14 ms·

Waymo’s Backseat Drivers: Confidential Data Reveals Self-Driving Taxi Hurdles

142 points by ballmers_peak 7y ago

mdorazio 7y ago
Interesting analysis using comments/rider scores. 10% improvement over 6 months is actually pretty good, though. Even if that rate slows as trickier edge cases get handled, It still implies early 2020s for true driverless readiness in select locations. That's pretty much on track with overall predictions for the autonomous market.
- bertil 7y ago
  I think standard car manufacturers and insurances would be in awe if their car safety improved at that rate.
  - onlyrealcuzzo 7y ago
    Well, there's a ceiling of 100% safety, so it's diminishing returns after a certain point.
- mactrey 7y ago
  10 percentage points, 25% relative improvement (40% non-5 stars to 30%), seems quite good to me. Although I wonder if seasonality is important here - obviously Phoenix isn't Minnesota but could driving conditions have been worse in Q1 compared to this summer, from the perspective of a self-driving car?
  - mdorazio 7y ago
    Definitely could be a factor. Other drivers might also be more used to seeing them on the road, which could have a positive impact in some way. I think the overall trend is probably pretty strong at that level of improvement, though.
- IshKebab 7y ago
  That's assuming their rate of improvement is constant and there isn't a long tail of difficulty. But there is. The long tail of difficult driving situations is very very long. An example from my most recent drive: I drove through a common where cows graze. I doubt a driverless car is programmed to slow down near cows. The only solution I can see is to whitelist roads and start with the simplest (motorways/highways) then gradually expand to more and more roads. I guess that's kind of what they're doing - suburban America is very easy (although you still have pedestrians and cyclists to deal with unlike motorways), and maybe whitelisting is the cause of the routing complaint.
- jacques_chester 7y ago
  If you get nagged to answer questions every time you give a 4/5 you eventually stop giving 4/5 and 5/5 everything because you just want to get on with your day. The whole "if it's not amaaaaazing it's terrible" thing is idiotic.
netinstructions 7y ago
I wonder how many people are upset that the vehicles had to drop them off in a valid passenger load/unload zone as opposed to the usual Lyft / Uber tactic of parking in a no parking / no stopping zone, bike lane, crosswalk, etc because it's most convenient for the drivers and passengers (at the safety and expense of everyone else sharing the space)... That quote at the end: > I guess Lyft has me spoiled. I like getting dropped off in front of the place im going too [sic] not just in the parking lot....
- bertil 7y ago
  It’s mainly a different lobbying tactic: - Uber and Lyft and more toe-stepping and will encourage their contractors to paint outside the line, deal with the consequence once the administration has caught up with them and is presented with the fait-accompli that this is voters’ expectations now. - Google/Waymo has better relationships with local authorities and can obtain the permit to drop people off after they’ve proven they are playing within the line — and can wait for, and eventually finance urban furniture changes. Both use people’s expectations, but differently.
- btrettel 7y ago
  Cyclist here. In my experience, the majority of the time a driver stops or parks in the bike lane in an urban area in the US (e.g., I live in Austin), there's a legal parking/stopping spot within a reasonable walking distance, often within 50 to 100 feet. (If this isn't true where you live, consider the difference in the location. There are probably exceptions too. I'm told that legal parking isn't typically close in SF.) Then again, my idea of "reasonable walking distance" seems longer than most people's. Having spoken to many drivers who have parked in the bike lane, I'm amazed by how negatively some have reacted to me recommending that they park as little as 50 feet away. In some cases the non-bike-lane spot is closer but the convenience of pulling to the side of the road rather than doing a more complicated maneuver seems irresistible. If Waymo follows the law, good for them. Makes me more likely to be a customer of theirs in the future.
  - jldugger 7y ago
    I often see waymo vans near San Antonio and El Camino in Mountain View. It's kind of a nightmare drive for all parties. Curbside parking is allowed, there are no demarcated bike lanes, and much of the road is in suboptimal condition. There is often construction going along sidewalks and buildings, and uber dropoffs are common. You occasionally see cyclists, though I suspect most stick to a side street. What I suspect people are complaining about is that Waymo doesn't do curbside dropoffs at locations with a parking lot -- not common biking routes. I bet Waymo doesn't have the data to know whether a curb is painted yellow, blue, or red, and just avoids them, while a Lyft driver would probably put on hazards and drop people off at yellow curbs and bus stops.
ttul 7y ago
The slow progress is not an indictment of self driving. This is one of the toughest engineering challenges ever mounted by humanity.
- Judgmentality 7y ago
  I think it parallels nuclear fusion in a lot of ways. The impact is unquestionable, but the feasibility is an ongoing debate.
- dumbfoundded 7y ago
  Yeah, that's a bit of an overstatement. We created nukes, landed on the moon, took sludge out of the ground and used it to power the world. We connected this world with wires and glass fibers to build a real-time global communication system that also gathered all of the world's information in a singular and immediately accessible place. Building a self driving car is hard but really not as tough. How quickly do think we'd get self driving cars if the USA spent 4% of the federal budget on it like NASA received in the 60s. (That's about $40B a year for a decade).
  - paxys 7y ago
    All great achievements, but none of these is in the same ballpark of difficulty as creating a general artificial intelligence (which is pretty much what self driving cars need to achieve true level 5 autonomy).
    - dumbfoundded 7y ago
      Level 5 driving doesn't require AGI at all. It's an extremely narrow set of problems that have to be solved incredibly well. It mostly just comes down to creating an accurate 3d representation of the world from a bunch of sensors. You also have to correctly segment and label each object in that 3d representation. If you did those two things extremely well, the actual driving logic can be hardcoded. The problem is that each of these systems has problems so they all have to improve and compensate for each other.
damq 7y ago
> The data provide the most detailed view yet of the passenger experience inside vehicles from Waymo, which doesn’t disclose the information even though it uses public roads as its testing grounds. What's up with this statement? Should I be forced to publicize my phone calls just because I made them while driving on public roads? It's utterly bizarre that the author thinks he is entitled to see Waymo's data.
- moomin 7y ago
  No, because you get the use of the roads as part of paying your taxes. These are special licenses, providing resources for private research for free with the costs being borne by the public.
  - lallysingh 7y ago
    What costs?
    - gowld 7y ago
      Passengers, who are property of the government, I guess?
    - mandelbrotwurst 7y ago
      There's a flaw in parent's comment in that the employees are themselves tax-paying citizens, but to answer your question anyway: land use, congestion, pollution, wear and tear, etc.
- ForrestN 7y ago
  I'm not sure if I agree with the writer's statement here, and I'm not sure the right outcome is for Waymo to share data, but I don't think your analogy works. Your individual phone calls are made as part of your general participation as an individual in society. One of the largest companies in the world, which does its best to avoid paying taxes, is using public roads as a fundamental part of the infrastructure for a project to generate data. Maybe a better analogy is people who grow large amounts of marijuana in national parks? Yes, it's true that the growers are part of the public, and that the public owns the land, but... I wouldn't have a problem with stuff like this if corporations were taxed at reasonable rates, and didn't participate so wholeheartedly in efforts to corrupt our democracy. Google donates to many truly vile, despicable politicians in order to shirk accountability, hamper regulation, and accomplish just this sort of de-facto subsidy and others like it.
lallysingh 7y ago
I wonder if they spread out to more areas - where there's more variety of driving conditions - if they could start rating difficulty well enough to estimate when they would be able to operate in each one. So maybe they could start working in 2 years in sprawled suburbs in hot areas where you don't have many cyclists or pedestrians? Or is that Phoenix already and it's still too hard?
avocado4 7y ago
10500 rides, 70% are rated 5 stars (perfect). 10% increase from last year.
- scarmig 7y ago
  5 stars isn't perfect. 5 stars is "there's nothing so bad that I want to take the time to complain about it." Waymo employees, who are encouraged to be especially tough, give reviews that are 47% negative. That's likely closer to the metric for perfect. We don't get perfection from human drivers either, of course. Though part of the promise of Waymo is much better performance than humans, which it apparently isn't close to yet. And this data is for relatively common cases: for the long tail, one-in-a-million case, the conventional wisdom is that humans would do better.
  - falcor84 7y ago
    If I were "encouraged to be especially tough", I can promise you that the taxis I ride in would get a much worse score than 47% negative.
    - DiogenesKynikos 7y ago
      For entirely different reasons, though. Your taxi driver is not going to randomly freak out because they mistook a bush for a pedestrian or didn't understand what traffic cones mean. The bad taxi driver is just going to drive too fast while talking on their cellphone.
      - BigJono 7y ago
        > or didn't understand what traffic cones mean You'd be surprised. I've had a taxi driver turn up train tracks before.
CloudYeller 7y ago
When are they going to just seal off an area so it's self-driving vehicles only, and test them there? In a 100% self driving environment, edge cases are probably much easier to reason about. And if the future is 100% self driving, they should probably get some training data from that environment anyway.
- icelancer 7y ago
  >> When are they going to just seal off an area so it's self-driving vehicles only, and test them there? Talk about your all-time overfitting problem.
- tlb 7y ago
  Waymo has quite a large autonomous-only test facility. https://www.theatlantic.com/technology/archive/2017/08/inside-waymos-secret-testing-and-simulation-facilities/537648/ https://www.theatlantic.com/technology/archive/2017/08/insid... They also use it to populate simulations with realistic traffic models.
- berdario 7y ago
  They already ran plenty of tests in a controlled environment: https://www.wired.com/story/google-waymo-self-driving-car-castle-testing/ https://www.wired.com/story/google-waymo-self-driving-car-ca...
- javagram 7y ago
  With bicycles and pedestrians sharing the common roads, a 100% self driving environment anywhere but the freeway is never going to happen though. They need to be able to handle the chaos.
xivzgrev 7y ago
Despite the complaints, this is fucking amazing. I remember just earlier this decade I was pissed off at a taxi company for saying a taxi will come in 20 minutes...maybe, dispatcher can’t promise. Now we got Uber/lyft, and then...DRIVERLESS vehicles?? Always seemed like science fiction but the future is here... The bitch of this business is the long tail of possible scenarios - before people have confidence you need to solve the long tail which is hard because not as much data / much less predictable. Sounds like from article though they are making headway!
moomin 7y ago
This might just be me, but it strikes me that the biggest problem self-driving cars has is that no-one actually needs them. Don’t get me wrong, you try to sell me a car that has the (fully capable) feature I’d definitely want it. But need it? Well, you’re not going to solve the problem of too many cars and the slow traffic that results with cars. It’d be hard to get special treatment for cars with a premium feature. Then there’s the taxi model, where the aim of the game is to undercut people that Uber has already managed to push below minimum wage. I mean, there’s a margin there, but is it really “Next Google” sized? Even with dedicated vehicles, that’s a lot of hardware to maintain. Now let’s assume it’s successful and sustainable. Let’s say we also manage to get it to work for freight. What have we achieved? Well, we’ve put a couple more of the jobs available to low-skilled workers on the scrap heap. There will be consequences to doing that, but I doubt Waymo will be footing that bill.
- gpm 7y ago
  There are 3.5 million truck drivers in the US [0]. They are paid at least minimum wage. Alphabet has 148.299B revenue per year [1]. A conservative estimate of the revenue waymo could earn just by replacing truck drivers is 7.25 (minimum US minimum wage) * 40 (working hours per weeks) * 50 (working weeks per year) * 3.5 million = $50 billion a year. I think once you expand past the US, and past trucks, this is definitely "Next Google sized". [0] https://www.trucking.org/News_and_Information_Reports_Industry_Data.aspx https://www.trucking.org/News_and_Information_Reports_Indust... [1] https://www.macrotrends.net/stocks/charts/GOOG/alphabet/revenue https://www.macrotrends.net/stocks/charts/GOOG/alphabet/reve...
  - Noos 7y ago
    That's a pyrhhic victory though. You'd essentially convert trucking into retail, making a semi-skilled workforce as well as small entrepreneurial one into a just-in time near contract workforce with high barriers to entry to start businesses.
- josephagoss 7y ago
  Disabled people need them, very old people need them. Drunk people need them. Lots of people could do with a driverless car.
gok 7y ago
What a strange thing to have leaked. Not details how the tech works, but the database of customer ride reviews?
- yifanl 7y ago
  I imagine ride reviews are considerably less confidential than training data.
  - m3at 7y ago
    Yes, also a few hundreds of terabytes of data are much less convenient to upload and share.
scarmig 7y ago
What's HN's stance on publishers posting links directly to HN? To be clear, this isn't meant as a blast against The Information or ballmers_peak; they're transparent about their affiliation and that this is an article white listed for readers referred by Hacker News. It's also a totally appropriate article for HN. That said, I kind of wish there was some in-line indicator of a motivated submission source. I guess a downside would be unethical publishers just laundering their submissions through ostensibly unaffiliated accounts, but I'd feel more comfortable knowing that we did have some kind of system here to encourage transparency.
- minimaxir 7y ago
  HN doesn't have any rules against self-promotion as long as it's not spammy/low quality content. Additionally, this is an interesting case as the custom link is an official way to bypass the paywall, which is the more frequent HN complaint: https://news.ycombinator.com/item?id=20414141 https://news.ycombinator.com/item?id=20414141 Comments are a different story, e.g. if a startup employee that works for the startup comments on a post about the startup, they should disclose it.
- abtinf 7y ago
  As long as HN has effective vote ring detection, I don’t see any issue with directly posting links. The crowd helps us here and paternalistic rules can only prevent it from functioning correctly. If the post is good, it will move up. If not, it won’t.
- deminature 7y ago
  The Information is some of the highest quality and well-researched tech journalism out there. They frequently break stories months in advance of other publications, and are one of the few tech outlets that actually warrant having a paywall, and are not just rehashes of press releases. I personally have no issue with publication staff posting content.
Animats 7y ago
That's useful info. I wonder where they had trouble with erratic steering. One Cruise video shows that where driving is alongside parked cars on a narrow street and the parking is irregular. The self driving vehicle is trying to stay in its half of the street unless absolutely necessary. Waymo still has to use "safety drivers". So far, nobody seems to have full self driving without "safety drivers". Which makes it useless. That's the big milestone to look for. When Waymo can get rid of the safety drivers, which they tried briefly last year, they're getting close to something useable. Some products take a long time to bring to market. Xerography - first copy made in 1939, commercial success 1959. Although we're now 15 years after the DARPA Grand Challenge, so we're coming close to that 20 year wait. Television - first broadcast, 1928, commercial success, around 1948. So 20 years agaon. Roller bearings - Timken founded 1898, Timken bearings in 80% of US cars by the 1920s. 20 years again. (Although it took a really long time for them to convert railroads. First locomotive with roller bearings, 1923. In 1949, they were struggling to get railroads to put roller bearings on freight cars.[1] In 1991, roller bearings became mandatory for US inter-line interchange of freight cars.) Air brakes and automatic couplers only took 7 years, but that's because the U.S. Congress made railroads convert and standardize between 1893 and 1900. What took a really long time, in post-1900 technology? (Before 1900, manufacturing infrastructure wasn't really ready for fast deployment.) [1] https://www.youtube.com/watch?v=R-1EZ6K7bpQ https://www.youtube.com/watch?v=R-1EZ6K7bpQ
- wpietri 7y ago
  Yeah, I think this is going to be a long one. Rodney Brooks, AI researcher and Roomba cofounder, says he doesn't expect a real robot taxi service before 2035: https://rodneybrooks.com/my-dated-predictions/ https://rodneybrooks.com/my-dated-predictions/ > When Waymo can get rid of the safety drivers [...] they're getting close to something useable. I suspect that they can cheat this one a little. As long as they're in an area with good connectivity and the cars are smart enough to pull over if they don't get human guidance, I expect they'll move the backup drivers to a central location. Call it SAE level 3.5. It wouldn't be good enough to sell cars, but it would be workable for a ride service, and would allow them to undercut Uber, etc, quite handily.
  - 7y ago
rammy1234 7y ago
why autonomous cars get so much attention ? Why they are always on news , why they always trend. Apparently for a common man why we are excited when we should be skeptical and more careful. so anything other than excited. But for some reason these driverless cars do get most attention.
jeffk_teh_haxor 7y ago
A week ago Lyft driver who drove like a maniac. Will he get a saucy exposé too?
- DiogenesKynikos 7y ago
  If you're planning on cloning him millions of times and putting him behind the wheel of every car, then yes, he does deserve an exposé!
dmix 7y ago
> The data provide the most detailed view yet of the passenger experience inside vehicles from Waymo, which doesn’t disclose the information even though it uses public roads as its testing grounds. Why would it riding on public roads matter when it's about what happens inside a private vehicle during a high-value R&D experiment? They got permission from Arizona to run the tests and it will likely benefit Arizona's economy in return being the first to get the most training data.
- jessriedel 7y ago
  That phrase is just thrown in there to stoke the reader and create a story out of nothing. It indicates a lack of journalistic integrity.
  - calcifer 7y ago
    > It indicates a lack of journalistic integrity. Or, the author is suggesting that public resources should be used for public, not private benefit?
    - rowanG077 7y ago
      Self driving car technology has huge public benefit.
      - tracker1 7y ago
        If people aren't sharing the vehicle there's the same number of vehicles on the same number of roads, just less need for parking spaces. Of course, the potential for reduced accidents should help. Unfortunately, part of why AZ was probably chosen is that it's mostly sunny/clear weather during the year.
        rowanG077 7y ago
        Reduced accidents. Increased time for people using self driving cars. Companies no longer needing to pay for staff that pilots cars. Just the time and cost savings alone for private people and companies is hard to be overstated.

142 points by ballmers_peak 7y ago

Interesting analysis using comments/rider scores. 10% improvement over 6 months is actually pretty good, though. Even if that rate slows as trickier edge cases get handled, It still implies early 2020s for true driverless readiness in select locations. That's pretty much on track with overall predictions for the autonomous market.

bertil 7y ago
I think standard car manufacturers and insurances would be in awe if their car safety improved at that rate.
- onlyrealcuzzo 7y ago
  Well, there's a ceiling of 100% safety, so it's diminishing returns after a certain point.
mactrey 7y ago
10 percentage points, 25% relative improvement (40% non-5 stars to 30%), seems quite good to me. Although I wonder if seasonality is important here - obviously Phoenix isn't Minnesota but could driving conditions have been worse in Q1 compared to this summer, from the perspective of a self-driving car?
- mdorazio 7y ago
  Definitely could be a factor. Other drivers might also be more used to seeing them on the road, which could have a positive impact in some way. I think the overall trend is probably pretty strong at that level of improvement, though.
IshKebab 7y ago
That's assuming their rate of improvement is constant and there isn't a long tail of difficulty. But there is. The long tail of difficult driving situations is very very long. An example from my most recent drive: I drove through a common where cows graze. I doubt a driverless car is programmed to slow down near cows. The only solution I can see is to whitelist roads and start with the simplest (motorways/highways) then gradually expand to more and more roads. I guess that's kind of what they're doing - suburban America is very easy (although you still have pedestrians and cyclists to deal with unlike motorways), and maybe whitelisting is the cause of the routing complaint.
jacques_chester 7y ago
If you get nagged to answer questions every time you give a 4/5 you eventually stop giving 4/5 and 5/5 everything because you just want to get on with your day. The whole "if it's not amaaaaazing it's terrible" thing is idiotic.

netinstructions 7y ago

I wonder how many people are upset that the vehicles had to drop them off in a valid passenger load/unload zone as opposed to the usual Lyft / Uber tactic of parking in a no parking / no stopping zone, bike lane, crosswalk, etc because it's most convenient for the drivers and passengers (at the safety and expense of everyone else sharing the space)... That quote at the end: > I guess Lyft has me spoiled. I like getting dropped off in front of the place im going too [sic] not just in the parking lot....

bertil 7y ago
It’s mainly a different lobbying tactic: - Uber and Lyft and more toe-stepping and will encourage their contractors to paint outside the line, deal with the consequence once the administration has caught up with them and is presented with the fait-accompli that this is voters’ expectations now. - Google/Waymo has better relationships with local authorities and can obtain the permit to drop people off after they’ve proven they are playing within the line — and can wait for, and eventually finance urban furniture changes. Both use people’s expectations, but differently.
btrettel 7y ago
Cyclist here. In my experience, the majority of the time a driver stops or parks in the bike lane in an urban area in the US (e.g., I live in Austin), there's a legal parking/stopping spot within a reasonable walking distance, often within 50 to 100 feet. (If this isn't true where you live, consider the difference in the location. There are probably exceptions too. I'm told that legal parking isn't typically close in SF.) Then again, my idea of "reasonable walking distance" seems longer than most people's. Having spoken to many drivers who have parked in the bike lane, I'm amazed by how negatively some have reacted to me recommending that they park as little as 50 feet away. In some cases the non-bike-lane spot is closer but the convenience of pulling to the side of the road rather than doing a more complicated maneuver seems irresistible. If Waymo follows the law, good for them. Makes me more likely to be a customer of theirs in the future.
- jldugger 7y ago
  I often see waymo vans near San Antonio and El Camino in Mountain View. It's kind of a nightmare drive for all parties. Curbside parking is allowed, there are no demarcated bike lanes, and much of the road is in suboptimal condition. There is often construction going along sidewalks and buildings, and uber dropoffs are common. You occasionally see cyclists, though I suspect most stick to a side street. What I suspect people are complaining about is that Waymo doesn't do curbside dropoffs at locations with a parking lot -- not common biking routes. I bet Waymo doesn't have the data to know whether a curb is painted yellow, blue, or red, and just avoids them, while a Lyft driver would probably put on hazards and drop people off at yellow curbs and bus stops.

ttul 7y ago

The slow progress is not an indictment of self driving. This is one of the toughest engineering challenges ever mounted by humanity.

Judgmentality 7y ago
I think it parallels nuclear fusion in a lot of ways. The impact is unquestionable, but the feasibility is an ongoing debate.
dumbfoundded 7y ago
Yeah, that's a bit of an overstatement. We created nukes, landed on the moon, took sludge out of the ground and used it to power the world. We connected this world with wires and glass fibers to build a real-time global communication system that also gathered all of the world's information in a singular and immediately accessible place. Building a self driving car is hard but really not as tough. How quickly do think we'd get self driving cars if the USA spent 4% of the federal budget on it like NASA received in the 60s. (That's about $40B a year for a decade).
- paxys 7y ago
  All great achievements, but none of these is in the same ballpark of difficulty as creating a general artificial intelligence (which is pretty much what self driving cars need to achieve true level 5 autonomy).
  - dumbfoundded 7y ago
    Level 5 driving doesn't require AGI at all. It's an extremely narrow set of problems that have to be solved incredibly well. It mostly just comes down to creating an accurate 3d representation of the world from a bunch of sensors. You also have to correctly segment and label each object in that 3d representation. If you did those two things extremely well, the actual driving logic can be hardcoded. The problem is that each of these systems has problems so they all have to improve and compensate for each other.

damq 7y ago

> The data provide the most detailed view yet of the passenger experience inside vehicles from Waymo, which doesn’t disclose the information even though it uses public roads as its testing grounds. What's up with this statement? Should I be forced to publicize my phone calls just because I made them while driving on public roads? It's utterly bizarre that the author thinks he is entitled to see Waymo's data.

moomin 7y ago
No, because you get the use of the roads as part of paying your taxes. These are special licenses, providing resources for private research for free with the costs being borne by the public.
- lallysingh 7y ago
  What costs?
  - gowld 7y ago
    Passengers, who are property of the government, I guess?
  - mandelbrotwurst 7y ago
    There's a flaw in parent's comment in that the employees are themselves tax-paying citizens, but to answer your question anyway: land use, congestion, pollution, wear and tear, etc.
ForrestN 7y ago
I'm not sure if I agree with the writer's statement here, and I'm not sure the right outcome is for Waymo to share data, but I don't think your analogy works. Your individual phone calls are made as part of your general participation as an individual in society. One of the largest companies in the world, which does its best to avoid paying taxes, is using public roads as a fundamental part of the infrastructure for a project to generate data. Maybe a better analogy is people who grow large amounts of marijuana in national parks? Yes, it's true that the growers are part of the public, and that the public owns the land, but... I wouldn't have a problem with stuff like this if corporations were taxed at reasonable rates, and didn't participate so wholeheartedly in efforts to corrupt our democracy. Google donates to many truly vile, despicable politicians in order to shirk accountability, hamper regulation, and accomplish just this sort of de-facto subsidy and others like it.

lallysingh 7y ago

I wonder if they spread out to more areas - where there's more variety of driving conditions - if they could start rating difficulty well enough to estimate when they would be able to operate in each one. So maybe they could start working in 2 years in sprawled suburbs in hot areas where you don't have many cyclists or pedestrians? Or is that Phoenix already and it's still too hard?

avocado4 7y ago

10500 rides, 70% are rated 5 stars (perfect). 10% increase from last year.

scarmig 7y ago
5 stars isn't perfect. 5 stars is "there's nothing so bad that I want to take the time to complain about it." Waymo employees, who are encouraged to be especially tough, give reviews that are 47% negative. That's likely closer to the metric for perfect. We don't get perfection from human drivers either, of course. Though part of the promise of Waymo is much better performance than humans, which it apparently isn't close to yet. And this data is for relatively common cases: for the long tail, one-in-a-million case, the conventional wisdom is that humans would do better.
- falcor84 7y ago
  If I were "encouraged to be especially tough", I can promise you that the taxis I ride in would get a much worse score than 47% negative.
  - DiogenesKynikos 7y ago
    For entirely different reasons, though. Your taxi driver is not going to randomly freak out because they mistook a bush for a pedestrian or didn't understand what traffic cones mean. The bad taxi driver is just going to drive too fast while talking on their cellphone.
    - BigJono 7y ago
      > or didn't understand what traffic cones mean You'd be surprised. I've had a taxi driver turn up train tracks before.

CloudYeller 7y ago

When are they going to just seal off an area so it's self-driving vehicles only, and test them there? In a 100% self driving environment, edge cases are probably much easier to reason about. And if the future is 100% self driving, they should probably get some training data from that environment anyway.

icelancer 7y ago
>> When are they going to just seal off an area so it's self-driving vehicles only, and test them there? Talk about your all-time overfitting problem.
tlb 7y ago
Waymo has quite a large autonomous-only test facility. https://www.theatlantic.com/technology/archive/2017/08/inside-waymos-secret-testing-and-simulation-facilities/537648/ https://www.theatlantic.com/technology/archive/2017/08/insid... They also use it to populate simulations with realistic traffic models.
berdario 7y ago
They already ran plenty of tests in a controlled environment: https://www.wired.com/story/google-waymo-self-driving-car-castle-testing/ https://www.wired.com/story/google-waymo-self-driving-car-ca...
javagram 7y ago
With bicycles and pedestrians sharing the common roads, a 100% self driving environment anywhere but the freeway is never going to happen though. They need to be able to handle the chaos.

xivzgrev 7y ago

Despite the complaints, this is fucking amazing. I remember just earlier this decade I was pissed off at a taxi company for saying a taxi will come in 20 minutes...maybe, dispatcher can’t promise. Now we got Uber/lyft, and then...DRIVERLESS vehicles?? Always seemed like science fiction but the future is here... The bitch of this business is the long tail of possible scenarios - before people have confidence you need to solve the long tail which is hard because not as much data / much less predictable. Sounds like from article though they are making headway!

moomin 7y ago

This might just be me, but it strikes me that the biggest problem self-driving cars has is that no-one actually needs them. Don’t get me wrong, you try to sell me a car that has the (fully capable) feature I’d definitely want it. But need it? Well, you’re not going to solve the problem of too many cars and the slow traffic that results with cars. It’d be hard to get special treatment for cars with a premium feature. Then there’s the taxi model, where the aim of the game is to undercut people that Uber has already managed to push below minimum wage. I mean, there’s a margin there, but is it really “Next Google” sized? Even with dedicated vehicles, that’s a lot of hardware to maintain. Now let’s assume it’s successful and sustainable. Let’s say we also manage to get it to work for freight. What have we achieved? Well, we’ve put a couple more of the jobs available to low-skilled workers on the scrap heap. There will be consequences to doing that, but I doubt Waymo will be footing that bill.

gpm 7y ago
There are 3.5 million truck drivers in the US [0]. They are paid at least minimum wage. Alphabet has 148.299B revenue per year [1]. A conservative estimate of the revenue waymo could earn just by replacing truck drivers is 7.25 (minimum US minimum wage) * 40 (working hours per weeks) * 50 (working weeks per year) * 3.5 million = $50 billion a year. I think once you expand past the US, and past trucks, this is definitely "Next Google sized". [0] https://www.trucking.org/News_and_Information_Reports_Industry_Data.aspx https://www.trucking.org/News_and_Information_Reports_Indust... [1] https://www.macrotrends.net/stocks/charts/GOOG/alphabet/revenue https://www.macrotrends.net/stocks/charts/GOOG/alphabet/reve...
- Noos 7y ago
  That's a pyrhhic victory though. You'd essentially convert trucking into retail, making a semi-skilled workforce as well as small entrepreneurial one into a just-in time near contract workforce with high barriers to entry to start businesses.
josephagoss 7y ago
Disabled people need them, very old people need them. Drunk people need them. Lots of people could do with a driverless car.

gok 7y ago

What a strange thing to have leaked. Not details how the tech works, but the database of customer ride reviews?

yifanl 7y ago
I imagine ride reviews are considerably less confidential than training data.
- m3at 7y ago
  Yes, also a few hundreds of terabytes of data are much less convenient to upload and share.

scarmig 7y ago

What's HN's stance on publishers posting links directly to HN? To be clear, this isn't meant as a blast against The Information or ballmers_peak; they're transparent about their affiliation and that this is an article white listed for readers referred by Hacker News. It's also a totally appropriate article for HN. That said, I kind of wish there was some in-line indicator of a motivated submission source. I guess a downside would be unethical publishers just laundering their submissions through ostensibly unaffiliated accounts, but I'd feel more comfortable knowing that we did have some kind of system here to encourage transparency.

minimaxir 7y ago
HN doesn't have any rules against self-promotion as long as it's not spammy/low quality content. Additionally, this is an interesting case as the custom link is an official way to bypass the paywall, which is the more frequent HN complaint: https://news.ycombinator.com/item?id=20414141 https://news.ycombinator.com/item?id=20414141 Comments are a different story, e.g. if a startup employee that works for the startup comments on a post about the startup, they should disclose it.
abtinf 7y ago
As long as HN has effective vote ring detection, I don’t see any issue with directly posting links. The crowd helps us here and paternalistic rules can only prevent it from functioning correctly. If the post is good, it will move up. If not, it won’t.
deminature 7y ago
The Information is some of the highest quality and well-researched tech journalism out there. They frequently break stories months in advance of other publications, and are one of the few tech outlets that actually warrant having a paywall, and are not just rehashes of press releases. I personally have no issue with publication staff posting content.

Animats 7y ago

That's useful info. I wonder where they had trouble with erratic steering. One Cruise video shows that where driving is alongside parked cars on a narrow street and the parking is irregular. The self driving vehicle is trying to stay in its half of the street unless absolutely necessary. Waymo still has to use "safety drivers". So far, nobody seems to have full self driving without "safety drivers". Which makes it useless. That's the big milestone to look for. When Waymo can get rid of the safety drivers, which they tried briefly last year, they're getting close to something useable. Some products take a long time to bring to market. Xerography - first copy made in 1939, commercial success 1959. Although we're now 15 years after the DARPA Grand Challenge, so we're coming close to that 20 year wait. Television - first broadcast, 1928, commercial success, around 1948. So 20 years agaon. Roller bearings - Timken founded 1898, Timken bearings in 80% of US cars by the 1920s. 20 years again. (Although it took a really long time for them to convert railroads. First locomotive with roller bearings, 1923. In 1949, they were struggling to get railroads to put roller bearings on freight cars.[1] In 1991, roller bearings became mandatory for US inter-line interchange of freight cars.) Air brakes and automatic couplers only took 7 years, but that's because the U.S. Congress made railroads convert and standardize between 1893 and 1900. What took a really long time, in post-1900 technology? (Before 1900, manufacturing infrastructure wasn't really ready for fast deployment.) [1] https://www.youtube.com/watch?v=R-1EZ6K7bpQ https://www.youtube.com/watch?v=R-1EZ6K7bpQ

wpietri 7y ago
Yeah, I think this is going to be a long one. Rodney Brooks, AI researcher and Roomba cofounder, says he doesn't expect a real robot taxi service before 2035: https://rodneybrooks.com/my-dated-predictions/ https://rodneybrooks.com/my-dated-predictions/ > When Waymo can get rid of the safety drivers [...] they're getting close to something useable. I suspect that they can cheat this one a little. As long as they're in an area with good connectivity and the cars are smart enough to pull over if they don't get human guidance, I expect they'll move the backup drivers to a central location. Call it SAE level 3.5. It wouldn't be good enough to sell cars, but it would be workable for a ride service, and would allow them to undercut Uber, etc, quite handily.
- 7y ago

rammy1234 7y ago

why autonomous cars get so much attention ? Why they are always on news , why they always trend. Apparently for a common man why we are excited when we should be skeptical and more careful. so anything other than excited. But for some reason these driverless cars do get most attention.

jeffk_teh_haxor 7y ago

A week ago Lyft driver who drove like a maniac. Will he get a saucy exposé too?

DiogenesKynikos 7y ago
If you're planning on cloning him millions of times and putting him behind the wheel of every car, then yes, he does deserve an exposé!

dmix 7y ago

> The data provide the most detailed view yet of the passenger experience inside vehicles from Waymo, which doesn’t disclose the information even though it uses public roads as its testing grounds. Why would it riding on public roads matter when it's about what happens inside a private vehicle during a high-value R&D experiment? They got permission from Arizona to run the tests and it will likely benefit Arizona's economy in return being the first to get the most training data.

jessriedel 7y ago
That phrase is just thrown in there to stoke the reader and create a story out of nothing. It indicates a lack of journalistic integrity.
- calcifer 7y ago
  > It indicates a lack of journalistic integrity. Or, the author is suggesting that public resources should be used for public, not private benefit?
  - rowanG077 7y ago
    Self driving car technology has huge public benefit.
    - tracker1 7y ago
      If people aren't sharing the vehicle there's the same number of vehicles on the same number of roads, just less need for parking spaces. Of course, the potential for reduced accidents should help. Unfortunately, part of why AZ was probably chosen is that it's mostly sunny/clear weather during the year.
      - rowanG077 7y ago
        Reduced accidents. Increased time for people using self driving cars. Companies no longer needing to pay for staff that pilots cars. Just the time and cost savings alone for private people and companies is hard to be overstated.