Autonomous Car Progress

[diplomat33]

We talk about groups critical of robotaxis in SF but it is not all naysayers. San Francisco's LightHouse for the Blind and Visually Impaired is among a group of community organizations urging state regulators to approve Waymo's permit that would enable the self-driving car company to operate around-the-clock in San Francisco.

https://www.axios.com/local/san-francisco/2023/07/21/disability-advocates-support-waymo-self-driving-cars

 

[nativewolf]

@diplomat33 without a doubt this is the greatest benefit to robotaxi, that and elderly- I think like Facebook older people will utilize robotaxi more than most.

 

[diplomat33]

Looks like the hearing with Waymo and Cruise to get more data is on hold. Brad Templeton posted this on Reddit:

In a rather strange turn of events, while I wrote previously about the interesting set of questions the CPUC had planned for Waymo and Cruise, a sign of success in SF's quest to slow things down and get data at California PUC To Quiz Waymo & Cruise About Their Problems With The City Of San Francisco, reportedly this hearing was never supposed to have been posted and is cancelled. The text of the hearing request has been removed. I got this statement from the CPUC "The Ruling was inadvertently uploaded to the Docket and was therefore pulled. Commissioner Shiroma is doing further research and assessment."

I don't understand it as the hearing was not a ruling.

https://www.reddit.com/r/SelfDrivingCars/comments/1564fl5

 

[stopcrazypp]

Looks like the hearing with Waymo and Cruise to get more data is on hold. Brad Templeton posted this on Reddit:

https://www.reddit.com/r/SelfDrivingCars/comments/1564fl5

On the possibilities:
1) The hearing was supposed to be private and not on record until at least after occurred.
2) The release was a draft that was not supposed to be public.

On why it may be a "ruling":
The suggested hearing reverses the ruling by CPUC that SF overreached in requesting this information, as such, just the request for information (instead of just approving the expansion, as should had been done already if they abided by the arguments in the previous January ruling) may be considered a ruling in itself. It is acknowledgement of the SF argument that the circumstances have changed sufficiently to require more information or changes before proceeding, instead of sticking with requirements drafted before driverless operation was widespread.

Brad brings up the worry of setting precedent of individual cities interfering with the process, but so far the concerns brought up by multiple cities are largely the same (SF is just the most visible, but LA brought up pretty much the same concerns). Also, cities feeling they have a voice in the process reduces the chances of going to extreme measures to block AV companies (there are many shenanigans that can be deployed that can make things very difficult for the AV companies if cities wanted to play dirty).

 

[diplomat33]

Some info on Cruise from GM's Q2 2023 Financials:

Cruise has now reached the 3 million driverless miles milestone in just 49 days which is nearly 2x faster than it took to get to 2 million.

Right now, the service operates in select parts of San Francisco, CA, Phoenix, AZ and Austin, TX and is currently doing around 10,000 rides a week.

Cruise recently hit a peak of 390 driverless AVs running concurrently across all operating cities.

They achieved their first million driverless miles safety report shows 54% fewer collisions overall vs human drivers in a comparable driving environment.

The service has seen an ~15% decrease in cost per mile each month, for the last 6 months, led by optimizations in infrastructure, process improvements, and automation.

Cruise is now targeting costs below $1 per mile.

GM's Cruise division doing 10,000 rides per week, now targeting $1 per mile - techAU

Looking for the latest news, reviews, and insights on technology in Australia? Look no further than our tech website! From the newest gadgets to the latest software releases, we've got you covered with up-to-date information and expert analysis. Stay ahead of the curve and explore the world of...

techau.com.au

For comparison, according to the article, the average cost per mile for transportation in the US is estimated at between $0.58 and $0.75 per mile for a gas-powered car, or $0.30 – $0.45 per mile for an EV. So a target cost below $1/mile is good.

 

[diplomat33]

Kyle Vogt on GM Earnings Call with some info on Cruise's plans for SF and more:

Yes, good question. So on the San Francisco side, so right now, as I said earlier, a significant portion of our fleet is operating 24x7, and that service is open to employees. So we are not far from opening that up to the general public. I can't give specific dates. But basically, we're operating that service to employees. Things are looking pretty good. So that is coming pretty soon.

And as for what it would take to blanket a city like San Francisco, our goal is, as I think I said on previous calls is to make sure that we ramp up manufacturing capacity. We've got a variety of markets to absorb those vehicles. And there are practical reasons to ramp up gradually in the city, just to make sure it acclimates as it's transitioning to a new form of mobility.

So it's not our intention just to sort of vehicles and sort of direct them all into a single city. That's our perspective. There's over 10,000 human ride hill drivers in San Francisco, potentially much more than that, depending on how you count it. Those drivers, of course, aren't working 20 hours a day like a robotaxi could. So it does not make a very high number to generate significant revenue in a city like San Francisco. But certainly, there's capacity to absorb several thousand per city at minimum.

 

[diplomat33]

https://twitter.com/x/status/1684027030722195456

 

[willow_hiller]

https://twitter.com/x/status/1684027030722195456

Correct me if I'm wrong, but that actually sounds like very low utilization. From the April recall we know there are at least 300 Cruise vehicles on the road. 10,000 rides per week for 300 vehicles works out to just over 33 rides per vehicle per week. Or ~5 rides per vehicle per day. Or each vehicle being ridden once every ~5 hours.

And if there are more than 300 vehicles, the math gets worse. If it's grown to 500 vehicles, that's one ride every ~8 hours.

 

[diplomat33]

Correct me if I'm wrong, but that actually sounds like very low utilization. From the April recall we know there are at least 300 Cruise vehicles on the road. 10,000 rides per week for 300 vehicles works out to just over 33 rides per vehicle per week. Or ~5 rides per vehicle per day. Or each vehicle being ridden once every ~5 hours.

And if there are more than 300 vehicles, the math gets worse. If it's grown to 500 vehicles, that's one ride every ~8 hours.

I think you are correct. And Cruise has said that their peak is 390 vehicles at the same time. So unless they only do like 50 vehicles most of the same time and only briefly peak to 390, it would be low utilization indeed. I think this explains how they are able to say that they are growing at 49% per month. It's easy to grow fast when you are starting at almost no utilization. And since utilization is low, there is lots of room to scale up number of rides per week. But I think that 49% growth per month will not be sustainable. As they get higher utilization, they will hit a plateau. They can expand the geofence or expand the hours of operations to give themselves more room for growth and I think they will continue to growth but not at 49% every month.

 

[willow_hiller]

Waymo is taking a step back from autonomous trucking to focus on the Waymo Driver for passenger vehicles:

Doubling down on Waymo One

At Waymo, we’re developing a generalizable, fully autonomous Driver – the Waymo Driver – designed to integrate with a variety of vehicle platforms and commercial applications including ride-hailing, trucking, local delivery, and personal car ownership. Given the tremendous momentum and...

waymo.com

 

[Bitdepth]

Waymo is taking a step back from autonomous trucking to focus on the Waymo Driver for passenger vehicles:

Doubling down on Waymo One

At Waymo, we’re developing a generalizable, fully autonomous Driver – the Waymo Driver – designed to integrate with a variety of vehicle platforms and commercial applications including ride-hailing, trucking, local delivery, and personal car ownership. Given the tremendous momentum and...

waymo.com

To be honest it reads more like cutting cost and firing people. But that's just me and being cynical.

 

[diplomat33]

To be honest it reads more like cutting cost and firing people. But that's just me and being cynical.

Kirsten Korosec of Tech Crunch says the vast majority of the trucking division will be absorbed into the ride-hailing division:

https://twitter.com/x/status/1684296075392475138

 

[diplomat33]

Waymo is taking a step back from autonomous trucking to focus on the Waymo Driver for passenger vehicles:

Doubling down on Waymo One

At Waymo, we’re developing a generalizable, fully autonomous Driver – the Waymo Driver – designed to integrate with a variety of vehicle platforms and commercial applications including ride-hailing, trucking, local delivery, and personal car ownership. Given the tremendous momentum and...

waymo.com

I think it is a logical decision. Waymo has been making nice progress with the Waymo Driver and with scaling their ride-hailing in SF and Phoenix. And SF, LA and Phoenix have a total combined population of 5M+. So ride-hailing in LA, SF and Phoenix definitely has a lot of profit potential. As Waymo continues to improve the capabilities of the Waymo Driver, they can scale more in SF, LA and Phoenix and offer a really meaningful and eventually profitable service that benefits the general public. And I see this as a good sign that Waymo will get into personal car ownership. They specifically mention building a generalizable driver for personal car ownership in the first sentence. And the improvements in the Waymo Diver to safely handle city driving and highway driving in all weather can be directly applied to consumer cars. Autonomous trucking, on the other hand, might help businesses deliver goods safer and more efficiently but does not really benefit the general public. Ride-hailing and personal car ownership does help the general public. So going all-in on autonomous ride-hailing does seem to offer more benefits than autonomous trucking. It is also smart because it means they won't be splitting their resources between two different business models. They can focus more resources on ride-hailing and make faster progress.

So I think Waymo's strategy will be to continue to improve the Waymo Driver capabilities. Use those improvements to scale robotaxis to bigger geofences and more cities. And then once the Waymo Driver is proven capable and safe enough in a big enough ODD (city, suburb, rural, highway, all weather), they can put it on consumer cars or lease their existing robotaxis from their fleet as personal cars.

 

[Doggydogworld]

To be honest it reads more like cutting cost and firing people. But that's just me and being cynical.

Agreed. IMHO it also means highways are farther in the future than thought and/or the Teamsters have "Union Joe" is their pocket.

Absent regulatory blockades, trucking would be a bigger market than ride-hail in the intermediate term. Robotaxis could be bigger long-term if they can displace personal car ownership. It'll be years before costs drop that much, though, and it's not clear how many people are even willing to give up the convenience of personal cars.

Robotaxis can generate a lot more hype and probably lead to a much bigger IPO. That could be the deciding factor. Google has funded Waymo for almost 15 years now, it's only natural to want a payoff at some point.

 

[stopcrazypp]

I think it is a logical decision. Waymo has been making nice progress with the Waymo Driver and with scaling their ride-hailing in SF and Phoenix. And SF, LA and Phoenix have a total combined population of 5M+. So ride-hailing in LA, SF and Phoenix definitely has a lot of profit potential. As Waymo continues to improve the capabilities of the Waymo Driver, they can scale more in SF, LA and Phoenix and offer a really meaningful and eventually profitable service that benefits the general public. And I see this as a good sign that Waymo will get into personal car ownership. They specifically mention building a generalizable driver for personal car ownership in the first sentence. And the improvements in the Waymo Diver to safely handle city driving and highway driving in all weather can be directly applied to consumer cars. Autonomous trucking, on the other hand, might help businesses deliver goods safer and more efficiently but does not really benefit the general public. Ride-hailing and personal car ownership does help the general public. So going all-in on autonomous ride-hailing does seem to offer more benefits than autonomous trucking. It is also smart because it means they won't be splitting their resources between two different business models. They can focus more resources on ride-hailing and make faster progress.

So I think Waymo's strategy will be to continue to improve the Waymo Driver capabilities. Use those improvements to scale robotaxis to bigger geofences and more cities. And then once the Waymo Driver is proven capable and safe enough in a big enough ODD (city, suburb, rural, highway, all weather), they can put it on consumer cars or lease their existing robotaxis from their fleet as personal cars.

I guess that is one way to see it, but as another mentioned, the cynic would say this is Alphabet putting the whip on Waymo to get results (very similar to what Cruise was forced to do). I mentioned in the past that Alphabet has a history of suddenly pulling projects when they lose patience and it seems the trucking division has reached that point.

 

[diplomat33]

I guess that is one way to see it, but as another mentioned, the cynic would say this is Alphabet putting the whip on Waymo to get results (very similar to what Cruise was forced to do). I mentioned in the past that Alphabet has a history of suddenly pulling projects when they lose patience and it seems the trucking division has reached that point.

Oh certainly. I don't think the two views are mutually exclusive. I could definitely see Alphabet telling Waymo that the trucking part was dead weight, cut it and make the ride-hailing division work. But it is also true that focusing on ride-hailing only does have benefits and could help Waymo be more successful.

And we all know that investors were not going to write blank checks forever. They were patient while autonomous driving was in development but at some point, they will want to see returns. So there is huge pressure on robotaxi companies like Waymo and Cruise to show results. We saw that Argo got the axe because they did not show results fast enough. As you said, GM is pressuring Cruise to show results. Hence why Cruise is now focused on scaling faster, announcing new cities, even while they have some reliability issues. So it is not surprising that Waymo would be feeling the pressure too. But this could be a good thing if it means AVs scale faster. Maybe these companies do need some pressure put on them to speed things up?

 

[diplomat33]

Robotaxis could be bigger long-term if they can displace personal car ownership. It'll be years before costs drop that much, though, and it's not clear how many people are even willing to give up the convenience of personal cars.

I don't think the plan is for robotaxis to displace personal car ownership. Frankly, that's not realistic. There are 278M+ cars in the US. Waymo would need tens of millions of robotaxis to even start to put a dent in that. And even if Waymo somehow did manage to displace personal car ownership in one city like SF or Phoenix, it would still be a tiny blip compared to total personal car ownership in the US. Rather, I think the plan is to use robotaxis to improve the driverless tech and do both robotaxis and personal cars. That seems to match the line in the blog about building a generalizable driver for ride-hailing, trucking, delivery and personal car ownership. It hints that Waymo wants to adapt the Waymo Driver eventually to personal cars, not replace them with robotaxis. At least that's my take. And I think licensing the Waymo Driver to consumer cars is far more realistic than trying to displace pesonal cars with robotaxis.

On a related note, Alphabet's CFO, Ruth Porat, just got promoted to President & Chief Investment Officer, with powers to oversee various Alphabet projects including Waymo. I don't think it is a coincidence. This decision to shift away from trucking and focus on ride-hailing likely comes from Ruth Porat.

Source: Google’s CFO just got a promotion

 

[SidetrackedSue]

From today's Guardian: San Francisco’s Safe Street Rebel group has waged a war against robotaxis in an effort to end vehicle dominance in the city and promote public transport - by putting traffic cones on the hoods of robotaxis to paralyse them.

The rebel group stopping self-driving cars in San Francisco – one cone at a time

The Safe Street Rebel group has waged a war against robotaxis in an effort to end vehicle dominance in the city and promote public transport

www.theguardian.com

 

[diplomat33]

From today's Guardian: San Francisco’s Safe Street Rebel group has waged a war against robotaxis in an effort to end vehicle dominance in the city and promote public transport - by putting traffic cones on the hoods of robotaxis to paralyse them.

The rebel group stopping self-driving cars in San Francisco – one cone at a time

The Safe Street Rebel group has waged a war against robotaxis in an effort to end vehicle dominance in the city and promote public transport

www.theguardian.com

Old news.

 

[MP3Mike]

I guess we know why Waymo gave up on their trucking product and doesn't currently drive on the highway now:

https://twitter.com/x/status/1684360467802910721

Because apparently you can't do "safe highway speed driving" with current Lidar technology. You have to have the new FirstLight Lidar-on-a-chip unit to be able to do that. It just isn't possible with the large expensive Lidar that is available now.