Autonomous Car Progress

[stopcrazypp]

Its called "Urban driving".
This is pathetic and sad, just like I expected.
And the BMW Traffic light feature is live in the locales that BMW enabled it in.
There are dozens of features like that that is only available in some locales.
The same is the case with Tesla.

But not surprised you will duck everything. Deny and Deflect.
Why? Because it directly destroys the myth that has been pushed by the Tesla community for years.
I have yet to see a Tesla fan address it because they can't come to terms with it.

I haven't been following all this but instead of posting examples of where BMW's traffic lights feature is obviously disabled or restricted (which doesn't prove anything, only the opposing argument that it's not working properly yet), why not just post a single example of it being used in a owner vehicle? Unfortunately this caveat has to be put due to the fact that previous features like Audi's L3 had been demoed extensively, but ultimately was not available for use for a single real owner, only in Audi's demo fleet. That should settle things conclusively.

Edit: a quick Google found a lot of people are hacking the car to enable the feature ahead of official release, so I guess that complicated things. This isn't possible with a Tesla (some power boost updates have been hacked but even those are not before release, none of the ADAS features had been hacked to date to enable outside of Tesla authorization).

 

[rxlawdude]

I haven't been following all this but instead of posting examples of where BMW's traffic lights feature is obviously disabled or restricted (which doesn't prove anything, only the opposing argument that it's not working properly yet), why not just post a single example of it being used in a owner vehicle? Unfortunately this caveat has to be put due to the fact that previous features like Audi's L3 had been demoed extensively, but ultimately was not available for use for a single real owner, only in Audi's demo fleet. That should settle things conclusively.

Edit: a quick Google found a lot of people are hacking the car to enable the feature ahead of official release, so I guess that complicated things. This isn't possible with a Tesla (some power boost updates have been hacked but even those are not before release, none of the ADAS features had been hacked to date to enable outside of Tesla authorization).

Hacking a car to get an unsupported feature? Why, this sounds like people hacking the Autopilot safety features to demonstrate they can do stupid things.

 

[Bladerskb]

I haven't been following all this but instead of posting examples of where BMW's traffic lights feature is obviously disabled or restricted (which doesn't prove anything, only the opposing argument that it's not working properly yet), why not just post a single example of it being used in a owner vehicle? Unfortunately this caveat has to be put due to the fact that previous features like Audi's L3 had been demoed extensively, but ultimately was not available for use for a single real owner, only in Audi's demo fleet. That should settle things conclusively.

Edit: a quick Google found a lot of people are hacking the car to enable the feature ahead of official release, so I guess that complicated things. This isn't possible with a Tesla (some power boost updates have been hacked but even those are not before release, none of the ADAS features had been hacked to date to enable outside of Tesla authorization).

This is actually possible with Tesla.

 

[S4WRXTTCS]

This is actually possible with Tesla.

The embedded video shows no information that it was hacked in.

You only know it was hacked in if you look at who posted, and when they posted it (before Traffic light response was added). Where you have go to the YouTube link to see the description.

But, yeah its possible for some individuals like Green to enable some features that are currently not enabled. It's not currently possible for general end users like myself to hack any ADAS features or settings.

 

[rxlawdude]

But, yeah its possible for some individuals like Green to enable some features that are currently not enabled. It's not currently possible for general end users like myself to hack any ADAS features or settings.

Not to mention, from a liability standpoint, driving a modded/hacked Tesla is not wise.

 

[Bladerskb]

The embedded video shows no information that it was hacked in.

You only know it was hacked in if you look at who posted, and when they posted it (before Traffic light response was added). Where you have go to the YouTube link to see the description.

Video description:
"Ok, this is not perfect yet, but this is production firmware 19.8.3 and it does a lot of new tricks. Obviously you cannot yet enable it in GUI, and for a good reason that I'll list below."

But, yeah its possible for some individuals like Green to enable some features that are currently not enabled. It's not currently possible for general end users like myself to hack any ADAS features or settings.

Neither will you or any general end users be able to code/hack any features on BMW. It takes detailed walkthroughs and multiple specific softwares and days/weeks long attempts, most people end up paying others to do it.

 

[Bladerskb]

@stopcrazypp

Instead of rehashing the past. If you want to have an objective focused discussion/tracking (seeing we are in the 'Autonomous Car Progress' thread).
We can evaluate who is the leader in autonomous driving and actually quantify it especially as ADAS that blur the line with L4 feature capabilities are on their way.

Currently as you know Tesla fans will say that Tesla ALREADY won autonomous driving. That its game over. That they are 5-10 years ahead and there are no competition. However when you ask for quantification. For a laid out logic and deductive reason that can be applied to all companies universally. You hear crickets followed by "Tesla has billions of miles of data" and that's it.

No evidence, no proof. Nothing factual. Nothing tangible. No actual comparison using deductive reasoning.

1. Not only that, but their argument change constantly, first it was that Tesla is building a HD map moat, however when Elon came out against HD map it became, HD map is a useless crutch.
2. Another argument was that Tesla created Radar that was better than Lidar, then when Elon came out against radar it became that radar is a useless crutch.
3. Finally I will mention the argument initially in AP2.0 days was that a car with 10 Tops with no deep learning accelerator while using 100 Watts is all you need. Anything more is Nvidia trying to sell you chips in their own interest. Tesla is the only one with a 10 Tops / 100 watts board in a car and cameras which you need to collect data so Tesla has immense lead. However after Tesla announced HW3, it became you need a FSD computer with atleast 144 TOPs to do autonomous driving or collect data and anything less than that is worthless.

The thesis was that Tesla with acouple thousand cars with 10 Tops / 100 watts computer with no DLA and cameras/radar was years ahead of everyone.
So why isn't Xpeng or several other companies with 30 Tops /30 watts computer, 360 camera, 360 radar also years ahead of everyone?

All the argument revolved around whatever Tesla was doing/says and any company who does the same thing or better doesn't matter because they don't have the Tesla logo.

The only option is to come up with a logic of full proof logic/criteria from deductive reasoning that is white-labeled and doesn't change based on what new direction a company is going or their latest statement.
This logic/criteria will be able to determine what being in the lead looks like and how many years in the lead.

Logic will be based on quantifiable evidence not "data data data".
Logic must be based on current production features/capability and not future updates.

We can create a spreadsheet.

For example:

• Miles per safety disengagement in urban environment.
• Miles per safety disengagement in suburban environment.
• Total safety disengagement per mile.
• Total intervention per mile.
• Complexity of maneuvers the system can do.
• Complexity of driving environment and traffic the system can handle.
• Numbers of scenarios covered (ex: stopping for emergency vehicle, emergency personnel or when a peds waves to go or stop, parking, garage, yielding, reverse, u-turn, 3-point turn, etc)
• Simple constructions
• Complex constructions and detours.
• ODD limits
• Geofences
• Driverless ability (Fully autonomous) / L4

 

[stopcrazypp]

This is actually possible with Tesla.

Ok, I stand corrected on there being absolute zero, but from the description, that's verygreen, which we know is someone special in the community and he does things that no one else have done (no one had replicated what he did in the video back then, everyone else had to wait until official release).

From what I found for BMW, the situation is completely different, the hack is readily available to many people, not just one single person. The only thing similar in the Tesla community are the power boost hacks, which you can easily purchase.

 

[stopcrazypp]

@stopcrazypp

Instead of rehashing the past. If you want to have an objective focused discussion/tracking (seeing we are in the 'Autonomous Car Progress' thread).
We can evaluate who is the leader in autonomous driving and actually quantify it especially as ADAS that blur the line with L4 feature capabilities are on their way.

Currently as you know Tesla fans will say that Tesla ALREADY won autonomous driving. That its game over. That they are 5-10 years ahead and there are no competition. However when you ask for quantification. For a laid out logic and deductive reason that can be applied to all companies universally. You hear crickets followed by "Tesla has billions of miles of data" and that's it.

No evidence, no proof. Nothing factual. Nothing tangible. No actual comparison using deductive reasoning.

1. Not only that, but their argument change constantly, first it was that Tesla is building a HD map moat, however when Elon came out against HD map it became, HD map is a useless crutch.
2. Another argument was that Tesla created Radar that was better than Lidar, then when Elon came out against radar it became that radar is a useless crutch.
3. Finally I will mention the argument initially in AP2.0 days was that a car with 10 Tops with no deep learning accelerator while using 100 Watts is all you need. Anything more is Nvidia trying to sell you chips in their own interest. Tesla is the only one with a 10 Tops / 100 watts board in a car and cameras which you need to collect data so Tesla has immense lead. However after Tesla announced HW3, it became you need a FSD computer with atleast 144 TOPs to do autonomous driving or collect data and anything less than that is worthless.

The thesis was that Tesla with acouple thousand cars with 10 Tops / 100 watts computer with no DLA and cameras/radar was years ahead of everyone.
So why isn't Xpeng or several other companies with 30 Tops /30 watts computer, 360 camera, 360 radar also years ahead of everyone?

All the argument revolved around whatever Tesla was doing/says and any company who does the same thing or better doesn't matter because they don't have the Tesla logo.

The only option is to come up with a logic of full proof logic/criteria from deductive reasoning that is white-labeled and doesn't change based on what new direction a company is going or their latest statement.
This logic/criteria will be able to determine what being in the lead looks like and how many years in the lead.

Logic will be based on quantifiable evidence not "data data data".
Logic must be based on current production features/capability and not future updates.

We can create a spreadsheet.

For example:

• Miles per safety disengagement in urban environment.
• Miles per safety disengagement in suburban environment.
• Total safety disengagement per mile.
• Total intervention per mile.
• Complexity of maneuvers the system can do.
• Complexity of driving environment and traffic the system can handle.
• Numbers of scenarios covered (ex: stopping for emergency vehicle, emergency personnel or when a peds waves to go or stop, parking, garage, yielding, reverse, u-turn, 3-point turn, etc)
• Simple constructions
• Complex constructions and detours.
• ODD limits
• Geofences
• Driverless ability (Fully autonomous) / L4

It seems you are continually obssessed with others "winning" over Tesla or Tesla "winning" over others in autonomy. I started this thread back then so people can post news on how other autonomous (and semi-autonomous) cars are progressing, not for it to be a measuring contest between the different players, for which there likely will never be a consistent agreed upon metric on how to measure which one "won".

Given many of the players that are working on L4/L5 don't have any immediate plans to sell vehicles to consumers (or even to give the public access to such vehicles), there isn't really a practical reason to do spreadsheet comparisons at the moment. This is different from other car comparisons where people are making real purchasing decisions based on comparing different specs, so there is some value to doing so.

We can point out highlights/milestones like for example Honda being the first to start leasing a L3 car to consumers (albeit a small 100 vehicle fleet), but it's hard to say if that is "winning", as there are many possible milestones ahead that can leapfrog that. Kind of like milestones in EVs like the original Roadster being the first 200+ mile production EV, Nissan Leaf the first affordable mass market EV, the Bolt the first affordable 200+ mile EV, and so on, but would people really consider those models the "winners" in the EV market?

 

[powertoold]

I have a feeling the BMW traffic light feature will never be officially released, especially not in the whole USA.

 

[Bladerskb]

Ok, I stand corrected on there being absolute zero, but from the description, that's verygreen, which we know is someone special in the community and he does things that no one else have done (no one had replicated what he did in the video back then, everyone else had to wait until official release).

From what I found for BMW, the situation is completely different, the hack is readily available to many people, not just one single person. The only thing similar in the Tesla community are the power boost hacks, which you can easily purchase.

Green isn't the only one there are tons of people in the tesla community who mod their cars but don't publicize the info.

Example..

It seems you are continually obssessed with others "winning" over Tesla or Tesla "winning" over others in autonomy.

I don't get it. I follow the ADAS and AV industry closely and almost every fan who happens to be a Tesla fan response to me is that Tesla already won, they have solved self driving, they are decades ahead. Sandy Munro who has been on so many platform said before for a long time that Tesla already solved L5 and won't deploy it because they are waiting for other OEMs to catch up. But if I respond to that or try to find ways to educate and provide independent information. I'm the obsessed one? But they are free to mislead?

I started this thread back then so people can post news on how other autonomous (and semi-autonomous) cars are progressing, not for it to be a measuring contest between the different players, for which there likely will never be a consistent agreed upon metric on how to measure which one "won".

You can't determine progress if you don't have a measuring metric and a way of keeping track of who can do what and how reliably.
Its not a measure of who "won" or "winning". Like I said, its to quantify what exactly is progress or "winning" and provide independent data. Tesla fans will dislike this because I have yet to see anyone put in paper a logic that computes Tesla being "5-10 years ahead". So its easier to avoid providing any logic or deductive reason and just repeat the mantra.

Given many of the players that are working on L4/L5 don't have any immediate plans to sell vehicles to consumers (or even to give the public access to such vehicles), there isn't really a practical reason to do spreadsheet comparisons at the moment. This is different from other car comparisons where people are making real purchasing decisions based on comparing different specs, so there is some value to doing so.

Actually this is changing and will continue to change in the future. A number of SDC developers are deploying their L4 systems as door to door full scenario L2 systems. First starting with Huawei, Mobileye and Tesla in 2021. Then Biadu, NIO, Xpeng later (2022+). This is the perfect time to come up with white-labeled reasonable logic and deductive reason without bias.

We can point out highlights/milestones like for example Honda being the first to start leasing a L3 car to consumers (albeit a small 100 vehicle fleet), but it's hard to say if that is "winning", as there are many possible milestones ahead that can leapfrog that. Kind of like milestones in EVs like the original Roadster being the first 200+ mile production EV, Nissan Leaf the first affordable mass market EV, the Bolt the first affordable 200+ mile EV, and so on, but would people really consider those models the "winners" in the EV market?

Not to claim "winning" but to quantify "winning" according to people who claim a specific company is winning.

 

[diplomat33]

Podcast on the future of self-driving cars with founders of Zoox, Cruise, Aurora, Voyages as well as ex-Uber and TechCrunch experts:

undefined

Websites for creators

constine.club

The podcast is 1hr40mn long but it is an awesome podcast. I highly recommend.

 

[Cheburashka]

Baidu rolls out paid driverless taxi service in Beijing

Apparently without a safety driver but with a laughable geofence of 1.2 square miles.

 

[stopcrazypp]

Green isn't the only one there are tons of people in the tesla community who mod their cars but don't publicize the info.

Example..

I don't get it. I follow the ADAS and AV industry closely and almost every fan who happens to be a Tesla fan response to me is that Tesla already won, they have solved self driving, they are decades ahead. Sandy Munro who has been on so many platform said before for a long time that Tesla already solved L5 and won't deploy it because they are waiting for other OEMs to catch up. But if I respond to that or try to find ways to educate and provide independent information. I'm the obsessed one? But they are free to mislead?

Sure you are free to respond, but it's pretty clear they aren't buying your argument they are "wrong" and it seems unlikely to change anytime soon. My point is more that if you focus less on which players are "winners" (which tends to be a subjective evaluation, based on metrics chosen by the given individual) things may be less antagonistic and instead focus on the facts and news. Once things turn into a measuring contest, emotions tend to get involved.

You can't determine progress if you don't have a measuring metric and a way of keeping track of who can do what and how reliably.
Its not a measure of who "won" or "winning". Like I said, its to quantify what exactly is progress or "winning" and provide independent data. Tesla fans will dislike this because I have yet to see anyone put in paper a logic that computes Tesla being "5-10 years ahead". So its easier to avoid providing any logic or deductive reason and just repeat the mantra.

Again, feel free to do that sheet if you want, but just know that it is unlikely to sway opinion, as plenty of people will likely disagree with the metrics chosen and which ones are the most important. Heck, people don't even do such detailed comparisons on the L2 systems in use today and those people can actually buy and use right now (so it might have some real world practical use, not just something to use to win an internet argument).

Actually this is changing and will continue to change in the future. A number of SDC developers are deploying their L4 systems as door to door full scenario L2 systems. First starting with Huawei, Mobileye and Tesla in 2021. Then Biadu, NIO, Xpeng later (2022+). This is the perfect time to come up with white-labeled reasonable logic and deductive reason without bias.

I again caution on counting eggs before they hatch. I looked back on the start of the thread, and a lot of the original plans back then (from Ford, Nissan, Volvo, etc, and yes Tesla) have not exactly panned out as planned. Heck even Cruise (which has made some of the most progress, which I personally have observed, seeing their cars all the time in person) has delayed their plans. Still a long road ahead for all players.

Not to claim "winning" but to quantify "winning" according to people who claim a specific company is winning.

 

[mark95476]

I watched bits of Bjorn livestreaming the BMW iX3 and it wasn't able to follow clear day-time roads at times. I'm really surprised that it's 2021 and following slight bends on clearly lined roads isn't reliable in the BMW. BMW and their vendor have work ahead of them to get minimal basic functionality, let alone anything more advanced.

I have a feeling the BMW traffic light feature will never be officially released, especially not in the whole USA.

Yup!

Sure you are free to respond, but it's pretty clear they aren't buying your argument they are "wrong" and it seems unlikely to change anytime soon.

 

[powertoold]

I watched bits of Bjorn livestreaming the BMW iX3 and it wasn't able to follow clear day-time roads at times. I'm really surprised that it's 2021 and following turns on clear roads isn't reliable in the BMW. BMW and their vendor have work ahead of them to get minimal basic functionality, let alone anything more advanced.

Mobileye is the biggest joke of the fsd developers. I've seen enough of Amnon. He's better off being a politician.

I'm so glad Tesla broke up with Mobileye in the way they did: rub salt on your own wound, become stronger.

This is all just my opinion. I'm very confident Mobileye won't deploy anything meaningful in the fsd space.

 

[JHCCAZ]

Podcast on the future of self-driving cars with founders of Zoox, Cruise, Aurora, Voyages as well as ex-Uber and TechCrunch experts:

undefined

Websites for creators

constine.club

The podcast is 1hr40mn long but it is an awesome podcast. I highly recommend.

Thanks I just finished listening to this whole thing. (I note that I found a top-ten list of self-driving podcasts, not including this one just out, so there's a lot more to be heard).

Here are some summary points and quick reactions:

• Panel is made up of host, several RoboTaxi developers, ex-Uber "scientist" and TechCrunch journalist. No current Tesla, no MobilEye, no ICE-car manufacturer.
• Specific to Tesla interest, this site's focus:
  • A lead-in commentary from about 27:00 to 30:00 by the Zoox guy I think, not naming Tesla but obviously opining on Tesla's irresponsible approach that can hurt the industry.
  • At 30 min., the host names Tesla and mentions the recent deaths while "running Autopilot". Zoox guy clarifies that we don't know that, but still worries about public perception damage from lax feature rollout.
  • Aurora / ex-Tesla guy expresses some agreement but endeavors to make two points:
    • Great respect for Tesla developers working on FSD
    • Tesla is solving a "different problem" than the panel members, focusing on Assistance not Autonomy and requiring hardware cost economies to be able to sell cars to individuals, not fleet owners.
  • Voyage guy is more blatantly critical of Tesla silence and irresponsibility, considering AP-prank culture
  • Discussion moves on to difficulty of maintaining needed focus in L2 scenario
• Overall, the panelists agree among themselves that
  • autonomy requires an extensive & expensive sensor package
  • L2 is fundamentally problematic due to Zoning Out tendency, L3 is maybe unrealistic. L4 / L5 is where it's at.
  • The future with Autonomy will be great by
    • Giving humans more quality time
    • Reducing parking and personal-car clogging of cities*
    • Of course, reducing injury and fatalities
    • Enabling the desirable and eventual end of humans driving personally-owned vehicles.
    • This is necessarily an oversimplification of the discussion but I think a reasonable summary.
  • Personal ownership of self-driving cars will be rare, expensive and makes little sense.
    • Especially because of the required super-sensor-hardware
    • This will have cultural impact in several ways, including discussion of teenage experiences and relationships
  • The required remaining improvements (the last fractional-percent of use-cases) are very hard, will take time, will make vehicles expensive
• My take very briefly:
  • The panelists are RoboTaxi developers, little to no willingness to consider that non-L4+ and/or affordable personal cars can be autonomous.
  • *The viewpoints and future-imagining are very city-dweller-centric
  • Though there may be some professional respect for Tesla's team, there is basically no respect for Tesla's ideas regarding timing, camera-centric hardware, and no coverage at all of Tesla's own L4/L5/RoboTaxi plans.
  • Having broken the ice by dissing Tesla specifically, the host should have further prodded the panelists to explain why the collegially-respected team of Karpathy et. al. would continue to work towards Elon's unrealistic goals, and to cooperate with their firm's purportedly irresponsible support of both
    • intentional AP pranking
    • and unintentional human-nature inability to correctly handle L2.

 

[Steerpike]

A fully self-driving car has to have redundancy and failsafe provided by at least two separate systems. My Tesla often warns me that it can't see properly because the sun is shining or there is heavy rain. For an autonomous car to be legally permitted it must be significantly safer than humans for close to 100% of the time. Recent Tesla announcements that suggest that cameras alone can provide this functionality and level of reliability lack credibility. "Really Cool Advanced Driver Assist" perhaps but "Full Self Driving" never.

 

[diplomat33]

Thanks I just finished listening to this whole thing. (I note that I found a top-ten list of self-driving podcasts, not including this one just out, so there's a lot more to be heard).

Here are some summary points and quick reactions:

• Panel is made up of host, several RoboTaxi developers, ex-Uber "scientist" and TechCrunch journalist. No current Tesla, no MobilEye, no ICE-car manufacturer.
• Specific to Tesla interest, this site's focus:
  • A lead-in commentary from about 27:00 to 30:00 by the Zoox guy I think, not naming Tesla but obviously opining on Tesla's irresponsible approach that can hurt the industry.
  • At 30 min., the host names Tesla and mentions the recent deaths while "running Autopilot". Zoox guy clarifies that we don't know that, but still worries about public perception damage from lax feature rollout.
  • Aurora / ex-Tesla guy expresses some agreement but endeavors to make two points:
    • Great respect for Tesla developers working on FSD
    • Tesla is solving a "different problem" than the panel members, focusing on Assistance not Autonomy and requiring hardware cost economies to be able to sell cars to individuals, not fleet owners.
  • Voyage guy is more blatantly critical of Tesla silence and irresponsibility, considering AP-prank culture
  • Discussion moves on to difficulty of maintaining needed focus in L2 scenario
• Overall, the panelists agree among themselves that
  • autonomy requires an extensive & expensive sensor package
  • L2 is fundamentally problematic due to Zoning Out tendency, L3 is maybe unrealistic. L4 / L5 is where it's at.
  • The future with Autonomy will be great by
    • Giving humans more quality time
    • Reducing parking and personal-car clogging of cities*
    • Of course, reducing injury and fatalities
    • Enabling the desirable and eventual end of humans driving personally-owned vehicles.
    • This is necessarily an oversimplification of the discussion but I think a reasonable summary.
  • Personal ownership of self-driving cars will be rare, expensive and makes little sense.
    • Especially because of the required super-sensor-hardware
    • This will have cultural impact in several ways, including discussion of teenage experiences and relationships
  • The required remaining improvements (the last fractional-percent of use-cases) are very hard, will take time, will make vehicles expensive

Thanks for writing this up. Good job!

A few additional things:

They also mentioned consolidation. They made 2 points that consolidation is happening because of the high cost of developing FSD. Some companies simply cannot afford to develop FSD on their own. Also, deploying FSD requires expertise in 3 areas, Networks (logistics, ride-hailing), FSD tech itself, and production of vehicles. Companies are unlikely to be well positioned to do all 3 on their own so it is natural to partner with other companies with complimentary skills.

He also mentioned that we are unlikely to see new start-ups in FSD because of the high cost. He sees this as a bad thing because start-ups could provide new ideas and a fresh perspective.

For L3, it was more than just "maybe unrealistic". The guy said that L3 doesn't work and was only included to make the SAE nomenclature look more continuous. He explained that it takes about 30 seconds for a human to fully re-engage again with the road after not paying attention for awhile but avoiding accidents often requires split second intervention. So basically, a driver in a L3 car won't be able to intervene in time to prevent most crashes. And if a system is able to foresee a problem 30 seconds in advance in order to properly warn the driver, then your system can probably handle the situation without the driver so it is on its way to L4, so you might as well as skip L3 and just focus on L4. So either way, L3 does not make sense.

On solving that last 0.1%, they mentioned that humans tend to overestimate how good FSD is. We look at 99.9% FSD and assume it is already 100% FSD. He mentioned that humans can watch FSD videos of the system handling an edge case really well, and assume that the FSD is almost done when it is not.

 

[ItsNotAboutTheMoney]

@stopcrazypp

Instead of rehashing the past. If you want to have an objective focused discussion/tracking (seeing we are in the 'Autonomous Car Progress' thread).
We can evaluate who is the leader in autonomous driving and actually quantify it especially as ADAS that blur the line with L4 feature capabilities are on their way.

Currently as you know Tesla fans will say that Tesla ALREADY won autonomous driving. That its game over. That they are 5-10 years ahead and there are no competition. However when you ask for quantification. For a laid out logic and deductive reason that can be applied to all companies universally. You hear crickets followed by "Tesla has billions of miles of data" and that's it.

No evidence, no proof. Nothing factual. Nothing tangible. No actual comparison using deductive reasoning.

1. Not only that, but their argument change constantly, first it was that Tesla is building a HD map moat, however when Elon came out against HD map it became, HD map is a useless crutch.
2. Another argument was that Tesla created Radar that was better than Lidar, then when Elon came out against radar it became that radar is a useless crutch.
3. Finally I will mention the argument initially in AP2.0 days was that a car with 10 Tops with no deep learning accelerator while using 100 Watts is all you need. Anything more is Nvidia trying to sell you chips in their own interest. Tesla is the only one with a 10 Tops / 100 watts board in a car and cameras which you need to collect data so Tesla has immense lead. However after Tesla announced HW3, it became you need a FSD computer with atleast 144 TOPs to do autonomous driving or collect data and anything less than that is worthless.

The thesis was that Tesla with acouple thousand cars with 10 Tops / 100 watts computer with no DLA and cameras/radar was years ahead of everyone.
So why isn't Xpeng or several other companies with 30 Tops /30 watts computer, 360 camera, 360 radar also years ahead of everyone?

All the argument revolved around whatever Tesla was doing/says and any company who does the same thing or better doesn't matter because they don't have the Tesla logo.

The only option is to come up with a logic of full proof logic/criteria from deductive reasoning that is white-labeled and doesn't change based on what new direction a company is going or their latest statement.
This logic/criteria will be able to determine what being in the lead looks like and how many years in the lead.

Logic will be based on quantifiable evidence not "data data data".
Logic must be based on current production features/capability and not future updates.

We can create a spreadsheet.

For example:

• Miles per safety disengagement in urban environment.
• Miles per safety disengagement in suburban environment.
• Total safety disengagement per mile.
• Total intervention per mile.
• Complexity of maneuvers the system can do.
• Complexity of driving environment and traffic the system can handle.
• Numbers of scenarios covered (ex: stopping for emergency vehicle, emergency personnel or when a peds waves to go or stop, parking, garage, yielding, reverse, u-turn, 3-point turn, etc)
• Simple constructions
• Complex constructions and detours.
• ODD limits
• Geofences
• Driverless ability (Fully autonomous) / L4

• Money spent
• Money earned
• Cost per mile