Autonomous Car Progress

[Knightshade]

The posting I was referring to above, dates back from Oct. 9, 2017.
Cruise' experimenting in SF is interesting. Not the easiest infrastructure there.
Then again, the robo-taxis are performing services only between 11 PM and 5 AM.
So, no school kids on the road to avoid.

FWIW the few times I've used FSDBeta in those hours I've also typically had 0 intervention drives, usually longer than the typical SF cab ride too.

That doesn't mean RTs are ready, it means it's much easier to drive when there's few cars or pedestrians out.

 

[diplomat33]

The posting I was referring to above, dates back from Oct. 9, 2017.
Cruise' experimenting in SF is interesting. Not the easiest infrastructure there.
Then again, the robo-taxis are performing services only between 11 PM and 5 AM.
So, no school kids on the road to avoid.

Yeah, we knew that Cruise picked SF because they felt it would present a challenge. And Cruise is not the only company to do this. Other companies have also argued that they picked a specific city for testing because it would offer unique edge cases to solve. I think Waymo has also said that they picked SF because it presents some unique FSD challenges to solve (for example, tight and hilly streets, dense fog). Also, Waymo and Mobileye have also pointed to the difficult driving in NYC and why they are testing there. I think Mobileye even had a marketing line about "if you can drive in NYC, you can drive anywhere".

If you look at the most recent CA DMV report, Cruise reported 800k autonomous miles and only 21 disengagements. And all the disengagements were precautionary. And if you've watched the videos that Cruise has shared, their AVs seem to handle some tight spots pretty smoothly. Also, the "Under the Hood" presentation showed that Cruise has very advanced perception, prediction and planning stack. I would say that Cruise seems to have really good autonomous driving in SF.

I think that main reason Cruise is limiting their driverless rides to 11pm to 5am is just to be extra safe as they roll out their robotaxi service to the public. So they are starting with an easier ODD to minimize risk. I think Cruise will probably expand the ODD as they gain more confidence that their AVs can drive safely.

 

[EVNow]

But what basis do you have for the claim that Tesla FSD is a "9" in geography? Indeed what basis do you have for it being more than a zero? Is there any geography in which it shows signs of working? Working means "taking the driver out" with full self-driving.

You are all over the place.

Let me put it in another way. Tesla works the same way in all geographic locations where EA FSD Beta is allowed now and that is a very large geographic area. That is why they are at "9" on that.

Waymo works well but in a very tiny geographic area / ODD. Infact the ODD was not even expanded from parts of a suburb of Pheonix to entire Pheonix. So, obviously they have geographic scaling issues.

You do understand the the tech that Tesla uses they did not invent -- many elements of it were invented by people at Google, who share their stuff with Waymo.

Yes - but Google didn't invent all the tech either. They got it from elsewhere. Thats how the industry works, if you haven't noticed

They are not independent elements at all. Quality has a threshold to it. If you aren't near that threshold, it doesn't matter how many places you are also not near that threshold.

That makes little sense. The quality steadily increases. Afterall 60k people are using FSD Beta now. So, what exact "threshold" are you talking about ? Do you have a miles per disengagements threshold ? What will you say when Tesla crosses that threshold ?

BTW, your arguments are quite silly and can easily be made against Waymo just changing a few terms, like I've done below.

Not saying Tesla doesn't have some assets and some innovating things they are doing. But you have to understand where they sit in the hierarchy of teams, and it doesn't matter how many places they drive unacceptably badly. All the others could also drive unacceptably badly on all roads should they want to follow that path, but they think it a foolish path.

"Not saying Waymo doesn't have some assets and some innovating things they are doing. But you have to understand where they sit in the hierarchy of teams, and it doesn't matter how well they drive in an unacceptably small area. All the others could also drive very well in a tiny area should they want to follow that path, but they think it a foolish path."

You are now basically talking like GM/Merc execs who said Tesla can't survive because they had bad "panel gaps". Dogma is strong.

 

[diplomat33]

Cruise is seeking NHTSA approval to deploy the Origin:

https://twitter.com/x/status/1494745190749421572

 

[Daniel in SD]

Tesla works the same way in all geographic locations where EA FSD Beta is allowed now and that is a very large geographic area.

This is almost certainly false. Most people claim FSD Beta works better in some locations than others.

That makes little sense. The quality steadily increases. Afterall 60k people are using FSD Beta now. So, what exact "threshold" are you talking about ?

ROBOTAXIS!

 

[Doggydogworld]

Apparently Waymo drives well in some parts of Chandler but not in rest of Phoenix.

There is nothing special that makes Chandler easier than the rest of Phoenix. It's simply that there's no reason to expand because they'd fail to attract customers in the rest of Phoenix just as badly as in Chandler.

That's why Waymo is testing in different areas like SF, LA and now in NYC. Waymo is gaining driving experience in different environments to expand their driving beyond Chandler.

Waymo has tested in other areas for almost a decade. It's not some "develop the ability to expand beyond Chandler" initiative.

Waymo is pivoting to dense urban. The technological problem is much more difficult in dense cities, but that's where the customers are. It's the clearest signal you can get that the business model failed in Chandler, not the technology. Waymo is also pivoting to trucking, The regulatory hurdles to scaling there are much more daunting. It's another clear case of abandoning an "easy" segment because the business failed and pivoting to a tough segment because the economics are better.

Waymo "solved" the technology problem years ago. Dolgov said so himself. They can't scale until they find a business model that works. And they seem to have no clue how. I know everyone is tired of me harping on this, but it's the key issue.

 

[EVNow]

This is almost certainly false. Most people claim FSD Beta works better in some locations than others.

Based on what ? WholeMars video as ?

We have had this discussion a lot. FSD doesn’t handle some peculiarities of certain regions for sure - but it works to the same degree everywhere.

 

[EVNow]

There is nothing special that makes Chandler easier than the rest of Phoenix. It's simply that there's no reason to expand because they'd fail to attract customers in the rest of Phoenix just as badly as in Chandler.

They avoided the busiest areas of even Chandler. They specifically left out that one important street there that had all the big box retailers.

Even in SF their service area is the exact opposite of area where Uber makes most trips.

In terms of business - why would expanding to entire Phoenix not be ok ? Surely there are a lot of Uber drivers there ? Why not cover Phoenix airport ..

 

[Daniel in SD]

Based on what ? WholeMars video as ?

We have had this discussion a lot. FSD doesn’t handle some peculiarities of certain regions for sir sure - but it works to the same degree everywhere.

Only Tesla has the data but I bet disengagement rates vary by location significantly.
For Waymo we have some data. For 2020 they had a disengagement rate of .033 per 1000 miles but when they started testing in SF in 2021 it went up to .126 per 1000 miles. I'm willing to bet that Tesla's disengagement rate varies by more than .093 disengagements per 1000 miles between different cities.

 

[EVNow]

Only Tesla has the data but I bet disengagement rates vary by location significantly.
For Waymo we have some data. For 2020 they had a disengagement rate of .033 per 1000 miles but when they started testing in SF in 2021 it went up to .126 per 1000 miles. I'm willing to bet that Tesla's disengagement rate varies by more than .093 disengagements per 1000 miles between different cities.

But Waymo won’t even start in my neighborhood. That’s the point.

I mean if Waymo people aren’t even willing to acknowledge the obvious - Tesla FSD beta works publicly all over US and Waymo restricts to 2 tiny areas - there is no basis for productive discussion in this thread.

Like in partisan politics, we would have 2 different sets of “facts”.

 

[Daniel in SD]

But Waymo won’t even start in my neighborhood. That’s the point.

I mean if Waymo people aren’t even willing to acknowledge the obvious - Tesla FSD beta works publicly all over US and Waymo restricts to 2 tiny areas - there is no basis for productive discussion in this thread.

Like in partisan politics, we would have 2 different sets of “facts”.

But we don't know where Waymo works, we only know where they test. The fact is that they started testing in a new city and it seems to work there too. You're only speculating that it wouldn't work better than FSD Beta in your neighborhood.
It just doesn't seem to me that variation between cities is the hard part of the self-driving problem, I think that's our fundamental disagreement.

 

[ThomasD]

These autonomous robotaxi companies tout they are safer than a regular car. Will save thousands of lives . Cheaper than owning your own car and more convenient. But yet will only operate in the dense part of cities where car ownership is lower. Waymo has been at this a long time. Yet won't expand to an entire metro area. Waymo is in Detroit and Miami and have been for a while but no videos of passenger pickups. Cruise vehicles can transport up to 6 people. Unless they are going to deploy millions of vehicles in cities, suburbs and semi rural areas millions of people will still own private cars 20 years from now

 

[diplomat33]

I mean if Waymo people aren’t even willing to acknowledge the obvious - Tesla FSD beta works publicly all over US and Waymo restricts to 2 tiny areas - there is no basis for productive discussion in this thread.

Well, to use your dimensions: Tesla starts with geography and then works on quality. Waymo starts with quality and then works on geography. So Tesla has a system that works everywhere but it is not autonomous and Tesla hopes to make it reliable autonomous driving. Waymo has reliable autonomous driving but geofenced for now and Waymo is expanding the reliable autonomous driving to more areas over time.

Waymo is pivoting to dense urban. The technological problem is much more difficult in dense cities, but that's where the customers are. It's the clearest signal you can get that the business model failed in Chandler, not the technology. Waymo is also pivoting to trucking, The regulatory hurdles to scaling there are much more daunting. It's another clear case of abandoning an "easy" segment because the business failed and pivoting to a tough segment because the economics are better.

I am not convinced that Chandler was ever supposed to be the end-all-be-all business model. Chandler was a "test case", a "proof of concept". It showed that Waymo could do robotaxis and it gave Waymo experience for the logistics of a robotaxi service. It allowed them to learn about a driverless robotaxi service in an easy area in preparation for when they expand to more difficult areas. The real business model was always dense urban areas.

Waymo "solved" the technology problem years ago. Dolgov said so himself. They can't scale until they find a business model that works. And they seem to have no clue how. I know everyone is tired of me harping on this, but it's the key issue.

Waymo has a business model: they charge $4.99 + $0.80 per mile for a robotaxi to pick you up, drive you to your destination and drop you off. I do agree with you that Waymo needs dense urban areas to find enough customers to make their business model work. But that is why they are testing in dense urban areas. They need to "solve" dense urban driving so that they can deploy driverless robotaxis where the customers are. So I disagree that the business model is preventing scale. What is preventing scaling is "solving" dense urban driving. Do that and Waymo can deploy where the customers are and their business model will work.

 

[voyager]

I love the discussion that's taking place here. Two points are quintessential:
1.the business model. I can remember a WSJ journalist writing that ADS (FSD) is sooo costly
that having a human driver operate a taxi is cheaper.
2. An AV at your disposal will replace car ownership, you might say the longer-term goal of
many ADS developers: get rid of the excess rolling hardware that's clogging up our streets.

Both points plead IMHO for the same thing that I have mentioned in another thread.
People will want a vehicle with the carrying capacity needed at the moment of request.

Since the average car occupancy is around 1.2 person, we can all and should all work towards
having as little car mass and size as possible on the road. It will benefit traffic throughput
and safety of more vulnerable road users (cyclists, pedestrians).

You will seldom need the 6-seat capacity of the Cruise Origin. Cruise may want to reconsider
deploying a robo-taxi that will carry maybe 1-2 passenger on an average ride-hail trip.

Smaller (sleeker) AVs may require less complicated ADS equipment and a lot less batteries
(kWh) to still have a decent range. That will make grid demand more manageable.

Below: robo-taxis may return to the helmet-on-wheels that Waymo introduced a decade ago...

Still want to hang on to having your own car, which is to be expected in more rural areas,
then you might want to complete your order for Tesla's Cybertruck.

 

[Knightshade]

I love the discussion that's taking place here. Two points are quintessential:
1.the business model. I can remember a WSJ journalist writing that ADS (FSD) is sooo costly
that having a human driver operate a taxi is cheaper.

I expect they were assuming robotaxis that cost like 100k plus due to huge expensive lidar arrays and such?

2. An AV at your disposal will replace car ownership, you might say the longer-term goal of
many ADS developers: get rid of the excess rolling hardware that's clogging up our streets.

I continue to think this trend is grossly overstated.

It's non-zero... some families might well give up a 2nd or 3rd car to use RTs once they're both cheap and ubiquitous.... but people REALLY like owning a car.

And I use NYC as an example. Owning a car there is INSANELY expensive.... and doesn't make much sense given you have massive access to excellent public transit, tons and tons of actual taxis, plus uber and lyft where your wait is never more than a few minutes, AND it's a very walkable city for reasonable distances too.

And yet a shockingly high % of people there own cars given the conditions.

Those people are going to stop owning them because robotaxi.

In places owning a car is less onerous, or where density makes couple-minutes-pickup-times less likely, they're even less likely to stop owning them.

 

[voyager]

I continue to think this trend is grossly overstated.

And yet a shockingly high % of people there own cars given the conditions.
Those people are going to stop owning them because robotaxi.

In places owning a car is less onerous, or where density makes couple-minutes-pickup-times less likely, they're even less likely to stop owning them.

I agree. "Will replace" should be replaced by "may replace". It all depends on the circumstances.
For instance in Amsterdam the majority of households does not own a car, thanks to pricey
parking tariffs and permits, great Public Transport and the availability of shared EVs.

 

[diplomat33]

I don't think robotaxis will replace personal cars any time soon. IMO, here are a few reasons:
1) Robotaxis will be geofenced at least at first. So people will still want or need a personal car to go places that the robotaxis can't go.
2) Deployment of robotaxis will take time. While that happens, people will still want or need personal cars to get around. People living in rural areas will want a personal car as robotaxis will likely start in major cities first.
3) Some people will just want the freedom of owning a personal car. They can travel anywhere, anytime and not be dependent on having to summon and wait for a robotaxi. Plus, people can make a car their own, it's their own personal space. And some people just enjoy driving too.

Having said, I think there will be some advantages to robotaxis:
1) Personal car ownership can be expensive and inconvenient in big cities. So just hailing a robotaxi for a few bucks will be cheaper and more convenient.
2) There are situations where driving is unsafe, for example, when you are tired or had too many drinks. Being able to hail a robotaxi will be safer.
3) Robotaxis can free up time and make commutes more relaxing. Instead of being stuck in stressful traffic on a long commute, you can chill in the back seat of a robotaxi or get work done.

So IMO, I think robotaxis will meet a need and fill a gap but won't completely replace personal cars for a long time.

 

[diplomat33]

You will seldom need the 6-seat capacity of the Cruise Origin. Cruise may want to reconsider
deploying a robo-taxi that will carry maybe 1-2 passenger on an average ride-hail trip.

You are right that you would not need the 6 seat Origin just to carry 1-2 people. That would be very inefficient. But that is not how the Origin will be used. The Origin will be used like a city shuttle. The Origin will carry up to 6 people, making multiple stops per trips. So it might pick up 3 people at the first stop, drop off 1 at the next stop and pick up 3, drop off 2 at the next stop and pick up 1, pick up 2 and drop off 2 at the next stop, etc... That would be a more efficient use of the vehicle and maximize the usage of the space. And I think that model would be more profitable as you can charge each person and collect a lot more fees per trip. With the 1-2 person pod , you can only charge 1-2 fees per trip. So you will collect less money per trip.

Also, don't underestimate the times when you do need to carry more than 1-2 people. For example, the family of 4 that wants to go out to the movies on a Saturday night or the group of 5 college friends that want to go out on the town on Friday night. The 1-2 person pod won't work for those situations. Maybe they don't happen as frequently as the 1-2 person trips, but they do happen and probably happen a lot in cities. I don't think it is smart business to ignore those opportunities, especially since you can collect more fees from those trips. Why would you restrict yourself to just the trips where you can only collect 1-2 fees per trip, when you could 4-6 fees per trip?

 

[voyager]

Sounds cumbersome and may be unnecessarily time-consuming to organize routes around pickups
and dropoffs, just to keep all seats occupied. But I guess time will tell what's more economical.

What I am trying to say is that a ride-hail provider may want to differentiate vehicle sizes according to needs,
which may vary considerably. Could be a combination of both. Don't forget that sleek pods will get you
on your destination faster than an over 2 meter wide Cruise Origin can.

 

[diplomat33]

What I am trying to say is that a ride-hail provider may want to differentiate vehicle sizes according to needs,
which may vary considerably.

Yes, ride-hail providers will have different autonomous vehicles for different needs. There is no one size fits all. Cruise will add the Origin to the fleet to meet the needs of high capacity routes where they can take 6 people per trip. But Cruise will still keep their Bolt vehicles for those 1-2 people trips. Likewise, Waymo has the i-Pace which is good for 1-2 people trips, but they also have the bigger Pacifica that can fit more people, and lastly they are partnering with Geely to develop an autonomous shuttle. So Waymo will also have multiple vehicles for different needs.