Tesla unlikely to Survive (Vol. 3)

[RichardM5]

I'll not follow up on everything, I'd be here all day, but concentrate on the wipers as they are my #1 day to day annoyance.

The rain sensitive wipers on my E39 BMW M5 are vastly superior to the Tesla, it's over 20 years old. They are not perfect, but they have two features that make them infinitely less frustrating to live with in our Yorkshire climate :-

- You can adjust the sensitivity, easily, without looking via a dial on the stalk.
- You can turn them off, or on, easily, without looking via the stalk.

Removing that basic functionality that would have minimum cost is, in my opinion, a joke on a £60k (Base Model 3 Performance price, best part of £70k with options) car.

And one other point you mentioned, I can chose not to buy if I don't like the features, I won't be buying a Highland due to the indicators which, in my opinion and probably very specific circumstances (a bit like the Full Self Driving), are down right dangerous. With a few easily made changes, or even options, I would still seriously consider one when I come to replace my current daily Model 3. So that has cost them a customer, I suspect I won't be the only one.

[h0b0]

I have now seen 3 or 4 cyber trucks driving around. There is one local to me. The novelty has worn off and now kids are pointing and laughing.

The cyber truck is going to be studied in the future as the symbol of Teslas destruction. I am convince Elon didn’t think about it getting this far into his con to drum up the stock price. I also think he is sabotaging the company.

But, for me the cyber truck is just the final nail. The death of Tesla has been on going for years. There has been no innovation. Just lies. A $30k model 3 would be amazing. That was never going to happen. Neither was the model 2. The model S is barely different from 12 years ago.

I wonder how successful Tesla would be if the US government had’t killed Chinese competition with 25-100% tariffs on top of the billions in subsidies.

[skwdenyer]

RichardM5 said:

I'll not follow up on everything, I'd be here all day, but concentrate on the wipers as they are my #1 day to day annoyance.

The rain sensitive wipers on my E39 BMW M5 are vastly superior to the Tesla, it's over 20 years old. They are not perfect, but they have two features that make them infinitely less frustrating to live with in our Yorkshire climate :-

- You can adjust the sensitivity, easily, without looking via a dial on the stalk.
- You can turn them off, or on, easily, without looking via the stalk.

Removing that basic functionality that would have minimum cost is, in my opinion, a joke on a £60k (Base Model 3 Performance price, best part of £70k with options) car.

And one other point you mentioned, I can chose not to buy if I don't like the features, I won't be buying a Highland due to the indicators which, in my opinion and probably very specific circumstances (a bit like the Full Self Driving), are down right dangerous. With a few easily made changes, or even options, I would still seriously consider one when I come to replace my current daily Model 3. So that has cost them a customer, I suspect I won't be the only one.

Re the wipers, I've not owned a car yet with adjustable sensitivity for auto wipers - the options are just on/off, functionally unacceptable IMHO Very specifically, the problem I find is getting used to them working *sometimes* and then having to intervene. Hence I just turn them off now - I learned to drive using wipers I controlled myself, and don't feel I'm missing anything by continuing to exercise that control But I may just be a relic... Your BMW sounds good in this regard. At least in Yorkshire there's a good chance any test drive would feature rain...

In terms of the Highland indicator buttons, I don't disagree with you BTW. On a Cybertruck with variable-ratio steering it is fine IMHO, but when the wheel needs twirling then not so much. Probably a function of Tesla not caring too much about roundabouts. I gather there are third party stalk solutions available now (see https://tesstudio.us/products/tesstudio-tesla-mode... for instance) and (because those exist) I'd consider a Highland Model 3 now. Whether an aftermarket solution is good enough for those who need it is an interesting question - speaking only personally, if the rest of the car was right for my needs (a Model 3 isn't right now), I probably wouldn't worry about spending a few hundred $ to fix an interface dislike - but I completely get that others wouldn't see it like that.

[TheDeuce]

skwdenyer said:

RichardM5 said:

I'll not follow up on everything, I'd be here all day, but concentrate on the wipers as they are my #1 day to day annoyance.

The rain sensitive wipers on my E39 BMW M5 are vastly superior to the Tesla, it's over 20 years old. They are not perfect, but they have two features that make them infinitely less frustrating to live with in our Yorkshire climate :-

- You can adjust the sensitivity, easily, without looking via a dial on the stalk.
- You can turn them off, or on, easily, without looking via the stalk.

Removing that basic functionality that would have minimum cost is, in my opinion, a joke on a £60k (Base Model 3 Performance price, best part of £70k with options) car.

And one other point you mentioned, I can chose not to buy if I don't like the features, I won't be buying a Highland due to the indicators which, in my opinion and probably very specific circumstances (a bit like the Full Self Driving), are down right dangerous. With a few easily made changes, or even options, I would still seriously consider one when I come to replace my current daily Model 3. So that has cost them a customer, I suspect I won't be the only one.

Re the wipers, I've not owned a car yet with adjustable sensitivity for auto wipers - the options are just on/off, functionally unacceptable IMHO Very specifically, the problem I find is getting used to them working *sometimes* and then having to intervene. Hence I just turn them off now - I learned to drive using wipers I controlled myself, and don't feel I'm missing anything by continuing to exercise that control But I may just be a relic... Your BMW sounds good in this regard. At least in Yorkshire there's a good chance any test drive would feature rain...

In terms of the Highland indicator buttons, I don't disagree with you BTW. On a Cybertruck with variable-ratio steering it is fine IMHO, but when the wheel needs twirling then not so much. Probably a function of Tesla not caring too much about roundabouts. I gather there are third party stalk solutions available now (see https://tesstudio.us/products/tesstudio-tesla-mode... for instance) and (because those exist) I'd consider a Highland Model 3 now. Whether an aftermarket solution is good enough for those who need it is an interesting question - speaking only personally, if the rest of the car was right for my needs (a Model 3 isn't right now), I probably wouldn't worry about spending a few hundred $ to fix an interface dislike - but I completely get that others wouldn't see it like that.

I can't remember ever owning a car with auto wipers where the sensitivity wasn't adjustable. You just use the same dial which selects frequency of intermittent wipe to control the sensitivity (or level of response more accurately) to different levels of rain.

Accord
Focus
330d
Golf
430d
iPace
i4

All the same

[skwdenyer]

TheDeuce said:

I can't remember ever owning a car with auto wipers where the sensitivity wasn't adjustable. You just use the same dial which selects frequency of intermittent wipe to control the sensitivity (or level of response more accurately) to different levels of rain.

Accord
Focus
330d
Golf
430d
iPace
i4

All the same

Some specific instances:

Current family steed: a Peugeot 5008 GT line (2023). Auto wipers are just “auto;” intermittent wipers are speed-sensitive - no user adjustment possible of either.

My Disco 3 also has non-adjustable sensitivity.

Our previous Alhambra *did* in fairness have adjustable sensitivity, but no adjustment of intermittent. In the Dales we found the sensor continually reacting to insects (vast numbers of midges where we are at many times of year) and to salt in winter - both triggering smearing. Along with auto headlights that constantly triggered with reflective road signs (same problem on our Peugeot), the auto systems were so dumb we turned them off.

I’m not saying Tesla’s system is great; only that I believe a camera system + continuous software improvement provides the best path to potential success for a vision based problem.

[RichardM5]

skwdenyer said:

Re the wipers, I've not owned a car yet with adjustable sensitivity for auto wipers - the options are just on/off, functionally unacceptable IMHO Very specifically, the problem I find is getting used to them working *sometimes* and then having to intervene. Hence I just turn them off now - I learned to drive using wipers I controlled myself, and don't feel I'm missing anything by continuing to exercise that control But I may just be a relic... Your BMW sounds good in this regard. At least in Yorkshire there's a good chance any test drive would feature rain...

I don't disagree, if the rain sensitive wipers don't work well I'm happy to switch to manual - but in the Tesla the wiper setting are in a menu on the touch screen, and that is just a 'joke' for a control you might use many times on a journey and have to take your eyes off the road to adjust.

skwdenyer said:

In terms of the Highland indicator buttons, I don't disagree with you BTW. On a Cybertruck with variable-ratio steering it is fine IMHO, but when the wheel needs twirling then not so much. Probably a function of Tesla not caring too much about roundabouts. I gather there are third party stalk solutions available now (see https://tesstudio.us/products/tesstudio-tesla-mode... for instance) and (because those exist) I'd consider a Highland Model 3 now. Whether an aftermarket solution is good enough for those who need it is an interesting question - speaking only personally, if the rest of the car was right for my needs (a Model 3 isn't right now), I probably wouldn't worry about spending a few hundred $ to fix an interface dislike - but I completely get that others wouldn't see it like that.

Interesting, I was thinking someone would come up with something like that pretty soon but had not seen it yet.

[durbster]

skwdenyer said:

I’m not saying Tesla’s system is great; only that I believe a camera system + continuous software improvement provides the best path to potential success for a vision based problem.

So what is shown to be the worst solution is actually the best because... *optimism*? Come off it.

Ask anyone who works in film and they'll tell you that rain is notoriously difficult to capture on camera. They've been trying to solve this problem for what must be a century now and they still have to use tricks with lighting to get the camera to pick it up at all.

The problem is that if the light is wrong, rain is essentially invisible to a camera. If it's difficult to achieve in a controlled filming environment, how is it ever going to work on a car?

There's no way Tesla don't know this but that's fine. All they need to do is keep telling their customers to believe that it'll all be great, one day.

[skwdenyer]

durbster said:

skwdenyer said:

I’m not saying Tesla’s system is great; only that I believe a camera system + continuous software improvement provides the best path to potential success for a vision based problem.

So what is shown to be the worst solution is actually the best because... *optimism*? Come off it.

Ask anyone who works in film and they'll tell you that rain is notoriously difficult to capture on camera. They've been trying to solve this problem for what must be a century now and they still have to use tricks with lighting to get the camera to pick it up at all.

The problem is that if the light is wrong, rain is essentially invisible to a camera. If it's difficult to achieve in a controlled filming environment, how is it ever going to work on a car?

There's no way Tesla don't know this but that's fine. All they need to do is keep telling their customers to believe that it'll all be great, one day.

No rain sensors actually sense rain (I'm assuming you knew that, but just in case...). They sense water droplets (or similar, e.g. insects) on the glass. Conventional sensors bounce IR light off of whatever is on the screen (if nothing then little or no reflection). So long as a Tesla (or other) cameras can resolve that there is an obstruction on the glass, they should be as good as a conventional sensor. From what I can see, the focal lens of the cameras is sufficient for that.

If Tesla are trying to capture rain in flight then, I agree, they're doomed to failure.

As regards hope, no, not at all; merely my analysis of the problem and the range of possible solutions. Conventional sensors have nowhere to go, no development path available to them. Camera-based systems are just starting. As Alan Staniforth liked to say, the last of the old will beat the first of the new - at first... I believe (and you clearly do not) that the camera-based system has better *potential* to improve through software alone. I may be wrong, and don't mind if I am

[Durzel]

h0b0 said:

I have now seen 3 or 4 cyber trucks driving around. There is one local to me. The novelty has worn off and now kids are pointing and laughing.

The cyber truck is going to be studied in the future as the symbol of Teslas destruction. I am convince Elon didn’t think about it getting this far into his con to drum up the stock price. I also think he is sabotaging the company.

But, for me the cyber truck is just the final nail. The death of Tesla has been on going for years. There has been no innovation. Just lies. A $30k model 3 would be amazing. That was never going to happen. Neither was the model 2. The model S is barely different from 12 years ago.

I wonder how successful Tesla would be if the US government had’t killed Chinese competition with 25-100% tariffs on top of the billions in subsidies.

This tends to be my thinking too.

A switched on CEO would have had Model 2 in production, scaling up Semis, maybe even actually get Roadster 2.0 built as a halo model. I'd also argue a truck derived from a Model Y would have been first to market, long before Ford, Rivian and co had got their models out.

Instead we have a CEO who is asleep at the wheel as far as Tesla is concerned. He is utterly obsessed with Twitter to the detriment of everything else. His personal crusade now is all about "free speech" (in heavy quotes). He pushed through Cybertruck as a concept-made-real in spite of it being a nightmare to manufacture, littered with issues, and failing to meet any of the promised specs (this isn't new from Musk). It is emblematic of the state of Tesla nowadays, I think. Every time it breaks down in bizarre "college project" ways in public Tesla are shamed. Panels falling off, unclipping and bending, etc.

If as is likely shareholders grant Musk his share options then it is a certainty in my opinion that he will liquidate a decent amount of them to prop up Twitter - a completely irrelevant thing to Tesla and its mission. He has done it previously, so there's no reason that woukdn't continue.

I also happen to think that Tesla will not crack Optimus and Robotaxis in any meaningful timeframe, or at all if they continue on this trajectory.

[Durzel]

skwdenyer said:

durbster said:

skwdenyer said:

I’m not saying Tesla’s system is great; only that I believe a camera system + continuous software improvement provides the best path to potential success for a vision based problem.

So what is shown to be the worst solution is actually the best because... *optimism*? Come off it.

Ask anyone who works in film and they'll tell you that rain is notoriously difficult to capture on camera. They've been trying to solve this problem for what must be a century now and they still have to use tricks with lighting to get the camera to pick it up at all.

The problem is that if the light is wrong, rain is essentially invisible to a camera. If it's difficult to achieve in a controlled filming environment, how is it ever going to work on a car?

There's no way Tesla don't know this but that's fine. All they need to do is keep telling their customers to believe that it'll all be great, one day.

No rain sensors actually sense rain (I'm assuming you knew that, but just in case...). They sense water droplets (or similar, e.g. insects) on the glass. Conventional sensors bounce IR light off of whatever is on the screen (if nothing then little or no reflection). So long as a Tesla (or other) cameras can resolve that there is an obstruction on the glass, they should be as good as a conventional sensor. From what I can see, the focal lens of the cameras is sufficient for that.

If Tesla are trying to capture rain in flight then, I agree, they're doomed to failure.

As regards hope, no, not at all; merely my analysis of the problem and the range of possible solutions. Conventional sensors have nowhere to go, no development path available to them. Camera-based systems are just starting. As Alan Staniforth liked to say, the last of the old will beat the first of the new - at first... I believe (and you clearly do not) that the camera-based system has better *potential* to improve through software alone. I may be wrong, and don't mind if I am

It's worth noting that the cameras on the car are neither high resolution, nor do they have any night vision capability.

Rain sensors that detect IR refractions work in any conditions, are tried and tested and reliable. How individual manufacturers handle sensitivity options using them is up to them, some do it better than others. For what it's worth the experience I've had of cars 10 years older than my Model 3 - they were a "set and forget" thing, they just worked without any conscious thought.

Apocryphal but I've found my Model 3 auto wipers to be "ok", mainly because I don't do a lot of miles and the circumstances where I've been caught in the rain the conditions have been somewhat ideal (consistent rainfall, no glare from the sun). They have been atrocious at night.

Edited by Durzel on Wednesday 22 May 18:23

[ShortBeardy]

On the Model Y you can program the left steering wheel button to also control the wipers. You do this before setting off and it is stored in your profile and remains in memory. A dab on the end of the stalk will initiate a single wipe and then the button on the steering wheel button can be pushed left or right to change the wiper setting.
You can do something similar on the new Model 3 using the right wheel button instead of the (missing) stick. I'm not saying it's a good way of doing it, but you don't have to look at the screen.
You can drive it manually, and you can turn the beeping off and all the driver aids if you want.
At least you can on US models, not sure about the UK...

[Gone fishing]

The wiper control is laughable. You can single wipe then use the scroll button left and right through the settings,, but the wonders of OTA means it’s all changed you now do up and down on, oh, and they’ve changed the order so my old way of setting to off, it starts to drizzle, do a single swipe and go to slow intermittent with one click, bit more repeat to fast intermittent etc doesn’t work, I now need to either start in auto which is where the problem of dry wiping or frantic sweeps is or start with them off and scroll to the end of the list.. thanks for that.

The cameras were presumably set up specifically not to be obscured by dirt on the windscreen, be crazy to have cameras responsible for self driving compromised by a bit of dirt or rain so choice of camera focal length, depth of field etc all chosen to see past the glass .. so it’s correct a rain sensor doesn’t detect rain as such, it detects the windscreen needs wiping, exactly what the Tesla system can’t really do,

As for adjusting the sensitivity, I can’t think of a car other than Tesla that isn’t adjustable, maybe there are, but driving yesterday in the rain, I had to both manually turn on the headlights in our Tesla as well as manually control the wipers.. every other car turns on the headlights automatically if the wipers do anything more than the occasional wipe, but not the ‘safest car you can buy’ Tesla, and then when you stop in traffic, the wiper speed doesn’t automatically slow down like our bmw did. Lucky it wasn’t petrol or I imagine I’d have a manual choke too.

Anyway, give superchargers are increasingly open to all (and musk doesn’t value them), lots of other rapids now at motorway service stations, and an incoming iX 50 with a 300 mile range, our Model Y has been chopped in, end of Tesla for me after 8 years of owning one. I simply don’t want to be associated with a brand with Musk at the helm, a design ethos which values fart noises over driving function, and cars which fail on the attention to detail.

[durbster]

skwdenyer said:

As regards hope, no, not at all; merely my analysis of the problem and the range of possible solutions. Conventional sensors have nowhere to go, no development path available to them. Camera-based systems are just starting. As Alan Staniforth liked to say, the last of the old will beat the first of the new - at first... I believe (and you clearly do not) that the camera-based system has better *potential* to improve through software alone. I may be wrong, and don't mind if I am

I used to think that too but years of unfulfilled promises and finding out more about Musk's character has made me deeply sceptical. I won't believe Tesla's claims until I see some evidence.

It seems to me that a system of sensors that are each tuned to do a specific job is more likely to produce a better result than one that tries to do everything, but I'm also happy to be wrong because the other solutions I've tried are far from perfect. The best one I've used was in a Skoda but it still needed adjusting occasionally (thankfully easy to do on the stalk).

I'm just struggling to find a credible basis for your optimism. How can a low-resolution camera figure out the exact rate of rainfall based on the distortion of rain drops falling on the tiny surface area of its lens. Especially when that camera is also the sensor dedicated to analysing its feed to achieve automated driving. They are conflicting requirements so you have to compromise one or the other.

The scenario I find more plausible is much simpler: it's cheaper to build the cars without the other sensors and make hollow promises about the future potential so people will still buy them at the same price.

[skwdenyer]

durbster said:

skwdenyer said:

As regards hope, no, not at all; merely my analysis of the problem and the range of possible solutions. Conventional sensors have nowhere to go, no development path available to them. Camera-based systems are just starting. As Alan Staniforth liked to say, the last of the old will beat the first of the new - at first... I believe (and you clearly do not) that the camera-based system has better *potential* to improve through software alone. I may be wrong, and don't mind if I am

I used to think that too but years of unfulfilled promises and finding out more about Musk's character has made me deeply sceptical. I won't believe Tesla's claims until I see some evidence.

It seems to me that a system of sensors that are each tuned to do a specific job is more likely to produce a better result than one that tries to do everything, but I'm also happy to be wrong because the other solutions I've tried are far from perfect. The best one I've used was in a Skoda but it still needed adjusting occasionally (thankfully easy to do on the stalk).

I'm just struggling to find a credible basis for your optimism. How can a low-resolution camera figure out the exact rate of rainfall based on the distortion of rain drops falling on the tiny surface area of its lens. Especially when that camera is also the sensor dedicated to analysing its feed to achieve automated driving. They are conflicting requirements so you have to compromise one or the other.

The scenario I find more plausible is much simpler: it's cheaper to build the cars without the other sensors and make hollow promises about the future potential so people will still buy them at the same price.

The camera generates a video feed. It doesn't care how many processes - or CPUs - are consuming that feed. The camera doesn't get overloaded by demand.

The camera has to be kept clear in order to operate. At a distance of (say) 46mm from the camera lens to the windscreen, with a field of view of 90 degrees (say), and a horizontal resolution of 1280 pixels (HW3; HW4 is far higher), a rain drop (2mm diameter in flight, more like 5mm minimum on the screen) is occupying about 6 degrees of view, and obscuring 88 pixels. That's material, and should be easy to train an algorithm to detect well.

Since the focal length of the medium-zoom camera (HW3; HW4 drops the widest angle I believe) is 46mm, there's no reason why the raindrop won't be in focus. It will also be obscuring a lot of view. Therefore it *should* be used as an input to trigger the wipers.

As regards nght vision, the distortion caused by water droplets should be sufficient to train by. Whether that work is happening, of course, is a very different question