past year.

 

[mhan00]

FSDBeta 9.0 - 2021.4.18.12 - 5 way stop signs with Drone view

So it doesn’t know how to deal with the intersection. I’m guessing the issue is that there is no stop line painted at that stop sign ahead of it. Some nice Looking turns when approaching that intersection from another angle.

 

[Daniel in SD]

I think the problem here is that you're splitting; you want the argument to be between "FSD was wrong and unsafe", and "FSD was fine". My contention is that "FSD was wrong, but the basic safety features were working and the degradation was clean". Not every failure is a "safety" problem of the same magnitude and it's important not to make arguments of the form "FSD should never have done that!" because you're setting an impossible bar.

At the end of the day all autonomy systems are going to fall down in traffic like this, because the truth is that traffic signage isn't an objectively solvable problem even for people.

Should it have failed here? Probably not. They should fix the median detection. Most of the arguments about "It never should have turned at all!" though are bunk.

I don't think it's bunk to say the system shouldn't turn left when there's a red left turn arrow. I guess I'm a lot more optimistic than you because this seems like the easy part of autonomous driving.

 

[n.one.one]

Well, maybe this will become a regular Friday night thing.

 

[Chazman92]

FSDBeta 9.0 - 2021.4.18.12 - Honey Jar Run and Apiary Tour

 

[mhan00]

FSDBeta 9.0 - 2021.4.18.12 - Honey Jar Run and Apiary Tour

20:40- fairly major disengagement. Car got confused at some railroad tracks on a left turn light. Car stopped shot I’m between lanes. Driver took over to get to the proper stopping spot and re-engaged. Car decided for some reason to try to cut in front of the car adjacent to it and proceeded to try to do so on a red. I’m assuming it didn’t know it was in a turn lane and was in The opposite lane and was trying to rectify. Very bad decision.

23:10 - disengagement. truck backing out of driveway ahead. Looked like AP would have tried to pass behind the still backing up truck so driver hit the brake. It might have stopped by itself, but the driver rightfully didn’t wait.

 

[mikes_fsd]

honestly don’t understand, what is the benefit from a system that “REQUIRES alertness and readiness to take over at any time.”. Isn’t it much more stressful than driving without it?

Current Autopilot suite has always required alertness and readiness to take over at anytime.... Yet ALL my trips have been less stressful because of Autopilot!

 

[aronth5]

FSDBeta 9.0 - 2021.4.18.12 - Honey Jar Run and Apiary Tour

First really enjoy your videos.
Chazman92 one of the improvements I'm hoping for is the transition from on ramp to off ramp. I often have to take over just before the exit if the exit speed is too high which is pretty common in my area. The off ramp transition is just not very smooth now. Merging is also problematic when getting on the highway. Will FSD with the addition of City driving significantly improve these two transitions (on/off) even if I elect not continue FSD city streets enabled once your completing off the high way ramp? That alone would be helpful.