This is probably true. But the company that develops the level 3 and beyond computer vision that controls the car will initially probably not be the same as the maker of the car and its components. I would rather Tesla partner with a separate company, or buy out the company when the time is right. There are no guarantees that Tesla vision will ever achieve level 3 to an acceptable safety level. And I would rather decouple the fate of Tesla from the fate of Tesla vision.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Autopilot nag
- Thread starter alfagiulia
- Start date
Knightshade
Well-Known Member
"similar"
I mean- yeah... a BMW 3 series is "similar" to a Honda accord... they both use gas, they both offer a 6 cylinder with 4 doors...
but they're pretty damn far from the same.
Caddys system for example- is only available on one model, it's not one I'd want to own, and it only works on a very specific set of pre-mapped roads. And only will ever work on those.
Lexus for another example offers all-speeds radar cruise- but unlike the Tesla system what happens in actual use is it follows too far behind at even the "nearest" setting- so you're constantly being cut off- and then the car slows down even MORE because now someone is too close in front of you...so then you get cut off again and it slows down more... and so on. It's "similar" to TACC- but in actual use it's crap in comparison.
The Mercedes system does more than the Lexus one, but doesn't do it nearly as well as Teslas system (and has the same issue with letting everyone cut you off)
So...similar. But worse.
Except for the caddy system they're generally so much worse they're not really worth using. Teslas is.
Caddys is actually FANTASTIC...arguably better than Teslas as long as you're on a mapped road where it works.
But the other 90% of US roads it doesn't work on are roads I use occasionally... so it loses to Tesla there too.
Last edited:
@alfagiulia read this. what @Knightshade described is my exact experience but with infiniti, and toyota. the features are advertised and available (from other manufacturers) but they're FAR from tesla (incredibly functional) implementation. if tesla is a experienced student driver, everyone else is standing in line waiting for driving school to start. Supercruise and pro-pilot being notable exceptions and could be compared as another student driver.
I don't disagree. The Tesla ADAS is probably slightly better than any other right now. But is it much better now than when Tesla still partnered with Mobileye? And who will have the best system in the future? Tesla collects a lot of data, but do they have the best algorithms and hardware components?
I live in Germany. The German automakers are very risk averse. They are developing the tech. but won't deploy it until it's really safe. The Germans are great at automation and hardware, but fortunately not quite as great with software. If they manage to achieve par or get ahead then all the AP fanboys are going to have to look somewhere other than Tesla for state of the art autonomy. Tesla can't play catchup because of their competitive disadvantage in the auto industry. They HAVE TO stay ahead. I hope they pick their battles wisely.
chronopc
Active Member
Good info. Do you have any knowledge of Audi's offering? Someone on this thread said it was close to parity with EAP.
I think this answers at least some of your questions. Do they have the algorithm? They say they do now. At this point, they're working to implement it based on collecting and processing of data from each of their cars. Hence Elon saying we will see massive improvements this year (I.e. because it's easier to train system than rewrite algorithm).
Article: Tesla’s AI director gives insights into Autopilot’s computer vision and neural net development
Too bad Tesla wasted so much money developing ADAS. Not much to show for it at this point. Maybe they wouldn't have had to lay off so many people if they had stuck to building great cars and charging infrastructure. They'll probably need another round of financing next year too.
I'm driving 21.9 and nags are perfect as long as I hold steering wheel and pay attention. I get fewer nags now vs 18.13
*The lightest of torque application, not the lightest of touches. I rest my hands on the steering wheel, but still get nags very frequently, since apparently Tesla thinks I don't know how to hold a steering wheel.
Is it just me, or have the rest of you gotten fed up with the crap alfagulia continuously spews in this forum?
Why are you even here? If you don't like Tesla, fine. Keep driving your alfa and go away.........
Why are you even here? If you don't like Tesla, fine. Keep driving your alfa and go away.........
I'm surprised everyone has been so reasonable about not hitting that disagree button on those posts. I once posted I didn't like model 3 steering wheel and got disagreed into Oblivion. They must have thought I liked the steering wheel on model 3. It turned out ok now that I'm using it but could have been much nicer steering wheel.
chronopc
Active Member
My wife is kind of the same. But with Auto Pilot, you can turn it off if the nagging is too much. With the wife, um....
Elon... It's not quite the crux. It's also a lack of user system knowledge. If people understood AP confidence levels in different driving conditions, they'd pay more attention as conditions dictate it.
For example (SWAG),
Straight Highway, centerlane = 99.99% confidence.
Curvy highway @ 55 mpg = 95%.
Curvy highway @ 75 mph = 90% as it may pop out on the turn for excessive wheel turn at given speed!
Curvy highway @ 75 mph = 90% with a dip on a curve = 70% because it has to correct for the dip causing the steering to exceed the limit (and it will do it every time repeatedly).
Giving no indication of AP confidence levels will teach most techno-peasants that it's all equally safe which couldn't be further from the truth. I think people need to understand what's going on. That way, if they want to play the lottery they know their odds. The software has an aggregate confidence number at all times (otherwise it wouldn't know when to slow or when not to allow AP). Either display that number, or teach people through this phase of development.
One way to explain how this is safer is to also show statistics of the same conditions for the average human. Might be difficult to come up with this second set of numbers, but you get the idea maybe pick common areas. But just saying "AP is safer than human driver" is so shallow. At present, there are very few cases where the car can drive better than this driver, IMO. However, it's way better at other things like peripheral awareness and response time. Explain that too. I'm still amazed at how our M3 brakes before I see the brake lights in front. That's a clear radar advantage. And explain the sensors too. Explain how fog is better... how?
I would even recommend getting certified before using it. Isn't that how you do it in the factory? Or do you just suit up and put people to work around all those robots? Why is the customer different here?
For example (SWAG),
Straight Highway, centerlane = 99.99% confidence.
Curvy highway @ 55 mpg = 95%.
Curvy highway @ 75 mph = 90% as it may pop out on the turn for excessive wheel turn at given speed!
Curvy highway @ 75 mph = 90% with a dip on a curve = 70% because it has to correct for the dip causing the steering to exceed the limit (and it will do it every time repeatedly).
Giving no indication of AP confidence levels will teach most techno-peasants that it's all equally safe which couldn't be further from the truth. I think people need to understand what's going on. That way, if they want to play the lottery they know their odds. The software has an aggregate confidence number at all times (otherwise it wouldn't know when to slow or when not to allow AP). Either display that number, or teach people through this phase of development.
One way to explain how this is safer is to also show statistics of the same conditions for the average human. Might be difficult to come up with this second set of numbers, but you get the idea maybe pick common areas. But just saying "AP is safer than human driver" is so shallow. At present, there are very few cases where the car can drive better than this driver, IMO. However, it's way better at other things like peripheral awareness and response time. Explain that too. I'm still amazed at how our M3 brakes before I see the brake lights in front. That's a clear radar advantage. And explain the sensors too. Explain how fog is better... how?
I would even recommend getting certified before using it. Isn't that how you do it in the factory? Or do you just suit up and put people to work around all those robots? Why is the customer different here?
No point explaining anything. They detail most of it in the manual. But people don’t read it then complain about things they don’t understand or make stupid videos doing stupid things which the press inflates and we end up with annoying restrictions like extra nags.
Remember the videos of creative people getting into the back seat etc. ??
Tesla so called fans have done this to themselves then complain about it.
Remember the videos of creative people getting into the back seat etc. ??
Tesla so called fans have done this to themselves then complain about it.
charliepmayer
Member
The wheel is looking for force trying to effect r or l steering, not simply a touch.
The system cannot recognize any grip, it only recognizes resistance to steering wheel movement.
This is poorly described in Tesla literature.
Any force that tries to turn the wheel will be recognized and kill the nags:
*a weight at the 3 o'clock or 9 o'clock position. Your hand works well for this.
*a rubber band or bungee cord trying to pull the wheel in a clockwise or anti clockwise position
*your knee applying slight pressure trying to deflect the wheel right or left
*your suggestion here
If the force is too high it will disable Autosteer and wake up your passengers.
If the force is too low you will get the nag.
The system cannot recognize any grip, it only recognizes resistance to steering wheel movement.
This is poorly described in Tesla literature.
Any force that tries to turn the wheel will be recognized and kill the nags:
*a weight at the 3 o'clock or 9 o'clock position. Your hand works well for this.
*a rubber band or bungee cord trying to pull the wheel in a clockwise or anti clockwise position
*your knee applying slight pressure trying to deflect the wheel right or left
*your suggestion here
If the force is too high it will disable Autosteer and wake up your passengers.
If the force is too low you will get the nag.
Where is AP % confidence explained anywhere? Who actually knew that a bump in a turn could kick it out of AP? I had to discover that myself which is what we're all doing - randomly. Some are not so good at it.
I'm going to draw from experience on a problem I saw at Intel. I gathered human related excursion data x-factory for 3 years to create a pareto of "training" issues on a single operation. Long story short, people knew they weren't supposed to do things, but they were never told why. When reasons were explained, there was a measurable upside in yield as reported by Production. Most Technicians cut corners to save time, but had no clue how this impacted yield as pre/post testing showed. It was not a controlled experiment, and true, they knew people were watching them... but I'm a firm believer in an understanding of things and not just blind rules. The only question should be to balance safety and knowledge against Intellectual Property. But I doubt there's anything new here until Tesla creates its own silicon. It's all OpenAI.
Fully agree with you - but sadly those of us who want to understand how things work are in the diminishing minority
I learned to drive outside of the US, so I learned to drive properly. By that I mean by driving around corners, understanding how the car works, what over/under-steer is and more importantly what each feels like. So I love finding out what its doing.
For the majority though they just attribute magical capabilities to AP and believe all the press headlines that constantly proclaim another "Tesla self driving car" in an accident.
sadly those of us who want to understand how things work are in the diminishing minority
Everything can be made simple. Don't give up on Humans.
Similar threads
