V11 - what’s your verdict?

[Nick61]

A slider control to adjust the auto wiper’s sensitivity would solve the problem.

Most other cars cars I’ve owned have had 3 or more sensitivity settings.

 

[Nick61]

I can confidently say that the V11 UI was not designed by a 60 year old with varifocal lenses & an astigmatism who’s daily commute includes 15 miles of unlisted & B roads.

 

[MrT3]

A slider control to adjust the auto wiper’s sensitivity would solve the problem.

Most other cars cars I’ve owned have had 3 or more sensitivity settings.

It would certainly help but I'm not sure it will completely solve the issue. I notice a big difference between daylight and night conditions for example which I never get in other vehicles so you might need two sensitivity settings.

I also experience the Tesla being incredibly slow to react to light rain/spray, to the point of being not able to see before it wipes, and then a few moments later as you pass a truck on the motorway the wipers go on full speed and continue to do so for 60 seconds after the conditions have cleared - so it is both not sensitive enough and too sensitive at the same time. Again not something I experience with other vehicles and not something I can see a simply sensitivity setting would solve.

 

[hingus2000]

The more I think about it the more not having an option to view an odometer while driving is utterly baffling. I’ve never ever driven a car without a visible trip computer. For Tesla to take this away and bury it in a sub-menu is bizarre.

A smaller point - but they’ve taken the compass off the map when navigating to a destination. It still appears when no designation is selected. Just seems a weird bifurcation. I used to enjoy occasionally glancing/confirming our heading against the compass.

 

[Teddy69]

With the wipers it's not unusual to need to press a button/lever to get an extra wipe, that's easily accessed.

What's really difficult is when it decides to do a few minutes of wiping with no rain, causing bad smearing and so on. So you turn the things off then next day get a pretty angry spouse who doesn't have a clue how to put the wipers back on auto.

 

[Wol747]

FWIW my take is that the wipers have got better this last few months at detecting rain. Unfortunately they are far too sensitive, and the scraping of rubber over dry glass offends my ears so I end up switching them off and giving a quick stalk press every few seconds.
Not cutting edge, but it works.
Regarding the comment from our friend in Cambridgeshire, I think we are all aware just how complex automatic driving is (I'm beginning to understand that the folks in Fremont are only just beginning to understand it, too) but I don't recall any vehicular advance that's been made in the past where potential buyers have not only had to pay years in advance to develop the system but have become unpaid - and possibly uninsured - guinea pigs for it!

 

[pgkevet]

I'd guess that trying to develop a driving AI that's significantly better than a human using a vast swathe of data from a pool of drivers is inherently doomed. Considering that half those drivers are below average and Tesla aim to be 10x better then they really should only use the data from the top few percentiles - those with special advanced training. At the very least they should weed out all those whose cars record speeding, fast cornering, hard acceleration or hard braking and toss them from the pool.

 

[spdpsba]

FWIW my take is that the wipers have got better this last few months at detecting rain. Unfortunately they are far too sensitive, and the scraping of rubber over dry glass offends my ears so I end up switching them off and giving a quick stalk press every few seconds.
Not cutting edge, but it works.
Regarding the comment from our friend in Cambridgeshire, I think we are all aware just how complex automatic driving is (I'm beginning to understand that the folks in Fremont are only just beginning to understand it, too) but I don't recall any vehicular advance that's been made in the past where potential buyers have not only had to pay years in advance to develop the system but have become unpaid - and possibly uninsured - guinea pigs for it!

You know you still have a choice if you have not committed your next one to be a Tesla

 

[GRiLLA]

FWIW my take is that the wipers have got better this last few months at detecting rain. Unfortunately they are far too sensitive, and the scraping of rubber over dry glass offends my ears so I end up switching them off and giving a quick stalk press every few seconds.
Not cutting edge, but it works.
Regarding the comment from our friend in Cambridgeshire, I think we are all aware just how complex automatic driving is (I'm beginning to understand that the folks in Fremont are only just beginning to understand it, too) but I don't recall any vehicular advance that's been made in the past where potential buyers have not only had to pay years in advance to develop the system but have become unpaid - and possibly uninsured - guinea pigs for it!

Remember those terms you voluntarily accepted when turning on AutoPilot on a new car.

I'd guess that trying to develop a driving AI that's significantly better than a human using a vast swathe of data from a pool of drivers is inherently doomed. Considering that half those drivers are below average and Tesla aim to be 10x better then they really should only use the data from the top few percentiles - those with special advanced training. At the very least they should weed out all those whose cars record speeding, fast cornering, hard acceleration or hard braking and toss them from the pool.

This is exactly what's happening with the FSD Beta in the US, drivers have to achieve a high 'Safety Score' and maintain that to be allowed to be part of the Beta. The Safety Score is determined from your last month of driving. People need to have pretty much avoided :-
- Forward Collision Warnings
- Hard Braking
- Aggressive Turning
- Unsafe Following
- Hands on the Wheel

Realistically though, that's largely about Tesla maintaining safety for it's Beta and avoiding headlines from driving morons. The car isn't learning directly from your driving, the car has no learning capability. Tesla AP Team tell the cars the kind of sights and scenarios they are interested to collect and the cars provide them with clips. These are then reviewed by people and systems to label them to train the vision model to better know what it's seeing. The side of the AI that determines what it's going to do based on what it sees is different, and from what I recall isn't trained by people's actual driving, there is no need, it's easy to determine what is the right action to take.

 

[Wol747]

You know you still have a choice if you have not committed your next one to be a Tesla

Yes, I do - there's plenty of competition either here or coming soon but I'm of a "certain age" and can't see myself buying another car in this life.

 

[pgkevet]

Realistically though, that's largely about Tesla maintaining safety for it's Beta and avoiding headlines from driving morons. The car isn't learning directly from your driving, the car has no learning capability. Tesla AP Team tell the cars the kind of sights and scenarios they are interested to collect and the cars provide them with clips. These are then reviewed by people and systems to label them to train the vision model to better know what it's seeing. The side of the AI that determines what it's going to do based on what it sees is different, and from what I recall isn't trained by people's actual driving, there is no need, it's easy to determine what is the right action to take.

I assumed, presumably erroneously, that AI would be making its determination of what to do and then 'learn' from the difference in it’s action and that of a human (safe) driver. or at least that would be my naive approach to teaching an AI.
The obvious faults at the moment are a lack of looking far enough ahead and anticipating bends and traffic instead of being reactive to near instant situations and a complete failure to integrate that into things like headlight function - which is why headlights dip on chevrons, roadside reflectors and parked unattended vehicles and why they reman dipped for 3 secs even when there is no new on-coming.

 

[pdk42]

I still think using a handful of cheapo web cams, exposed to the elements, is a flawed strategy for controlling a car at typical road speeds - even with a sufficiently developed AI computer model (which is also some way off). Bottom line, FSD is still a gimmick to impress friends, but with limited practical use.

 

[Wol747]

I still think using a handful of cheapo web cams, exposed to the elements, is a flawed strategy for controlling a car at typical road speeds - even with a sufficiently developed AI computer model (which is also some way off). Bottom line, FSD is still a gimmick to impress friends, but with limited practical use.

I wouldn't put it quite as negatively as that, but the "easy" parts have largely been done and dusted (Motorways etc).
It may appear from the myriad videos of beta that city driving has also been solved, but most of them have little traffic and are (obviously) in the US with very different roads and traffic behaviour.
I think Tesla is coming up against an insuperable problem of trying to fit in with high traffic density environments with millions of humans who can negotiate with gestures and flashing lights and who can know when they have to get out of a situation. We do this sort of thing without even thinking but so-called AI is not up to it IMO. I would be surprised if FSD as we understand it is accomplished any time soon although I would like to be wrong.

 

[pdk42]

I wouldn't put it quite as negatively as that, but the "easy" parts have largely been done and dusted (Motorways etc).
It may appear from the myriad videos of beta that city driving has also been solved, but most of them have little traffic and are (obviously) in the US with very different roads and traffic behaviour.
I think Tesla is coming up against an insuperable problem of trying to fit in with high traffic density environments with millions of humans who can negotiate with gestures and flashing lights and who can know when they have to get out of a situation. We do this sort of thing without even thinking but so-called AI is not up to it IMO. I would be surprised if FSD as we understand it is accomplished any time soon although I would like to be wrong.

If they could just do motorway/highway driving without the endless phantom braking and without needing to touch the wheel every 15 seconds then I'd be happy. City driving seems a long way off.

 

[Timcat]

I don't think Tesla has any excuse for not having effective automatic headlights and wipers; the technology to do both is tried, tested and available off the shelf at a low price. The same applies to TACC or adaptive cruse control (using well established radar based systems would eliminate phantom braking almost entirely).

As for the development of AI for FSD. It doesn't wash: we bought cars, not places in a software development program.

Having said this, I do think that the lights and wipers are better with the latest update, but phantom braking is unpredictable, frankly dangerous, and renders Autopilot only safe to use in light traffic.

 

[Doudeau]

Apologies if I missed this already being raised when I was reading this thread.

Yesterday, I put on the indicator moments after there was an alert for an incoming call. The blind spot camera view totally covered the pop up window showing who was calling and the options for taking or declining the call. I’m guessing that scenario wasn’t tested (either). I needed to back out of the overtake to turn off the indicator and take the call. Is there a workaround?

Over the last 2+ years of Tesla ownership I’ve really enjoyed the evolution of the FSD preview screen - I recall driving slower on the next “bin day” to see them being rendered on the screen. But now, I’ve concluded that I don’t believe the FSD preview content offers me any benefit. If it rendered cars in the blind spot then maybe. Am I missing a good reason why Tesla think it’ll be useful to see that content full time and not rely on audio warnings and a pop-up message when alerting for danger? Right now I’d happily lose that content to get trip data back and the blind spot camera higher up on the screen.

 

[spooksman]

I needed to back out of the overtake to turn off the indicator and take the call. Is there a workaround?

If it helps, you can swipe down on the blind spot display to get rid of it, leaving the call display...

You can actually swipe down on it anytime to remove it if you want.

 

[MrT3]

I don't think Tesla has any excuse for not having effective automatic headlights and wipers; the technology to do both is tried, tested and available off the shelf at a low price. The same applies to TACC or adaptive cruse control (using well established radar based systems would eliminate phantom braking almost entirely).

A traditional car will have a sensor for AC solar loading, one for auto lights, another for auto wipers etc. and each one controlled by a separate autonomous ECU made by a different OEM. Tesla's philosophy is to have a single system operating the vehicle, using as few sensors as possible for multiple tasks - making the whole vehicle software upgradable to do things that weren't envisaged when first designed.

The exception was radar, as they saw it's advantages in being able to 'see' in low visibility so it was in addition to vision. But either the computing power or their software couldn't cope with conflicting information from vision and radar, what Tesla is calling 'noise', and so they are doing away with radar - which is a shame because in the future the vision system could become as aware as a human, but not necessarily better.

So unfortunately they are redesigning the wheel on a lot of these systems which is frustrating for owners because they typically come from other vehicles where these systems work almost perfectly, and the speed at which these systems are improving appears slow to non-existent because it seems Tesla put their development focus on full self-driving in nice controlled, wide, well sign posted American streets, rather than perfecting the basics.

 

[Timcat]

A traditional car will have a sensor for AC solar loading, one for auto lights, another for auto wipers etc. and each one controlled by a separate autonomous ECU made by a different OEM. Tesla's philosophy is to have a single system operating the vehicle, using as few sensors as possible for multiple tasks - making the whole vehicle software upgradable to do things that weren't envisaged when first designed.

Agreed, but why not use the tried and tested sensors with a single operating system? I can't see why this should be too difficult for a company like Tesla, given the well known natures of the sensors concerned.

 

[Nick77]

Agreed, but why not use the tried and tested sensors with a single operating system? I can't see why this should be too difficult for a company like Tesla, given the well known natures of the sensors concerned.

Because Tesla aren't interested in using products other than their own. They want to use their own AI to control every aspect of the car.