Autonomous Car Progress

[diplomat33]

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.

Waymo does not say I am wrong. LOL. I am literally going with what a Waymo employee told JJ Ricks in his video. He said it is the same Waymo Driver in SF and downtown Phoenix. And Waymo has said repeatedly that the trucks use the same Waymo Driver as the I-Pace. So yes, Waymo knows. I am quoting Waymo. Stop gaslighting.

You don't understand how Waymo works. Waymo has one Waymo Driver. It is the same everywhere. They simply take the same Waymo Driver to different areas to train it in different areas, so it becomes better overall. It is generalized because the same software can handle lots of situations. This has been Waymo's entire approach. First, they trained the Waymo Driver in Chandler so it could handle suburb driving well. They took the same Waymo Driver to SF, so now the same Waymo Driver works in both suburb and tighter streets. They take the same Waymo Driver to downtown Phoenix. So now, the same Waymo Driver can handle suburbs, denser suburbs and urban. They take the same Waymo Driver to NYC. Now, they have the same Waymo Driver that works in suburb, urban and more dense urban. Then they take the same Waymo Driver to Bellevue, WA. Now the same Waymo Driver can drive in suburb, urban, and rain better. The Waymo trucks also use the same Waymo Driver as the I-Pace. So now their trucks can handle suburb or urban or highways or rain. The lessons learned on the trucks can be applied to the I-Pace since it is the same Waymo Driver. So now the Waymo Driver on the I-pace can handle suburb, urban, rain and highways better. Next, they are deploying in LA so that it can learn from there.

The experience gained is cumulative over time. So the time to test in LA will be less since the Waymo Driver will already be able to handle a lot, based on everything it has already learned from driving in Chandler, downtown Phoenix, Bellevue, and SF. The goal is to eventually get one Waymo Driver that can safely and reliably handle all four domains (suburb, urban, highways and weather) so that Waymo can scale the Waymo Driver to any vehicle, any geography. Eventually, the Waymo Driver will be so good "everywhere" that they can just drop it in a new city. That is why scaling will accelerate as the Waymo Driver becomes more experienced.

 

[Knightshade]

Waymo does not say I am wrong. LOL. I am literally going with what a Waymo employee told JJ Ricks in his video. He said it is the same Waymo Driver in SF and downtown Phoenix.

That's great.

But then I cited Waymo saying it's NOT going to be the same in Seattle. Because the weather is different.

That they're now testing there specifically to address that difference.

So yes, Waymo says you're wrong. The thing working today in SF is NOT the thing that'll be working when they deploy driverless in Seattle for example- because the thing in SF can't safely operate in Seattle.

You don't understand how Waymo works. Waymo has one Waymo Driver. It is the same everywhere. They simply take the same Waymo Driver to different areas to train it in different areas, so it becomes better overall. It is generalized because the same software can handle lots of situations. This has been Waymo's entire approach. First, they trained the Waymo Driver in Chandler so it could handle suburb driving well. They took the same Waymo Driver to SF, so now the same Waymo Driver works in both suburb and tighter streets. They take the same Waymo Driver to downtown Phoenix. So now, the same Waymo Driver can handle suburbs, denser suburbs and urban. They take the same Waymo Driver to NYC. Now, they have the same Waymo Driver that works in suburb, urban and more dense urban.

If you keep changing a thing each time it goes somewhere different, then it's not really the same thing each time.

That's kinda what the word change means

The goal is to eventually get one Waymo Driver that can safely and reliably handle all four domains (suburb, urban, highways and weather) so that Waymo can scale the Waymo Driver to any vehicle, any geography.

Sure. But that doesn't exist yet. And there'll certainly be more domains than just those 4 as they try and move to additional cities that are different from the 2 they've managed to deploy anything but a couple test cars in... (for example left side of the road driving would be quite different just to cite the most obvious one)


Hence why they have been the opposite of quick to scale

Tesla also has goals that have also not been quick to be realized, they're just of a different nature.

 

[diplomat33]

That's great.

But then I cited Waymo saying it's NOT going to be the same in Seattle. Because the weather is different.

That they're now testing there specifically to address that difference.

So yes, Waymo says you're wrong. The thing working today in SF is NOT the thing that'll be working when they deploy driverless in Seattle for example- because the thing in SF can't safely operate in Seattle.

If you keep changing a thing each time it goes somewhere different, then it's not really the same thing each time.

That's kinda what the word change means

Sure. But that doesn't exist yet. And there'll certainly be more domains than just those 4 as they try and move to additional cities that are different from the 2 they've managed to deploy anything but a couple test cars in... (for example left side of the road driving would be quite different just to cite the most obvious one)

Hence why they have been the opposite of quick to scale

Tesla also has goals that have also not been quick to be realized, they're just of a different nature.

No. You don't understand. They are not customizing the Waymo Driver for each city. They are training the Waymo Driver. It's the iterative process of machine learning. They train new NN to handle rain better and test it in Bellevue. If the NN perform better, it goes into the Waymo Driver that gets pushed to the entire fleet. The I-Pace's in SF will get the new Waymo Driver version with the better NN. It's no different than when Tesla trains FSD Beta on new data and pushes a new software version. Waymo does the same thing. So yes, the software changes as it is trained. But the fleet in SF or downtown Phoenix or LA get the same software update with the latest version.

If you still want to believe they can't scale, be my guest. I know they started in only Chandler and now have driverless in downtown Phoenix and all of SF and will be adding parts of LA soon. That looks like scaling to me.

 

[nativewolf]

Oh it is scaling. Address a market if 10k, then a few million, then 10 million. All in a couple of years. By the end of 2023 they should be addressing 10% of most profitable tam. Not bad from almost nothing 2 years ago.

 

[Knightshade]

No. You don't understand. They are not customizing the Waymo Driver for each city. They are training the Waymo Driver.

It's not rocket surgery dude.

The driver SW they're using in SF is not capable of being "the same" that works in Seattle.

So they're gathering data and testing in Seattle to CHANGE THE SOFTWARE so it can work safely there.

Thus, again, your claim what they use in SF today is what they will use in Seattle (eventually) is factually wrong.


That doesn't mean they won't take this NEW version and feed the improvements back to the car in SF.

But it DOES mean they do not have drop in any city software

Which you even admitted, yet somehow want to keep arguing about.

If you still want to believe they can't scale, be my guest. I know they started in only Chandler and now have driverless in downtown Phoenix and all of SF and will be adding parts of LA soon. That looks like scaling to me.

Again you are making up strawmen.

I never claimed they can't scale at all-- in fact I cited specifically the speed at which they DID SCALE (glacial).

So what I said was they can't scale quickly

Which based on all evidence so far, they can't.

 

[diplomat33]

The driver SW they're using in SF is not capable of being "the same" that works in Seattle.

Stop making stuff up. Yes, it is the same. The Waymo Driver is generalized. Why would it not be capable of working in Seattle?

So they're gathering data and testing in Seattle to CHANGE THE SOFTWARE so it can work safely there.

No, that is a myth you want to believe. they are gathering data in Seattle so the Waymo Driver can handle rain better everywhere.

Thus, again, your claim what they use in SF today is what they will use in Seattle (eventually) is factually wrong.

Well, the software they end up using in Seattle will be whatever version they are at when they launch in Seattle. But it will be the generalized Waymo Driver that they use everywhere.

But it DOES mean they do not have drop in any city software

Which you even admitted, yet somehow want to keep arguing about.

You are twisting my words. Waymo does have "drop in any city" software. Waymo could drop in any city but it would likely require a safety driver and some testing in the beginning. Waymo wants to be able to scale very quickly so they want the software to be good enough where they can drop in any city and go straight to public driverless. So they do have "drop in any city" software, it just needs to be better. Just like everybody. Everybody has autonomous driving that could be dropped in any city but nobody has solved autonomous driving where it would work safely and reliably in any city right out of the box.

I never claimed they can't scale at all-- in fact I cited specifically the speed at which they DID SCALE (glacial).

I am saying the scaling is not glacial. Try to keep up.

So what I said was they can't scale quickly

Which based on all evidence so far, they can't.

Wrong. All evidence points to they can scale quickly. They literally just scaled to all of SF and doubled their geofence in downtown Phoenix in a matter of months. You just refuse to accept it. You want to start counting scaling from 2009 so that you can pretend that it takes them years when it does not. You cannot accept that Waymo is scaling quickly because you have to hold on to your Tesla myths. Good luck with that.

 

[Knightshade]

Stop making stuff up

Stop accusing others of this when it's clearly you that's wrong.

. Yes, it is the same. The Waymo Driver is generalized. Why would it not be capable of working in Seattle?

Because.

It.

Doesn't.

Work.

Safely.

In.

Wet.

Weather.

That is why they're deploying a tiny number of safety-driver vehicles with no passengers there my dude.

As you've had explained to you like 10 times now including by Waymo themselves

They hope to CHANGE the Waymo driver so it does work safely there.

But today- using the one they use in SF, it does not


Which part, specifically, is causing you to wrongly accuse others of making stuff up?

No, that is a myth you want to believe. they are gathering data in Seattle so the Waymo Driver can handle rain better everywhere.

You just asked why the current SF version can't work in Seattle.

Then just told us why.


Holy cow.

Wrong. All evidence points to they can scale quickly. They literally just scaled to all of SF and doubled their geofence in downtown Phoenix in a matter of months.

YM matter of years.

Because that's how long it was between "dropped into the city with testing and safety drivers" and "driverless paid robotaxis"

HTH!

 

[diplomat33]

Because.

It.

Doesn't.

Work.

Safely.

In.

Wet.

Weather.

Dolgov says it does.

https://twitter.com/x/status/1599150826584702977

That is why they're deploying a tiny number of safety-driver vehicles with no passengers there my dude.

As you've had explained to you like 10 times now including by Waymo themselves

They hope to CHANGE the Waymo driver so it does work safely there.

But today- using the one they use in SF, it does not

First, stop pretending Waymo said it when Waymo did not say it. Show me the quote from Waymo.

Dolgov said it does work safely in rain and wet roads.

Second, they did not hope to change the Waymo Driver just to work safely in Seattle. They tested in Seattle to improve rain performance everywhere.

You just asked why the current SF version can't work in Seattle.

Then just told us why.

What are you talking about? I did not ask that. And Waymo never said that. Again, show me the quote where they said that. You keep claiming Waymo said stuff but don't provide any quotation. In fact, Dolgov says the software can handle rain and wet roads safely. So the current SF version would work in Seattle.

 

[Knightshade]

Dolgov says it does.

https://twitter.com/x/status/1568288678975533057

That's Waymo, directly quoted.

If the system could ALREADY drive safely there they'd be offering rides- instead of test-deploying 6 cars, with safety drivers, for the express purpose of ENABLING THEM IN THE FUTURE to drive safely with riders there.

Because that is a thing the system can not do today

First, stop pretending Waymo said it when Waymo did not say it. Show me the quote from Waymo.

Just did.

Second, they did not hope to change the Waymo Driver just to work safely in Seattle. They tested in Seattle to improve rain performance everywhere.

Ok. Which still means the version they were running in AZ/SF had poor enough rain performance they could NOT drop in that system someplace with a lot of rain- thus they needed a better version to drive safely in such places like Seattle.

Though- again- they picked Bellvue, specifically, because the OTHER local factors were similar to the very few places the system already works safely, so they ONLY had to solve for the weather there.

They will have to solve for something ELSE in other places with different roads. Because their "scaling" involves tailoring each time they encounter anything that wasn't in an existing service city.


Here- let me cite a source on THAT too

Waymo to test its latest self-driving car technology on the wet, wide streets of Bellevue, Wash.

Winter is coming, and Waymo is here for it. Specifically, the autonomous driving company is in Bellevue, Wash., where it has started testing the fifth generation of its Waymo Driver… Read More

www.geekwire.com

the company chose Bellevue specifically because the roads and overall driving environment are similar to other places where Waymo operates, specifically Phoenix and San Francisco.

You keep claiming Waymo said stuff but don't provide any quotation. In fact, Dolgov says the software can handle rain and wet roads safely. So the current SF version would work in Seattle.

Sorry- I thought you were the guy who actually followed all the stuff Waymo says so didn't figure you'd need the sources.

My apologies, they've now been provided to, once again, correct you.

 

[diplomat33]

https://twitter.com/x/status/1568288678975533057

That's Waymo, directly quoted.

If the system could ALREADY drive safely there they'd be offering rides- instead of test-deploying 6 cars, with safety drivers, for the express purpose of ENABLING THEM IN THE FUTURE to drive safely with riders there.

Because that is a thing the system can not do today





Just did.



Ok. Which still means the version they were running in AZ/SF had poor enough rain performance they could NOT drop in that system someplace with a lot of rain- thus they needed a better version to drive safely in such places like Seattle.

Though- again- they picked Bellvue, specifically, because the OTHER local factors were similar to the very few places the system already works safely, so they ONLY had to solve for the weather there.

They will have to solve for something ELSE in other places with different roads. Because their "scaling" involves tailoring each time they encounter anything that wasn't in an existing service city.

Here- let me cite a source on THAT too

Waymo to test its latest self-driving car technology on the wet, wide streets of Bellevue, Wash.

Winter is coming, and Waymo is here for it. Specifically, the autonomous driving company is in Bellevue, Wash., where it has started testing the fifth generation of its Waymo Driver… Read More

www.geekwire.com

Sorry- I thought you were the guy who actually followed all the stuff Waymo says so didn't figure you'd need the sources.

My apologies, they've now been provided to, once again, correct you.

Uh, all the Waymo tweets and articles about training in Bellevue for rain was BEFORE the Dolgov tweet. The Bellevue tweet is marked Sept. The Dolgov tweet is 3 months later, in Dec. You are quoting from BEFORE they got the system working in rain. So you are wrong when you say it cannot do it today since all those quotes and articles are from BEFORE they got the system working in rain. I was right. Yes, the system used to be poor in rain before. After they trained in Bellevue, now the system is good in rain. So nice try but fail.

And please stop with the snippy "I thought you knew Waymo" when you are the one twisting words and using old tweets and ignoring the more recent tweets to try to a push a false narrative.

 

[diplomat33]

Aurora just 3 months from "feature complete".

https://twitter.com/x/status/1604839144475488262

And update:

https://twitter.com/x/status/1604846010760663041

 

[Bitdepth]

Driving safely and being safe enough to deploy an autonomous taxi are two different safety cases. Nobody is going to push out an autonomous taxi service without first testing and validating for each region before deploying, that's just foolishness. The fact that there is not a single company doing that today, should tell you there is no basis in reality to expect a company to just deploy an autonomous taxi service nationwide without first testing and validating in each region.

There is "One" Waymo Driver that can be deployed in most region and should be able to handle it safely and everything learned in each region is pushed out to all the fleet. To be more precise, what i mean is the core underlying technology and software running on it is the same or very similar regardless of vehicle and platform but of course there and 10s if not hundreds of versions at any given time both in vehicle and simulator. Does FSD beta not show you this, where you can literally see different versions being pushed out and tested simultaneously? They are not just deploying every version to all vehicles. You have to have different branches of software and deploy carefully to make sure things aren't broken.

Testing and validating for each region is how they deploy safely and make sure the system can handle the different nuances and driving behavior in each region. That is also how they make the safety case for getting regulatory approval to operate these services.

Drago Anguelov gave a very nuanced answer on this at Weights and Biases if anyone cares to watch. The entire interview is very interesting but i bookmarked the specific part pertinent to this topic.

 

[diplomat33]

Driving safely and being safe enough to deploy an autonomous taxi are two different safety cases. Nobody is going to push out an autonomous taxi service without first testing and validating for each region before deploying, that's just foolishness. The fact that there is not a single company doing that today, should tell you there is no basis in reality to expect a company to just deploy an autonomous taxi service nationwide without first testing and validating in each region.

There is "One" Waymo Driver that can be deployed in most region and should be able to handle it safely and everything learned in each region is pushed out to all the fleet. To be more precise, what i mean is the core underlying technology and software running on it is the same or very similar regardless of vehicle and platform but of course there and 10s if not hundreds of versions at any given time both in vehicle and simulator. Does FSD beta not show you this, where you can literally see different versions being pushed out and tested simultaneously? They are not just deploying every version to all vehicles. You have to have different branches of software and deploy carefully to make sure things aren't broken.

Testing and validating for each region is how they deploy safely and make sure the system can handle the different nuances and driving behavior in each region. That is also how they make the safety case for getting regulatory approval to operate these services.

Drago Anguelov gave a very nuanced answer on this at Weights and Biases if anyone cares to watch. The entire interview is very interesting but i bookmarked the specific part pertinent to this topic.

Thank you!

 

[Knightshade]

Uh, all the Waymo tweets and articles about training in Bellevue for rain was BEFORE the Dolgov tweet. The Bellevue tweet is marked Sept. The Dolgov tweet is 3 months later, in Dec. You are quoting from BEFORE they got the system working in rain.

But WAS working in SF. Meaning the one they'll eventually deploy in Seattle isn't the one they offered service in SF with that wasn't capable of wet performance.

Also, if they were "done" making it work in the rain, they'd be offering driverless rides in Seattle.

Which, of course, they're not. And haven't announced any date at all to do so.

I know facts are pesky things, but maybe try sticking to them?

 

[Knightshade]

Driving safely and being safe enough to deploy an autonomous taxi are two different safety cases. Nobody is going to push out an autonomous taxi service without first testing and validating for each region before deploying, that's just foolishness.

Absolutely.

Which is why trying to cite "Waymo demoed a no-public-customers video driving in the rain" is not "they have a drop in any city system that is scaling quickly!"

Same is true for Tesla-- I've had many trips where it drove me safely between A and B with no intervention or action on my part besides being ready to take over. But it's not remotely ready for general autonomous RT service.

The fact that there is not a single company doing that today, should tell you there is no basis in reality to expect a company to just deploy an autonomous taxi service nationwide without first testing and validating in each region.

But this isn't just testing-- it's deploying vehicles to a new city because the current system is known not to be able to handle that city.

So they deploy vehicles- years in advance in the case of say SF or LA, to do local mapping.... and in seattles case to make the SW capable of handling the weather there.

They'll need to do this in advance every time they try and expand to a new city that has anything significantly different from existing service areas (other side of the road driving being one obvious example I cited)



Thanks for the video BTW.... one quote Diplomat might want to key on is where he tells us scalability is a problem and there is a lot of local particularities in every location

Both of which are exactly what I've been saying.

They will need to test and update the software to tailor to anything new in each new city they deploy to.

Maybe he'll believe it now after watching the video and seeing Waymo telling him the same thing I am.

 

[diplomat33]

But WAS working in SF. Meaning the one they'll eventually deploy in Seattle isn't the one they offered service in SF with that wasn't capable of wet performance.

Also, if they were "done" making it work in the rain, they'd be offering driverless rides in Seattle.

Which, of course, they're not. And haven't announced any date at all to do so.

I know facts are pesky things, but maybe try sticking to them?

Stop with the snarkiness. I have given you nothing but facts. Of course, Waymo is not "done". No, Waymo is not offering driverless in Seattle yet. Nobody said that they were. That is a strawman. Waymo is continually improving their system. As Waymo improves their system and validates safety, they deploy in more and more areas. In Sept, they announced testing in Bellevue to improve performance in rain. 3 months later, Dolgov tweeted update that they are making progress and showed clips showing good performance in rain. That is fact.

Once, they finish validation in rain and are confident with safety, then they will deploy driverless in rain. Obviously, they are not there yet. But it seems you just want to win points by arguing that Waymo has not deployed driverless in Seattle yet therefore you are right about their slow scaling. Gotta love how Tesla fans move the goalposts. They argued Waymo will never get out of Chandler. That was wrong. They argued Waymo can't scale in Phoenix. That was wrong. They argued that Waymo can't scale in SF. That was wrong. So now they just pick another target "well they haven't scaled yet in Seattle." I am sure once Waymo does do driverless in rain in Seattle, you will pick another target to cling to your myth that Waymo can't scale fast. You will probably start from 2009 so you can claim it took them almost almost 20 years to scale to Seattle.

Frankly, I think Tesla fans just want to distract from the fact that Tesla has not scaled driverless in even one city yet. Plus, the TSLA stock is down 63% YTD so they are losing their life savings.

 

[Knightshade]

Stop with the snarkiness. I have given you nothing but facts. Of course, Waymo is not "done". No, Waymo is not offering driverless in Seattle yet. Nobody said that they were. That is a strawman.

It's not though.

If it worked fine in Seattle weather (which you mistakenly assumed from a demo video of it driving in wet weather) then they'd be deploying it to customers there. They're not because it doesn't.

Waymo is continually improving their system. As Waymo improves their system and validates safety, they deploy in more and more areas. In Sept, they announced testing in Bellevue to improve performance in rain. 3 months later, Dolgov tweeted update that they are making progress and showed clips showing good performance in rain. That is fact.

I'm glad you've walked back your claim from "it does work safely in wet weather" to "making progress" on getting to that goal--- that's.... progress on your part

Once, they finish validation in rain and are confident with safety, then they will deploy driverless in rain. Obviously, they are not there yet.

[

Yes... Like I told you several posts ago and you tried to claim otherwise.

The system they were using in SF was NOT the system they'd deploy in Seattle, because that system wasn't safe in wet weather.

Instead they'll spend some amount of time (months possibly, years likely based on past history) from when they start working on adapting their software to a new city/scenario until they can widely deploy it to paid driverless robotaxis.

Like I've been saying the whole time.

They argued Waymo will never get out of Chandler.

Now THAT is a strawman-- I never made any such claim.

They argued Waymo can't scale in Phoenix.

Never made any such claim of that either.

They argued that Waymo can't scale in SF.

IBID

I am sure once Waymo does do driverless in rain in Seattle, you will pick another target to cling to your myth that Waymo can't scale fast. You will probably start from 2009 so you can claim it took them almost almost 20 years to scale to Seattle.

On the contrary- I absolutely expect they'll eventually do driverless in Seattle.

But taking years between adding each new city is NOT SCALING QUICKLY.

It's the opposite of that.

Frankly, I think Tesla fans just want to distract from the fact that Tesla has not scaled driverless in even one city yet.

Tesla doesn't even have driverless software, so that's a weird comparison. They have a very good L2 ADAS system whose ODD is explicitly not capable of higher operation.

They do plan to offer higher more capability SW in the future, but that future isn't here yet.

Plus, the TSLA stock is down 63% YTD so they are losing their life savings.

Yeah, the folks who bought at IPO are only up 11,416.02% now.... practically broke!

 

[Dewg]

Since there is so much angst over easily quantifiable terms such as fast and slow, I encourage you all to do so before continuing this argument. What is "slow" and what is "fast" in terms of scaling? Be precise. Your quantifiable definition will be used later to support, refute, or mock your arguments.

As for Waymo's scaling - here are the facts:

2009 - Google began testing AV in SF (this is an early prototype and used for gathering information on sensor requirements)
2014 - Google began road testing in Mountain View, CA (limited speeds to 25MPH)
2015 - Began testing in TX
2015 - OCT, first driverless ride was 10 mins around Austin, TX
2016 - Began testing in Phoenix, AZ and Kirkland, WA (for wet climate testing)
2016 - AUG, Waymo cars had driven 126,000 miles in full autonomy that month alone
2017 - OCT, Waymo begins weather testing in MI
2017 - NOV, Waymo removed safety drivers from AZ cars
2018 - JAN, Waymo had tested its system (Waymo Driver) in 6 states and 25 cities over 9 years, logging over 5 million miles on public roads
2018 - OCT, CA finally grants permits to test AV without safety drivers (Waymo was first company to receive this permit)
2018 - DEC, Waymo begins taking fares in AZ (with safety drivers)
2019 - NOV, Waymo is the first company in the world operating AVs without safety drivers for fared rides
2020 - COVID lockdown paused operations for several months
2021 - FEB, Expanded operations in SF with riders (previously only employees were permitted)
2021 - AUG, General public allowed to ride in SF (with safety drivers)
2022 - MAR, SF offers rides without a safety driver

I think an important fact, and one that surprised me given all the talk about PH and SF, is that by 2018 Waymo was testing in 25 cities and 6 states. I didn't know they were testing in that many locales.

 

[diplomat33]

Since there is so much angst over easily quantifiable terms such as fast and slow, I encourage you all to do so before continuing this argument. What is "slow" and what is "fast" in terms of scaling? Be precise. Your quantifiable definition will be used later to support, refute, or mock your arguments.

As for Waymo's scaling - here are the facts:

2009 - Google began testing AV in SF (this is an early prototype and used for gathering information on sensor requirements)
2014 - Google began road testing in Mountain View, CA (limited speeds to 25MPH)
2015 - Began testing in TX
2015 - OCT, first driverless ride was 10 mins around Austin, TX
2016 - Began testing in Phoenix, AZ and Kirkland, WA (for wet climate testing)
2016 - AUG, Waymo cars had driven 126,000 miles in full autonomy that month alone
2017 - OCT, Waymo begins weather testing in MI
2017 - NOV, Waymo removed safety drivers from AZ cars
2018 - JAN, Waymo had tested its system (Waymo Driver) in 6 states and 25 cities over 9 years, logging over 5 million miles on public roads
2018 - OCT, CA finally grants permits to test AV without safety drivers (Waymo was first company to receive this permit)
2018 - DEC, Waymo begins taking fares in AZ (with safety drivers)
2019 - NOV, Waymo is the first company in the world operating AVs without safety drivers for fared rides
2020 - COVID lockdown paused operations for several months
2021 - FEB, Expanded operations in SF with riders (previously only employees were permitted)
2021 - AUG, General public allowed to ride in SF (with safety drivers)
2022 - MAR, SF offers rides without a safety driver

Thanks for the timeline.

You could add:

2022 - DEC, doubled Phoenix downtown area without a safety driver, removed safety for Phoenix airport run, expanded to all of SF without a safety driver. Applied for CPUC permit to charge fares.

 

[ThomasD]

So any car should be able to operate in any city with Waymo driver since the updates are pushed to all cars no matter where they operate? You could take a car from San Francisco and have it drive to L A and drive around L A once L A is mapped?