Yes, there are more driverless Waymos in S.F | Hacker News Reader

Yes, there are more driverless Waymos in S.F

sfchronicle.com

29 pointseklitzke55 comments

I rode my first Waymo last weekend. It's _significantly_ better than an Uber. You get an almost 'white-glove' experience getting inside. Then, you have complete privacy. I could get work done in the car guilt-free. I was very surprised with the experience

> I could get work done in the car guilt-free.

...have rideshare drivers guilt-tripped you about working?

Or is this just some kind of reflexive response?

If the addition of a human per se produces guilt, I want to gentle suggest that work on boundaries is in order, and that no amount of automation is likely to fix the underlying problem.

It's OK to meet your wants and needs (such as getting some work done - a want I think nearly all of us share, many of us somewhat compulsively) in the presence of other people. No shame in it.

Can’t work if the driver is making small talk.

Rarely take taxis but when I get a car to the airport (which is admittedly probably more professional than the average taxi driver) I've sometimes had some enjoyable small talk when I've been in the mood but never had an issue if I wanted to relax or do some work/read.

There's a reason that limousines and Rolls-Royce's have dividing walls between the driver and passengers.

And police cars.

This is just (neurotypical?) human programming to be considerate of others. The driver feels this, too. I'd assume that the CEO of Uber Dara can manage his emotions and boundaries, well. But he mentions

> Some experiences made him feel slighted, such as when riders discussed personal problems and company secrets on speakerphone, as if there was no one else present.

https://archive.is/2023.04.17-151927/https://www.wsj.com/amp... (WSJ article)

Odd that you felt you had complete privacy. I would imagine you are monitored six ways to Sunday.

Video recording is always on, but audio recording is only when you call for customer service.

does that mean you're a waymo employee with access to their audio data retention management, or are you referencing their public claims? either way, simply by virtue of the video recording alone (which is truly a drop in the bucket), you have at best the same privacy as in a highly monitored human driven vehicle.

the absence of other humans is immaterial with regard to privacy.

I base this on the information that is displayed in the vehicle.

> the absence of other humans is immaterial with regard to privacy.

I think my attitude with Uber v. Waymo as very similar to AirBnB v. hotel chains. My trust in AirBnB hosts to not install hidden cameras is much lower than hotel chains.

The best thing about waymos is the screen that shows the machine vision and object identification. It sees all the cars, bikes, pedestrians, buses in a full 360 degree circle around the vehicle. It can make good predictions about what each one is doing.

This visual makes it clear how much better it is at driving than humans. There aren't blind spots, it stops at stop signs, it doesn't dangerously speed around corners.

Human drivers create a lot of dangers for pedestrians and cyclists in the city.

This object identification and prediction capability can be easily overloaded. Last night, coming home from Summer Symphony (Idaho's big improvement on Frisco's Stern Grove), I encountered 30 or so drivers behaving rather erratically due to a small roadside meadow containing at least 75 deer, all of which could decide to cross the roadway en masse within a few seconds. I successfully navigated all this, but I would NOT want to be in a Waymo for this trip. And this was in very good weather; obviously bad weather complicates this scenario even further.

Waymo vehicles have definitely encountered roads with 75+ people by the side of the road, and have definitely encountered deer. I would feel very safe in a Waymo in the circumstance you described.

A human is likely better than a computer at predicting the movement and behaviour of a single object.

A computer is going to be way better than any human at doing the same for a large number of objects simultaneously.

In my experience in SF, a Waymo would handle this just fine. This isn’t really a good example of how it “can be easily overloaded” because there are no Waymo in Idaho?

"Waymo operates about 300 driverless cars in the city, up from the roughly 250 robotaxis it used to start commercial service last August."

Sounds like a relatively small increase in actual cars over the last year despite a significant increase in use of those cars.

I regularly walk by one of their parking lots (14th St & Shotwell), which used to be typically relatively full, and now it's generally pretty empty. They're all out and about ... but on the other hand, a significant proportion of the waymo vehicles I see on the street are empty. I would be curious to see the difference in miles driven without a passenger per vehicle relative to Uber/Lyft. Are those empty miles about distance between drop-off and next pick-up? Or are they still doing a lot of extra miles just to collect data (e.g. going down streets that fares typically don't go down)?

I wonder how far they are from profitability on a cost per ride basis. The biggest long term risk to self driving cars is probably that number being unreachable.

The super expensive base vehicles don't help either. They need lidar hats for the Kia Forte or whatever

Just a LIDAR on top is not NEARLY the instrumentation requirement. There are cameras and LIDARs all around, and many other changes made to the car. Designing a strap-on self driving kit that would actually work is likely impossible to make cost-efficient without the car manufacturer's help.

Replacing ~minimum wage drivers in conventional cars with expensive cars backed by expensive software developers and operations staffs is almost certainly not a business model. There may be (probably will be) a business model behind autonomous driving technology in some undefined future. But there's no guarantee Waymo won't be another killed by Google in the interim. The thought from a decade ago that the average teenager wouldn't need to learn to drive is basically a fantasy.

If there were a fleet of a million Waymo cars, 2,000 support staff, and 2,000 engineers keeping a stable technology running, and vehicle wear and tear was low (possibly lower than a human driver running an ICE car ragged)...it could work. But they're going to be unprofitable for a long, long time.

It's hard for me to believe that Alphabet/Google even wants to be this huge robo-taxi service in the long run. I have to believe they want to be a technology supplier. Which means they need to enable robo-taxis-are-us to be a cheaper taxi service than Uber et al.

Price needs to be comparable, but not necessarily cheaper. Waymo rides are safer, cleaner, and more private than Uber rides.

The want it because time you spend driving is time you could have spent browsing the web and therefore looking at Google ads

2000 support staff handling troubleshooting, cleaning, and maintenance on 500 cars per person? Not likely.

Uber itself has 30,000 employees and 7M drivers and is highly profitable. Waymo could support a lot more staff than you suggest and be profitable.

They just poured $5B into it, that's a decent runway.

https://techcrunch.com/2024/07/23/alphabet-to-invest-another...

Could the hardware price come down with scale, or is it inherently expensive?

If the former, I think a viable model could be to eventually start selling the hardware (along with ongoing software updates, cloud services, maintenance, etc.). Let individuals across the world buy self-driving cars and run them as small businesses through whatever rideshare services they want in whatever locations they want. If there's a way, people will make the economics work. Over time, autonomous vehicle operators could eventually price human drivers out of the market on Uber and Lyft.

I see the current phrase of centrally managing all this as essentially a public beta test. Once the tech is proven and the regulatory environment is mature, the situation will change drastically. The Waymo app/network would no longer be needed at that point, unless it happened to become a big enough cash cow to be worth keeping around.

Presumably the incremental cost comes down over time and selling the tech to car manufacturers could become an interesting business. (I'm skeptical it would be a direct to consumer business.) I assume that is how it will play out but it will probably be over a fairly long time horizon.

I'm also skeptical that the "let others use your car when you're not" ever becomes a major model.

That's a good point, agreed, I don't think Waymo would need to be the vendor of the actual vehicle. It makes more sense that existing manufacturers would sell "Waymo Inside" models that include their tech, and then services like Uber and Lyft would build integrations on Waymo's SDK to be able to manage such vehicles on behalf of the owners.

What value is the owner providing here? For traditional cabs they're doing (1) driving, (2) customer acquisition, (3) maintenance, (4) capital provision. For Uber/Lyft the customer acquisition is centralized, but they're still doing 1, 3, 4. For a Waymo-like service the owner is basically just investing cash and doing occasional cleaning/maintenance work.

Is capital so scarce that this is needed for services to operate? And is decentralized maintenance/cleaning really going to be more economical than centralized operations that benefit from economies of scale?

In my mind, it would be controlling both pricing and geographic allocation of resources in a market-driven way. The idea is to be a strategy for efficiently rolling out the service at scale in a way that distributes the risks and profits, like how Uber and Lyft grew by outsourcing to independent drivers rather than buying a bunch of cars and hiring a bunch of drivers in particular locations.

I don't know how the numbers work out, but if the parent commenter is correct that hardware costs are going to be a bottleneck, it seems reasonable to me that essentially crowdfunding a global deployment would be a way to address that. It could accelerate the rollout and provide economies of scale more quickly than Waymo alone footing the bill, and wouldn't preclude them from still additionally investing their own capital.

Aside from that, it seems like there would be some demand for ownership of autonomous vehicles from people who either don't want to rent them out or only want to do so part-time, and selling the vehicles directly to tap into that would help further increase the scale of both production and robotaxiing.

My hypothesis is that in the early days the key to success for robotaxi service companies is going to be safety and consistency, in order to convince regulators that huge fleets of robotaxis are safe. "Early days" doesn't just mean the current technical workup period -- it also covers the time period where robotaxi services start to eat into the market for privately-owned cars. That's the point where you'll see the heaviest anti-robotaxi lobbying from entrenched competitors, which will include not just taxi services but also politically-connected interests like car dealerships, manufacturers, service centers, etc. These folks will see their entire livelihood disappearing, and "robotaxis aren't safe" is the obvious argument they'll turn to.

During this period, it seems like a huge risk for a robotaxi company to play it "fast and loose" with car quality or maintenance. You'd want to maintain exceptional standards on anything safety-critical, and that means you don't want to be seen as outsourcing critical hardware purchases and maintenance to random sole-proprietor businesses. (Maybe somewhere down the line this will change, when the tech is so advanced and accepted that anyone can offer it.)

Obviously I might be wrong. Maybe capital is going to become so tight that it'll be hard for the Waymos of the world to scale up quickly? But at least for the first, say, 7 years of major expansion it feels like safety and quality are going to dominate the conversation.

That makes sense, I could definitely see things play out that way too.

I am curious how much support staff this fleet has, and how much work they do. The article mentions nothing about that.

I’d like to see the financial analysis.

This job competes directly with San Franciscans who pay taxes and commuting drivers who spend money on food and gas.

When we wipe them out and replace it with a tax-dodging multinational corporation what happens to our local economy?

When was the last time you had an Uber driver that actually lived in San Francisco? Most of the drivers I talk to were driving in, from as far away as San Diego.

Doubtful. I've done that car trip multiple times, it's a 7-9 hour drive.

Yeah, it was extraordinarily far, which is why I remember the conversation. But he drove up to SF and slept in his car for the weekend because the trips were a lot more valuable here.

Are you saying a driver commuted 8 hours north to drive for Uber? and then back when he was done?

Yeah that’s not a thing anymore and it hasn’t been lucrative to do that in quite some time.

Me, my neighbor, my other neighbor, two guys down the block from me…

There are at least 3 people in my building that do it. I also know a handful of cabis that all live in the city.

Waymo pays 3.25% to SF and when the new tax inevitably passes in November it will be 7.75%.

This seems strongly to me like a broken window argument. Taken another way, you're paying people that could likely be doing more productive things to sit in a car and burn gasoline.

If money draining out of the economy is a major concern, there could just be a tax on AI rideshares (which given the political environment seems quite likely)

Are you in favor of UBI?

Generally, yes. But implementing UBI by wasting people's time is a bad idea.

You have to read carefully the article to extract the information. This is like the opposite of journalism.

tldr:

* 903,000 vehicle miles traveled during commercial driverless ride-hailing in May (57% increase from April).

* Waymo increased from 250 taxis last August to 300 in May.

So there's almost no new taxis, the increase comes from an increased utilization of the fleet.

A 2016 Muni analysis said that daily VMT in SF was 6 million and that Uber+Lyft were 570k of that, so given that scale Waymo would be about 5% of TNC traffic and 0.5% of all VMT.

The 903k miles driven is across all of California while the 20% month over month increase in Waymo cars from 250 to 300 was just in San Francisco.

Waymo isn't going to revolutionize transportation until it costs less than an Uber or Lyft. Right now it's just a novell way for high salaried folks to get around or to show off when family is in town. In my experience it's been 25-50% costlier than the competition.

Sure, there are a lot of engineers to pay, but the same applies to Uber and Lyft.

Since Uber and Lyft make no money it's not a reasonable comparison, they are only cheap because they are financed by VC money. Uber and Lyft only exists today because there's an expectation that in a not so distant future they'll be able to replace all their pesky drivers that demand compensation and benefits for Waymos.

Uber is significantly profitable.

Possibly only because drivers don't realize the hidden cost of insurance, maintenance, accidents, fuel, etc. It's anyone's guess if there is a limitless supply of desperate drivers, or if they will wise up.

My recent experiences puts them at +$4-8 the cost, so similar price after tip. The biggest limiting factor for me is service range and +20% ride times

Reply on news.ycombinator.com