Second thoughts on FSD

[MrBadger]

4. Improving speed control when driving in AP along winding single carriageways. At the moment it is woefully poor at deciding what speed to enter and exit bends. Makes the idea that we are on the verge of proper FSD laughable.

So a use case outside the current functionality. I agree that this needs to be sorted, but using it as a reason why FSD (or whatever portion may or may not be implemented/legislated for) is not going to happen any time soon is a bit like saying that someone is not going to run soon because they have not walked yet - my son could run pretty much at the same time he could walk.

These things can happen suddenly, although 100% FSD and Robotaxi as some are imagining, ie L5, is, imho not going to happen as soon as some are making out, although I think they may get to L4 (which is where I would like FSD to get to in UK) in some jurisdictions within life time of some peoples lease/hire period but for this to happen in Europe, lots of things will need to change outside Tesla's control. I was in an CAV (connected autonomous vehicles) vendors meeting probably 2-3 years back, and it was stated that in Australia, for full autonomy, 10,000 laws would need to be repealed. Limited L4 Robotaxi may well happen, but I'm not convinced they are going to get rid of the driver for quite a while, especially in Europe - and their route will be very limited, but that may well cover a significant proportion of a city journey - it wouldn't be surprised if a robotaxi service started off as a hybrid autonomous/chauffeur service where some journeys could start off autonomous, those ones where taxi progress is slow so driver cost is proportionally high, but then a driver take over to finish off the out of town parts that the car could not do by itself. That would still be a huge cost saving driver wise, albeit not 100% driverless.

imho.

 

[tess19]

I can only admire someone who is able to get 20,000 people to part with $500.00 for "Not a flame thrower"', you can't blame him for trying to get that much for FSD

Or to convince someone to pay 20k for ludicrous mode

 

[WannabeOwner]

Or to convince someone to pay 20k for ludicrous mode

Although:

You try upgrading an ICE from 3.5s to 2.5s 0-60 for less than £20K

 

[Peteski]

So a use case outside the current functionality. I agree that this needs to be sorted, but using it as a reason why FSD (or whatever portion may or may not be implemented/legislated for) is not going to happen any time soon is a bit like saying that someone is not going to run soon because they have not walked yet - my son could run pretty much at the same time he could walk.

Another way of looking at it is that it's an easy excuse to say that it is "outside current functionality" wherever it falls down. Why does it even bother to attempt speed control at all if it is outside its remit? Why does it even allow AP at all on such roads? Having driven with EAP for 25k miles now, I have a pretty good idea of its limitations and have witnessed it's snail pace development over the last 18 months, with Elon tweeting about FSD being just around the corner every couple of months along the way. Of course the definition of FSD now appears to have been watered down to "FSD features" rather than actual FSD.

So I don't personally think it will suddenly take some massive leap forward. In fact I believe every step from here to actual FSD becomes progressively more difficult to achieve as you rely less and less on human input. The leap from current AP to a system that could safely and efficiently drive itself through a busy uk city is truly enormous. Make whatever largely irrelevant analogies you like, it's a huge task ahead that nobody has really come close to nailing yet. If it was easy, the tech giants would have cracked it by now before the likes of Tesla. But you have to hand it to Elon for hyping up Tesla's FSD game! There are plenty of people around who already think my car can drive itself!

 

[WannabeOwner]

There are plenty of people around who already think my car can drive itself!

That may be my fault I do casually drop into conversation "Of course it drives itself most of the time" ... apologies to anyone here who might have been at the dinner table when i was there

 

[Yev000]

I recon actual "FSD" will go up in price - when it actually works, but you will be able to get things like auto lane change back in with regular AP.

People will vote with their wallet, and Elon needs a lot of cars with "FSD" to get those neural networks to improve, if only say 20% of the buyers cough up they will lower the price or move features around to get more cars to teach the neural network.

Now all cars TACC for free, before it was an option.

I'm sure lane change on TACC will be free in a year or two. They will call it AP+ or something so that "FSD" stays more expensive as promised.

Ultimately the pipe dream of having a robo-taxi is a software lock away, you could have full self driving for yourself, personal use only, but to join Tesla taxi fleet he could charge 50k for a 'license'. multiple ways to skin that cat. But I think they will try and get more cars on FSD to teach the AI which will keep prices down in the near term.

 

[interbear]

I think I must be a rare soon to be Tesla owner who couldn’t actually give a damn about FSD. I’m buying my Model 3 for the zero emissions goodness, range and driving experience, which all reviews are positive about. I don’t really want to pay thousands to delegate that driving experience to a piece of software that will probably work, most of the time. So yes, there’s a trust issue there also. I’ll be happy enough with Auto Pilot which seems to me to be clever cruise control. That’s quite enough automation for me.

Equally, after paying close to £50k for a car I’ll be damned if I let it out of my sight as a “robo taxi” when I’m not using it. No chance. And I think that’s years away in any case. Even if the tech can surprise us in terms of how quickly it gets to market, I think it’ll be so tied up in regulatory and safety issues with governments worldwide that it’s actual introduction will be much later.

 

[Peteski]

I’ll be happy enough with Auto Pilot which seems to me to be clever cruise control. That’s quite enough automation for me.

It's actually a very clever cruise control that can also steer safely and effectively on motorways and dual carriageways. The only thing it's missing (deliberately to encourage buyers to pay another £6k for FSD) is the auto-lane change function (where you still have to indicate manually and FSD then steers into the next lane for you). Having to steer manually across lanes is no big deal (actually preferable in many dynamic situations), but it does mean you have to disengage auto-steer every time you need to change lanes. That wouldn't even be a problem if it wasn't for the annoying bing-bongs and media volume suppression you get each time! Fortunately if you just stay in TACC (Traffic Aware Cruise Control) without auto-steer you don't get the bing-bong warnings and are free to change lanes without switching TACC off. Once you get onto a long stretch of motorway in the same lane you can just flick on auto-steer for full AP.

The fact that Tesla feel the need to cripple AP by excluding simple auto-lane change is a sign that FSD is still not very far ahead and they are now scraping the barrel to make some meaningful differentiation.

 

[Roy W.]

It's actually a very clever cruise control that can also steer safely and effectively on motorways and dual carriageways. The only thing it's missing (deliberately to encourage buyers to pay another £6k for FSD) is the auto-lane change function (where you still have to indicate manually and FSD then steers into the next lane for you). Having to steer manually across lanes is no big deal (actually preferable in many dynamic situations), but it does mean you have to disengage auto-steer every time you need to change lanes. That wouldn't even be a problem if it wasn't for the annoying bing-bongs and media volume suppression you get each time! Fortunately if you just stay in TACC (Traffic Aware Cruise Control) without auto-steer you don't get the bing-bong warnings and are free to change lanes without switching TACC off. Once you get onto a long stretch of motorway in the same lane you can just flick on auto-steer for full AP.

The fact that Tesla feel the need to cripple AP by excluding simple auto-lane change is a sign that FSD is still not very far ahead and they are now scraping the barrel to make some meaningful differentiation.

The implementation on my Kona is better as far as I’m concerned. It is TACC plus LKA, and I believe it’s a Mobileye system that Hyundai have bought in, so very similar to AP1 hardware.

I know what you mean about the annoyance of disengaging AP to change lanes, I experienced it during my test drive on Monday. On the Kona, indicating to overtake disengages the LKA, but as soon as you are in the new lane it automatically re-engages, so it’s a lot more seamless than the Tesla AP. Don’t get me wrong though, the Kona has plenty of quirks of its own, and I’m sure the Tesla will be far better!

 

[Peteski]

The implementation on my Kona is better as far as I’m concerned. It is TACC plus LKA, and I believe it’s a Mobileye system that Hyundai have bought in, so very similar to AP1 hardware.

I know what you mean about the annoyance of disengaging AP to change lanes, I experienced it during my test drive on Monday. On the Kona, indicating to overtake disengages the LKA, but as soon as you are in the new lane it automatically re-engages, so it’s a lot more seamless than the Tesla AP. Don’t get me wrong though, the Kona has plenty of quirks of its own, and I’m sure the Tesla will be far better!

Your Kona is more in line with other manufacturer systems like VW. Unfortunately Tesla AP without auto-lane change is a poor implementation. It would be extremely easy for Tesla to tweak the software to allow manual lane change without fully disengaging auto-steer (just like in your Kona), but they are crippling it deliberately to up-sell FSD.

 

[Durzel]

Another way of looking at it is that it's an easy excuse to say that it is "outside current functionality" wherever it falls down. Why does it even bother to attempt speed control at all if it is outside its remit? Why does it even allow AP at all on such roads? Having driven with EAP for 25k miles now, I have a pretty good idea of its limitations and have witnessed it's snail pace development over the last 18 months, with Elon tweeting about FSD being just around the corner every couple of months along the way. Of course the definition of FSD now appears to have been watered down to "FSD features" rather than actual FSD.

I don't really understand why the terms "Autopilot" and particularly "Full Self Driving" are used at all.

Another way of looking at it is that it's an easy excuse to say that it is "outside current functionality" wherever it falls down. Why does it even bother to attempt speed control at all if it is outside its remit? Why does it even allow AP at all on such roads? Having driven with EAP for 25k miles now, I have a pretty good idea of its limitations and have witnessed it's snail pace development over the last 18 months, with Elon tweeting about FSD being just around the corner every couple of months along the way. Of course the definition of FSD now appears to have been watered down to "FSD features" rather than actual FSD.

So I don't personally think it will suddenly take some massive leap forward. In fact I believe every step from here to actual FSD becomes progressively more difficult to achieve as you rely less and less on human input. The leap from current AP to a system that could safely and efficiently drive itself through a busy uk city is truly enormous. Make whatever largely irrelevant analogies you like, it's a huge task ahead that nobody has really come close to nailing yet. If it was easy, the tech giants would have cracked it by now before the likes of Tesla. But you have to hand it to Elon for hyping up Tesla's FSD game! There are plenty of people around who already think my car can drive itself!

A lot comes from the shrewd (or irritating, depending on your perspective) use of "Autopilot" and "Full Self Driving" for things that are not that at all, especially the latter.

I get why it's done - "autopilot" is something almost everyone will understand, they've heard of autopilot on planes (which nowadays is fully autonomous flight with a pilot just sat ready to intercept). But "autopilot" in a Tesla is not the same thing at all. "Assistive driving" is more accurate, but nowhere near as cool.

"Full Self Driving" is way beyond suggestive though, it's false advertising. One day it will become that, but the whole "my car can drive itself" is several years away both in technological/deep learning and legislation, but for the forseeable future it's not that at all.

 

[Durzel]

Well, I messed that post up and can't edit it

 

[davidmc]

Wonder when Toyota come out with the full self driving self charging hybrid!

 

[MrBadger]

Actually autopilot in planes is something that most people misunderstand, hence confusion as to what people think Tesla autopilot does. Its actually pretty close and if anything, Tesla autopilot does more than on a plane because it often bundles more features under the single term.

An autopilot is a system used to control the trajectory of an aircraft without constant 'hands-on' control by a human operator being required. Autopilots do not replace human operators, but instead they assist them in controlling the aircraft. This allows them to focus on broader aspects of operations such as monitoring the trajectory, weather and systems.[1]

The autopilot is often used in conjunction with the autothrottle, when present, which is the analogous system controlling the power delivered by the engines.

I agree partly about FSD though as it doesn't actually specify what level autonomy it is so can be interpreted in many ways. Especially with robotaxi thrown into the mix, people assume Level 5, ie fully driverless, but my guess is that depending on where, its going to be a mix of level 3 initially (possibly quite soon) then level 4 (in an ever increasing set of controlled environments) and a very broad interpretation of level 5.

 

[tess19]

Autopilot in the air is completely differently to city streets and highway with idiots on the road

 

[Roy W.]

Autopilot in the air is completely differently to city streets and highway with idiots on the road

Perhaps we all need to be talking to “Road Traffic Control”

 

[Peteski]

I agree partly about FSD though as it doesn't actually specify what level autonomy it is so can be interpreted in many ways.

"Full Self-Driving" can really only be interpreted in one way. But I have noticed that Tesla are not as bold today as they used to be in advertising what FSD can actually achieve. Although the descriptions of its features are still a bit misleading for uk driving. There is no fully automatic lane change and your car will not come and find you anywhere in a carpark!

 

[electricspace]

People will vote with their wallet, and Elon needs a lot of cars with "FSD" to get those neural networks to improve, if only say 20% of the buyers cough up they will lower the price or move features around to get more cars to teach the neural network.

But I think they will try and get more cars on FSD to teach the AI which will keep prices down in the near term.

It's my understanding that the training happens with every mile travelled by a Tesla. Older variants will perhaps contribute less data than the newer FSD enabled models can.

 

[fuzzywuzzy]

I can't see the point is spec'ing FSD either, sure if it's just a tick-box option for someone that doesn't care about another £5k here or there but otherwise it really doesn't make a lot of sense to me. Maybe if you were a gambler, purchasing out-right and intending to keep the car for 10+ years it might work out in the end but otherwise I'd just save the money.
The YouTube clips of yanks thinking their car is going to start earning them $300k a year from next year are comedy gold

 

[Peteski]

The YouTube clips of yanks thinking their car is going to start earning them $300k a year from next year are comedy gold

That's when you start moving into the realms of blind faith and fantasy, lol!