Autonomous Car Progress

[diplomat33]

I am confused why people say that Tesla FSD "works" anywhere, let alone everywhere. I have not heard of any place it is said to be even close to working.

When people say that FSD Beta works everywhere, I think they simply mean that it can be used everywhere. They are not looking at how good or bad it is, just that it can be used everywhere.

 

[bradtem]

When people say that FSD Beta works everywhere, I think they simply mean that it can be used everywhere. They are not looking at how good or bad it is, just that it can be used everywhere.

Sure, but so can cruise control. So I am being a bit rhetorical in that I know they think that the fact you can turn it on and get some basic stuff out of it everywhere somehow implies it works everywhere, except for the fact it doesn't work.

I believe this confusion comes because Autopilot (ADAS) is super different from self-driving. With ADAS, you judge it by what it's able to do, like "Look, it was able to do this turn!" That makes sense because it is not intended to do the whole task, and you are there to do the things it can't do. So Autopilot works on all highways, even though you have to watch and intervene.

With a self driving system, "works" means it can actually self-drive. That you could be asleep. You judge a self-driving system on what it can't do, not what it can. If it does the turn 999 times and fails it once, then it can't do the turn. That's a bit foreign to those not used to the problem, but they have to learn to adjust the vocabulary. As such there is nowhere that Tesla FSD works as a self-driving system, and nowhere that it is even remotely close to being able to see a self-driving system far off in the distance. So yes, it tries to do that everywhere.

MobilEye in their demonstration of a nearly-working self-driving system, sent 2 people and a car to Munich for 2 weeks, and it was nearly working there too, or so they claim. That's a better claim on being close to working everywhere, if they can really pull that off. Well, everywhere that there are cars driving the roads with EQ4s, which is much of the world.

FSD doesn't work everywhere, it fails everywhere. But something makes people not understand the difference.

 

[FloridaJohn]

Humans don’t have had maps they have very sparse, imprecise maps. I can’t tell you with cm resolution what the wife every road or lane I drive daily. I can’t even do that with my bedroom.

It's an analogy, not a comparison.

The point is that you are a better human driver once you have been to a location, and an even better human driver if you have driven a route many times.

HD maps for autonomous cars are like having all that knowledge "pre-loaded" without first having to drive every route 1000 times.

Here comes another analogy: Think of it like Neo in the Matrix when he gets downloaded with knowledge of King-Fu.

 

[OncomingStorm]

Sure, but so can cruise control. So I am being a bit rhetorical in that I know they think that the fact you can turn it on and get some basic stuff out of it everywhere somehow implies it works everywhere, except for the fact it doesn't work.

I believe this confusion comes because Autopilot (ADAS) is super different from self-driving. With ADAS, you judge it by what it's able to do, like "Look, it was able to do this turn!" That makes sense because it is not intended to do the whole task, and you are there to do the things it can't do. So Autopilot works on all highways, even though you have to watch and intervene.

With a self driving system, "works" means it can actually self-drive. That you could be asleep. You judge a self-driving system on what it can't do, not what it can. If it does the turn 999 times and fails it once, then it can't do the turn. That's a bit foreign to those not used to the problem, but they have to learn to adjust the vocabulary. As such there is nowhere that Tesla FSD works as a self-driving system, and nowhere that it is even remotely close to being able to see a self-driving system far off in the distance. So yes, it tries to do that everywhere.

MobilEye in their demonstration of a nearly-working self-driving system, sent 2 people and a car to Munich for 2 weeks, and it was nearly working there too, or so they claim. That's a better claim on being close to working everywhere, if they can really pull that off. Well, everywhere that there are cars driving the roads with EQ4s, which is much of the world.

FSD doesn't work everywhere, it fails everywhere. But something makes people not understand the difference.

thats a very semantical argument tho, whether it works or works simply depends on the context in which it's used. ie saying FSD beta works for testing purposes is correct, saying it works flawlessly everywhere is incorrect. and saying that it doesn't "work" because its not full level 5 yet is also facetious since it's a work in progress, hell even once they give it some kind of public release at best it will be level 2/3, so you kind of have to say if it "works" for what its current intentions are and by that measure yes it works lol.... of course this is not the same thing as "works well" or even "is good" lol, but just that it technically works... the same thing could be said about the Zune so it's not a ringing endorsement anything.

and FWIW the fact that it fails in various situations is kind of by design to an extent (while its in limited access BETA). as a whole the system actually needs lots of failures in order to learn where the software breaks down since obviously those scenarios where not properly accounted for. that is basically the whole reason for a beta program, to find the bugs and fix them, not to have a finished polished product for mass consumption. also keep in mind that tests like the one by MobilEye are done by trained drivers not random people who aren't there to test the car but actually use it so those results will be skewed in that kind of comparison, and by their definition of "nearly working" lol.

just my 2 cents

 

[Daniel in SD]

context

Look at the thread title. In the context of this thread "Work" = performance good enough for driverless operation.

 

[OncomingStorm]

I am confused why people say that Tesla FSD "works" anywhere, let alone everywhere. I have not heard of any place it is said to be even close to working.

Look at the thread title. In the context of this thread "Work" = performance good enough for driverless operation.

that might be true in the general context of the thread but per your post you were making a very subjective statement about the word "works" and how it applies to FSD Beta. and i stand by my argument that FSD beta works, it might not work well or in every case, but the fact that it doesn't just drive every person straight into a wall proves it technically works lol, its just flawed as of yet. and furthermore, we cannot make direct comparisons with other companies and how well they "work" since the only thing we have to go by is their own materials in most cases. so it could very well be the case that they all work at the same level or that one is vastly superior but since FSD beta is the only one any of US can actually use, i'd say its a pretty safe bet tesla's is better when all context is accounted for, otherwise you'd see a mad rush of auto makers picking up that specific thing instead of the extremely limited trickle of pseudo-tesla features you see now. remember you will never likely see geofenced robotaxi hardware/software on a production car you can buy, so if my goal is a self driving car that tech does not "work" for me.

 

[diplomat33]

the fact that it doesn't just drive every person straight into a wall proves it technically works lol, its just flawed as of yet.

That's an absurdly low standard for defining "works". I think we should hold FSD beta to a higher standard than that.

furthermore, we cannot make direct comparisons with other companies and how well they "work" since the only thing we have to go by is their own materials in most cases. so it could very well be the case that they all work at the same level or that one is vastly superior but since FSD beta is the only one any of US can actually use

This is absurdly false. We have more than "marketing materials". We have safety data on Waymo, Cruise and others. We know they can go thousands of miles without human intervention. And FSD Beta is not the only one that people can use. The public is actually riding in Waymo and Cruise robotaxis with no safety driver at all. And we can see all the disengagements and problems that FSD Beta has. So yes, we can compare them to FSD Beta and we can say with confidence that Waymo, Cruise work much better than Tesla's FSD Beta. In fact, based on disengagement data, Waymo and Cruise work about 1000x better than FSD Beta. It is not even close.

 

[Daniel in SD]

that might be true in the general context of the thread but per your post you were making a very subjective statement about the word "works" and how it applies to FSD Beta. and i stand by my argument that FSD beta works, it might not work well or in every case, but the fact that it doesn't just drive every person straight into a wall proves it technically works lol, its just flawed as of yet. and furthermore, we cannot make direct comparisons with other companies and how well they "work" since the only thing we have to go by is their own materials in most cases. so it could very well be the case that they all work at the same level or that one is vastly superior but since FSD beta is the only one any of US can actually use, i'd say its a pretty safe bet tesla's is better when all context is accounted for, otherwise you'd see a mad rush of auto makers picking up that specific thing instead of the extremely limited trickle of pseudo-tesla features you see now. remember you will never likely see geofenced robotaxi hardware/software on a production car you can buy, so if my goal is a self driving car that tech does not "work" for me.

It sounds like you're in the wrong thread. This is the thread where FSD Beta is the beta version of FSD Robotaxi. There are other threads where FSD Beta is the beta version of a driver assist feature.
Also, FSD Beta is not the only robotaxi software that we can use. Cruise in San Francisco and Waymo in Chandler allow the public to use their robotaxis. I agree that FSD Beta is the only robotaxi software that allows the public to test it anywhere and see that it does not work well enough to be a robotaxi anywhere.

 

[OncomingStorm]

It sounds like you're in the wrong thread. This is the thread where FSD Beta is the beta version of FSD Robotaxi. There are other threads where FSD Beta is the beta version of a driver assist feature.
Also, FSD Beta is not the only robotaxi software that we can use. Cruise in San Francisco and Waymo in Chandler allow the public to use their robotaxis. I agree that FSD Beta is the only robotaxi software that allows the public to test it anywhere and see that it does not work well enough to be a robotaxi anywhere.

thats actually my bad then, sorry this is a long thread lol, i didn't realize the ongoing conversation was strictly limited to FSD Beta as a robotaxi service, that completely changes my argument lol in that sence then no it doesn't work and the geofenced companies will rise up much faster than tesla's service.

and yes of course i'm backward on the whole "testing" thing since it's pretty much tesla's you can't test in that way haha i recant there as well.

 

[OncomingStorm]

That's an absurdly low standard for defining "works". I think we should hold FSD beta to a higher standard than that.



This is absurdly false. We have more than "marketing materials". We have safety data on Waymo, Cruise and others. We know they can go thousands of miles without human intervention. And FSD Beta is not the only one that people can use. The public is actually riding in Waymo and Cruise robotaxis with no safety driver at all. And we can see all the disengagements and problems that FSD Beta has. So yes, we can compare them to FSD Beta and we can say with confidence that Waymo, Cruise work much better than Tesla's FSD Beta. In fact, based on disengagement data, Waymo and Cruise work about 1000x better than FSD Beta. It is not even close.

in the context of the point i was trying to make yes thats an extremely low standard lol, but that was also kind of the point. a point which is very moot now.

to your second point, do we actually have verified data on this? or just the reports they publish? i'll be honest it's hard to keep up with everything but so far as i was aware there is no solid independent sourcing of that kind of data, merely what they report. cause i'd wager if you ask any of them they could show you data about how they are better lol. am i wrong there? (serious question lol) i'd be very interested in following that group/reports if so!

 

[EVNow]

thats actually my bad then, sorry this is a long thread lol, i didn't realize the ongoing conversation was strictly limited to FSD Beta as a robotaxi service ….

It is not … that comment was a joke.

There are some Tesla haters in this thread, though, who are very dogmatic. They are like those auto industry “experts” who laughed at Tesla because they thought panel gaps were not good …

I’ve no idea why they come and post in TMC if they think Tesla effort is worthless.

ps : apparently not a joke. So, in the dogmatic bucket.

pps : Anyone who thinks Tesla FSD beta doesn’t belong here go open a new thread in a non-Tesla related sub-forum.

Cars and Transportation

Discussion about any form of transportation, including those which may include fossil fuels.

teslamotorsclub.com

 

[diplomat33]

to your second point, do we actually have verified data on this? or just the reports they publish? i'll be honest it's hard to keep up with everything but so far as i was aware there is no solid independent sourcing of that kind of data, merely what they report. cause i'd wager if you ask any of them they could show you data about how they are better lol. am i wrong there? (serious question lol) i'd be very interested in following that group/reports if so!

Well, we have the reports that they must submit to the CA DMV every year. Those are official reports that they submit to a government entity. If they were found to be lying on those reports, the CA DMV would take away their testing permits. So they have every reason to be truthful. And even if you think that they are tweaking the numbers a bit, I doubt that they are tweaking the numbers by a factor of 1000x which is what would be required for their performance to be as bad as Tesla's FSD Beta. Also, they can't hide accidents because those require police reports. So we know the accident reports are correct. Furthermore, the public is actually riding in driverless robotaxis in SF and Chandler. They are independent sources and you can't fake that! It takes a lot of reliability to allow people to ride in a robotaxi with no safety driver.

 

[Daniel in SD]

It is not … that comment was a joke.

There are some Tesla haters in this thread, though, who are very dogmatic. They are like those auto industry “experts” who laughed at Tesla because they thought panel gaps were not good …

I’ve no idea why they come and post in TMC if they think Tesla effort is worthless.

It was not a joke. Tesla's effort is not worthless, they are making progress towards robotaxis. I'm just saying that in the context of this thread an autonomous vehicle that "works" is one that can operate safely without a driver. If Waymo kills someone in Chandler next week I will be the first to say that their system doesn't work (or maybe they got extremely unlucky).

 

[OncomingStorm]

Well, we have the reports that they must submit to the CA DMV every year. Those are official reports that they submit to a government entity. We can trust those reports because the CA DMV is a government regulator. If they were found to be lying on those reports, the CA DMV would take away their testing permits. So they have every reason to be truthful. And even if you think that they are tweaking the numbers a bit, I doubt that they are tweaking the numbers by a factor of 1000x which is what would be required for their performance to be as bad as Tesla's FSD Beta. Also, they can't hide accidents because those require police reports. So we know the accident reports are correct.

fair enough yeah those would presumably be fairly on par with reality. that said tho, it would be really easy to skew those numbers in favor of the company in a geofenced setup. considering you dont need to take the most direct route, it would be cake to simply disallow routing on roads that cause the system issues thereby altering the actual results. and honestly i wouldn't be surprised if they did just that, in their situation is almost better to map out the roads that are perfect for your system and minimize any travel on roads or intersections that are trickier, with the intention of utilizing those roads more as the system matures. whether that matters for a robotaxi system or not would be an open debate as well.

i'm ALWAYS suspect to data where i cannot examine the source material, the DMV reports are interesting tho, i'm going to have to dig into those numbers a bit, it would be nice if we could get more, i'd love to see a map of the traveled roads coded by disengagement rate or something along those lines. i'd wager good money the area's where these systems fail is likely similar across the board, its just presented in different contexts so its hard to correlate (obv just opinion there).

 

[AKAlex]

fair enough yeah those would presumably be fairly on par with reality. that said tho, it would be really easy to skew those numbers in favor of the company in a geofenced setup. considering you dont need to take the most direct route, it would be cake to simply disallow routing on roads that cause the system issues thereby altering the actual results. and honestly i wouldn't be surprised if they did just that, in their situation is almost better to map out the roads that are perfect for your system and minimize any travel on roads or intersections that are trickier, with the intention of utilizing those roads more as the system matures. whether that matters for a robotaxi system or not would be an open debate as well.

i'm ALWAYS suspect to data where i cannot examine the source material, the DMV reports are interesting tho, i'm going to have to dig into those numbers a bit, it would be nice if we could get more, i'd love to see a map of the traveled roads coded by disengagement rate or something along those lines. i'd wager good money the area's where these systems fail is likely similar across the board, its just presented in different contexts so its hard to correlate (obv just opinion there).

I have a similar suspicion of some of the data that is coming out of those companies. Without the context around disengagements, it's very difficult to draw any conclusions. As far as I can tell, Tesla has been more transparent than any other company in this realm so far. I mean, I have the software, I can poke and prod it any way I want.

 

[Daniel in SD]

I have a similar suspicion of some of the data that is coming out of those companies. Without the context around disengagements, it's very difficult to draw any conclusions. As far as I can tell, Tesla has been more transparent than any other company in this realm so far. I mean, I have the software, I can poke and prod it any way I want.

Tesla doesn't report disengagements or collisions at all whereas all the other companies testing in my state do.
The disengagement reports do provide a description of the reason for disengagement (though admittedly it's often not very enlightening) whereas with FSD Beta Tesla provides no explanation for what went wrong though users can speculate.

 

[AKAlex]

that might be true in the general context of the thread but per your post you were making a very subjective statement about the word "works" and how it applies to FSD Beta. and i stand by my argument that FSD beta works, it might not work well or in every case, but the fact that it doesn't just drive every person straight into a wall proves it technically works lol, its just flawed as of yet. and furthermore, we cannot make direct comparisons with other companies and how well they "work" since the only thing we have to go by is their own materials in most cases. so it could very well be the case that they all work at the same level or that one is vastly superior but since FSD beta is the only one any of US can actually use, i'd say its a pretty safe bet tesla's is better when all context is accounted for, otherwise you'd see a mad rush of auto makers picking up that specific thing instead of the extremely limited trickle of pseudo-tesla features you see now. remember you will never likely see geofenced robotaxi hardware/software on a production car you can buy, so if my goal is a self driving car that tech does not "work" for me.

There are a bunch of ways that you can classify "does it work?". FSD is capable of operating on ~98% of roads in the US with limited reliability. Waymo can operate on ~0.1% of roads in the US with high reliability. However, the real question in my head is "Who is going to dominate the American robotaxi market at the end of this decade?".

 

[AKAlex]

Tesla doesn't report disengagements or collisions at all whereas all the other companies testing in my state do.
The disengagement reports do provide a description of the reason for disengagement (though admittedly it's often not very enlightening) whereas with FSD Beta Tesla provides no explanation for what went wrong though users can speculate.

Yea this is totally true, it would be great if they did release official stats, I'm curious. Although, we do have some informal comments indicating that there have not been any accidents in the Beta program. I really wish I could see their internal stats. What is the data that Elon has seen that backs up his assertion of level 4 by the end of the year?

From a transperancy stand point, the amount of information that is currently available about the Beta program to an average joe like me is unsurpassed. There is literally 100X more information about Beta available than any of the other AV projects (Videos, personal experience and now crowd sourced data). Sure, this data is noisy, but it is available.

 

[Daniel in SD]

There are a bunch of ways that you can classify "does it work?". FSD is capable of operating on ~98% of roads in the US with limited reliability. Waymo can operate on ~0.1% of roads in the US with high reliability. However, the real question in my head is "Who is going to dominate the American robotaxi market at the end of this decade?".

I think it's undeniable that Tesla will dominate if their system works well enough on HW3 for robotaxi operation.
I don't think it will but I'm told I should never underestimate Elon!
A big question I have about Tesla is how they're going to manage the transition to new hardware if it proves necessary.

 

[OncomingStorm]

I have a similar suspicion of some of the data that is coming out of those companies. Without the context around disengagements, it's very difficult to draw any conclusions. As far as I can tell, Tesla has been more transparent than any other company in this realm so far. I mean, I have the software, I can poke and prod it any way I want.

saying tesla is more transparent than anything made me chuckle lol... i was waiting for the punchline but you are right! lol