Catch up on stories from the past week (and beyond) at the Slashdot story archive

 


Forgot your password?
Close
typodupeerror
×
AI Transportation

'How Close Are We to Self-Driving Cars, Really?' (slate.com) 370

Posted by EditorDavid from the great-expectations dept.
Chris Urmson helped pioneer self-driving car technology at Google before founding Aurora (which sells self-driving car software to automakers, and this week announced a new partnership with Chrysler and a new round of investment by Hyundai). In a new interview, Urmson "says he expects that in about five to 10 years, Americans will start seeing robots cruising down the road in a handful of cities and towns across the country," reports Slate.

"It will be about 30 to 50 years, he says, until they're everywhere. " I think within the next five years we'll see small-scale deployment. That'll be a few hundred or a few thousand vehicles. Really this is the, it's Silicon Valley speak, this is the zero-to-one moment of proving that the technology actually works, understanding how customers want to use it, convincing ourselves that -- and when I say ourselves, I mean as a society -- that these are sufficiently safe, that we trust them on the roadway, and that's that first phase... [W]hen the technology actually starts to become scaled, then we can ask the question what have we learned, what are the ways that we can make this a little bit safer, a little bit incrementally more efficient, and that's what I think local and state governments and federal government would invest in infrastructure...

The statistic I heard was 30 percent of traffic in San Francisco is people looking for parking. I heard a more alarming statistic that was 80 percent of traffic in Paris was people looking for parking. So imagine you have automated vehicles that take you to a location, you hop out, then it just drives down the block and picks up the next person and takes them where they're going. Suddenly, you've alleviated a massive chunk of the congestion in a city. Similarly, if you look at the floor plan of a city today, somewhere between 30â"40 percent of cities is dedicated to parking and roads. And so again, if you have automated vehicles operating as a transportation service, whether it's private or public transportation networks in the city, you don't need that real estate to be dedicated for parking. That real estate now can be recaptured, and it can be used for park space, it can be used for residential space, yeah, it can be used for mixed residential-commercial office space... Certainly for urban centers, I think it's much more likely that this technology is a shared platform that people get on and get off. It's an even more convenient version of a bus or of a taxi service.
This discussion has been archived. No new comments can be posted.

'How Close Are We to Self-Driving Cars, Really?' More

'How Close Are We to Self-Driving Cars, Really?'

Comments Filter:

  • Prediction timelines (Score:4, Interesting)

    by grasshoppa ( 657393 ) on Sunday June 16, 2019 @03:50AM (#58770392) Homepage

    Be wary of prediction timelines which extend beyond the life span of the predictor.

  • Perhaps in the future, as energy becomes scarce, driving will once again be seen as a rare privilege to be enjoyed, rather than a chore to be delegated.

    • Perhaps in the future, as energy becomes scarce, driving will once again be seen as a rare privilege to be enjoyed, rather than a chore to be delegated.

      Or it will be a crime like in the lyrics to Red Barchetta.

      I'm happy to live in a time when I can enjoy driving. Thankfully it will probably not be outlawed in my lifetime.

    • Re: (Score:3)

      by jwhyche ( 6192 )

      Perhaps in the future, as energy becomes scarce, driving will once again be seen as a rare privilege to be enjoyed, rather than a chore to be delegated.

      When this happens then it will be a privilege enjoyed only by the wealthy and elite. Those that rule your society. The most likely the time the rest of you will be in a car is when you are being transported to a local re-education camp or your execution. So, perhaps its best that it never becomes a rare privilege.

  • "It will take another 50 years before ... (Score:5, Insightful)

    by Qbertino ( 265505 ) <moiraNO@SPAMmodparlor.com> on Sunday June 16, 2019 @03:53AM (#58770408)

    ... humans can fly."

    Wilbur Wright, 2 years before his first flight.

    It takes 5 minutes in a festive robot car to convince anyone that the era of driving yourself is over.

    Point in case/anecdotal example: Two weeks ago I had a short conversation with a Spanish woman using only Googles instant translation feature on my cheap Moto G7. I don't speak Spanish and she needed help getting the next train to the airport.

    A lady following this was utterly amazed. "I want that app too!" she said. "You already have it on your Smartphone, it's Google."

    This would've been science fiction just a handful of years ago. In a year this will be normal. Same with robot cars. Once they can cruise the Autobahn safer than humans - something that is Just a handful of years away and mostly due to missing infrastructure today - the manually controlled car will quickly become extinct.
    It took new York a year to transition from coaches to cars entirely. As soon as people learn they can sleep write to their car is driving them, they will do it. And that's gonna happen sooner than this guy thinks.

    • It does seem like if he thinks it is that far out he is holding it wrong. I am sure that there are places where the rules of the road are too ambiguous to automate driving for along time (southeast Asia comes to mind), but other places the problem is “easy” to solve.

      I am curious when a self-driving car will know to interpret facial expressions and gesticulations, or to honk the horn at a pedestrian or manual car.

    • Re:"It will take another 50 years before ... (Score:4, Informative)

      by gravewax ( 4772409 ) on Sunday June 16, 2019 @04:29AM (#58770500)
      At the time flight was a completely unknown problem with no way to really tell how close or how far away they were. Self driving is a VERY well understood set of problems that various people are working on the answers too. however it is still incredibly unpredictable with many of the vision problems a long way from being perfected, we are still decades away before they are mainstream and 30-50 years before non manually controlled cars have any hope of being extinct. I think basically we are much closer than the pessimists believe but much much further away than optimists like you believe.

      • Wright quote is suspecious but interesting (Score:5, Interesting)

        by aberglas ( 991072 ) on Sunday June 16, 2019 @06:10AM (#58770682)

        Because the Wrights were very aware that Otto Linenthal was making controlled flights 20 years earlier. They built one of his gliders. Then reworked the design and added an engine. So reference please.

        What is really interesting about flight is that despite Linenthal 's success he got very little interest. And then once the Wrights built a practical aeroplane about 2 years from the first powered flight, nobody was interested. The Wrights could not get the army to see the value of being able to spot an enemy. One year (1906?) the Wrights gave up entirely, no flying was done.

        Then about 1908 they went to Paris, and the world went plane mad. By the beginning of WW1 planes were very competent. But still not airliners.

        For self driving cars, the big question is how autonomous. Self driving down the freeway is almost there. Self driving with supervision a few years away. And the uptake will be very quick.

        Completely autonomous, nobody in the car at all, that might take 10 years to mature.

        But as Bill Gates said, we tend to overestimate what can be achieved in one year, but underestimate what can be achieved in ten years.

      • Comment removed (Score:5, Informative)

        by account_deleted ( 4530225 ) on Sunday June 16, 2019 @10:50AM (#58771442)
        Comment removed based on user account deletion

    • Re:"It will take another 50 years before ... (Score:5, Insightful)

      by goose-incarnated ( 1145029 ) on Sunday June 16, 2019 @04:36AM (#58770508) Journal

      ... humans can fly."

      Wilbur Wright, 2 years before his first flight.

      You're cherry picking here. There were more people who said "humans won't fly soon" and were right than people who said "humans won't fly soon" and were wrong.

      It takes 5 minutes in a festive robot car to convince anyone that the era of driving yourself is over.

      Yeah, right! We've heard experts sing the "it's five years out" song since 2012. The problem with all of the predictions is that they are using advancements in hardware as an indication of how close we are to self-driving.

      Unfortunately hardware is not the limiting factor, so it doesn't matter how good the sensors are if we don't have software anywhere close to the level of autonomy that humans have.

      Over twenty years ago the state of the art in self-driving cars was "human intervention needed 9% of the time." Now, using around 10000x more resources, it is "human intervention needed 8% of the time". That isn't progress.

    • Sorry I'm confused what app you used to translate voice to voice? I know google can recognise my voice, pretty darn well, but I don't know about automating the translation into another language and then saying it out loud?

      I'm sure it could be done? Via the web interface on translate.google.com - but for the correct flow of a conversation, how does this all work?

      • I know some very basic Russian, enough too "sometimes" catch enough words to understand the gist of the conversation, not enough to join in though. Yet even I cringe at some of the translations that google comes up with and I find myself correcting it for very basic stuff, would hate to think of the pigs breakfast it would actually make of a full conversation.

        • I found the URL he's talking about and I was surprised, it is relatively easy to use on the web and probably engage in a very basic conversation with gesturing, fairly quickly - it's overall impressive (all things considered)

          I suspect it would be enough for an emergency like "help, lost keys hotel room"

          Etc.

      • Sorry I'm confused what app you used to translate voice to voice? I know google can recognise my voice, pretty darn well, but I don't know about automating the translation into another language and then saying it out loud?

        Not sure if the GP's solution is automated but in Google translate all you have to do is press a single button and it will speak the translation out loud.

    • ... humans can fly."

      Wilbur Wright, 2 years before his first flight.

      "We'll have fully self-driving cars next year."
      -- Elon Musk, 10+ years ago.

    • Re: (Score:2, Interesting)

      by srichard25 ( 221590 )

      After 5 minutes in an autodriving Tesla, I'm convinced that the era of driving myself is nowhere near over. The car made me extremely nervous. It keep moving from one side of the lane to the other, making me feel like it was about the change lanes. The drivers around me must have been concerned as well because it looked like the car was about to slowly move into their lane.

    • ... humans can fly."

      Wilbur Wright, 2 years before his first flight.

      It takes 5 minutes in a festive robot car to convince anyone that the era of driving yourself is over.

      Then why when we have buses, do we still have cars?

      A lady following this was utterly amazed. "I want that app too!" she said. "You already have it on your Smartphone, it's Google."

      This would've been science fiction just a handful of years ago. In a year this will be normal.

      Or you could just learn Spanish, French, whatever. I've used those translators. They are okay in a pinch, But a nuisance if trying to have a real coversation. If you are using say, English to Russian and back, it gets pretty strange.

      Anyhow if the SD cars are like Google translate, I'm not interested other than a similar emergency use.

      Perhaps some people's lifestyle is very simple. I have various cases with equipment that I keep in my vehicle. I have t

    • Once they can cruise the Autobahn safer than humans - something that is Just a handful of years away and mostly due to missing infrastructure today - the manually controlled car will quickly become extinct.

      You were doing well until you got to here.

      First, AVs which depend on smart roads are a total non-starter. There's nothing wrong with adding sensors and merging the intelligence from those sensors, but you cannot rely on them. They could fail, they could be malicious... you can only use them as advisory input.

      Second, the Autobahn is not the standard. Freeway driving is easy compared to local driving. Correctly, your statement would have begun "Once the AVs can handle puttering around local streets with dogs

  • Imagine... (Score:5, Insightful)

    by enriquevagu ( 1026480 ) on Sunday June 16, 2019 @04:11AM (#58770458)

    So imagine you have automated vehicles that take you to a location, you hop out, then it just drives down the block and picks up the next person and takes them where they're going

    If we ever invent this, we might call it "taxi". It would be a huge improvement for traffic congestion. Can you imagine...

    • If we ever invent this, we might call it "taxi". It would be a huge improvement for traffic congestion. Can you imagine...

      Statistically nobody would own a car if it were cheaper to get picked up in a Taxi, and due to economies of scale (fleet ownership vs. private vehicle ownership) it would be, except for the need to pay drivers. We could stagger work start times and let passengers split fares when carpooling to permit commuters to get to work, although if we're designing transport systems for commuters, trains make more sense.

      By the same token, buses only exist because we need drivers to operate automobiles. If you have self

  • The insurance companies will demand it (Score:4, Insightful)

    by petes_PoV ( 912422 ) on Sunday June 16, 2019 @04:24AM (#58770486)
    As soon as there is clear, actuarial, evidence that self-driving cars are safer than manually driven ones, the cost of insurance (and the restrictions placed on it) for people to drive will start to rise. Ultimately what will push every car to being an AV will be the unaffordability of insurance. Especially as the AVs will have cameras everywhere and be able to provide irrefutable proof that the cause of a collision was a human driver.

    After that will come laws mandating that humans cannot drive in densely populated areas, or at speed, or under ever-more restrictive conditions. Until the opportunity and cost simply pushes everyone to being a passenger. Although since the safety angle of AVs will be well-known and widespread, there would be little desire from ordinary people (leaving aside the petrol-heads and those irrationally opposed to using seat-belts) to ever travel in a manually operated vehicle.

    It will be commercial pressures, not technological ones, that will drive the change. And once it starts, it will move very quickly.

  • I heard a more alarming statistic that was 80 perc (Score:4, Informative)

    by mapkinase ( 958129 ) on Sunday June 16, 2019 @04:30AM (#58770502) Homepage Journal

    I heard a more alarming statistic that was 80 percent of traffic in Paris was people looking for parking.

    It actually 67.8%. The same amount of statistics is made up, by pure coincidence.

  • then we'll be almost there.

    This is a promise.

    • Once we understand how a bird wing works, then we'll be almost ready to make an airplane. This is a promise.

      • Yes, that is exactly what happened.

        • Re:Once we understand how the brain actually works (Score:4, Informative)

          by religionofpeas ( 4511805 ) on Sunday June 16, 2019 @05:35AM (#58770614)

          Except that it isn't. People have used bird wings as early inspiration, but the further development into finding the best shapes was done without looking back at birds. Airplanes don't use their wings to generate thrust, and they have completely different construction. Despite not looking like a bird, a fully loaded 747 can carry a lot more coconuts than a swallow.

          Similarly, early neural nets were inspired by brain cells, but most of the development is just done directly on artificial neural nets, without trying to stay close to actual brains. There already is a long list of things where machine learning outperforms human brains for the same task. A simple example is the AlphaGo program, which taught itself to play the Go board game, starting with nothing but the rules, and then beat the best human players. All of this was done without understanding the brain of a Go player. In fact, we don't even really understand how the machine works.

          • Re: (Score:2, Insightful)

            by Mr. Dollar Ton ( 5495648 )

            Except that it isn't.

            Of course it is, ignoramus. People built 747s because they understood the physics of wings in general, and then proceeded to make the wings they were interested in, applying their understanding of the general principles to a specific case.

            The same will apply to real "intelligence" as opposed to the statistical games that are called "AI" today.

            And yes, it is quite obvious that you don't understand how and why a lot of the shit around you works.

            • No you dolt. A plane wing and a bird wing aren't remotely similar. The aerodynamics of feathers and flapping is completely different from a rigid, fixed wing. That's why most birds can do VTOL and hover for short periods of time while pretty much zero fixed wing aircraft can.

              We engineered fixed wings with pretty much no crossover from bird wings other than the realization that a light structure with a large wing area is a requirement for flight. Bird wings ARE THE PROPULSION, while fixed wings can either only glide, or need an additional power source.

              And yes, it is quite obvious that you don't understand how and why a lot of the shit around you works.

              The irony is hilarious.

  • Reinventing the wheel? (Score:3)

    by 15Bit ( 940730 ) on Sunday June 16, 2019 @04:45AM (#58770520)
    "So imagine you have automated vehicles that take you to a location, you hop out, then it just drives down the block and picks up the next person and takes them where they’re going. Suddenly, you’ve alleviated a massive chunk of the congestion in a city"

    That sounds a lot like buses and taxis to me. Which we already have. If people aren't using those now, then why would they use self driving versions of them in the future?

  • Nodoby knows, especially not this guy (Score:3)

    by locater16 ( 2326718 ) on Sunday June 16, 2019 @04:52AM (#58770530)
    The halting problem: The end point of any computer program cannot be determined faster than running the program and waiting for it to end.

    Which is to say, we'll only know when self driving cars will be around, when we produce a program that can drive a car. We have everything else, billions of dollars just slammed into the industry year after year has produced better than adequate sensors and chips to run whatever this program looks like. We just don't have the program itself, and by the very nature of this we can't predict exactly what it is until we get it. And thus it's fantastically hard to say when we'll get it either. Maybe it'll be five years from now, but that will just be a coincidence. Maybe two months from now Elon Musk will do some dance onto a stage, shitpost on twitter, and demo a Tesla that can safely drive itself.

  • It will take only a little bit longer than fusion energy, which should be fairly soon after the year of the Linux desktop.

    As far as I can tell there's been relatively little progress in recent time in understanding natural intelligence, and all the advances in artificial intelligence seem to be 20-year-old algorithms working on larger data sets, more powerful hardward, and better distributed computing platforms, not any advance in the actual science.

    What car makers call self-driving will undoubtedly become

    • What car makers call self-driving will undoubtedly become better, undoubtedly with significant practical benefits

      Significant practical benefits, that get better every year. What more do we need ? If that can be done with old science, then this means the science is good enough.

  • and one that talks to me like KIT

  • The thing that I have noticed with pretty much all of these autonomous driving tests is that they take place near where the developers live. For example, Google has been testing their cars for years near their headquarters in Silicon Valley. They do this because it is convenient for them and allows them to test near where they prefer to live. The problem with this is that there hundreds of millions of people who live somewhere where they get a winter with snow and ice.

    The snow changes your own vehicles hand

  • We should realize that this is all push and no pull. The public is not clamoring for self driving cars. What will happen is that industry and government will attempt to cram such an environment down our throats while sweeping away with grand visions the very real problems that will arise. And when they do arise, industry and government will simply deny them or propose simple minded solutions that are not solutions at all. We saw the same thing happen with the Space Shuttle - the same thinking is developing.
  • If you define it to mean that a driver will not be needed for routine and non-routine driving on any legal road in the Untied States, then we are far, far away from self-driving cars. If you define it with a lesser requirement, such that the requirement can be satisfied by the current crop of self-driving cars, then we are there already. Reality is somewhere in between, just where is not known. Those who want to sell you a self-driving car think closer to "we are there already." Those who have concerns

  • While I can admire the "Star Trek" like utopia that the guy selling the vision promotes, I'm fearful of what it will really be like.

    Rolling around town, people in these things will be largely un-monitoried. I imagine I'd call up a taxi-bot to pick us up in front of a mall or event center and when we get it, it's a mess of trash, a half eaten hamburger left behind, sticky kid handprints on every window, sticky sex leftover fluids on the seats and the smell of it all baking in the summer sun.

    I think the visi

  • The statistic I heard was 30 percent of traffic in San Francisco is people looking for parking. I heard a more alarming statistic that was 80 percent of traffic in Paris was people looking for parking.

    Sounds like what is really needed are just a bunch of automated parking garages.

  • Well, we have been about 10 years away from general purpose AI for about 60 years now, so I figure we will get self driving cars sometime after that.

  • Watch the Tesla Autonomy Day video. The whole thing - it's a treat for nerds.

    Tesla's claim, and that's backed by both Elon and his hyper-nerd AI guy from Stanford - is that they've reached the point of exponential gains on self-driving now.

    You may dispute this claim - hopefully with data - but that's the claim on the table. If the claim is true, we're looking at super-human FSD within two years, not thirty.

    Remember your hockey sticks.

  • And The Next Step? (Score:4, Insightful)

    by ytene ( 4376651 ) on Sunday June 16, 2019 @01:33PM (#58772000)
    As we seem to be having some difficulty in predicting exactly when autonomous vehicles will become the default mode of transport, how about our attempting to predict the next few of steps?

    First, authorities and governments around the world moving to outlaw "manually operated" transportation, on the grounds that such vehicles are "dangerous".

    Second, authorities and governments around the world mandating that "autonomous vehicles" must, by law, be equipped with a "state intervention mechanism" which will allow law enforcement or other government service to forcibly disable a vehicle or cause it to pull over to the side of the road.

    Third, the criminal element figure out how to compromise those remote control and security systems, which lead to a series of vehicle-based kidnappings...

    And if you wondering if all of this sounds a bit tin-foil-hat and far-fetched... well tell me that the authorities really don't want to have this sort of control... And maybe go read a copy of Heinlein's "Stranger in a Strange Land" to see what happens to Ben Caxton when he challenges authority from the back seat of an automated mini-cab...

  • Driving is a hard AI problem (Score:3)

    by HuguesT ( 84078 ) on Sunday June 16, 2019 @05:16PM (#58772936)

    Autonomous driving is a hard AI problem with a lot of constraints and the capacity to actually kill people, both in the car and outside, so it must be taken very seriously. The mere idea of privately-developed, close-source software driving tons of hardware at highway speed without supervision is enough to evoke nightmares.

    AI has made a lot of progress lately. Somewhat hardish AI problems that we can solve today are in the order of facial recognition, language translation, categorising the content of images, helping with the diagnosis of some illnesses like skin cancer, etc. These are incredibly useful, but notice that they still have a non-zero error rate. The current state of the art on ImageNet for picture recognition is in the ordre of 3%, and they are relatively easy to fool. If you have ever used google translate, you must know its limitations.

    Autonomous cars need to have a precise image of their surrounding up to a few hundred meters in all conditions (day, night, wet, etc), very frequently updated, and they need to be able to anticipate other drivers as well as other hazards on the road. The technology is simply not there yet, even with the 70k$ sensors that Waymo uses.

    We all want this technology to be here and to be better than humans at driving. This sounds easy since most of us drivers think everyone else drives like an idiot, however, the fatality rate in cars is in the order of 1 death per 100 million miles travelled. This is actually very low. This presents a challenge to the autonomous driving research community, because to certify that their system is actually better than humans, they will have to travel significantly more than hundreds of millions of miles, in real, not simulated or recorded conditions. The cost of doing this is astronomical, and it must be done everytime a new version comes out.

    Trusting this software is going to be very hard in practice.

    Source: RAND corp [rand.org].

  • We've heard that before (Score:3)

    by HangingChad ( 677530 ) on Sunday June 16, 2019 @07:36PM (#58773326) Homepage

    So imagine you have automated vehicles that take you to a location, you hop out, then it just drives down the block and picks up the next person and takes them where they're going.

    Uber made the exact same argument. That it would eliminate cars on the road but none of the traffic promises held true. In fact, Uber increased traffic in congested areas.

    It's goofy to think self-driving cars will do what Uber could not.

Slashdot Top Deals

I have hardly ever known a mathematician who was capable of reasoning. -- Plato

Close