Make your robotaxi predictions for the 8/8 reveal

So Elon says that Tesla will reveal a dedicated robotaxi vehicle on 8/8. What do you think we will see? Will it look like this concept art or something else?

I will say that while this concept drawing looks super cool, I am a bit skeptical if it is practical as a robotaxi. It looks to only have 2 seats which would be fine for 1-2 people who need a ride but would not work for more than 2 people. I feel like that would limit the robotaxis value for a lot of people. Also, it would likely need a steering wheel and pedals for regulatory reasons even if Tesla did achieve eyes-off capability.

So I think this is concept art for a hypothetical 2 seater, cheap Tesla, not a robotaxi.

Could the robotaxi look more like this concept art but smaller? It could look a bit more like say the Zoox vehicle or the Cruise Origin, more futuristic box like shape IMO and seat 5-6 people.

Or maybe the robotaxi will look more like the "model 2" concept:

Other questions:
- Will the robotaxis be available to own by individuals as a personal car or will it strictly be owned by Tesla and only used in a ride-hailing network?
- What will cost be?
- Will it have upgraded hardware? Radar? Lidar? additional compute?
- Will Elon reveal any details on how the ride-hailing network will work?

Thoughts? Let the fun speculation begin!

 

[DanCar]

I see tons of Waymo vehicles driving around San Francisco every weekend. How can it be argued that they don’t have a nice sized lead over Tesla (that have none) at this time? Any argument otherwise feels specious, at best.

One can try to project what will happen in the future. One scenario is that Waymo continues to deploy slowly. At some point maybe 5 years from now, Tesla solves the real full self driving opportunity. Then Tesla is able to cover the entire country quickly, while Waymo is still only serving a relatively small area with expensive equipment.

Yes, Waymo is ahead today. That doesn't mean they will be ahead some years from now. Besides Tesla there are others that are competing, such as Mobileye.

https://www.timesofisrael.com/autonomous-vehicles-expected-on-the-road-in-next-2-years-says-mobileye-ceo/

 

[spacecoin]

One can try to project what will happen in the future. One scenario is that Waymo continues to deploy slowly. At some point maybe 5 years from now, Tesla solves the real full self driving opportunity. Then Tesla is able to cover the entire country quickly, while Waymo is still only serving a relatively small area with expensive equipment.

Yes, Waymo is ahead today. That doesn't mean they will be ahead some years from now. Besides Tesla there are others that are competing, such as Mobileye.

https://www.timesofisrael.com/autonomous-vehicles-expected-on-the-road-in-next-2-years-says-mobileye-ceo/

That’s about as likely as Tesla solves AGI before Google Deepmind. Tesla has to my knowledge ”solved” zero computer vision or ML research problems. If anyone is aware of anything coming out of Tesla that is a new state-of-the-art approach, please let me know what that is/was.

 

[mborkow]

That’s about as likely as Tesla solves AGI before Google Deepmind. Tesla has to my knowledge ”solved” zero computer vision or ML research problems. If anyone is aware of anything coming out of Tesla that is a new state-of-the-art approach, please let me know what that is/was.

That’s what Elon is going to announce on 8/8; they have solved P v NP! I think that’s about as likely as them putting an L4 robotaxi on the road this year or next

 

[JulienW]

That’s what Elon is going to announce on 8/8; they have solved P v NP! I think that’s about as likely as them putting an L4 robotaxi on the road this year or next

Hope he doesn't announce that Grok is ready to figure out FSD L4. That Grok must be on some Heavy Acid Trip.

 

[spacecoin]

An old friend of Elon’s telling it like it is:

https://twitter.com/x/status/1780943993406169534

 

[aataskin]

Tesla does not need to be L5. That may be Tesla's goal, but for Tesla to be successful, they need deployable L4, but with a much , much wider ODD than what Waymo currently operates in.

Put another way, there is a potentially HUGE difference in applicability in different L4 systems. Waymo is currently very narrow ODD. How broadly Waymo is planning to expand that ODD is not clear. Tesla (or anyone else) can achieve a much, much more applicable system, and still "only" be L4.

Quoting from diplomat's post:

• L4 is designed to perform all driving tasks and fallback on its own but only in an artificially limited ODD.
• L5 is designed to perform all driving tasks and the fallback on its own in a wide ODD of a typical human driver (ie everywhere, day, night, all drivable road and weather conditions).

L4 essentially describes Waymo, and it's there already, technology-wise. It won't work everywhere but it works (not perfectly, but well enough already) in the geofenced mapped areas.

Tesla's Vision FSD does not have the concept of artificially limited ODD, if I understand it correctly. It either works everywhere or it doesn't. Therefore for Tesla to be a robotaxi it needs to be L5. If it is L4 it will most likely mean that it's limited to simpler rural areas or highways which is where the current version of FSD seems to struggle the least - but then this is probably not worth the effort, this market is too small. Therefore, with its reliance on Vision only, Tesla needs to be an L5.

 

[diplomat33]

Tesla's Vision FSD does not have the concept of artificially limited ODD, if I understand it correctly. It either works everywhere or it doesn't. Therefore for Tesla to be a robotaxi it needs to be L5. If it is L4 it will most likely mean that it's limited to simpler rural areas or highways which is where the current version of FSD seems to struggle the least - but then this is probably not worth the effort, this market is too small. Therefore, with its reliance on Vision only, Tesla needs to be an L5.

Tesla's approach is very much towards eventually being L5. That's why Tesla relies on vision-only and does not use HD maps or geofences. Tesla did not want anything to restrict the ODD. They wanted to "solve" FSD everywhere.

But if Tesla waits for safety to be " near perfect" everywhere before removing supervision everywhere, they are going to wait a long time. It is inevitable that FSD will work better in some areas than others. You are not going to solve every edge case everywhere.

So I think Tesla has 2 choices if they want to remove supervision "soon":
1) Get to where FSD is "good enough". Remove supervision everywhere knowing that in some areas, FSD will be less safe, but the average safety is "good enough". This has the benefit that Tesla could declare L5 but with the cost of accepting some areas will be less safe than others.
2) Add in geofences. This way, they can overfit the data and get FSD to driverless in the geofences quicker. Essentially, forget the dream of L5, and make FSD into L4.

 

[aataskin]

Tesla's approach is very much towards eventually being L5. That's why Tesla relies on vision-only and does not use HD maps or geofences. Tesla did not want anything to restrict the ODD. They wanted to "solve" FSD everywhere.

But if Tesla waits for safety to be " near perfect" everywhere before removing supervision everywhere, they are going to wait a long time. It is inevitable that FSD will work better in some areas than others. You are not going to solve every edge case everywhere.

So I think Tesla has 2 choices if they want to remove supervision "soon":
1) Get to where FSD is "good enough". Remove supervision everywhere knowing that in some areas, FSD will be less safe, but the average safety is "good enough". This has the benefit that Tesla could declare L5 but with the cost of accepting some areas will be less safe than others.
2) Add in geofences. This way, they can overfit the data and get FSD to driverless in the geofences quicker. Essentially, forget the dream of L5, and make FSD into L4.

I suppose 2) will not arrive before 1) in any case. And L3 will have to arrive before that.

Also, I am not sure it's for Tesla to declare itself an L4 or L5, it will be up to the regulators.

So I think my original point is correct then. Tesla needs to be an L5 for its version of robotaxi to make sense while Waymo will remain at L4 and just keep mapping more places where it makes economic sense to do so.

 

[diplomat33]

Also, I am not sure it's for Tesla to declare itself an L4 or L5, it will be up to the regulators.

In the US, it is up to the manufacturers to declare the SAE level. We use a self-certification system. Regulators can deny or approve the deployment like we see in CA. But the manufacturer declares what the SAE level is. In fact, that is how Tesla has avoided the CA DMV regulations: they just declare that FSD is L2 and therefore exempt.

 

[mborkow]

An old friend of Elon’s telling it like it is:

https://twitter.com/x/status/1780943993406169534

This is 100% true; it’s the 80/20 law. That last 20% is the hard part (and the last quartile of that part is the hardest).

 

[aataskin]

In the US, it is up to the manufacturers to declare the SAE level. We use a self-certification system. Regulators can deny or approve the deployment like we see in CA. But the manufacturer declares what the SAE level is. In fact, that is how Tesla has avoided the CA DMV regulations: they just declare that FSD is L2 and therefore exempt.

So Tesla can identify as an L5 but it won't mean much in reality because as long as the CA DMV is unhappy with the technology it won't be able to deploy robotaxis in California.

In this case I think the bet between spacecoin and Joey D needs to be modified. It's not what Tesla calls itself (L4, L5 Supervised, L6++, etc), it's what jurisdictions allow it. The SAE codes are absolutely irrelevant.

 

[spacecoin]

In this case I think the bet between spacecoin and Joey D needs to be modified. It's not what Tesla calls itself (L4, L5 Supervised, L6++, etc), it's what jurisdictions allow it. The SAE codes are absolutely irrelevant.

Driverless (no safety driver) deployment in any US city, open to anyone in the public, in a geo area approx 225 sq miles or more was what we settled on I guess, by end of 2027. Ten years after Waymo. I give it 20% chance.

 

[JulienW]

So Tesla can identify as an L5 but it won't mean much in reality because as long as the CA DMV is unhappy with the technology it won't be able to deploy robotaxis in California.......

Declaring L5 is absolutely obtuse. You can be L4 and have as few ODDs as you want but still have an ODD(s) fallback. L5 is a "no going back" trap and not likely to see this until we see a compony make a L4 and work there way until they have removed almost all ODDs. This would probably be a decade after that.

 

[diplomat33]

So Tesla can identify as an L5 but it won't mean much in reality because as long as the CA DMV is unhappy with the technology it won't be able to deploy robotaxis in California.

Well, declaring FSD to be L5 would essentially be Tesla communicating that they consider that FSD can do all driving tasks and can be activated everywhere. So it would not be meaningless. But you are correct that to deploy driverless in CA, whether L4 or L5, Tesla would need to get regulatory approval in CA. Specifically, they would first need to get a driverless testing permit from the CA DMV. This would allow Tesla to deploy driverless in a testing capacity but not charge money for it. Then Tesla would need to get a commercial driverless permit from the CPUC before they could charge money for driverless rides. It is a lengthy process as we saw with Cruise and Waymo. And as we saw with Cruise, if Tesla were to have incidents with driverless, there would likely to be pressure to revoke their deployment permits. So Tesla could lose the robotaxi permits if FSD caused too many incidents while driverless.

 

[brkaus]

That’s about as likely as Tesla solves AGI before Google Deepmind. Tesla has to my knowledge ”solved” zero computer vision or ML research problems. If anyone is aware of anything coming out of Tesla that is a new state-of-the-art approach, please let me know what that is/was.

Robots?

Oops. Nevermind.

 

[spacecoin]

Robots?

Oops. Nevermind.

Yeah, love that video. Impressive electro-mechanics (beats the sh!t of out of Optimus) but won't still pass the Coffee test in 15 years.

 

[jabloomf1230]

I see tons of Waymo vehicles driving around San Francisco every weekend. How can it be argued that they don’t have a nice sized lead over Tesla (that have none) at this time? Any argument otherwise feels specious, at best.

By Waymo's own statements, they have 250 driverless cars in SF, of which approximately 100 are on the road at any given time. In contrast, the average number of ride share vehicles on the streets in SF is 5,700 with a peak exceeding 6,500. There also hundreds of traditional taxis.

 

[mborkow]

By Waymo's own statements, they have 250 driverless cars in SF, of which approximately 100 are on the road at any given time. In contrast, the average number of ride share vehicles on the streets in SF is 5,700 with a peak exceeding 6,500. There also hundreds of traditional taxis.

I get they are a much lower density than Uber drivers but all I know is that every time I’m out in the avenues at my favorite shabu-shabu restaurant I see them all over the place…and I haven’t seen even one driverless Tesla.

 

[stopcrazypp]

Quoting from diplomat's post:

• L4 is designed to perform all driving tasks and fallback on its own but only in an artificially limited ODD.
• L5 is designed to perform all driving tasks and the fallback on its own in a wide ODD of a typical human driver (ie everywhere, day, night, all drivable road and weather conditions).

L4 essentially describes Waymo, and it's there already, technology-wise. It won't work everywhere but it works (not perfectly, but well enough already) in the geofenced mapped areas.

Tesla's Vision FSD does not have the concept of artificially limited ODD, if I understand it correctly. It either works everywhere or it doesn't. Therefore for Tesla to be a robotaxi it needs to be L5. If it is L4 it will most likely mean that it's limited to simpler rural areas or highways which is where the current version of FSD seems to struggle the least - but then this is probably not worth the effort, this market is too small. Therefore, with its reliance on Vision only, Tesla needs to be an L5.

Nope incorrect. You can have a SAE L4 system that has no geofences, but it is still SAE L4. One example is one that doesn't work in fog that humans can still drive in. Such a car can have no geofence and drive everywhere in the US, but it'll squarely be SAE L4 given it can't handle fog (which a human can). Of course, most people may colloquially consider that as "L5", but it's not SAE L5.

 

[stopcrazypp]

Robots?

Oops. Nevermind.

As impressive as the Boston Dynamics robots are, they are very much a parallel to Waymo or other fleet robotaxies, in which the price tag is way too high for people in the general public to own widely. For example Spot is $75k. I imagine Atlas's price tag is going to be significantly higher than that.

Tesla's probably trying to have Optimus be more like the Unitree Go1 analog (which launched at $2700). It may be less capable than Spot in some regards, but it can do most of what Spot can do and is affordable enough for widespread use.
This $2,700 robot dog will carry a single bottle of water for you