Follow Slashdot blog updates by subscribing to our blog RSS feed

 



Forgot your password?
typodupeerror
Businesses Transportation Technology

Uber Nowhere Close to Having a Fully Autonomous Vehicle, Its Self-Driving Cars Need a Lot of Human Help (recode.net) 87

Uber may see self-driving cars as "existential" to its future, but the company is nowhere close to having a fully autonomous vehicle. According to internal documents obtained by Recode, during the week ending March 8, Uber's self-driving cars traveled, on average, just 0.8 miles on their own before a human had to take over, in a process known as "disengagement." From the report: As a whole, Uber's self-driving system is putting on many more miles than it did in January. Last week, the company's 43 active cars drove 20,354 miles autonomously, according to the documents. This is only the second time since late December 2016 that its cars have driven more than 20,000 miles in a week. In January, the cars only drove 5,000 miles. At that point, however, the company only had about 20 active vehicles, mainly in Pittsburgh. By February, the company's cars were driving themselves around 18,000 miles a week. Uber passengers took around 930 rides in these autonomous cars in Pittsburgh last week and around 150 rides in Phoenix. To be clear, these vehicles still had a driver at the wheel to take over if needed. In Pittsburgh, where Uber launched its commercial self-driving pilot in September, the company has been performing around 800 or more UberX trips per week in semi-autonomous mode since the middle of February.
This discussion has been archived. No new comments can be posted.

Uber Nowhere Close to Having a Fully Autonomous Vehicle, Its Self-Driving Cars Need a Lot of Human Help

Comments Filter:
  • by elrous0 ( 869638 ) on Friday March 17, 2017 @05:26PM (#54062027)

    We're still a lot farther away from truly autonomous cars than most people tend to think. Sure, driver assist and suped-up cruise control is coming in, and will be great on clearly marked and standardized interstates. But good luck trying to get a computer to navigate the old backroads of some city or country backwater.

    Hell, Alexa still can't even understand a lot of basic questions I ask her. I'm sure as shit not about to let that bitch drive.

    • by phantomfive ( 622387 ) on Friday March 17, 2017 @05:36PM (#54062105) Journal
      It would be nice if we had similar numbers from Google. How long do their cars go on average before a human needs to take over?
      That number is a lot more useful than "total miles driven" or "accident per /1000 miles"
      • by s.petry ( 762400 )

        When they do 15-20 miles per hour on rural roads with a speed limit of 35, I don't call it a fair test. I despise these traffic blockers with a passion.

        • You appear to be mistaking farm equipment for automated cars.
          • by s.petry ( 762400 )

            No, I happen to live in Mountain View so am near the big fleet. From the old hand rigged rides to the little toy cars they drive all around Mountain View.

            I don't mind them testing and understand the need, but rush hour on El Camino, Central, Shoreline, and Middlefield is already bad. Then you have to try and get around these crawling vehicles if you can. They should drive out of rush hours, or go to an area with less traffic for traffic testing if they can't get within 15-20 miles of the speed limit.

            • No, I happen to live in Mountain View so am near the big fleet.

              Where do you find rural roads in Mountain View?

          • by mjr167 ( 2477430 )
            Farm equipment doesn't go that fast.
      • by Arkh89 ( 2870391 )

        Is it really? Are all of these events caused by improper decision of the AI or is it more slight over-corrections imposed by the humans?
        Also, I guess Uber's vehicles operates mostly in the dense, chaotic traffic of the inner-cities rather than say speedway. 0.8 miles between take-over on speedway would be much more alarming.

        • 0.8 miles between take-over on speedway would be much more alarming.

          Yes it would. Do you know what Google's actual numbers are for take-over on the speedway? (Or freeway here in California)? These kinds of numbers are kind of important for evaluating the quality of self-driving cars.

          • Do you know what Google's actual numbers are for take-over on the speedway?

            I don't know about Google, but my Tesla can drive on US101 from south San Jose to Palo Alto without me touching the steering wheel or brake even once.

            Instead of developing their own tech, Uber should team up with someone further along, such as Tesla, or Google, or ???.

            • It's not about the tech for Uber. It's about looking like you're going something trendy and special so you can get that next round of funding.

        • There's surely some heavy duty urban driving. And in city driving the human drivers will probably intervene often when they don't really have to. OTOH, Pittsburgh's nightmarish topography will likely result in a fair amount of expressway-like driving. That's pretty easy and should result in few to no interventions for miles. It's not like driving in Manhattan, San Francisco, or (the mind boggles) Boston.

      • Re: (Score:2, Troll)

        by Guillermito ( 187510 )
        We do have similar numbers from Google and other companies. Google numbers for 2016 are 0.20 disengagements per 1000 miles driven. https://www.dmv.ca.gov/portal/... [ca.gov]
      • by Hadlock ( 143607 )

        I was crossing the street in mountain view on... monday I want to say... The driver had to take over manual control doing a right turn on to central expressway from castro st/moffett blvd. The car decided that it was time to pull on to the highway. There was a car in front of it waiting to turn right as well. It's kind of a weird intersection though, as there's a double set of train tracks right next to central expressway, which you cross, then come to a stop again, then turn right at a light. Add to the fa

    • by Anonymous Coward on Friday March 17, 2017 @06:07PM (#54062323)

      I don't see why this is so difficult to understand. Only real AI can hope to cope with the infinite number of variables that arise while driving. What does a rudimentary AI do when a wind storm has blown down a tree branch in the middle of the road? Does it know that it's OK to go around the obstacle even if it means briefly driving into an oncoming lane? What does it do when a piece of newspaper blows around in traffic on the freeway? Does it swerve to avoid collision with a harmless object? What does it do when it approaches a vehicle that has stalled? Does it wait for ever? Does it go around when the oncoming traffic has cleared? Does it do a 3-point turn and pick a new route? What does the rudimentary AI do when it's Friday night and the passengers are requesting to be dropped off at a night club in a part of the city that has scores of drunk pedestrians that are jay walking all over the place and are not sticking to the sidewalks. (eg 10th and Pike in Seattle). Maybe it drops the passengers off 2 blocks away instead.

      Fully autonomous vehicles right now are nothing more than a hair-brained-scheme. Once (if ever) real AI is developed, teaching it to drive can be done over night. So to speak. Until then, I would not be investing a single dollar into companies like Uber or robotics companies. All of this talk about robots taking over wreaks of investor propaganda.

      • What does it do when a piece of newspaper blows around in traffic on the freeway?

        A piece of what?

      • Only real AI can hope to cope with the infinite number of variables that arise while driving.

        It doesn't have to handle everything perfectly. It just has to do better than a human.

        What does a rudimentary AI do when a wind storm has blown down a tree branch in the middle of the road? Does it know that it's OK to go around the obstacle even if it means briefly driving into an oncoming lane?

        I know for a fact that Teslas will NOT go around the obstacle. They will just stop. But they are not trying to be "level 5".

        What does it do when a piece of newspaper blows around in traffic on the freeway?

        I know several people that work/worked on Google's SDC. They have object recognition software that is trained to look for blowing/floating objects, and their system has been trained on many thousands of examples. They handle the "blowing plastic bag" problem better than humans, with earlier recog

        • Think of the total amount of scrapes, dings, fender benders and worse that people get into today. When we switch to automated driving, those little accidents are no longer the driver's fault so I sure hope manufacturers plan to pay with almost all accidents on the road. Furthermore, if they are not significantly less safe, insurance coverage isn't going down in price either so I hope manufacturers plan to cover all that as well.

          If the point of this is to save lives then, "only a little safer than a human
      • by mjwx ( 966435 )
        A common scenario here in England is crossing over to drive on the American (or incorrect) side of the road due to parked cars. A lot of places in England were built before roads were a thing, so there are a lot of houses with no driveways on narrow streets. Obviously people park on the street. So to go through you have to go onto the wrong side of the road. There is a bit of an issue when you have two cars at opposite ends both wanting to get through. After a while you learn who has to give way to who and
    • by lorinc ( 2470890 )

      We're still a lot farther away from truly autonomous cars than most people tend to think. Sure, driver assist and suped-up cruise control is coming in, and will be great on clearly marked and standardized interstates. But good luck trying to get a computer to navigate the old backroads of some city or country backwater.

      Hell, Alexa still can't even understand a lot of basic questions I ask her. I'm sure as shit not about to let that bitch drive.

      Inferring the performances of a vision based vehicle control software from the performances of a natural language processing software is about as relevant as saying that all hammers are flawed because your screwdriver is not functioning properly.

    • Hell, Alexa still can't even understand a lot of basic questions I ask her.

      That is a silly comparison. Natural language processing is a MUCH harder problem than navigation and lane control. Computers will be able to drive long before they can have an unstructured conversation with a human.

      • Navigation and lane control is easy. The hard part is understanding how to drive around unplanned things and knowing how to respond to them. Is that thing flying through the air a balloon or a rock? Is that thing on the side of the road going to run out into it unexpectedly? What is that construction equipment that isn't moving with the flow of traffic going to do? Should I pass in the oncoming lane the vehicle driving 20mph?

        That is harder than natural language detection.
  • ...does that include or exclude the 0.8 mile stretches to the next human interaction or not?

    Seems that you shouldn't start tallying self-driving miles until you exceed a mile.

  • by foxalopex ( 522681 ) on Friday March 17, 2017 @05:58PM (#54062261)

    If Google's lawsuit on theft of trade secrets and intellectual property and patent violations goes though, I suspect this will kill Uber's self-driving program. Last I recall Google was actually the farthest ahead on this "self-driving" technology and from the sounds of it the safest to rely on.

  • by thegarbz ( 1787294 ) on Friday March 17, 2017 @05:58PM (#54062267)

    Uber wasn't able to solve self-driving car problem in a few short months. What does this mean for the future? Is innovation dead? Read on for all the redundant words on the topic that no one is interested in below.

    • Do you think Google (or anyone) is close to getting level5 autonomous cars? Level 5 means you can basically sleep while driving.
      • Do you think Google (or anyone) is close to getting level5 autonomous cars? Level 5 means you can basically sleep while driving.

        I've ridden in a Google car. Yes, Google is very close, if not already there. In the hour I was in the car, the human driver never had to take over, and the vehicle dealt with busy city streets, including pedestrians and cyclists, many different sorts of intersections, road construction... other than driving like my grandmother (*extremely* cautious), it was flawless.

        • Driving would not be workable if every car drove like your grandmother.
          • Driving would not be workable if every car drove like your grandmother.

            Sure it would. In fact it would be a lot safer, if a bit slower. But it might confuse the human drivers the road, which is why they have since updated the algorithms to drive more "normally". My ride was about three years ago, and as I understand it that's mostly what they've been doing since then, tweaking the behavior to make the vehicles to drive like humans, including being more aggressive when the situation demands it.

            • If that is a better solution then they should just half the speed limit right now everywhere and be done with it.
              • If that is a better solution then they should just half the speed limit right now everywhere and be done with it.

                I've read that sentence three times now and I can't figure out what you're trying to say. The referent for "solution" seems to be my comment about making cars drive more like humans, but I see no connection with the speed of light.

                • You were saying driving slower would be safer, but the point is that people don't want to drive slower or they already would be. This doesn't change with AI. I don't want my ride to work to take longer in an automated car than it does manually. If that is the case then ai isn't really effective.
                  • You were saying driving slower would be safer

                    1. I never said that driving slower is safer. Nor did I ever imply it in any way.

                    2. I never said that self-driving cars must drive slowly. I said that Google cars drive very cautiously, which is a bit slower, but the whole point of the post which your hard-to-follow response followed is that there's no reason to expect that to be an inherent "problem" with self-driving cars in general, or Google's cars in particular.

                    3. You're dead wrong when you say that people wouldn't be willing to exchange a slower trip

                    • 1. "Sure it would. In fact it would be a lot safer, if a bit slower. "
                      3. It's going to be an hour at the most, hardly enough time to do anything productive. I can see if you are on a long road trip but not to work. Besides, everyone on the road is absolutely free to trade speed for safety now. In most places there are no minimum speed limits. They don't drive slower because they don't want to.
                      4. It's hard to comment on that since we're no where close to that yet. Yes they can see everything around the
      • Do you think Google (or anyone) is close to getting level5 autonomous cars? Level 5 means you can basically sleep while driving.

        Nope, and I will repeat the same snyde comment but replace Uber with Google and a few months with a few years. It is one of the most complicated problems we have, the idea that we solved this in a couple of years is laughable.

        That said Google's record is astonishing. Based on what I've read I would probably feel comfortable sleeping in one.

  • I think Lyft, Uber spending money on research and development of autonomous vehicles is awesome, and we should encourage them (and Google) to do even more. That said, I got the opinion that Uber's business leaders thought self-driving cars where not just far-off rnd, but rather a vital investment to get a return on within a few years so they could start replacing human drivers ASAP.

    This sends bad signals across the marketplace. It implies Uber's current business model is unsustainable (which may be true
    • This sends bad signals across the marketplace. It implies Uber's current business model is unsustainable (which may be true according to recent slashdot posts) It also implies that Uber doesn't care about their current drivers, or labor as a whole, it also reeks of desperation. Market signals are important, if they don't course-correct soon they could find themselves with angry investors, boycotting consumers, and no additional venture capital.

      That depends on what Uber's business model is. As far as I can tell they are pumping as much money as they can into undermining taxi operators and achieving regulatory capture in as many cities as possible. Once they have achieved that and crushed existing operators prices will skyrocket for a ride as they will have achieved a monopoly. Then existing drivers will be squeezed until automated vehicles are available, after all where else can they go?

      Let's not pretend Uber is a tech company either. Their app a

      • We are going from hundreds of taxi companies to a few taxi companies. Of course the price is going to go up. People are complaining about the taxi companies cornering the market but it's actually a lot better situation right now than we will end up with.
        • by MrKaos ( 858439 )

          We are going from hundreds of taxi companies to a few taxi companies. Of course the price is going to go up.

          I haven't done work in transport for some time so I haven't really looked lately. When I interacted with them by the time you got something past all of the committees they would change the chairs and you had to go through it all over again. It was pretty stultifying getting them to evolve their business model.

          People are complaining about the taxi companies cornering the market but it's actually a lot better situation right now than we will end up with.

          Exactly. They're duping people and they just lap it up.

          Uber looks like a Trojan Unicorn, IMHO.

      • This is a VERY interesting point of view that I haven't heard/thought of before. But do you have any evidence/articles that indicate regulatory capture? I certainly know of UBER fighting regulators that where previously captured by local taxi companies, and the regulations that gave those local taxi companies unfair market advantage. Other than getting regulators to create a new type of service (TNC or 'transportation networking companies') that regulates Uber/Lyft differently than local taxi's, i haven't s
  • by Guillermito ( 187510 ) on Friday March 17, 2017 @06:45PM (#54062561) Homepage
    This is the official Autonomous Vehicle Disengagement Report for California https://www.dmv.ca.gov/portal/... [ca.gov] Google is at 5000 miles before a disengagement is required, compared to the 0.8 miles reported for Uber. Google also logged over half a million miles, compared to a couple of thousand of some of the other companies. So at least for the companies doing autonomous vehicle testing in public roads in California, no one come even close to Google.
    • by ttsai ( 135075 )

      This is the official Autonomous Vehicle Disengagement Report for California

      https://www.dmv.ca.gov/portal/... [ca.gov]

      Google is at 5000 miles before a disengagement is required, compared to the 0.8 miles reported for Uber. Google also logged over half a million miles, compared to a couple of thousand of some of the other companies. So at least for the companies doing autonomous vehicle testing in public roads in California, no one come even close to Google.

      But are we comparing similar driving environments and challenges? I only see Google cars on El Camino, driving slowly and mostly in the middle lane, i.e., among the least challenging of all driving environments. It would be perhaps more comparable to see how well Google cars perform with arbitrary start and end locations and keeping up with the flow of traffic.

    • by nnull ( 1148259 )
      Uber thinks that throwing more money on the project will result in a fully autonomous vehicle. The fact is, Google had people interested in doing it and had an R&D program to do it, with the resources to do it. They had renown researchers they could take off from other projects to focus on this. They even made videos of their projects to try to encourage more smart people to join them. This is completely the inverse what Uber has.
  • I had no idea that "Uber Nowhere" existed, much less that they were so close to solving the problem of autonomous driving. It makes sense that its self-driving cars need a lot of human help, too, because humans know a lot about going nowhere.

"All the people are so happy now, their heads are caving in. I'm glad they are a snowman with protective rubber skin" -- They Might Be Giants

Working...