Find me a video of a Tesla driving off a cliff! I need to see that….
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Find me a video of a Tesla driving off a cliff! I need to see that….
Daniel in SD
(supervised)
I'm sure it would have stopped. Whole Mars Catalog just didn't have enough faith in the mission. This is great example of why it's easy to design a system that never gets stuck. People should think a little bit harder about why when you remove the driver you might want to accept some number of fallbacks to minimal risk condition.
powertoold
Active Member
I don’t know how many of you realize this but Cruise requires remote assistance (to plan new paths or allow to proceed) every 5 to 10 miles in SF. In the easier to drive parts of SF.
And when the car requires remote assistance, it’s still considered autonomous, so in the Cruise videos you watch, keep it in mind.
And when the car requires remote assistance, it’s still considered autonomous, so in the Cruise videos you watch, keep it in mind.
Remember when this thread was for FSD videos and discussion about those videos? Maybe move the arguments to a different thread...
powertoold
Active Member
We discuss disengagement rates often here, so I just wanted to give some perspective on how other fsd developers game their stats. They rarely disclose this type of information, but when they do (even then, it’s difficult to find), it’s often surprising (in a negative way).
We‘re in a unique position to see fsd beta essentially unfiltered.
Remember, as so many supporters will point out here, as someone that bought post April 2019, all Tesla EVER owes you is City Streets Autosteer as part of "FSD". They never sold/advertised you anything beyond that. Shame on you for seeing videos and such that made you assume you would get that, the checkout process and purchase contract is clear.
Meanwhile I have a 2017 and 2018 car, sold as full L4 capable, and Tesla wants $1K each to upgrade the HW in those cars to do what Tesla already said they could do.
Good luck fighting for HW4 is what most supporters here would say, Tesla never promised you that, and even when they did promise it to you, it still doesn't matter.
It would be a lot more unique if Tesla would release it like they keep claiming they will. I mean, who cares about betas that you can't get on your own car that you already paid for.
Why won't they release it? What's unacceptable about it?
stopcrazypp
Well-Known Member
You contradicted yourself right there in the end, saying the money you earn doesn't have to do with the type of content, which was my point exactly. Just because you make superfan content doesn't necessarily mean you make money, and if you make hater or neutral content doesn't mean you don't make money.
You have Tesla/Elon haters that get lots of views also, for example Adam Something (which just popped up in my recommend recently, I'm sure there are many examples I'm not aware of). It's mainly on how clickbait your titles, thumbnails are. For example Jalopnik writes a lot of articles critical on Tesla, but they attract fans and haters alike by making the title/graphics clickbaity. Some go to comments to speak against the content, some in support, but it doesn't matter as long as they got the click.
I paid for Full Self-Driving Capability, which means the car should be capable of fully self-drive. If City Streets is the last piece of it, then great, we know it runs on HW3. If they add more things under the FSD umbrella, I believe I'm owed that. If they release something like "Comfort+Entertainment Package", then yeah, I don't think I'm owed that. But again, "free" FSD hardware upgrades are precedented, so I don't think my assumption is too farfetched.
Did it drive off or no? No? Ok then
All I can say is that we're on the same page here, because that's what Tesla clearly showed with all their advertising.
However, supporters will remind you that "Full Self Driving" is just a name. What you were sold is the list above. Nothing more. They can release more, and only put it on newer cars, and you have zero recourse.
The "free" hardware upgrades have a precedent, that supporters will say comes only for HW2 to HW3, based on the April 2019 autonomy day. But there is also precedent for charging for upgrades, like they are doing for subscriptions right now. You know, cars that are supposed to have ALL THE HARDWARE NEEDED FOR FSD CAPABILITY but people will say "No, no, that was all the HW needed for FSD, assuming you buy FSD, not subscribe to it. FSD subscriptions are different than FSD purchases, despite the fact they are identically named, are identical software, and are sold in the App under identical click flows and descriptions.
Yeah. If you believe just because Tesla did something in the past that it will happen in the future you are going to be very, very disappointed, and there will always be someone that bought their car after that moment going "I got mine, why are you complaining!"?
It doesn’t even seem like it was close to driving off a cliff. Maybe through some rocks onto an old dirt trail, which makes sense since the car uses maps and a cliff isn’t on a map while fire access roads (the dirt road) would be.
Are you seriously suggesting FSD beta drivers should not take over when the car is not driving in a way they are comfortable?Did it drive off or no? No? Ok then
The instructions are literally to take over in this case. If FSD is risking you or your car, you take over, and that's a disengagement. FSD needs to not make drivers nervous and drive like real humans do in order to succeed.
...Or Tesla needs to pay for damages caused by FSD.
It'll be able to drive itself with constant supervision expected in a Level 2 ADAS. Tesla has straight up told regulators that the wider public release of City Streets and even a "final release" will continue being Level 2 -- "final release" could be interpreted in different ways but I'd take the pessimistic interpretation. That's not to say the FSD Suite won't have added Level 5 functions in the (distant?) future.
The system's OEDR has some significant limitations when it comes to unique obstacles, road signs, emergency vehicles, static objects etc and Tesla told regulators they don't expect significant OEDR enhancements that would allow shifting responsibility for the Dynamic Driving Task away from the human. And that is fleshing itself out in every successive beta update -- we've already seen that Beta 10 will still gladly drive into gaggles of Road Closed signs, temporary buildings erected on the road, or apparently off a cliff where a mapped road had been modified. There doesn't seem to be any prospect of these major issues being solved: they've existed for a long time, are still completely repeatable today, and they're not tied to how well the car navigates a left turn and the other tweaks we're seeing in Beta updates.
So it'll Fully Self-drive and control steering, braking, accelerating etc as long as your eyeballs are pointed forward and/or you have your hands on the wheel to take over when these limitations ^^ present themselves. Consumer vehicles that allow us to set a waypoint and stop paying attention / do other things / nap while the car drives, well that is a long ways away IMO.
Last edited by a moderator:
You obviously no nothing about the FDA. The FDA would be thrilled if you could show drugs that improved at 300% from one trial to another. THRILLED
Drugs dont do this. Software can. You are really going to be an unhappy camper in 6 months or so when they release this to the general public.
On a side note, when a drug shows simply fantastic performance and few side effects the FDA will quickly dispense with most of the placebo trails. This was recently the case with what looks to be the cure for Cystic Fibrosis (CF) which is maybe the most miraculous drug in years. On another note why in the hell are you bringing FDA into a thread on FSD. If anyone is doing big trials in real world it is Tesla.
Overall your comments are looking increasingly bizarre. I am setting aside everyones frustration (rightful frustration) regarding being sold a product that did not exist. OTOH- if you bought a tesla you could probably afford the investment in FSD. Thanks to all you funders. It looks like a quite useable product is on the cusp of rollout, you could literally be watching the world change...and what do you spend your time doing?
I've seen several comments by beta testers about taking certain road sections way faster than they would have done. Also seems common for testers to dial down the speed to coax fsd through difficult scenarios.
What is the current state of speed modulation? I don't see it discussed much but it feels to me like it should be an easential part of a safe self driving system.
IMO it needs to use both map data to determine when there is an upcoming road layout that might demand additional caution, as well as responding to its own local perception regarding parked cars, level of pedestrian activity, multiple cyclists etc.
Is there any evidence of consistent use of speed control below posted limits and responding to hazards in real time and in good time?
What is the current state of speed modulation? I don't see it discussed much but it feels to me like it should be an easential part of a safe self driving system.
IMO it needs to use both map data to determine when there is an upcoming road layout that might demand additional caution, as well as responding to its own local perception regarding parked cars, level of pedestrian activity, multiple cyclists etc.
Is there any evidence of consistent use of speed control below posted limits and responding to hazards in real time and in good time?
Yes, if you can SHOW it.
Tesla is not showing anything before release. They are using the release as a test, and then changing everything between each trial.
They can't say "because V9 worked, V10 (which is completely different) must be fine also." This is like saying because Aspirin works, Ibuprofen is fine too. I mean FSD V10 was a complete rewrite, right? And V10.1 will have yet another complete rewrite for highway code? When will they ever prove that any specific release is acceptable to go wider? What does V9 tell us about V10 if it's all new?
I've developed safety critical software for aircraft. If you change everything, you start over. A program which announces "We just re-wrote everything, try this version" is not on the cusp of a rollout, they're flailing if their goal is a stable, proven, safety critical system. But I think it's pretty clear FSD is not about safety at all (right now)....
Yep, been hearing this for 5 years.
Why hasn't Tesla released the current FSD? What is stopping them? What is the metric? What is the data they need? How are the current beta testers on V10 helping them understand if V10.1 can be released?
What's the use case for city streets autosteer? You must pay attention all the time, it will try to hit things, you are liable. What driver assistance function does it perform?
Nope. I'll be a subscriber because it looks fun to play with. (after Tesla loses the small claims case over the $1K FSD computer upgrade fee, which was included in my vehicle purchase)
Last edited:
It would be best to move the arguments to one of the many threads that already argue about this, and have one thread that just discusses the videos so that those of us that just want to see and discuss the progress don't have to wade through pages of arguments
I’m suggesting I want to see a Tesla drive off a cliff. For full self driving to be remotely useful it has to inspire confidence and not in a “holy *sugar* we made it” sort of a way. There is no such thing as an intervention that wasn’t needed.
favo
P3D+ owner
Similar threads
