Waymo

[Doggydogworld]

Dolgov has already said this in a few interviews

Krafcik said it six years ago and a bunch of times after that. It's always been their plan.

 

[spacecoin]

Krafcik said it six years ago and a bunch of times after that. It's always been their plan.

Yeah. The fully-owned robotaxi business is a proving ground for the scale-out and until the hardware and software is cheap and mature enough.

 

[JulienW]

[boring but]Just another Waymo spotting in ATL BUT this one has Texas plates. All the others (that I could identify) had Calf plates. Don't know that this means much of anything (from Austin???) and I'm sure it would be more definitive to spot one with GA plates. Also I'm guess it is being human driven since it is parked (LiDAR still spinning) in a parking spot. Do Waymos ever park like this (other than the depot)?

 

[enemji]

[boring but]Just another Waymo spotting in ATL BUT this one has Texas plates. All the others (that I could identify) had Calf plates. Don't know that this means much of anything (from Austin???) and I'm sure it would be more definitive to spot one with GA plates. Also I'm guess it is being human driven since it is parked (LiDAR still spinning) in a parking spot. Do Waymos ever park like this (other than the depot)?

View attachment 1053455

I don't see any of the sensors getting smaller...

 

[spacecoin]

I don't see any of the sensors getting smaller...

They wont change sensors for the iPace. Check out their 6th gen platform. Perhaps not a whole lot smaller, but scale means cheaper.

ZEEKR | Pure Electric Vehicle Brand

ZEEKR selects global premium EV segment markets in due course. With support from global partners, ZEEKR is able to produce the best-in-class vehicles with leading dynamics, connectivity, intelligence and shared functionality, bringing premium zero-emission transportation...

zeekrglobal.com

 

[diplomat33]

Waymo expands Phoenix area to 315 sq mi:

Starting today, Waymo One riders can now access an additional 90 square miles of Metro Phoenix, making the largest autonomous ride-hail territory in the United States even larger. Riders can enjoy the benefits of autonomy across a whopping 315 square miles of The Valley, with this expanded service area stretching further into North Phoenix, as far as Desert Ridge with its popular retail and music scene.

Employee-only curbside terminal pick up and drop offs at Phoenix Sky Harbor International Airport, coming soon to public:

We have recently given Waymo employees 24/7 access to curbside terminal pick up and drop offs at Phoenix Sky Harbor International Airport, with the intention of offering this service to public riders soon. Waymo One riders can currently access the terminals directly from 10 pm to 6 am or use the 24th St and 44th St PHX Sky Train® Stations at all hours of the day. Our strong partnership with the City of Phoenix and Sky Harbor remains critical to providing travelers and residents alike quality transportation options.

Source: Largest Autonomous Ride-Hail Territory in US Now Even Larger

 

[Doggydogworld]

Waymo expands Phoenix area to 315 sq mi:

I think I read Baidu has 1200 sq km in one city, which is 460 sq miles. That's consistent with Waymo's "largest in US" claim.

 

[diplomat33]

I think I read Baidu has 1200 sq km in one city, which is 460 sq miles. That's consistent with Waymo's "largest in US" claim.

Yeah, Waymo was careful to say "largest in the US", not "largest ever". Baidu has a bigger geofence than Waymo in China. But Waymo has the biggest in the US.

 

[ThomasD]

These companies have been so far off on when AVs would be able to take anyone anywhere.

Dude, where’s my self-driving car?

When it comes to autonomous vehicles, the landscape is littered with over-optimistic predictions and missed deadlines. What happened?

www.theverge.com

In Two Years, There Could Be 10 Million Self-Driving Cars on the Roads

A laboratory at Stanford is working madly to keep us safe in that future.

stanfordmag.org

 

[MichaelSte2022]

Lol, she's so Swedish. Cool design/engineering though (also Swedish)

I've traveled in Waymo now in Arizona State. They're really cool but they drive way too politically correct so I prefer speed with my Tesla.

I do not stop at all Stop signs. I slow down. However, Waymo does! No thank you

I am also Swedish.

 

[JulienW]

I've traveled in Waymo now in Arizona State. They're really cool but they drive way too politically correct so I prefer speed with my Tesla.

I do not stop at all Stop signs. I slow down. However, Waymo does! No thank you

I am also Swedish.

I'm 99.9% sure it will be the same in Europe but a L3 or up car MUST obey driving laws since there is no driver that can be held accountable.

 

[Doggydogworld]

The basic problem is that we don't actually know the number of interventions or the cost for each. What I see in Waymo videos is that it is very common to get the "We are working to get you moving again" message.

Nobody discloses these proprietary metrics, but we can make informed guesses. JJRicks made videos for >100 full rides, mostly in Phoenix back in 2020-21 plus a few more recent ones in Phoenix and SF. He tries to get Waymo stuck. Fleet Response was involved maybe 1 in 10 trips early, much less later.

Mayapapaya on Reddit has done 1000+ rides in SF. She posted a bunch of videos last year, but is back under NDA now. She mostly just used it as a normal taxi, without trying so hard to get the cars stuck. I'd guess Fleet Response got involved once every 25-50 of her rides. Here's some stuff she posted after 500 rides.

If FR costs $60/hr and the average time spent is 5 minutes (vast majority less than one minute), that's $5/response. So 50 cents per ride if 1 in 10 rides requires a response and 10 cents if it's 1 in 50.

Roadside Assistance is much more expensive, but less frequent.Two people plus a car for ~30 minutes could be $50-100 per call. I think Maya had a few in her first 800+ rides. Both metrics improve over time.

Once Tesla starts its service, it won't take long for the performance of Tesla's system to surpass Waymo. Tesla will win is because end-to-end is a better approach and Waymo will not be able to catch up. Waymo will be behind in both hardware cost and core driving technology.

Same story I've heard for a decade, just with new buzzwords.

I don't think Waymo will be able to get to 50k in 3 years because the losses will be too high to keep expanding.

Few bash Waymo's lack of business acumen more than me. That said, Ruth Porat axes GOOG's "Other Bets" when the numbers don't pan out. She also forced Waymo to walk away from trucking, at least for now. Yet she approved a 5x scaling in the last 12 months followed by another 5x in the works. That says to me the unit economics work.

Tesla will build the whole robotaxi vehicle at a much lower cost than Waymo will ever be able to acquire outfitted vehicles from a third party. I don't think that is in dispute.

That's one of the things we expect to see on 8/8. We will see a super low-cost, purpose-built robotaxi. Tesla will build them very fast and very cheap using the unboxed process. Nobody will be able to match it.

Unboxed is the new Alien Dreadnought. Vehicle cost is a 2030 battle, anyway. Many horses must come before that cart.

 

[nativewolf]

@Doggydogworld Thats interesting on Porat. I would add that Waymo has lots of interest from outside entities eager to purchase a larger slice of Waymo. I think google still has 90% but I am not sure.

 

[spacecoin]

@Doggydogworld Thats interesting on Porat. I would add that Waymo has lots of interest from outside entities eager to purchase a larger slice of Waymo. I think google still has 90% but I am not sure.

Correct, Alphabet owns about 90% of Waymo.

 

[alexxs88]

I would add that Waymo has lots of interest from outside entities eager to purchase a larger slice of Waymo. I think google still has 90% but I am not sure.

I don't see any gain from them selling stakes in Waymo to other companies. They have ample cash to sustain any development, there's no reason to give up the upside they obviously believe in.

 

[diplomat33]

Update on the pole accident:

Waymo is issuing a voluntary software recall after one of its driverless vehicles collided with a telephone pole in Phoenix, Arizona, last month, the company said. The vehicle was damaged, but no passengers or bystanders were hurt in the incident. The company is filing the recall with the National Highway Traffic Safety Administration (NHTSA) after completing a software update to 672 vehicles — the total number of driverless-capable vehicles in Waymo’s fleet. The update corrects an error in the software that “assigned a low damage score” to the telephone pole, and updates its map to account for the hard road edge in the alleyway that was not previously included.

Waymo issues software and mapping recall after robotaxi crashes into a telephone pole

It’s the Google spinoff’s second-ever recall.

www.theverge.com

 

[Daniel in SD]

Update on the pole accident:

Waymo issues software and mapping recall after robotaxi crashes into a telephone pole

It’s the Google spinoff’s second-ever recall.

www.theverge.com

Yikes. I guess this is related to how they deal with soft debris? i.e. tumbleweeds, balloons, plastic bags, etc. You would think it should be tuned to be super conservative...

 

[Matias]

Does this mean Waymo needs to hand code every curb and telephone pole?
If so, sounds difficult to scale.

 

[stopcrazypp]

Yikes. I guess this is related to how they deal with soft debris? i.e. tumbleweeds, balloons, plastic bags, etc. You would think it should be tuned to be super conservative...

Sounds like it. No mention of remote assistance playing any role, so it appears to be just an issue with the base code (as I speculated upthread).

 

[willow_hiller]

Yikes. I guess this is related to how they deal with soft debris? i.e. tumbleweeds, balloons, plastic bags, etc. You would think it should be tuned to be super conservative...

I can only guess that something like a telephone pole must be way out of sample for the network that was determining damage scores. If it was only trained on objects found in roadways, the shape could be mistaken for a traffic drum, although you wouldn't necessarily want a vehicle hitting those either.

So it's a swiss-cheese kind of failure, where all the systems that did make mistakes had to line up in just the right ways:

1. HD map did not contain the hard road edge.
2. System responsible for detecting road edges failed to recognize the unmapped curb.
3. Once over the curb and outside of the normal drive-able space, the system responsible for classifying obstacles failed to recognize the telephone pole as something hard and unyielding.