Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Tesla testing self driving in Arizona and Texas, not so much CA

This site may earn commission on affiliate links.
Elon said recently he is very confident FSD will be enabled by end of next year in some jurisdiction(s). In the past 5 years he has said two years away. Now he is saying one year away. Will we go another 5 years with it being one year away? Then switch to 6 months away for a few more years? :p
Most companies have been saying FSD is 5 years away for quite sometime. I dug up the quotes and included them in a post a while back. This includes companies like Google/Waymo, Honda, Nissan etc.

As Musk so cleverly said in a different context (ramping up 3 production) - what the heck does he know ? Its not like they have done this before ... :eek:
 
Do you have a definition of "prototype autonomous vehicle" that somehow includes City NOA but not Freeway NOA ?
Yes. Freeway NoA is not intended to do all driving tasks, beta FSD is. Also, Freeway NoA actually makes driving easier.
I agree that the dividing line is a little blurry but I fail to see how beta FSD is significantly different than what every other autonomous vehicle company is testing on public roads. In the end the question is if it's safe for untrained and unmonitored safety drivers to be testing autonomous vehicles? I don't think it is safe which is why many states have strict rules about testing. If Tesla releases it to everyone we're going to find out very quickly! Exciting times.
 
Beta FSD is a prototype autonomous vehicle not a driver assist. It's absurd to claim otherwise IMO.

I know others have already disagreed. and I don't mean to gang up on you

But I am going to hard disagree as well. Aside from the naming scheme and the comments from Musk, FSD beta is 100% a driver assist system.

By that logic then there is no such thing as as a prototype autonomous vehicle. It's simply a driver assist system until it is deployed.
.

Perhaps we need another term that's not driver assist. I would say a system like Waymo or Cruise while they still have a safety driver is an L2 system. but I guess driver assist isn't a good description. But FSD beta is not like that, FSD beta is a driver assist system, and even without the "beta" it is still a driver assist feature.

Yes. Freeway NoA is not intended to do all driving tasks, beta FSD is.

Disagree with this... can you explain why you believe this?

I agree that the dividing line is a little blurry but I fail to see how beta FSD is significantly different than what every other autonomous vehicle company is testing on public roads. In the end the question is if it's safe for untrained and unmonitored safety drivers to be testing autonomous vehicles? I don't think it is safe which is why many states have strict rules about testing. If Tesla releases it to everyone we're going to find out very quickly! Exciting times.

I do agree with this.

To add some clarity... with the exception of say California. In almost any state, a robotaxi company can hire safety drivers and test in any state without any permit or anything. Except, since it is a grey area, it may cause unwanted attention or media, so most companies refrain.
 
Last edited:
I would say a system like Waymo or Cruise while they still have a safety driver is an L2 system.
Except that the SAE standard says it is not (and all states with autonomous vehicle rules reference the SAE standard as far as I am aware).
It specifically addresses your claim:
SAE J3016 said:
The level of a driving automation system feature corresponds to the feature’s production design intent. This applies regardless of whether the vehicle on which it is equipped is a production vehicle already deployed in commerce, or a test vehicle that has yet to be deployed. As such, it is incorrect to classify a level 4 design-intended ADS feature equipped on a test vehicle as level 2 simply because on-road testing requires a test driver to supervise the feature while engaged, and to intervene if necessary to maintain safe operation.
Anyway, I will die on this hill. I have eyes and I can see that beta FSD is a prototype AV.

To add some clarity... with the exception of say California. In almost any state, a robotaxi company can hire safety drivers and test in any state without any permit or anything. Except, since it is a grey area, it may cause unwanted attention or media, so most companies refrain.
That's not quite true. You don't need a permit in AZ or FL but you do need to fill out some forms and prove that you can cover the liability.
 
Except that the SAE standard says it is not (and all states with autonomous vehicle rules reference the SAE standard as far as I am aware).
It specifically addresses your claim:

Anyway, I will die on this hill. I have eyes and I can see that beta FSD is a prototype AV.


That's not quite true. You don't need a permit in AZ or FL but you do need to fill out some forms and prove that you can cover the liability.

I think you are missing my main argument.

" level 4 design-intended ADS feature equipped on a test vehicle as level 2 simply"

My point is that Tesla FSD is not a "level-4 design intended ADS features that is being tested"

That's not quite true. You don't need a permit in AZ or FL but you do need to fill out some forms and prove that you can cover the liability.

I am not convinced... I know a lot of states do ask for insurance paperwork liability, however, I am saying that most of the time this is not necessary even when states do ask for this. And many companies do anyways just because it's easier to do so then to make a scene


NO! L2 has a very specific definition. Even with a safety driver, Waymo and Cruise are NOT L2. They don't fit the definition of L2. L2 has nothing to do with having a safety driver. Waymo and Cruise are L4.

I figured you would say this, and technically you are correct. However, I would argue that most companies testing robotaxi vehicles, with the exception some cruise and waymo and maybe some others. Are using platforms that will never and will never be intended for L4 use even in the future, and their design intention is only for testing with safety driver and therefore L2. And the majority of AV testing is like so.

The Waymo Pacificas and jaguars are L4 and the latest gen Cruise Bolts are L4 sure... but basically any other vehicle testing a robotaxi like system with a safety driver, those vehicles would be L2 systems.

L2 has nothing to do with having a safety driver. .

How can you say that?
 
Last edited:
I think you are missing my main argument.

" level 4 design-intended ADS feature equipped on a test vehicle as level 2 simply"

My point is that Tesla FSD is not a "level-4 design intended ADS features that is being tested"

I am not convinced... I know a lot of states do ask for insurance paperwork liability, however, I am saying that most of the time this is not necessary even when states do ask for this. And many companies do anyways just because it's easier to do so then to make a scene

I figured you would say this, and technically you are correct. However, I would argue that most companies testing robotaxi vehicles, with the exception some cruise and waymo and maybe some others. Are using platforms that will never and will never be intended for L4 use even in the future, and their design intention is only for testing with safety driver and therefore L2. And the majority of AV testing is like so.

The Waymo Pacificas and jaguars are L4 and the latest gen Cruise Bolts are L4 sure... but basically any other vehicle testing a robotaxi like system with a safety driver, those vehicles would be L2 systems.

How can you say that?

Because of what the SAE document says:

"The level of a driving automation system feature corresponds to the feature’s production design intent. This applies regardless of whether the vehicle on which it is equipped is a production vehicle already deployed in commerce, or a test vehicle that has yet to be deployed. As such, it is incorrect to classify a level 4 design-intended ADS feature equipped on a test vehicle as level 2 simply because on-road testing requires a test driver to supervise the feature while engaged, and to intervene if necessary to maintain safe operation."

I get 3 things from this paragraph:
1) Levels are based on the feature's production design intent. ie, What is the feature designed to do? Is it designed to do the entire DDT including the OEDR?
2) The vehicle does not need to be a production vehicle to count as L4. A test vehicle that is not intended to be deployed still counts as L4.
3) A L4 car does not get downgraded to L2 if there is a safety driver.

So based on the SAE statement, I do not think we can say that test vehicles with safety drivers are automatically L2.
 
In a nutshell:

L2 = a driving system that is only designed to do some driving tasks but not all of the OEDR. The human is expected to do the other parts and supervise (or guide) the system.

L4 Beta = a driving system that is designed to do ALL driving tasks including all of the OEDR in a limited ODD. A human may supervise for testing/safety purposes only.

L4 Production (aka driverless robotaxi) = a driving system that is designed to do ALL driving tasks including all of the OEDR in a limited ODD. It is safe and reliable enough not to require any human in the driver seat.
 
Last edited:
  • Informative
Reactions: EVortex
I wonder what CA DMV will say if Tesla is testing Autopilot Level 5 robotaxi functionality in Arizona/Texas? Tesla would tell CA DMV Autopilot is officially Level 2 to avoid regulations, but Level 5 implies it's intended to work in California too. I suppose a simple fix would be for Tesla to say officially Arizona/Texas testing is "only" geofenced Level 4 and not to worry. ;)
 
My point is that Tesla FSD is not a "level-4 design intended ADS features that is being tested"
I think it is. The software is literally called "Full Self Driving." and robotaxi level reliability is planned for late next year. If this isn't the prototype software then how could they possibly meet that deadline?
I'm not sure what you imagine a L5 prototype would look like.
 
I am not convinced... I know a lot of states do ask for insurance paperwork liability, however, I am saying that most of the time this is not necessary even when states do ask for this. And many companies do anyways just because it's easier to do so then to make a scene
What are you not convinced about? I'm just saying what the law is. Companies can and do break the law.
Are you arguing that you can get away with breaking the law until there is a problem? That's what I'm saying about Tesla's testing in CA.
 
Because of what the SAE document says:

"The level of a driving automation system feature corresponds to the feature’s production design intent. This applies regardless of whether the vehicle on which it is equipped is a production vehicle already deployed in commerce, or a test vehicle that has yet to be deployed. As such, it is incorrect to classify a level 4 design-intended ADS feature equipped on a test vehicle as level 2 simply because on-road testing requires a test driver to supervise the feature while engaged, and to intervene if necessary to maintain safe operation."

I get 3 things from this paragraph:
1) Levels are based on the feature's production design intent. ie, What is the feature designed to do? Is it designed to do the entire DDT including the OEDR?
2) The vehicle does not need to be a production vehicle to count as L4. A test vehicle that is not intended to be deployed still counts as L4.
3) A L4 car does not get downgraded to L2 if there is a safety driver.

So based on the SAE statement, I do not think we can say that test vehicles with safety drivers are automatically L2.

Let's break this down

1) Agree. but very very rarely is an test vehicles designed to handle to do the entire DDT... a test vehicle may be designed to drive point to point ... address to address... and handle all driving modes... but that does not mean it is designed to handle the entire DDT

2) The vehicle does not need to be a production vehicle to count as L4. A test vehicle that is not intended to be deployed still counts as L4. Agree, but if that vehicle is lacking certain elements necessary for L4, then it very quickly becomes L2

3) Agree again, but if a system is designed to Always operate with a safety driver... then it is L2. In practice this is almost always the case with test vehicles.

In a nutshell:
L4 Beta = a driving system that is designed to do ALL driving tasks including all of the OEDR in a limited ODD. A human may supervise for testing/safety purposes only.

Again these L4 test vehicles very rarely are designed to handle the full OEDR or full DDT.
What happens if a test vehicle does not have the software module to pull over for emergency vehicles, and they rely on the safety driver to that? does that make it L2 or L4?

I think it is. The software is literally called "Full Self Driving." and robotaxi level reliability is planned for late next year. If this isn't the prototype software then how could they possibly meet that deadline?
I'm not sure what you imagine a L5 prototype would look like.

I already agreed, that the name of the system disagrees with what I am saying, at the start of this conversation I said aside from the name and musks comments. which is well known to be controversial.

I'm just saying what the law is.

I am saying I need full context, please show me the exact language of the law

Are you arguing that you can get away with breaking the law until there is a problem?

No, Absolutely not

What are you not convinced about? I'm just saying what the law is. Companies can and do break the law.
Are you arguing that you can get away with breaking the law until there is a problem? That's what I'm saying about Tesla's testing in CA.
 
What are you not convinced about? I'm just saying what the law is. Companies can and do break the law.
Are you arguing that you can get away with breaking the law until there is a problem? That's what I'm saying about Tesla's testing in CA.

It wouldn't let me edit my previous post... so this is a continuation.

I am saying I am not convinced that Arizona, Florida or other states require a company do anything extra (anything more than just a normal vehicle with no automation) for testing an automated vehicle that can handle all modes of driving, (point to point) and with safety driver.

Say a company like Tesla wanted to test the FSD beta, or Uber/Aurora, or Argo.ai, Zoox, etc wanted to test their test vehicles... I am saying they could legally do so in most US states (including AZ, Florida, etc) with no additional requirements (permits, approval, extra insurance).
 
  • Like
Reactions: WattPwr
Anyway, I will die on this hill. I have eyes and I can see that beta FSD is a prototype AV.

That's fine, but please hear me out.

"prototype AV" is vague and unclear, and I'll let you have that, but that's only if noa is also a prototype AV.
You are suggesting that there is something different about Tesla FSD and Tesla Noa. Or that one of them is driver assist and one of them is not, and that is the idea I have a problem with. Let's go back to EVNow's quote

Do you have a definition of "prototype autonomous vehicle" that somehow includes City NOA but not Freeway NOA ?

and you said

Yes. Freeway NoA is not intended to do all driving tasks, beta FSD is.

You did not explain this, and I will explain why I have a problem with this idea.

First, when you are talking about driving tasks, we should be clear whether we are talking about what SAE refers to as dynamic driving task, or DDT, and which also includes full OEDR... which is different from all driving modes (parking, lane keeping, lane changes, intersections, etc)

Let's start with the first. It sounds like you already have some background on SAE levels and terms, so I am assuming you understand what I am talking about with the OEDR and DDT. I am saying that neither Noa nor Tesla FSD perform the entire DDT. For example they do not do things like pull over for emergency vehicles, they leave the driver to do that, they also leave the driver to monitor the situation to understand known system limitations and take over or pull over when necessary, they don't pull over when the weather becomes to adverse or other situation outside of ODD. But most importantly the driver is still responsible for full OEDR, and is required to always be monitoring, even if the ADS does much of the OEDR.

you may have a counter argument for this, but if you do first ask yourself if this argument also applies to NoA, or is it somehow unique to FSD.

And the second is driving modes, you may say the NoA only does some of the driving modes and FSD is eventually going to do all driving modes. However, performing all driving modes is not something that differentiates driver assist vs something that is not. For example, you can have a driver assist system that performs all driving modes (point to point). Tesla may be one of the first to do this, but many other OEMs are working on releasing products like such. And furthermore, you can also have an autonomous vehicle (not driver asssit) that is NOT capable of all driving modes, for example an L4 highway pilot, or a driverless truck that only goes from onramp to offramp.


You can't argue that Tesla NoA is driver assist and Tesla FSD is not. And if you do, please tell me where and what the line is. and why such a line would make any sense. For example, say NoA is ADAS. Then how about NoA +traffic light control? then NoA + traffic light control + auto lane change on city streets with nav? then how about adding roundabouts? and so on.

ADAS can be designed to make entire trips hands free, address to address, all modes, and many automakers are making such products and labeling very clearly labeling them as driver assist. Tesla is just another OEM making such product, only difference is the misleading naming and advertising from the CEO.

There is only 1 factor that differentiates whether a is product is driver assist or not, and it's simple, whether that system is designed to replace the driver or designed for the driver to be always in the loop and the responsible party.

And finally it sounds like you were suggesting you believe there is virtually no difference (in terms of legality or role between human and driver) between Tesla FSD and most other companies testing prototype AV systems on public roads with safety drivers, and I do agree with you on this.
 
  • Like
Reactions: MP3Mike and WattPwr
Let's break this down

1) Agree. but very very rarely is an test vehicles designed to handle to do the entire DDT... a test vehicle may be designed to drive point to point ... address to address... and handle all driving modes... but that does not mean it is designed to handle the entire DDT

2) The vehicle does not need to be a production vehicle to count as L4. A test vehicle that is not intended to be deployed still counts as L4. Agree, but if that vehicle is lacking certain elements necessary for L4, then it very quickly becomes L2

3) Agree again, but if a system is designed to Always operate with a safety driver... then it is L2. In practice this is almost always the case with test vehicles.

Again these L4 test vehicles very rarely are designed to handle the full OEDR or full DDT.
What happens if a test vehicle does not have the software module to pull over for emergency vehicles, and they rely on the safety driver to that? does that make it L2 or L4?

I think I see where you are coming from. To answer your question: if the test vehicle cannot handle the entire OEDR then it is L2. So if it does not have the module to pull over for emergency vehicles, then it is L2 IMO.

But you are describing test vehicles that were never L4 to begin with. They were always L2.

Here is the key distinction: a test vehicle that cannot do the full OEDR is testing a L2 system. It was L2 to begin with. It is not a L4 system that became L2 for testing purposes.
 
  • Like
Reactions: MP3Mike
You can't argue that Tesla NoA is driver assist and Tesla FSD is not. And if you do, please tell me where and what the line is. and why such a line would make any sense. For example, say NoA is ADAS. Then how about NoA +traffic light control? then NoA + traffic light control + auto lane change on city streets with nav? then how about adding roundabouts? and so on.
This a slippery slope argument and I don't think there is a bright line. NoA is certainly pushing the limits of what I would consider driver assist. To me the only thing that really matters is if it improves safety. NoA might improve safety (I'm not convinced but it is plausible) but I seriously doubt beta FSD improves safety.
And finally it sounds like you were suggesting you believe there is virtually no difference (in terms of legality or role between human and driver) between Tesla FSD and most other companies testing prototype AV systems on public roads with safety drivers, and I do agree with you on this.
So you're arguing that all autonomous vehicle testing regulations are completely unnecessary and can be easily circumvented? If you look at Uber's AV testing saga in CA they made the exact same argument and were shutdown by the DMV. Of course after that Uber killed someone in Arizona while testing their AV.
I think I see where you are coming from. To answer your question: if the test vehicle cannot handle the entire OEDR then it is L2. So if it does not have the module to pull over for emergency vehicles, then it is L2 IMO.

But you are describing test vehicles that were never L4 to begin with. They were always L2.

Here is the key distinction: a test vehicle that cannot do the full OEDR is testing a L2 system. It was L2 to begin with. It is not a L4 system that became L2 for testing purposes.
The problem with this definition is someone could just leave out a single module and test their system as "L2" and then add the module in right before deploying.
All this is moot if testing AVs is no more dangerous than regular driving. Hopefully that is the case but I suspect that Tesla knows it is not and will only release a very scaled back version of the software (i.e. autosteer on city streets).
 
This a slippery slope argument and I don't think there is a bright line. NoA is certainly pushing the limits of what I would consider driver assist. To me the only thing that really matters is if it improves safety. NoA might improve safety (I'm not convinced but it is plausible) but I seriously doubt beta FSD improves safety.

So you're arguing that all autonomous vehicle testing regulations are completely unnecessary and can be easily circumvented? If you look at Uber's AV testing saga in CA they made the exact same argument and were shutdown by the DMV. Of course after that Uber killed someone in Arizona while testing their AV.

The problem with this definition is someone could just leave out a single module and test their system as "L2" and then add the module in right before deploying.
All this is moot if testing AVs is no more dangerous than regular driving. Hopefully that is the case but I suspect that Tesla knows it is not and will only release a very scaled back version of the software (i.e. autosteer on city streets).

Well, I could be wrong. The reason I said that is because I consider pulling over for emergency vehicles to be part of the OEDR. And based on the definition of L2, if the system is missing some OEDR, then it is L2. So it would seem to fall under L2.

But it is worth nothing that the SAE document clearly says that levels are assigned, not measured, based on the design intent of the manufacturer:

Page 30: "Levels are assigned, rather than measured, and reflect the design intent for the driving automation system feature as defined by its manufacturer.

As a practical matter, it is not possible to describe or specify a complete test or set of tests which can be applied to a given ADS feature to conclusively identify or verify its level of driving automation. The level assignment rather expresses the design intention for the feature and as such tells potential users or other interested parties that the feature can be expected to function such that the roles of the user versus the driving automation system while the feature is engaged are consistent with the assigned level, as defined in this document."

So it is possible that if regulators classified a system as L4, then they would not allow the manufacturer to call it L2 just because they removed a piece. And regulators could argue that they are not allowed to deploy the system as L4 since it is not the same system that was tested.