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!

 

[petercc9]

I'd never buy stock in an auto company. Not even one that poses as a software one. You need to deploy insane amounts of capital and the only thing that comes out of it is that you can participate in a price war with a commodity product "the EV" that is more or less the same as everyone else's.

How is that any different than companies that make something as simple as paperclips. Almost every industry is a "run to the bottom". Only suckering people into brand loyalty is what gets you good margins when you are somebody like Skippy and can charge an extra dollar for peanutbutter vs a no-name brand.

 

[petercc9]

I disagree here. The majority of urban "taxi" rides are one or two people, point to point.

I agree. Sure rides by a family of tourists will likely require more than 3 seats but most taxi rides are not dealing with the Orlando crowd.

I'm sure Uber has plenty of reliable data on what is needed.

 

[petercc9]

Agree and disagree. If Waymo continues to rely on pre-mapped roads, I don't think it's reasonable think that can be scaled to anything more than a handful of cities. It's not just the initial mapping to worry about, but the constant change (maintenance) that has to be dealt with.

If at some point Waymo is able to ditch the pre-mapping pre-requisite? Sure.

Seems Huawei is throwing some serious shade on all the "LIDAR is the only way" fanboys. Maybe we know the real reason Waymo has been slow to expand.

Huawei scraps expensive precision maps in self-driving cars

Chinese tech companies seek alternative navigation systems

asia.nikkei.com

"High-definition maps are very expensive," said Huawei senior executive Yu Chengdong at Auto Shanghai show on April 16. "We have been collecting data in Shanghai for one or two years, but we still have not been able to cover all 9,000 kilometers of the city's roads."

The survey cost for a map with 10-cm accuracy is 10 yuan ($1.44) per kilometer, but for accuracy to 1 cm costs up to 1,000 yuan, according to a 2021 white paper by industry group China Industry Innovation Alliance for Intelligent and Connected Vehicles.

 

[spacecoin]

Seems Huawei is throwing some serious shade on all the "LIDAR is the only way" fanboys. Maybe we know the real reason Waymo has been slow to expand.

First remove driver, then scale down the system such as remove maps. That's the order.

 

[spacecoin]

How is that any different than companies that make something as simple as paperclips. Almost every industry is a "run to the bottom". Only suckering people into brand loyalty is what gets you good margins when you are somebody like Skippy and can charge an extra dollar for peanutbutter vs a no-name brand.

It isn't. In software, the cost to create a new "paper clip" is close to zero though, so it works great. Plus subscription model.

 

[Joey D]

First remove driver, then scale down the system such as remove maps. That's the order.

That is not the only valid approach, surely you agree?

"Don't require maps at all or require your vehicles to be outfitted with super high powered computers and sensors. Improve system to the point where you can ultimately remove the driver" is another valid approach.

Each approach has pros and cons. Fair enough?

 

[spacecoin]

That is not the only valid approach, surely you agree?

"Don't require maps at all or require your vehicles to be outfitted with super high powered computers and sensors. Improve system to the point where you can ultimately remove the driver" is another valid approach.

Each approach has pros and cons. Fair enough?

Let me put it like this: Everyone that has gotten to the level of safety needed to remove the driver uses expensive sensing and hd-maps.
Until someone actually gets to robotaxi levels of reliability using another approach, it is in my view not yet proven to be viable. It's more of an ambition, hope or idea at this point.

If you prioritise time to market and safety, I'm sure most agree with that you need to make the product deployable first, and save costs second.

Perhaps vision-only will pan out, but it's unclear if and when at this point.

In the meantime Waymo and a few others are commercially deployed across quite a few cities. 15+ I believe at this point, and since 2017.

 

[enemji]

Seems Huawei is throwing some serious shade on all the "LIDAR is the only way" fanboys. Maybe we know the real reason Waymo has been slow to expand.

Huawei scraps expensive precision maps in self-driving cars

Chinese tech companies seek alternative navigation systems

asia.nikkei.com

The survey cost for a map with 10-cm accuracy is 10 yuan ($1.44) per kilometer, but for accuracy to 1 cm costs up to 1,000 yuan, according to a 2021 white paper by industry group China Industry Innovation Alliance for Intelligent and Connected Vehicles.

For reference, the Tomahawk has an accuracy of 16 feet. Just saying.

 

[stopcrazypp]

Let me put it like this: Everyone that has gotten to the level of safety needed to remove the driver uses expensive sensing and hd-maps.
Until someone actually gets to robotaxi levels of reliability using another approach, it is in my view not yet proven to be viable. It's more of an ambition, hope or idea at this point.

If you prioritise time to market and safety, I'm sure most agree with that you need to make the product deployable first, and save costs second.

Perhaps vision-only will pan out, but it's unclear if and when at this point.

In the meantime Waymo and a few others are commercially deployed across quite a few cities. 15+ I believe at this point, and since 2017.

But the former approach is already deployable as a door-to-door L2 system (deployed to hundreds of thousands if not over a million vehicles already), which makes it commercially viable. In fact this approach likely is more net profitable than deploying robotaxies in extremely limited areas (due to scale and eliminating the cost of maintaining a fleet). That's what Huawei and the Chinese makes are looking to do first. Those consumer vehicles also allow data gathering to improve the NNs (akin to how Tesla sets triggers for specific events they want to gather and the cars will send it to them).

If the ultimate goal is L3/L4 with no geofences or L5, then being able to drive without relying on HD maps is something that has to be solved anyways. If it is unsolvable then you stay as a door-to-door L2 system.

On the flip side, I know the argument (as brought up above by others) is that it's technically possible to map the entire US (or whatever country you are talking about), but the problem is that when you design the system to place high confidence in HD maps (given it's designed as a prerequisite for operation) it's going to naturally rely on them (so when things like construction zones come up, it's going to be predisposed to drive through them). This is putting aside that there will always be roads where it is not commercially viable to map (for example dead end public roads that lead to only one or two residents).

 

[Madmolecule]

My prediction is on the 8/7 a fleet of about 20,000 of them will leave Austin for a US tour. Free rides for all influencers. He might even add cold air thrusters, cyber mega fart mode, disco lights, CatQuest San Andreas Edition, a Bot will be included in each one to open doors, and handle luggage. No tipping allowed. It will have a cyber structural battery pack that is half full. Good news is design is done and staff is 90% full. They even plan to add a couple new commercials on tunein. but this is just my single source of truth.
luckily I did not claim any robotaxi income on my taxes today for 2023. I tried to tell my accountant it's like buying horse, (except the horse can drive itself) but no kickstarter deduction. But I should plan on a bid 4th qtr, unless its just hype and a stock rescue

there will only be a kick the can on down the road on 8/8, if Elon is still at Tesla

 

[Daniel in SD]

But the former approach is already deployable as a door-to-door L2 system (deployed to hundreds of thousands if not over a million vehicles already), which makes it commercially viable. In fact this approach likely is more net profitable than deploying robotaxies in extremely limited areas (due to scale and eliminating the cost of maintaining a fleet). That's what Huawei and the Chinese makes are looking to do first. Those consumer vehicles also allow data gathering to improve the NNs (akin to how Tesla sets triggers for specific events they want to gather and the cars will send it to them).

If the ultimate goal is L3/L4 with no geofences or L5, then being able to drive without relying on HD maps is something that has to be solved anyways. If it is unsolvable then you stay as a door-to-door L2 system.

On the flip side, I know the argument (as brought up above by others) is that it's technically possible to map the entire US (or whatever country you are talking about), but the problem is that when you design the system to place high confidence in HD maps (given it's designed as a prerequisite for operation) it's going to naturally rely on them (so when things like construction zones come up, it's going to be predisposed to drive through them). This is putting aside that there will always be roads where it is not commercially viable to map (for example dead end public roads that lead to only one or two residents).

Are you saying Waymo is not safe enough?
Presumably they could program it to not drive through construction zones by seeing that the construction zone does not match the HD map. I don't think it's HD maps that cause it to drive through construction zones. It's that recognizing construction zones is extremely difficult. HD maps are just more data and the consensus is that more data is good for current AI technology.

 

[spacecoin]

But the former approach is already deployable as a door-to-door L2 system (deployed to hundreds of thousands if not over a million vehicles already), which makes it commercially viable.

Yes, you don't need a rocket to reach the tree tops. You can just climb the tree. But people in the tree tops find it that it's not a viable path to space.

If we are discussing "viable approaches", we need to talk about the same thing: driverless autonomous cars otherwise the discussion is quite pointless.

People have been selling ADAS features in different ODD sizes for a decade, so it is indeed a viable business for car makers.

I personally don't believe that a more and more robust ADAS is a good way to autonomy. People will over-trust the system due to automation complacency long before it's even 10% of the human safety level.

 

[stopcrazypp]

Are you saying Waymo is not safe enough?
Presumably they could program it to not drive through construction zones by seeing that the construction zone does not match the HD map.

That's the theory and it has some merit, but given it just happened with Waymo, reality doesn't seem to match theory.

I don't think it's HD maps that cause it to drive through construction zones. It's that recognizing construction zones is extremely difficult. HD maps are just more data and the consensus is that more data is good for current AI technology.

But when you have more data you necessarily have to weigh it. And when the maps are prerequisite, you most likely will place significantly higher weight on the map than if it was truly just optional.

I see some parallels with the MachE accident that happened where the car just slammed into a stationary car. Knowing now the road was actually illuminated, I suspect the fact that the system had radar and the car was stationary may have played a factor. Basically if the car was vision-only, it would either reduce the speed or disable the system if it had trouble seeing the road. Instead the car proceeded, likely because it has radar. I see a similar logic with how people like radar because it allows them to continue to operate ACC in dense fog even when they can't see the cars ahead, but the flip side is it's dangerous to do because there can be obstacles the radar miss and it gives you a false sense of security.

https://www.google.com/amp/s/insideevs.com/news/715922/ford-mustang-mach-e-bluecruise-ntsb-investigation/amp/

Also the other factor is that for the former approach, you don't make the system L3+ until it can detect such zones, but for the latter you do and rely on that HD map difference detection (which apparently doesn't solve the problem).

 

[spacecoin]

That's the theory and it has some merit, but given it just happened with Waymo, reality doesn't seem to match theory.

But when you have more data you necessarily have to weigh it. And when the maps are prerequisite, you most likely will place significantly higher weight on the map than if it was truly just optional.

I see some parallels with the MachE accident that happened where the car just slammed into a stationary car. Knowing now the road was actually illuminated, I suspect the fact that the system had radar and the car was stationary may have played a factor. Basically if the car was vision-only, it would either reduce the speed or disable the system if it had trouble seeing the road. Instead the car proceeded, likely because it has radar. I see a similar logic with how people like radar because it allows them to continue to operate ACC in dense fog even when they can't see the cars ahead, but the flip side is it's dangerous to do because there can be obstacles the radar miss and it gives you a false sense of security.

https://www.google.com/amp/s/insideevs.com/news/715922/ford-mustang-mach-e-bluecruise-ntsb-investigation/amp/

Also the other factor is that for the former approach, you don't make the system L3+ until it can detect such zones, but for the latter you do and rely on that HD map difference detection (which apparently doesn't solve the problem).

So if a Waymo fails or stalls 1 time in 10000 around certain types of construction, your conclusion is that HD-mapping doesn't contribute to reliability? I don't get it.

 

[Daniel in SD]

That's the theory and it has some merit, but given it just happened with Waymo, reality doesn't seem to match theory.

But when you have more data you necessarily have to weigh it. And when the maps are prerequisite, you most likely will place significantly higher weight on the map than if it was truly just optional.

I see some parallels with the MachE accident that happened where the car just slammed into a stationary car. Knowing now the road was actually illuminated, I suspect the fact that the system had radar and the car was stationary may have played a factor. Basically if the car was vision-only, it would either reduce the speed or disable the system if it had trouble seeing the road. Instead the car proceeded, likely because it has radar. I see a similar logic with how people like radar because it allows them to continue to operate ACC in dense fog even when they can't see the cars ahead, but the flip side is it's dangerous to do because there can be obstacles the radar miss and it gives you a false sense of security.

https://www.google.com/amp/s/insideevs.com/news/715922/ford-mustang-mach-e-bluecruise-ntsb-investigation/amp/

Also the other factor is that for the former approach, you don't make the system L3+ until it can detect such zones, but for the latter you do and rely on that HD map difference detection (which apparently doesn't solve the problem).

If you’re waiting until AI can deal with construction zones 100% of the time you’ll never deploy.
What if Waymo is much safer than the average human? They are racking up the miles.

 

[JulienW]

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 $.

 

[spacecoin]

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 $.

The problem is that Elon always bets "the entire farm". Eventually you lose. The likelihood of losing goes up by a lot when you put a few unsolved research problems in the middle of your trajectory.

I heard a story on the All-in pod that he doubled his stakes in a poker game until he won. Unfortunately there are no re-ups in business on this level. The average Joe will lose a big chunk of their pensions and savings if the AI story of Tesla doesn't deliver.

 

[petercc9]

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 $.

They have to get FSD working because they need some unique "shtick" to differentiate themselves from Chinese EV makers that are currently racing to the bottom with prices. They aren't going to survive on brand name alone. The US is only 33% of their market.

 

[Dracaris]

As things stand today, nobody should get into a robotaxi driven via the V12 stack. It's not even close to good enough for unsupervised use. Essentially, anything said on 8/8 about Tesla and robotaxis would have to be heavily forward looking and full of smoke, mirrors, and lasers to make it all look nice - without any specification of what year, after saying "March," or "Spring," or whatever.

 

[diplomat33]

The problem is that Elon always bets "the entire farm". Eventually you lose. The likelihood of losing goes up by a lot when you put a few unsolved research problems in the middle of your trajectory.

Yeah and it would be a huge unforced error IMO. That's because there is no reason to go all-in on level 5 at this time. Tesla is making progress on "FSD". And "FSD" could be a compelling L2 system that beats many of the other L2 systems out there. They should focus on improving the build quality of the vehicles, up the luxury in the interior, produce new, more affordable and exciting models. Basically, just make a great EV! And then when "FSD" actually is capable of eyes-off, then roll it out. And Tesla could focus on an eyes-off highway system and likely deliver that before anyone else. But if Tesla neglects quality and just tries to rush to level 5, they will end up with neither.

They have to get FSD working because they need some unique "shtick" to differentiate themselves from Chinese EV makers that are currently racing to the bottom with prices. They aren't going to survive on brand name alone. The US is only 33% of their market.

IMO, a good product is always better than a schtick. Tesla should focus on delivering a good product. You do that by focusing on build quality and producing affordable and exciting new models. Just make a good car! That is a big reason why Tesla did so well in the beginning IMO. The early Model S was a great car, better than other cars in its category. And the Model X and Model 3 were great when they came out, compared to other EVs. Yes, the Chinese are coming out with cheaper cars. Tesla probably cannot compete on price alone. So Tesla should compete on quality.