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!

 

[tonylee5566]

Tesla Going All In on Robotaxis ??

 

[aataskin]

Looks like we are seeing Elon's potential mistake of betting the entire farm on Robotaxis over a near sure fired hit "Model 2" first. Little doubt that a Model 2 would be HUGE and likely sell in the millions. People WANT this car. But Elon wants to gamble on L4 robotaxies first and this is a HUGE unknown full of unknowable unknowns. Now today 18 year Drew Baglino, probably the straightest arrow at Tesla is leaving. While some may say that it is just part of the 10% reduction this is NOT likely since you NEVER cut strategic value. He is probably unhappy with the direction Elon is forcing with such a high stakes gamble. We know there must be people high up unhappy since the story about the Model 2 was leaked to the media.

If Tesla beats the odds and perfects a wide ODD L4 robotaxi it will be a huge value. But if it turns into the same old "by the end of this year" year after year TSLA will tank. Tesla should have put the Model 2 out first so we could see MUCH NEEDED sales and then work towards a robotaxi WHILE selling millions of Model 2s and making billons of $.

Does anybody here understand the logic? One would think an L4 robotaxi is a natural bet if your existing cars are already L3 and there's reasonable expectation that you have the technology to get to an L4 in the near future. But this is not where Tesla is right now, not even in the ballpark. I think I must be missing something.

 

[JulienW]

Does anybody here understand the logic? One would think an L4 robotaxi is a natural bet if your existing cars are already L3 and there's reasonable expectation that you have the technology to get to an L4 in the near future. But this is not where Tesla is right now, not even in the ballpark. I think I must be missing something.

....but we are sure to have L4 FSD "by the end of the year".

And who the HELL wants an affordable $25,000 Tesla......other than Wall Street, 1M Americans, 1M Europeans, 1M Chinese, 1M Indians, 1M.........

 

[spacecoin]

Does anybody here understand the logic? One would think an L4 robotaxi is a natural bet if your existing cars are already L3 and there's reasonable expectation that you have the technology to get to an L4 in the near future. But this is not where Tesla is right now, not even in the ballpark. I think I must be missing something.

Yeah... That's what I'm saying.

In L3 the person in the car doesn't need to:
a) Pay attention to the road. You can read a book, watch a movie et.c
b) Take over when the car is driving. The car will request for you to start performing the OEDR (look at the road) and then take over after the driver is ready, typically 5-10 seconds later. This happen when the car is leaving the ODD (typically when the speed increases to more than the system is capable of, like over 30 mph or the car need to leave the highway).

When the system is driving (in L3 and above), the manufacturer is liable. In L2 the system is never formally driving. It's assisting.

To get from "oh a zero intervention drive, yay" to 10000+ drives in a row is not likely happening on this system, given that it took Tesla three year to get from 5-10 miles to 50 miles per DE.

 

[aataskin]

Yeah... That's what I'm saying.

In L3 the person in the car doesn't need to:
a) Pay attention to the road. You can read a book, watch a movie et.c
b) Take over when the car is driving. The car will request for you to start performing the OEDR (look at the road) and then take over after the driver is ready, typically 5-10 seconds later. This happen when the car is leaving the ODD (typically when the speed increases to more than the system is capable of, like over 30 mph or the car need to leave the highway).

When the system is driving (in L3 and above), the manufacturer is liable. In L2 the system is never formally driving. It's assisting.

To get from "oh a zero intervention drive, yay" to 10000+ drives in a row is not likely happening on this system, given that it took Tesla three year to get from 5-10 miles to 50 miles per DE.

So I am not missing anything then. This is crazy..

 

[Joey D]

It’s simple. General autonomy is most likely 10+ years away. Someone will “get there” someday.

That's possible, but I'm betting it's more like 2 years away. By that I mean I'm betting we will see a "general autonomy vehicle" on roads where the regulators allow it. (Think: Tesla vehicles will be capable of autonomous driving in North America, but not all jurisdictions will allow it.)

In the mean time you try to get to market with a driverless product that is deployable, safe, appreciated and profitable.

That is one approach...not the only approach. I would argue that approach Waymo is taking will lead to a dead end when you try to scale.

“From driverless nowhere to everywhere”, is that your path for Tesla? It sounds very unlikely to me. Walk me through it?

The path is not rocket science (by that I don't mean it will be easy to execute...but the path itself is obvious):

1) Now: Supervised everywhere.
2) Late 2024 / Early 2025: Specific fleet of vehicles approved for "Level 3 Self-Driving operation" in specific jurisdictions...(with a human behind the wheel as back-up).
3) 2026: Data is presented to regulators proving that human back-up is no longer needed (so can go to Level 4/5). (Data on accidents, interventions, disengagements, etc.)
4) 2025 + Repeat 2 and 3 additional more jurisdictions.
5) 2026 + Other jurisdictions begin to approve based on historic performance in other jurisdictions (or there is just blanket national approval)

 

[JulienW]

So I am not missing anything then. This is crazy..

Actually it was crazy (and fun) but now it is bordering on fiduciary foolishness. We ALL want FSD to succeed and ONE day we will all have L4 but........Elon is so insecure that another compony will beat Tesla to nation wide (ODD) L4 that he can't see anything else.

No one is about to beat Tesla and Tesla has by far the best L2 system. Also selling a Model 2 first actually gives Tesla an advantage. Big bucks coming into the bank and continue perfecting FSD"s NNs for a couple more years and THEN have Autonomy day when FSD is looking close.

That will STLL beat everyone else and the announcement would be from a position of strength and not look like desperation.

 

[Daniel in SD]

Never said anything about Waymo.

It's a good point about acceptance of what we as humans see as totally stupid errors. Can't see it happening myself until robot racecars (with G force limiters) can beat humans on sight unseen rally courses. That will mean they can read the road very very well.

Then they will need to learn assertion of rights to coexist with human drivers. Good luck with that. It's the end for humans driving when they get to the point they can coexist with one another.

Mapping is a red herring. Of course the car needs to know what is supposed to be there.

It's how it reacts to the unexpected that has yet to be solved.

It needs AI on the level of interpreting a Larsen cartoon to do the job.

My point is that you seem to be asking for super human performance in all aspects in driving when all that is actually required is superhuman safety. Waymos are driving around with no driver today. Assuming they actually are safe (they haven't driven enough miles to prove this IMO) then that is proof that it is possible to replace human drivers with current technology.

 

[spacecoin]

1) Now: Supervised everywhere.
2) Late 2024 / Early 2025: Specific fleet of vehicles approved for "Level 3 Self-Driving operation" in specific jurisdictions...(with a human behind the wheel as back-up).

So how is Tesla getting from 95% of the drives to 99.99999% of the drives without intervention in 6-12 months when it took 3 years to from 5-10 miles to here. What road types and speeds are you thinking for "eyes off". The human is not a backup for the system in L3, btw. The system needs to drive always in the ODD and never fail. Handover typically happens when the system leaves the ODD (eg exits the highway).

3) 2026: Data is presented to regulators proving that human back-up is no longer needed (so can go to Level 4/5). (Data on accidents, interventions, disengagements, etc.)

There is currently no legal framework for personally owned wide-ODD autonomous cars anywhere in the world.

In places that have narrow ODD L3 legislation, such as in Germany, the regulators haven't adressed the liability as it would require a major rewrite of the traffic laws. I'm thinking 2035 at the earliest. The liability problem exists in the US as well. If you're in in the driver's seat of an L3 in autonomous mode, you don't want to be charged with manslaughter just because the CAR MANUFACTURER's driving system kills a person.

Until then, Tesla will need to operate their own fleet in jurisdictions that allow for L4.

4) 2025 + Repeat 2 and 3 additional more jurisdictions.

5) 2026 + Other jurisdictions begin to approve based on historic performance in other jurisdictions (or there is just blanket national approval)

Where? Most jurisdictions are subject to UNECE regulations, and they can't even decide if hands-off L2 is a good idea.

 

[Joey D]

So how is Tesla getting from 95% of the drives to 99.99999% of the drives without intervention in 6-12 months when it took 3 years to from 5-10 miles to here.

In about 1 year, with FSD12 architecture, Tesla exceeded the prior multi-years of development that culminated with FSD 11. And that year was done mostly without the large increase in compute that Tesla now has.

Better model. More Data, More Compute, and efficient training pipeline. That's how.

Will they succeed? We'll see.

So tell me..."what is the path for Waymo to reach general autonomy" (if that is in fact their goal beyond limited city deployments.)

What road types and speeds are you thinking for "eyes off". The human is not a backup for the system in L3, btw. The system needs to drive always in the ODD and never fail. Handover typically happens when the system leaves the ODD (eg exits the highway).

I'm guessing they will start with city streets (lower speeds), but ultimately Tesla's internal data will tell them where they will start in terms of chance of success.

There is currently no legal framework for personally owned wide-ODD autonomous cars anywhere in the world.

This is why I said "specific fleet" of vehicles. Meaning: Tesla owned / managed in order to go through the first regulatory approvals and data collection.

Until then, Tesla will need to operate their own fleet in jurisdictions that allow for L4.

I agree. But it will be the same software that can be rolled out to the public. So when jurisdictions give approval for L4, anyone with a tesla will be able to utilize L4 software in that jurisdiction.

Where? Most jurisdictions are subject to UNECE regulations, and they can't even decide if hands-off L2 is a good idea.

US is not subject to UNECE.

This will be a US rollout first, obviously.

 

[aataskin]

In about 1 year, with FSD12 architecture, Tesla exceeded the prior multi-years of development that culminated with FSD 11. And that year was done mostly without the large increase in compute that Tesla now has.

Better model. More Data, More Compute, and efficient training pipeline. That's how.

Will they succeed? We'll see.

So tell me..."what is the path for Waymo to reach general autonomy" (if that is in fact their goal beyond limited city deployments.)


I'm guessing they will start with city streets (lower speeds), but ultimately Tesla's internal data will tell them where they will start in terms of chance of success.


This is why I said "specific fleet" of vehicles. Meaning: Tesla owned / managed in order to go through the first regulatory approvals and data collection.

I agree. But it will be the same software that can be rolled out to the public. So when jurisdictions give approval for L4, anyone with a tesla will be able to utilize L4 software in that jurisdiction.

US is not subject to UNECE.

This will be a US rollout first, obviously.

London is 25% of all taxis in England and Wales. So my guess is large cities are probably around 70% of all demand for this kind of technology. Maybe this will be just enough for Waymo?

Why are you focusing on Tesla reaching the L4 level? L4 is enough for Waymo because they are happy to stay within the geofenced area. For Tesla robotaxi, since it does not rely on geofencing, aren't they supposed to be L5 to be certified as robotaxis?

 

[JulienW]

....Why are you focusing on Tesla reaching the L4 level? L4 is enough for Waymo because they are happy to stay within the geofenced area. For Tesla robotaxi, since it does not rely on geofencing, aren't they supposed to be L5 to be certified as robotaxis?

L4 dosn't require geo fencing and the ODD can be as wide as you want. L5 on the other hand requires that there be NO ODD whatsoever. That bar is near impossibly high and likely WAY over a decade or 2 away.

 

[spacecoin]

In about 1 year, with FSD12 architecture, Tesla exceeded the prior multi-years of development that culminated with FSD 11. And that year was done mostly without the large increase in compute that Tesla now has.

Better model. More Data, More Compute, and efficient training pipeline. That's how.

Will they succeed? We'll see.

They are at the same place, more or less as with 11.4. There's not that much of a difference (like 10x), but whatever. At this pace they'll be autonomous by the year 2100. Every order of magnitude gets harder and costlier to solve.

So tell me..."what is the path for Waymo to reach general autonomy" (if that is in fact their goal beyond limited city deployments.)

I don't believe in science fiction. I think it's multiple decades out.

I'm guessing they will start with city streets (lower speeds), but ultimately Tesla's internal data will tell them where they will start in terms of chance of success.

This is why I said "specific fleet" of vehicles. Meaning: Tesla owned / managed in order to go through the first regulatory approvals and data collection.

If we see a Tesla robotaxi deployment in a single city before 2030 with more than 50 cars, I'd be amazed. They have zero cars capable of L4 as of today. 95% reliability is a joke. You need 99.99999%

Can you name one system in any vertical that evolved like that without a several phases of complete redesign. I can't think of anything.

I agree. But it will be the same software that can be rolled out to the public. So when jurisdictions give approval for L4, anyone with a tesla will be able to utilize L4 software in that jurisdiction.

Not the tesla's on the road today, zero* chance Tesla takes on liability. Also, imho, zero* chance they are technically able to.

US is not subject to UNECE.
This will be a US rollout first, obviously.

So when you say jurisdiction you mean different states in the US?

*) you know what I mean.

 

[HitchHiker71]

So I am not missing anything then. This is crazy..

It's a huge bet without any doubt - and with a lot of variables and unknowns in play without any doubt. Purportedly Musk has cancelled the NV9 program - the M2 - and prioritized a huge datacenter build in GigaAustin that he is pushing to have complete by end of August - tied to the robotaxi announcement upcoming on 8/8 - ten to one this date changes due to delays. This would seem to mean that, despite all claims that compute is no longer a constraint for FSD, additional compute is needed? I really do not understand why a multi-billion dollar company with a hoard of cash cannot do both. Why cancel the M2 - which is an almost certain win for traditional/normal consumers who buy up Corollas/Civics/Sentras/etc. by the millions? Tesla has always seemed like a one trick pony in some respects when it comes to major engineering projects. They have a small team of core engineers that works each major project - and that team cannot span across different projects as a result. What doesn't make sense to me is that building a robotaxi datacenter has nothing to do with nextgen manufacturing - two entirely different sets of skills involved. So IMHO there's more to this cancellation of NV9 than just resourcing/funding. Otherwise they would be doing both. Is the unboxed methodology not really feasible? Are the savings not going to materialize based upon due diligence and therefore it's not going to be feasible to bring a $25k vehicle to market? We're missing key information here IMHO.

IMHO there seems to be a belief within Tesla that TaaS is going to be the big game changer if they can get autonomous driving to actually work. But even if that's the case - people aren't going to stop buying cars overnight. It will take years for TaaS to take hold and for consumers to adjust. In the meantime, why not make tens of billions of dollars in revenues, even assuming low margins, selling an M2 vehicle? If TaaS does not take hold sooner rather than later, then you've got all bases covered. Personally, I think Drew left because there's a fundamental shift going on within Tesla away from vehicle manufacturing and toward software (TaaS/robotaxi) that Musk has pronounced and is now pushing hard toward, why would a power train expert want to stay with this in mind?

 

[MD70]

I hope someone has the courage to ask on the earnings call what will happen if FSD is not approved in the next 5-10 years and tesla has only two mass market vehicles.

 

[enemji]

until robot racecars (with G force limiters) can beat humans on sight unseen rally courses

I assume sight unseen applies to both human drivers and robocars.

 

[MD70]

My prediction is that Because the Reuters article forced Elons hand between now and the 8/8 the team will be frantically working to be able to show a product.

Elon will make many far fetched claims about what the car will be able to do and give unrealistic timelines.

 

[QtownTaxiGreg]

I really do not understand why a multi-billion dollar company with a hoard of cash cannot do both

Space travel is expensive too?

I assume sight unseen applies to both human drivers and robocars.

Yes. Fair is fair.

The robo driver has to rely on visuals/ radar for safety as it goes around the next bend as do humans.

I had an idea what the massive data centre could be aiming at. Real time reportage of map to actual footage discrepancies and driver interventions pinging back to other Teslas in the area. You can 'cheat' your way to a very much safer car in environments with lots of Teslas running around partially autopiloted. Those same cars are just as naive out in the country all on their own.

 

[Joey D]

London is 25% of all taxis in England and Wales. So my guess is large cities are probably around 70% of all demand for this kind of technology. Maybe this will be just enough for Waymo?

Europe and USA are completely different.

In USA, there would be high demand for Level 3+ driving not only within cities, but outside of cities. (Suburbs, etc.)

Why are you focusing on Tesla reaching the L4 level?

Because L4 does not require a driver to be present / available to take over. A L4 vehicle must be able to "get itself out of trouble". At worst, it must be able to pull over all on its own.

L4 is enough for Waymo because they are happy to stay within the geofenced area.

Geofencing does not have anything to do with autonomy levels. Waymo is likely limiting itself to cities for a few reasons:
1) Density of people "hailing" rides.
2) Lower speeds means less likely to have catastrophic results when mistakes happen.

For Tesla robotaxi, since it does not rely on geofencing, aren't they supposed to be L5 to be certified as robotaxis?

No, Level 4 or 5 are "can drive all on its own without a driver present.

I believe (have to double check) the primary difference is that Level 5 must be able to handle "all conditions", where Level 4 must be aware of what conditions it cannot drive in and "disable itself."

 

[Joey D]

They are at the same place, more or less as with 11.4.

Disagree. I have driven both 11.x and 12.x. No comparison IMO. Have you driven either?

I don't believe in science fiction. I think it's multiple decades out.

Agree to disagree here. A decade ago, people thought the "current state" of AI (LLMs, generating video, etc.) was also "multiple decades out." Disruption and innovation happens fast.

If we see a Tesla robotaxi deployment in a single city before 2030 with more than 50 cars, I'd be amazed.

I'd be surprised if that doesn't happen.

Can you name one system in any vertical that evolved like that without a several phases of complete redesign. I can't think of anything.

How many iterations do you think Tesla's autopilot has gone through...none? They have in fact had several completel overhauls...adding occultion networks, path predicting, from "static images" to "video"....with USS and radar and without....to end to end Nets....

Not the tesla's on the road today, zero* chance Tesla takes on liability. Also, imho, zero* chance they are technically able to.

You're entitled to your opinion of course.

So when you say jurisdiction you mean different states in the US?

Yes. Tesla cannot even turn on FSD in Europe yet. (They are collecting millions of miles data of course....)