Autonomous Car Progress

[nativewolf]

From being able to offer service to a few tens of thousands to a few million in 2 years, then to 10 million in a few more weeks. Yes that is scaling. You might hate it because you'd be wrong Ol Knightshade but wrong you be. By the end of the year they'll have most of the profitable TAM for ride sharing west of the rockies. Yes, that's scaling.

 

[diplomat33]

Expanding from zero cities to part of one city in roughly 8 years... then expanding to part of a second (while still not even fully serving the first) 3 years later, and a full 11 years after they began mapping/testing in that city, is "quickly" to you?

Again, this is false.

2016: Waymo was founded
2017: early rider in Chandler
2020: Full driverless to general public in Chandler
2022: full driverless to general public in all of SF and downtown Phoenix

 

[stopcrazypp]

Since we are talking about scaling, how should we measure the time it takes to scale to a new city? Do you start measuring from the first mapping vehicle? The first test vehicle with a safety driver? The first employee-only driverless ride? Or the first early rider driverless ride? And what if they pause during the process? For example, they do the mapping, stop in that city to focus on something else and then say a few months later decide to go back to testing in that city again. I don't think you can count the paused time as part of the scaling timeline.

Depends on what perspective you are looking at. If you are in one of the less populated cities, you would want to start the count on the first mapping vehicle (which for Waymo is the same vehicle as the test vehicle anyways), because this would be your first known indication of the company looking to start service in your area. For "pauses," a lot of that would be hard to determine publicly, so very hard to account for. If we presume they expand at the speed of Uber however, all of this will average out, and you can simply count how many cities per year they launch in to get an average number (or count mouths between batches of cities).

 

[cwerdna]

TuSimple and Navistar end deal to co-develop autonomous trucks | TechCrunch

TuSimple and Navistar had agreed to jointly develop self-driving trucks by 2024, but now the two companies have scrapped their deal.

techcrunch.com

TuSimple plans to layoff half its workforce next week - WSJ

Self driving trucking company TuSimple Holdings Inc plans to potentially cut half of its workforce next week, WSJ reported.

www.reuters.com

 

[2101Guy]

https://www.reddit.com/r/nextfuckinglevel/comments/zo4xcx

 

[Doggydogworld]

Since we are talking about scaling, how should we measure the time it takes to scale to a new city?

I don't place much value on "new city". Waymo could have offered non-viable Chandler-level services in a few dozen cities years ago. It wouldn't have moved them any closer to a scalable business model.

I think the best metrics at this point are systemwide. I'm more impressed by ten million driverless miles in one or two cities than 100k in each of twenty different cities. Maybe rides is a better metric, since empty miles aren't scalable. Or just total fare revenue. If the tech and business model work, fare revenue should scale 4-10x per year for quite a few years.

Doubling every year won't cut it. Starting from $1m in 2023 it would take until 2040 to reach Uber's current size (~125b annual gross bookings).

Again, this is false.

2016: Waymo was founded

2016 was just a name change and some legal paperwork. The same team kept doing the same work with the technology base they'd been building since 2009.

 

[nativewolf]

I don't place much value on "new city". Waymo could have offered non-viable Chandler-level services in a few dozen cities years ago. It wouldn't have moved them any closer to a scalable business model.

I think the best metrics at this point are systemwide. I'm more impressed by ten million driverless miles in one or two cities than 100k in each of twenty different cities. Maybe rides is a better metric, since empty miles aren't scalable. Or just total fare revenue. If the tech and business model work, fare revenue should scale 4-10x per year for quite a few years.

Doubling every year won't cut it. Starting from $1m in 2023 it would take until 2040 to reach Uber's current size (~125b annual gross bookings).


2016 was just a name change and some legal paperwork. The same team kept doing the same work with the technology base they'd been building since 2009.

Its ok Knightshade is the king of nitpicking. Diplomat33 is just giving it back to him. Just as useful. Nitpicking on this or that is quite useless when AI developments are clearly moving forward, how fast only history is going to describe. The date of Waymo or the exact size of service area in dec 2022 is meaningless. What's important is whether or not Waymo can provide service, can they increase that TAM in something of an exponential manner.

If Uber is anything to judge by they need to only offer service in 20 metros and they will address the vast majority of ride sharing markets. By the end of 2023 they should be well on their way to a significant portion of 3 of the 20 whereas in early in 2022 they had very partial service in a tiny bit of 1. Impressive

 

[Knightshade]

Again, this is false.

2016: Waymo was founded

Again it's not.


Google began developing self driving cars in January 2009, led by Sebastian Thrun and Anthony Levandowski.

Levandowskis own tech that he brought to the project (as well as several others) would be bought out fully by google in 2011. Hilariously, that tech was shown in -2008- (prior to google starting their program using it in 2009) self driving in... San Francisco. It wasn't until 12 year later, 2020, that Waymo actually let riders enjoy (a much updated obviously) version.

Anyway, the 2009 Google program ran quietly for a whle- ROAD TESTING SELF DRIVING CARS- until the NY Times made it public in late 2010.

Levandowski eventually left in 2015...and in 2016 the project, which was founded in 2009 was renamed Waymo, acting as its own company owned by Alphabet/Google.

In case you're still unclear they started in 2009 you can see the company history also correcting you here, where they ALSO say they've been doing this over a decade and continuously refer to the pre-rename-to-Waymo companies actions as "We" as in WE did X, WE did Y:

Autonomous Driving Technology - Learn more about us - Waymo

Waymo, formerly the Google self-driving car project, is creating safe, autonomous driving technology. Learn about our self-driving car company today.

waymo.com

If Uber is anything to judge by they need to only offer service in 20 metros and they will address the vast majority of ride sharing markets. By the end of 2023 they should be well on their way to a significant portion of 3 of the 20 whereas in early in 2022 they had very partial service in a tiny bit of 1. Impressive

Also untrue.

In early 2022 they already had it in parts of 2 markets-- as they'd been taking riders in both AZ and SF before then.

Oh and if Uber is "anything to judge by" that also means they'll almost continuously lose tons of money- not a great business model

 

[diplomat33]

Again it's not.


Google began developing self driving cars in January 2009, led by Sebastian Thrun and Anthony Levandowski.

Levandowskis own tech that he brought to the project (as well as several others) would be bought out fully by google in 2011. Hilariously, that tech was shown in -2008- (prior to google starting their program using it in 2009) self driving in... San Francisco. It wasn't until 12 year later, 2020, that Waymo actually let riders enjoy (a much updated obviously) version.

Anyway, the 2009 Google program ran quietly for a whle- ROAD TESTING SELF DRIVING CARS- until the NY Times made it public in late 2010.

Levandowski eventually left in 2015...and in 2016 the project, which was founded in 2009 was renamed Waymo, acting as its own company owned by Alphabet/Google.

In case you're still unclear they started in 2009 you can see the company history also correcting you here, where they ALSO say they've been doing this over a decade and continuously refer to the pre-rename-to-Waymo companies actions as "We" as in WE did X, WE did Y:

Autonomous Driving Technology - Learn more about us - Waymo

Waymo, formerly the Google self-driving car project, is creating safe, autonomous driving technology. Learn about our self-driving car company today.

waymo.com

I know the history of the Google self-driving car project. Yes, Google started working on self-driving in 2009. My point is that you cannot start counting scaling at 2009. That is not when scaling started. In 2009, they were just building and testing a basic prototype, not actually scaling yet. They had to invent the tech first. So you canot start counting scaling then. That would be like me saying it took Tesla 10 years to scale the Model 3 because I start at 2008 when Tesla built a few roadsters in a garage. You should start counting scaling when they actually started scaling. So start when the project was renamed as Waymo because that is when Google decided the project was worth being its own self-driving company that could scale. Or better yet, start counting scaling when Waymo actually had a public robotaxi because that is when scaling actually started. I just think it is super misleading to say that Waymo took 15 years just to scale to 2 cities when you leave out that most of that 15 years was early prototype development, before any scaling could even start. But starting when Google first started working on self-driving certainly helps you push your false narrative that Waymo is not scaling.

 

[nativewolf]

Not just that but Tesla was based on prototype work done by two groups started in 2004 if I remember correctly. So you know dial it all the way back. He also conveniently ignores the fact that they are scaling. Also ignores the fact that he has argued in other threads that Tesla can make profit with robotaxi where Uber couldn’t. He likes to argue and nitpick. He is fun though, don’t want anything else from him.

I think robotaxi will be a very slow rollout in areas uber has had difficulty making it. Waymo just needs the 20 metro regions . Can they do it? Who knows. Unlike Tesla who has vehicles, they need to deploy 100k vehicles from someone. Just watched the canoe video from Munro. That would be a great robotaxi.

Great pickup too. Also Panasonic 2170s.

 

[Knightshade]

I know the history of the Google self-driving car project. Yes, Google started working on self-driving in 2009. My point is that you cannot start counting scaling at 2009. That is not when scaling started. In 2009, they were just building and testing a basic prototype, not actually scaling yet. They had to invent the tech first.

As noted, most of the tech they began with they bought from someone else who had already driven around SF with it.... but if you wanna cherry pick when you "can" start counting from you can make or support any argument you want that way I suppose.

. So start when the project was renamed as Waymo because that is when Google decided the project was worth being its own self-driving company that could scale.

The problem is if they'd been testing in SF since 2009 (and they had) and only got to deploying to consumers in 2020, it seems you need to count all 11 years.

Otherwise you're just picking some arbitrary date you made up based on what best supports your claims, because you get to pretend they won't need years of FURTHER work for each new city.

Waymo has been testing in LA since 2019 for example, and HOPES to have consumers in cars, with safety drivers in 2023, and without drivers at some unknown future date.

By your thinking we don't count from 2019? We only count from, when, 2023?

The fact they keep needing years per new city unless you choose to ignore the development time in each city is showing how poorly they're scaling.

They don't have a "just drop it into a city" solution-- they have one they need to spend years tailoring for each city each time solution. That's the opposite of quick scaling.


Now, if they actually do driverless rides in LA by say 2024 I'll grant you taking "only" 5 years from start of testing to driverless available IS an improvement over 11 years in SF... but "quick" is not the term I'd use to describe either.

Not just that but Tesla was based on prototype work done by two groups started in 2004 if I remember correctly.

<citation required>

 

[kabin]

TuSimple and Navistar end deal to co-develop autonomous trucks | TechCrunch

TuSimple and Navistar had agreed to jointly develop self-driving trucks by 2024, but now the two companies have scrapped their deal.

techcrunch.com

TuSimple plans to layoff half its workforce next week - WSJ

Self driving trucking company TuSimple Holdings Inc plans to potentially cut half of its workforce next week, WSJ reported.

www.reuters.com

Sorry to hear about the massive layoffs but it seemed like a cluster $%^ outfit. They reportedly did the algorithm work in Tucson where there's lots of Raytheon defense know-how. Hopefully no one jumped ship and gave away defense secrets as that was probably one of Tusimple's main goals. They were recently accused of sharing technology with a Chinese startup. Good riddance Tusimple!

 

[diplomat33]

As noted, most of the tech they began with they bought from someone else who had already driven around SF with it.... but if you wanna cherry pick when you "can" start counting from you can make or support any argument you want that way I suppose.




The problem is if they'd been testing in SF since 2009 (and they had) and only got to deploying to consumers in 2020, it seems you need to count all 11 years.

Otherwise you're just picking some arbitrary date you made up based on what best supports your claims, because you get to pretend they won't need years of FURTHER work for each new city.

Waymo has been testing in LA since 2019 for example, and HOPES to have consumers in cars, with safety drivers in 2023, and without drivers at some unknown future date.

By your thinking we don't count from 2019? We only count from, when, 2023?

The fact they keep needing years per new city unless you choose to ignore the development time in each city is showing how poorly they're scaling.

They don't have a "just drop it into a city" solution-- they have one they need to spend years tailoring for each city each time solution. That's the opposite of quick scaling.


Now, if they actually do driverless rides in LA by say 2024 I'll grant you taking "only" 5 years from start of testing to driverless available IS an improvement over 11 years in SF... but "quick" is not the term I'd use to describe either.






<citation required>

I am not picking arbitrary dates. I am picking dates when Waymo actually started scaling because that makes sense. You are including early testing when Google was not scaling to make it look like it takes years to scale to each city. Waymo does not need years per city. If you look at when they first start ride-hailing with safety drivers or early riders to when the ride-hailing goes driverless and to general public, it's not years. They scaled to downtown Phoenix and SF in mere months, not years. And, Waymo will have driverless public ride-hailing in LA in 2023.

And Waymo does not tailor for each city. Waymo says it is the same autonomous driving software that they use in every city. Waymo has generalized FSD that works in any city. Waymo is building a "just drop it into a city" solution. That is the goal. That is why Waymo is building generalized FSD and working to make their validation/deployment as efficient as possible to reduce the time it takes to test in a new city. And it only took Waymo a few months to scale to downtown Phoenix. They are not there yet but they are getting closer to a "just drop it in a city" solution.

 

[nativewolf]

As noted, most of the tech they began with they bought from someone else who had already driven around SF with it.... but if you wanna cherry pick when you "can" start counting from you can make or support any argument you want that way I suppose.




The problem is if they'd been testing in SF since 2009 (and they had) and only got to deploying to consumers in 2020, it seems you need to count all 11 years.

Otherwise you're just picking some arbitrary date you made up based on what best supports your claims, because you get to pretend they won't need years of FURTHER work for each new city.

Waymo has been testing in LA since 2019 for example, and HOPES to have consumers in cars, with safety drivers in 2023, and without drivers at some unknown future date.

By your thinking we don't count from 2019? We only count from, when, 2023?

The fact they keep needing years per new city unless you choose to ignore the development time in each city is showing how poorly they're scaling.

They don't have a "just drop it into a city" solution-- they have one they need to spend years tailoring for each city each time solution. That's the opposite of quick scaling.


Now, if they actually do driverless rides in LA by say 2024 I'll grant you taking "only" 5 years from start of testing to driverless available IS an improvement over 11 years in SF... but "quick" is not the term I'd use to describe either.






<citation required>

If you don’t know the origin of Tesla go do your own homework. Enjoy

 

[Bladerskb]

People like KS would do/say anything to disparage anything that isn't Tesla and prop up anything that is Tesla.
Why stop at 2009? Since Google acquired the team and entire tech (Stanley) that was in the grand challenge.
Why not go all the way back to 2002. Then you can get an even nice number like "It took waymo 20 years to deploy driverless cars to 2 cities".

Heck why stop there, Chris Urmson who lead the Stanley team and stayed and grew the team all the way till 2016 worked on the ALVINN self driving car that started in 1984. And the tech in ALVINN and lessons learned over the years lead to Stanley. So actually its not 20 years. Using KS gainsaying logic, "It took waymo 38 years to deploy driverless cars to 2 cities".

Its funny because this is coming from same guy who believes that Tesla self driving development began on October 19, 2016. Ignoring all the AP2 prototyping and development from the year before, ignoring all the AP1 code that was basis on AP2 development and software, ignoring all the AP1 development from 2013 and 2014.

 

[nativewolf]

People like KS would do/say anything to disparage anything that isn't Tesla and prop up anything that is Tesla.
Why stop at 2009? Since Google acquired the team and entire tech (Stanley) that was in the grand challenge.
Why not go all the way back to 2002. Then you can get an even nice number like "It took waymo 20 years to deploy driverless cars to 2 cities".

Heck why stop there, Chris Urmson who lead the Stanley team and stayed and grew the team all the way till 2016 worked on the ALVINN self driving car that started in 1984. And the tech in ALVINN and lessons learned over the years lead to Stanley. So actually its not 20 years. Using KS gainsaying logic, "It took waymo 38 years to deploy driverless cars to 2 cities".

Its funny because this is coming from same guy who believes that Tesla self driving development began on October 19, 2016. Ignoring all the AP2 prototyping and development from the year before, ignoring all the AP1 code that was basis on AP2 development and software, ignoring all the AP1 development from 2013 and 2014.

And thinks Tesla started in 2003 when the prototypes were being built in 1997.

I think he’s entertaining, only when he tries to argue both sides of something (Uber being a good model for Tesla robotaxi and bad for Waymo robotaxi does it get irritating). He’s just done it.

I am impressed by waymo here. Now they just need a basic vehicle. Something like canoe.

 

[diplomat33]

Now they just need a basic vehicle. Something like canoe.

Waymo is working on their own driverless vehicle with Geely.

Waymo’s new robotaxi is an all-electric people mover with no steering wheel

The new AV made its debut at the LA Auto Show.

www.theverge.com

 

[diplomat33]

Interesting.

https://twitter.com/x/status/1604563431482425346

 

[Knightshade]

I am not picking arbitrary dates.

Of course you are.

There's no official "date we began to scale" so you're making one up that best supports your argument.

I am picking dates when Waymo actually started scaling

Heck, even YOU didn't agree with you on what that date was.... first you said it was when they changed their name to Waymo.... then you moved the goalposts even further ahead to when they start when they "actually had a public robotaxi"

And I'm sure if we tried to drill down you'd move THAT forward to to not count the time with safety drivers, or before they could charge fares.

And none of those argument are really any better than, say, "We should start counting from when they first start operating, at all, in a given city" to determine how long it takes them to scale TO that city.

In SF it was a REALLY long time...more than 10 years. In LA it's going to be at minimum 4 years (which is less than SF, but hardly quick).

They scaled to downtown Phoenix and SF in mere months, not years.

Ah, see, you're pushing those goalposts years forward again and ignoring all the time spent in Chandler I guess....

And Waymo does not tailor for each city. Waymo says it is the same autonomous driving software that they use in every city. Waymo has generalized FSD that works in any city.

I mean... WAYMO says you're wrong, but WTF do they know right? Because they specifically cited their testing deployments in Washington state recently as being for them to learn to operate in wet weather. So doesn't sound very generalized right now- nor does it sound like the "same" SF working in super dry and sunny Phoenix is gonna work in Bellevue Washington where they're testing.

They DID mention that one reason they picked that spot was because apart from weather it was really similar to the places they've already got something mostly working.... which further implies they'll need to tailor again when they try and expand to someplace with significantly different roads.

Not just that but Tesla was based on prototype work done by two groups started in 2004 if I remember correctly.

You don't. Seems to be a pattern with you.

Teslas founders liked someone elses prototype, and encouraged them to make a commercial version- and they didn't want to- so they started Tesla to make something like it.

Initially they licensed the someone elses patents- but ended up so dissatisfied with the real world results they redesigned everything themselves-- in the end the roadster used none the original intellectual property from the someone else.

If you don’t know the origin of Tesla go do your own homework. Enjoy

I do, you do not- hence why I asked you to support your claim with a source- which you were unsurprisingly unable to do.

And thinks Tesla started in 2003 when the prototypes were being built in 1997.

Also factually wrong.


What's epsecially weird is none of this founding-of-tesla stuff has anything, at all, to do with self driving since it'd be years after the founding they'd do anything at all related to it... so it's a really weird red herring for you to be wrong again about.

I think he’s entertaining, only when he tries to argue both sides of something (Uber being a good model for Tesla robotaxi and bad for Waymo robotaxi does it get irritating).

When, specifically, did I do that?

(though of course there's a substantial difference in vehicle cost to consider, beyond the other major differences like one solution being functional L4 today with a narrow ODD and scaling problems- and the other being L2 today with a very wide ODD but seeming years from L4- so there's that as additional reasons they're pretty different things rather than any direct comparison apples/apples you imply I made but won't be able to provide evidence of)

 

[EVNow]

Interesting.

https://twitter.com/x/status/1604563431482425346

Really ... that is a tricky scenario ? FSDb can handle it fine ....