Poll: Waymo wide or Tesla level 3 first?

[powertoold]

I don't think Tesla will ever release "level 3." If their system is good enough for level 3, they'll just go straight to FSD, which is what they're trying to do. That's why I think the SAE designation doesn't apply to Tesla's approach. Tesla knows that if vision becomes good enough for "level 3," then it's good enough for "level 5."

 

[diplomat33]

Like @diplomat33
The problem is how it can expand quickly the geofenced area to cover new intersections, environments... once one area is perfected.

A new area will not be 100% new scenarios though. It will contain some repeat scenarios already perfected in the previous area. So you are not starting from scratch every time. But it also depends on where the new area is. For example, if I perfect FSD in a very urban area and then I move to a mostly rural area, yes, I will encounter a ton of new cases never seen before. But if I perfect FSD in an urban area and move to another urban area that is very similar, I will probably encounter a ton of "old cases" that I can already handle. A lot of the NN and software that Waymo has already perfected will carry over to new areas. So the time it takes to perfect FSD in the new area will probably go down as you expand.

 

[ThomasD]

Keeping the sensors and cameras clean all the time would be difficult Avoid Grasshopper swarms

 

[Tam]

...Tesla knows that if vision becomes good enough for "level 3," then it's good enough for "level 5."

I agree that it's unlikely Tesla will release a Level 3, 4, or 5 in the years to come:

1) At level 2 and below, it's the driver's responsibility. From Level 3 and above, it's the driver's responsibility so I doubt Tesla is willing like to take on the responsibility at Level 3 and above while on beta.

2) At level 2 and below, Tesla doesn't need to file California Annual Disengagement Report because if an accident happens, it's those human drivers who have passed a driver license test will have to answer why they didn't override the system. No disengagement reports mean there's no bad reputation of whether human drivers overriding the system or not.

3) From Level 3 and above, the responsibility is now shifted to the machine but these machines have never passed a driver license test so California wants to know how many disengagements annually to verify whether it should still be Level 3 or be demoted to level 2.

4) Even when Tesla will release level 5 FSD, Tesla still wants it to be classified as level 2 with human "supervision" or "supervised FSD". Since it's level 2 "supervised FSD", there's no need to report annual disengagements either. That means all those smart summon on youtube without a human driver inside the car don't have to be reported to the government either.

 

[powertoold]

4) Even when Tesla will release level 5 FSD, Tesla still wants it to be classified as level 2 with human "supervision" or "supervised FSD".

That's why it's silly to fit all our thoughts about FSD into the SAE terms.

 

[diplomat33]

I don't think Tesla will ever release "level 3." If their system is good enough for level 3, they'll just go straight to FSD, which is what they're trying to do. That's why I think the SAE designation doesn't apply to Tesla's approach. Tesla knows that if vision becomes good enough for "level 3," then it's good enough for "level 5."

The SAE levels are agnostic. They don't care what approach you are using to get to FSD. Whether you are using lidar or just a camera-only approach, the SAE levels apply the same to both. They simply describe your FSD based on certain criteria. And if Tesla skips L3 and goes straight to L5, the SAE levels still have the same definitions.

 

[powertoold]

And if Tesla skips L3 and goes straight to L5, the SAE levels still have the same definitions.

Yes, SAE definitions don't change, but they're not great at describing Tesla's progress. It's the main reason many of us still think Tesla is behind.

 

[Tam]

Yes, SAE definitions don't change, but they're not great at describing Tesla's progress. It's the main reason many of us still think Tesla is behind.

The problem of business naming is they do not need to adhere to the common understanding of that name.

Tesla FSD is one of them.

When you buy FSD, you can read its footnote that humans still have to drive for the FSD! That's the opposite of the meaning of FSD!

It should say "Supervised FSD" or "Level 2 FSD" and not just a plain FSD.

No matter how inferior Tesla FSD is, it's still better in terms of practicality.

I would love to have Waymo L4 jus to drive from home to work, school, shopping and other predetermined paths only. However, the chance of paying for it is slim because it only wants to deal with big business like fleets for now.

As a consolation, I can buy Tesla system freely.

If owners are not nerdy type, they would be frustrated with so many imperfections from the Tesla system.

However, I love to be a beta tester so I have no problem with Tesla beta program and I do think it makes my driving much safer with each release of new features.

So although the real FSD will be a long way from now, but I'll take supervised FSD anytime and it's good to know that could be this year!

 

[DanCar]

I think if Waymo focused on trucks on limited access freeway, level 3 or 4, they could easily be successful in that market segment. Level 3 the driver has to be able to take control put doesn't have to pay attention, level 4 means the driver can sleep.