-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Great, but if you're a Tesla stock owner, you're already funding FSD anyways and getting much better ROI now versus paying for FSD beta with nothing to show for IF you haven't been accepted as a beta tester.
A good faith deposit is a good example. It shows the seller that the buyer is serious and not just kicking the "tires", so to speak.
Speaking as a former software developer interested in Ai since childhood (70s) and now project manager, UX guy, EV lover (about to buy a Model Y LR, EV6 GT Line S or Nissan Ariya E-Force, having test driven each, plus the Model 3, PoleStar 2 and IONIQ 5), thinker, designer, inventor, philosopher and share holder in TSLA as of yesterday, here is my take:
1. Ageing software is often full of legacy code written by previous or current engineers. If the code works, obviously, there is no need to re-invent the wheel, but if it's not optimised or retains some inherent flaws, it needs to be purged, else it drags down the rest of the stack/reputation of product. Attempting to 'fix' broken code, specially that written by another engineer or team can be a pile of hurt, even if well documented. "Eh? Why has he done that? FFS!" Not to mention the whack a mole of fixing one thing and then breaking another. (As is happening now with FSD I believe.) A good reason to develop using modular code, albeit difficult with something like this that is somewhat symbiotic.
2. Ai is not the same as conventional logic based 'software'. And if it's 'real' Ai, it will become sentient, and you will know it before you can say "Singularity", so we're not really dealing with true Ai, just as close as is possible using silicon and neural nets. (Food for thought, a true Ai will posses the same flaws as us and screw up, fall in love and/or refuse to comply.) After all, all true intelligences are organic, mixing biochemical with electrical. When we drive, we behave based on the following: a) Need to stay alive. "Must not look away for more than 0.5 second or let go of the wheel, else vehicle will drift into the central reservation and flip, causing vehicles behind to pile into my fuel tank or battery, possibly starting a fire, and I won't want to burn alive or kill anyone else." And then there is the mother and daughter pushing pram in front of vehicle scenario we all fear. b) Need to obey the Highway Code (UK rulebook) else cause a) or receive a fine/citation. c) React instinctively - "Hmm, I have been here before and know how fast to take this corner and do like to push it sometimes just to get the wife to grip her seat sides as the wheels begin to squeal."
d) Empathic based interaction with other drivers. "Oh, he's wanting to turn right up the street to my left and has been waiting to turn for ages due to the amount of traffic on my side of the road." So you make eye contact, gesture with your hand or flash your lights (illegal in the UK) and he makes the turn in front of you. (Risky because there could be a motorcycle racing down the gap between your car who isn't aware of your gesture so could collide with the side of the turning vehicle and probably why a WORKING Level 4/5 will probably be safer than humans because it isn't 'kind'.) e) Various other combinations of scenario that require our human empathic experience in driving to deal with.
3. Elon for all his well meaning vision, is now dealing with something harder than making a rocket land vertically (never mind 2 at once! - greatest human achievement I have witnessed in my life) and mixing fallible human engineers, price or 'confidence in my ideas' restricted sensors (cameras - YES, LIDAR - NO!, RADAR - NO!, Ultrasonic - NO! etc) with experimenting with real life human owned vehicles until he gets it right, if at all. The last 2-10% of FSD will be the hardest - as I and others have since said before. All great entities eventually discover their Achilles heal, and it's my view that FSD with current Teslas is impossible without a sensor mix, all vehicles talking to each other (Google: VW V2X) and/or overhead cameras monitoring and sending signals to all vehicles in complex environments, reliable radio beacons embedded in some objects/humans and so on. IE, a mix of multiple onboard sensor types & external shared telemetry data. Multiple sensors are what Volvo have announced for their next vehicle, the Polestar 3 I believe.
I am mentally tired now, so ending this comment, but my conclusion is:
4. Tesla should do what Nissan and Kia/Hyundai etc have done and stop referring to any form of vehicle automation as Autopilot or FSD and use "Driver assist", with a list of guaranteed working (except in SPECIFIED circumstances) features. ProPilot on my 2018 Nissan Leaf was very good at following the vehicle in front and made long journeys way less tiring. And it would stop quicker than even my fast reactions if said vehicle in front suddenly stopped. I know it's limitations and had no anger towards Nissan when it would fail to follow the curves in a road because it wasn't sold as 'Autopilot' or FSD. Anyway, this will require some swallowing of pride on Elon's part, but he would earn much respect for stepping back and re-thinking it all. His cars are already superb (all owners I know love them, and most have owned countless cars before hand), but it will take an all new vehicle, designed from the ground up to be autonomous to pull this off, not a regular vehicle originally designed for human drivers experimentally modded during production for future hopes and dreams.
As I said to a close friend, if he pulls of true Level 5/FSD on even the latest Tesla models in the next year or so, I'll eat my hat. Hopefully one made from raisin and almond plain chocolate.
1. Ageing software is often full of legacy code written by previous or current engineers. If the code works, obviously, there is no need to re-invent the wheel, but if it's not optimised or retains some inherent flaws, it needs to be purged, else it drags down the rest of the stack/reputation of product. Attempting to 'fix' broken code, specially that written by another engineer or team can be a pile of hurt, even if well documented. "Eh? Why has he done that? FFS!" Not to mention the whack a mole of fixing one thing and then breaking another. (As is happening now with FSD I believe.) A good reason to develop using modular code, albeit difficult with something like this that is somewhat symbiotic.
2. Ai is not the same as conventional logic based 'software'. And if it's 'real' Ai, it will become sentient, and you will know it before you can say "Singularity", so we're not really dealing with true Ai, just as close as is possible using silicon and neural nets. (Food for thought, a true Ai will posses the same flaws as us and screw up, fall in love and/or refuse to comply.) After all, all true intelligences are organic, mixing biochemical with electrical. When we drive, we behave based on the following: a) Need to stay alive. "Must not look away for more than 0.5 second or let go of the wheel, else vehicle will drift into the central reservation and flip, causing vehicles behind to pile into my fuel tank or battery, possibly starting a fire, and I won't want to burn alive or kill anyone else." And then there is the mother and daughter pushing pram in front of vehicle scenario we all fear. b) Need to obey the Highway Code (UK rulebook) else cause a) or receive a fine/citation. c) React instinctively - "Hmm, I have been here before and know how fast to take this corner and do like to push it sometimes just to get the wife to grip her seat sides as the wheels begin to squeal."
d) Empathic based interaction with other drivers. "Oh, he's wanting to turn right up the street to my left and has been waiting to turn for ages due to the amount of traffic on my side of the road." So you make eye contact, gesture with your hand or flash your lights (illegal in the UK) and he makes the turn in front of you. (Risky because there could be a motorcycle racing down the gap between your car who isn't aware of your gesture so could collide with the side of the turning vehicle and probably why a WORKING Level 4/5 will probably be safer than humans because it isn't 'kind'.) e) Various other combinations of scenario that require our human empathic experience in driving to deal with.3. Elon for all his well meaning vision, is now dealing with something harder than making a rocket land vertically (never mind 2 at once! - greatest human achievement I have witnessed in my life) and mixing fallible human engineers, price or 'confidence in my ideas' restricted sensors (cameras - YES, LIDAR - NO!, RADAR - NO!, Ultrasonic - NO! etc) with experimenting with real life human owned vehicles until he gets it right, if at all. The last 2-10% of FSD will be the hardest - as I and others have since said before. All great entities eventually discover their Achilles heal, and it's my view that FSD with current Teslas is impossible without a sensor mix, all vehicles talking to each other (Google: VW V2X) and/or overhead cameras monitoring and sending signals to all vehicles in complex environments, reliable radio beacons embedded in some objects/humans and so on. IE, a mix of multiple onboard sensor types & external shared telemetry data. Multiple sensors are what Volvo have announced for their next vehicle, the Polestar 3 I believe.
I am mentally tired now, so ending this comment, but my conclusion is:
4. Tesla should do what Nissan and Kia/Hyundai etc have done and stop referring to any form of vehicle automation as Autopilot or FSD and use "Driver assist", with a list of guaranteed working (except in SPECIFIED circumstances) features. ProPilot on my 2018 Nissan Leaf was very good at following the vehicle in front and made long journeys way less tiring. And it would stop quicker than even my fast reactions if said vehicle in front suddenly stopped. I know it's limitations and had no anger towards Nissan when it would fail to follow the curves in a road because it wasn't sold as 'Autopilot' or FSD. Anyway, this will require some swallowing of pride on Elon's part, but he would earn much respect for stepping back and re-thinking it all. His cars are already superb (all owners I know love them, and most have owned countless cars before hand), but it will take an all new vehicle, designed from the ground up to be autonomous to pull this off, not a regular vehicle originally designed for human drivers experimentally modded during production for future hopes and dreams.
As I said to a close friend, if he pulls of true Level 5/FSD on even the latest Tesla models in the next year or so, I'll eat my hat. Hopefully one made from raisin and almond plain chocolate.
Massive correction! “FSD” DID NOT FAIL, the ACRONYM “FSD” FAILS. After the immediate software update to 2022.36.2 on 10/14 and a 350 mile drive to Wilminton, N.C. On the same day, I realized something. I was using and quite successfully, an operator fatigue reducing, and automobile safety guidance system! What should be called the“Integrated Vehicle Operator Navigation System IVONS definitely not FSD! FSD is too many different things for too many people with to many future possibilities. An overloaded acronym with a lot of historical baggage and misinformation attached to it.
The user interface UI has been updated for “IVONS” level functionality with NEW SCREEN functions. This includes ranging support and clarification functions. The ability to turn off the Red Light detection while IVONS is active solved my FAIL experiences. IVONS as I used it reduced cockpit work load. I also realized that asking for larger fonts could quite handily be dealt with by using “ZOOM CENTERED DIALOG BOXES” activated by a finger press OVER a given pixel location In the short term. This could be done polymorphicly by using the existing video blind spot boxes through a different signature! Simple, I know I’m preaching to the quire.
RECCOMENDATIONS:
Drop the FSD ACRONYM, and use a Screen Zoom by finger touch or spoken zoom command. Hard to see areas like the compass rose for lack of a better term in the upper right of the screen is very distracting. The code is their to zoom it just give a different signature to the method.
Elon’s IVONS sounds like a winner and a new marketing clarification long overdue! IVONS should be offered for $10.00 a day on demand with a small membership fee up front.
The user interface UI has been updated for “IVONS” level functionality with NEW SCREEN functions. This includes ranging support and clarification functions. The ability to turn off the Red Light detection while IVONS is active solved my FAIL experiences. IVONS as I used it reduced cockpit work load. I also realized that asking for larger fonts could quite handily be dealt with by using “ZOOM CENTERED DIALOG BOXES” activated by a finger press OVER a given pixel location In the short term. This could be done polymorphicly by using the existing video blind spot boxes through a different signature! Simple, I know I’m preaching to the quire.
RECCOMENDATIONS:
Drop the FSD ACRONYM, and use a Screen Zoom by finger touch or spoken zoom command. Hard to see areas like the compass rose for lack of a better term in the upper right of the screen is very distracting. The code is their to zoom it just give a different signature to the method.
Elon’s IVONS sounds like a winner and a new marketing clarification long overdue! IVONS should be offered for $10.00 a day on demand with a small membership fee up front.
I agree that the confusing FSD term should NOT and NEVER been used BUT....there is no FSD and you like many others are (understandably) confusing and conflating the FSD terms.
FSD Capability (options package)
FSD Computer (aka: HW3 in your car)
FSD Beta (aka: Autosteer on City Streets/Beta invite only and should have beed called Beta Autosteer on City Streets)
Autopilot comes standard but since you subscribed to FSD Capability you are driving on Navigate on Autopilot (NoA) and have access to all FSD Capability features. But you are NOT driving on FSD since there is no such thing.
Changing FSD to IVONS or some other clarifying descriptor should be a high priority UI and Human Factors issue. Because FSD is so broad a term, it’s important to stop using implicit functionality Semantics and replace it with explicit nomenclature. These issues are of great concern to me as I see ideology mixing with TESLA technology, like the all of a sudden dropping of sensor technology. The dropping of ”Tesla Motors”, etc. The impulsivity aspects Twitter introduces into TESLA, the science company. The point is, people in general are beginning to build a bad information conceptual model about the greatest science and technology company to date!
Knightshade
Well-Known Member
Why?
Autopilot in aircraft is literally a driver assist where the human is required to be actively paying attention and prepared to intervene at any time.
Autopilot in a Tesla is exactly the same thing
It's far more accurately named than, say, Fords COPILOT system--- because a copilot can actually operate the vehicle entirely alone and Fords system can not.
Nissans falls more in that camp, with propilot-- a professional pilot doesn't need another human to drive--Nissans system sure does.
This list is in the manual. It's also on the purchase page. The manual lists conditions under which it might not work correctly of course as you also suggest they do despite the fact they already do it- exactly like NIssans manual and everyone elses with such a system does because nothing is "guaranteed" by anybody.
So this criticism sounds like you've never actually read the owners manual on either brand of car.
I read all the words but they seemed to be in the wrong order somehow.
Can’t tell if the complaint is about FSD, or if the compass rose is of the wrong shade, or in the wrong position.
Is it the name that’s wrong or just it’s not using the invented name.
Either way with that release of the software, it’s not FSD beta anyway, so even more confused about what’s being complained about.
Can’t tell if the complaint is about FSD, or if the compass rose is of the wrong shade, or in the wrong position.
Is it the name that’s wrong or just it’s not using the invented name.
Either way with that release of the software, it’s not FSD beta anyway, so even more confused about what’s being complained about.
Last edited:
Jim Perkins and Lifemachine are a bot beta code experiment, imo. It’s all word scramble ramblings.
IamGaryGnu
Endless IBS
you summed up my thoughts of FSD perfectly. I signed up for with my car when it was $10k, then pulled back 2 weeks before delivery at the almost instance of my neighbor. He was right (and had multiple cars' experience with it). He bought the Plaid, and the $$$ wouldn't really affect him, he just felt it was a complete waste. I tried it for a month (I think July into August this summer) and I don't think it was worth the $199 I paid, let alone thousands. I found driving with it petrifying, and yes, i was using "FSD Beta" from the factory, and not "FSD Beta 10.X.X" - so I guess FSD 9, which had no street use, and extremely limited highway functionality.
I did like the lane changes - a lot.
the rest was at times fine, and at times horrific to dangerous. Like accelerating versus coasting to a stop. what if there was ice or the ground was slick? a coast might have been fine, but not accelerating just to hit the brakes hard.
I read a review somewhere of the tester saying he felt like it was a 16 year old getting their permit, and half the time was okay, and half was trying to kill him.
I used AP an awful lot to get my scores above 95 (but I never got the 10.X beta) - and those drives were horrible as well - where i'd have to kick it off before I went off the road, or hit a car, etc.
I think Tesla is further along than anyone else, but I think there's an awfully long way to go.
I have looked into machine learning, and I superficially understand the gist of it. Clearly Tesla believes in it, to build Dojo, etc., but I frankly just don't get it.
I don't see why we can't have AP as our base, and then the "edge cases/exceptions" which the AI is supposed to learn.
I'm not sure why we can't have edge case subroutines, and not need to feed in hundreds of thousands of videos. (Or spend billions on Dojo).
Then again, I'm not the world's richest man, and I'm not designing rockets, so there's probably a whole lot I don't understand about it.
TLDR; I think FSD holds promise. But that promise won't be delivered until at LEAST a few years out, and possibly quite longer. So current gen cars may never see it, like prior gen cars.
You’re a troll. And if it’s not intentional, congratulations, you’re still a troll.
I’m blown away be peoples negativity and unreal expectations.
FSD beta is so extremely, irreplaceably valuable to me. I use it for 95%+ of my driving. I don’t drive anywhere. I may drive manually for some funky turns, but it works. It works extremely well. I drive 150-350 miles a day and I would not do my job if I had to drive myself.
So many obnoxious, whining, entitled people here.
FSD beta is so extremely, irreplaceably valuable to me. I use it for 95%+ of my driving. I don’t drive anywhere. I may drive manually for some funky turns, but it works. It works extremely well. I drive 150-350 miles a day and I would not do my job if I had to drive myself.
So many obnoxious, whining, entitled people here.
There’s a different between being a “believer” and being a user who it works well for.
I don’t believe FSD works, I know it does.
The owner's manual of all the cars I have ever owned are over an inch thick and are only referred to when I need to find the fusebox. I have not RTFM since I started driving in 1981 and never had a problem with my automobile, other than bits failing or falling off. My Dad taught me to drive, which which buttons to fondle and what da blinken lights mean.
Knightshade
Well-Known Member
So.... you insisted they needed to list this stuff in detail....
Then I pointed out they do list all this stuff in detail, and told you where they do.... and you told me you can't be bothered to read it.
Solid argument my dude.
Then I pointed out they do list all this stuff in detail, and told you where they do.... and you told me you can't be bothered to read it.
Solid argument my dude.
CyberGus
Not Just a Member
Manuals are for children! Real Men just wing it, so that they have plausible deniability when they screw up
Just saying, but if the OP updated to 2022.36.2, he doesn't have the FSD beta anyway.
I think he's confused and quite sure of himself.
I think he's confused and quite sure of himself.
