I am certainly making informed observations based on other EVs. My Kia Soul may not be to everyone's taste but they do get the fundamentals right. The auto wipers work fine, the auto lights work fine, the range they quote is the correct range, not an exaggeration. The ride is better and quieter than my Model Y. Irritatingly there are flaws with the locking system. Overall it is a better car than the Tesla just less performance and no superchargers.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
bjorn nyland's test of tesla vision
- Thread starter debrooker
- Start date
M109Rider
Active Member
But it wasn’t on topic to the person you replied to.
Again, so much for the reasonable argument.
Its an american car not designed for a screen on the drivers left!
Zilla91
Active Member
The person I replied to was going off topic in order to deflect.But it wasn’t on topic to the person you replied to.
Again, so much for the reasonable argument.
Why are you even on the UK & IE forum? You lot in North America wouldn’t know anything about good headlights.
Edit:
_Redshift_
Member
I find the most vocal bashers here own them. Not sure why, if the software and headlights suck so so badly.
It’s called experience.
Till 2017, Tesla service was beyond amazing. Techs were as incompetent as they are today, but they did not get into arguments with you. Or try to say ‘it’s all within spec’.
Now, not only do they make cars they should be ashamed of regarding build quality, software that doesn’t do what’s promised, weird stuff like the yoke and capacitive buttons on the wheel, just to name a few things, but they also argue with you that you should accept things the way they are.
Whats preventing an exodus out of Tesla? Competition. Not very credible right now. Mainly because of the shitty charging network that is not reliable. Plus, the competition is still mired in fossil fuel thinking. Thus, you get no frunk with some of the other EVs. Or stupid fake engine noise. Or charging ports that look like they were designed by multiple committees.
However, that’s going to change, once they figure out the game.
The cars don’t learn unless they have a software update. The shadow mode is to capture real world scenarios where the software wants to do something significantly different to the driver, the question that then needs processing back at the mother ship is why, and feed in the scenario to the retraining. So far we still have some massive issues which don’t appear to have materially benefitted from that approach.
Level 3 is an often misunderstood thing in my opinion. Many people look at the Tesla software and think because it tries to do more things than the competition (especially the US beta) it must be nearer level 3 than the competition. But that has nothing to do with Level 3. Level 3 is about accountability of the things it does and less can be more. It’s also about the interaction with the driver and how accountability passes backwards and forwards in a controlled manner. Tesla seem to do virtually nothing in this space, if AP runs into trouble, it disengages with a bong and the driver, in an instant, without warning, is having to do everything. Now if you’re paying attention like you should with L2 then that’s fine, but L3 says you don’t need to be paying much attention other than waiting to be told to take back control in a bit which is an entirely different thing. Until Tesla do something in this area they’ll never be given level 3. The driver interaction side with Tesla is the worst of all the systems I’ve seen.
On the Audi headlights, I think it’s been done to death now.
Ever since my 'bug report' bubble got burst with the realisation that in effect, it did nothing, I have seriously doubted that there is much if any 'learning' going on, especially in the UK.
At one time, it did appear that when I returned home from a journey with many AP disengagements, my home internet ground to a standstill, presumably while the car sent back some data to Tesla. I am not aware of even that happening any more.
In any case, given Tesla's ongoing issues with FSD Beta as well as long standing problems like auto headlights, is there any evidence of a slick (or even mediocre) data-driven OTA update process keeping cars in an elevated status of technical superiority and performance?
I play around with music software and for many years the software has included an indication of how much you are loading the processor to help avoid any sudden, unexpected dropouts. IF any Tesla AP / processor driven aids depend on available processing resources, it would be good to know how stretched your processing system is.
Similarly, I would expect the car to have a measure of confidence in what it thinks it sees. Regardless of intermediate steps, ultimately it has to decide 'brake or not to brake', 'wipe or not to wipe’, 'head beam or not head beam'? Once it make a decision either way, you'd think there should be a high level of confidence based on consistent observation over time, multiple sensor feedback or both. The glitchy, lurching, unreassuring feel to auto features suggests to me that the algorithms are happy ditching what they just saw and knew a fraction of a second ago in favour of what they see now, regardless of how feasible that could be, and the potential for throwing hands up at any moment 'over to you driver' both suggest that decisions are (still) made based on individual views rather than a temporal confidence bubble.
At least for FSD but likely for everything else too, I feel the car should share an indication that it at least believes itself to be well within its confidence zone and that all auto systems should therefore work dependably.
Especially with driver control interface changes requiring longer periods of driver distraction (hunting for controls), a human can only drive the car based on having a notional confidence bubble to gauge when it might be safe to try a more demanding task .... like turning on the heater or ac!
Last edited:
Wol747
Active Member
It’s certainly something they need to do. Phantom braking for instance would disappear (well, maybe reduce) if the car didn’t go over time..
Nothing to see just a bridge
Nothing to see just a bridge
Where did that artic across the road come from, panic!
Nothing to see just a bridge
I think the second part of level 3+ isn’t just ‘can I cope with now’, it’s ‘the road ahead looks fine, no junctions I can’t cope with for half a mile, speed of traffic looks to be slow enough etc’
The Merc system on motorways can see what’s going on and knows the conditions are within its envelope, and when things look like they’re going to exceed that (traffic speeds up, junction etc) it effectively says to the driver ‘I probably could drive this bit but I can’t be accountable for it’ - the driver becomes accountable and starts paying attention but the car carries in driving on auto lane keep and speed. There’s a good article somewhere that talks about it, the car can do all the driving between level 2 and level 3 and that doesn’t change, the only thing that could be changing is who’s accountable and as a direct consequence whether the driver needs to be paying attention.
The all powerful neural net, Dojo, etc is basically the answer to all of life’s ills, and is permanently coming soon to replace the old, knackered, no longer developed stack we have in our cars.
I would have expected by now to have seen improvement in some basic areas, but there has been none that have been palpable, at least not in Europe. The US at least has FSD beta which seems to get updates.
As I’m in IT I’m aware these problems aren’t easy, but at the same time I’m also a customer, so I don’t think it’s unreasonable not to be lied to about features that are coming, with years going by with no sign of them, and no tactile improvements (I don’t count farting Boombox and lightshows as progress).
I would have expected by now to have seen improvement in some basic areas, but there has been none that have been palpable, at least not in Europe. The US at least has FSD beta which seems to get updates.
As I’m in IT I’m aware these problems aren’t easy, but at the same time I’m also a customer, so I don’t think it’s unreasonable not to be lied to about features that are coming, with years going by with no sign of them, and no tactile improvements (I don’t count farting Boombox and lightshows as progress).
I may not be describing this brilliantly because I’m bleary eyed still this morning… but the software in the car is supposed to self correct for basic, preventable errors hence my phrasing.
I.e misidentification of an object, correcting and then using that correction in deciding it’s ‘confidence’ level on things like that in the future. This is kind of the whole point for them doubling down on vision only.
Whether we each think this is noticeable or successful is another question, but this is what their approach was created to do, according to Karpathy.
So whilst not “learning” in the way that the software is “trained” with data before release, the software running on the car side is billed as self-improving or self-maintaining (within the parameters of the release its running). It’s a shame this can’t safely be done to correct for very geographically and user specific challenges I.e phantom braking, but obviously they want to control progress on anything that can have a direct impact on vehicle control in safe releases.
Edit: that matches my understanding of the shadow mode processes running in the background too, yes. That includes data from all the driver inputs and vehicle controls too.
Last edited:
It could be an Emperor's New Clothes moment, but I suspect it'll just peter out to nothing.
I personally think it's more likely that Tesla will just get left behind by other manufacturers. Not as dramatic, but same net result - an erosion of the mindset that Teslas are the "only cars that drive themselves", etc. Other manufacturers are already there with ALKS and ADAS as good if not better than Tesla's autopilot, they just don't make a big and deceptive song and dance about it.
Regulations wise I think in Europe at least we're several years away from there being any discussion, let alone approval, of cars being permitted to drive themselves from A to B.
I wish I'd thought about this more before I bought FSD, because I naively didn't know much about Tesla at that point and took their word of "city streets driving coming later this year" as gospel. My own fault for not doing enough due diligence. I do wonder how many people will feel similarly burned by Tesla, for that and other reasons - i.e. the disparity between what people imagine Teslas can do, and what they actually do (and more importantly - what they are missing relative to other cars at a similar price point, or less) how many customers will be one time owners only. There are plenty of competent alternatives now, and they are getting better with each iteration, far faster than Tesla seem to be.
Mostly I'm aggreived that Tesla patently do not care about Europe, at all, beyond selling cars. They've done nothing with adaptive headlights over here (my car predates them, but still), autonomous tech has stood still or gone backwards, basic stuff like auto lights & wipers are not fit for purpose, all the while we're expected to buy into the notion that the all conquering neural net/Dojo will fix everything. If our cars can't see distant rear lights to turn off headlights, how realistic is that really?
I personally think it's more likely that Tesla will just get left behind by other manufacturers. Not as dramatic, but same net result - an erosion of the mindset that Teslas are the "only cars that drive themselves", etc. Other manufacturers are already there with ALKS and ADAS as good if not better than Tesla's autopilot, they just don't make a big and deceptive song and dance about it.
Regulations wise I think in Europe at least we're several years away from there being any discussion, let alone approval, of cars being permitted to drive themselves from A to B.
I wish I'd thought about this more before I bought FSD, because I naively didn't know much about Tesla at that point and took their word of "city streets driving coming later this year" as gospel. My own fault for not doing enough due diligence. I do wonder how many people will feel similarly burned by Tesla, for that and other reasons - i.e. the disparity between what people imagine Teslas can do, and what they actually do (and more importantly - what they are missing relative to other cars at a similar price point, or less) how many customers will be one time owners only. There are plenty of competent alternatives now, and they are getting better with each iteration, far faster than Tesla seem to be.
Mostly I'm aggreived that Tesla patently do not care about Europe, at all, beyond selling cars. They've done nothing with adaptive headlights over here (my car predates them, but still), autonomous tech has stood still or gone backwards, basic stuff like auto lights & wipers are not fit for purpose, all the while we're expected to buy into the notion that the all conquering neural net/Dojo will fix everything. If our cars can't see distant rear lights to turn off headlights, how realistic is that really?
@Durzel I'm honestly quite surprised there hasn't been a class action against Tesla in Europe over FSD. Sunk cost fallacy? Most people wouldn't admit they feel fobbed off like that I guess. I suppose some just sell the car.
Whilst Tesla can claim that they don't "advertise" they've certainly shaped the narrative around this in other, just as impactful ways. As a business for whom most customers start with (or at least visit) their website at some point in their buying journey, their presentation of facts and copywriting is much more important. And a certain twitter user has specifically promised things, every year for what... 5 years now?
I'm excited for the full release personally, just to see where the tech goes.. BRING ON 2019!
Whilst Tesla can claim that they don't "advertise" they've certainly shaped the narrative around this in other, just as impactful ways. As a business for whom most customers start with (or at least visit) their website at some point in their buying journey, their presentation of facts and copywriting is much more important. And a certain twitter user has specifically promised things, every year for what... 5 years now?
I'm excited for the full release personally, just to see where the tech goes.. BRING ON 2019!
I think a class action would be against the hardcore fanboys on social media who did and still do argue FSD is imminent as they're the ones that have done much of the pursuading. Tesla have been very careful with their wording on the website.
The only real angle is Musk and his tweets (oh the irony that its twitter), and the talk of coast to coast and robotaxi etc.
That said, there does seem to be the odd case going through
Court orders Tesla to buy back car from customer who felt Autopilot was like 'a drunk first-time driver'
A court in Germany has ordered Tesla to buy back a Model 3 vehicle from a customer who filed a...
electrek.co
Musk really hates playing by someone elese rules, and in some areas throwing the rule book away has been the cause of his success, but you can't ride roughshot over everything, especially safety, and buggy software releases while trying to argue with dubious math that your car is twice as safe isn;t going to wash with the regulators. If anything, it will get their back up as it shows a lack of corporate hygiene in the say they do things at times. Why do we have beeping alarms when we open the door? Why does Tidal disappear after an update and require a convoluted work around? Why do these issues vary by owner, surely we should all have the same experience, will the cars all drive the same or will we all have variable FSD performance and is that acceptable?
I think the lawsuit that could happen is if Tesla end up caving to LiDAR or other upgraded hardware in future models to achieve decent L3, then people who bought the aptly named FSD package would be entitled to a free retrofit upgrade.
JupiterMan
Active Member
there has been discussion.
Law commission in UK have had consultations on this subject.
Automated Vehicles | Law Commission
its a very in depth and interesting read if you have time.
This is now in the hands of government of course, but still it shows how regulations could be adopted in the UK that differ to the UNECE standard and in theory adopt full autonomy.
And of course the law/UNECE and the like deciding the what and how of all this is one thing, Tesla being able to meet whatever they come up with is another. Imagine if they say something as simple as "for law enforcement and wider driver awareness, any car operating under level 4 conditions needs to have a steady green light visible from all angles" - Tesla won't be able to comply without changes, at least not with the existing fleet. On a similar note, the number of people who thought we'd get adaptive headlights when the US approved their use was quite staggering, Tesla simply haven't got the capability to actually do it at the moment.
I'm sure lots of lobbying is going on over the capabilities and rules, what's certain is being on the inside and working with regulators and playing the game is going to be a politically smarter place to be than being on the outside trying to goad them into something.
I'm sure lots of lobbying is going on over the capabilities and rules, what's certain is being on the inside and working with regulators and playing the game is going to be a politically smarter place to be than being on the outside trying to goad them into something.
Similar threads
