Please create an account to participate in the Slashdot moderation system

 



Forgot your password?
typodupeerror
×
AI Transportation Technology

NVIDIA To Install Computers In Cars To Teach Them How To Drive 77

jfruh writes: NVIDIA has unveiled the Drive PX, a $10,000 computer that will be installed in cars and gather data about how to react to driving obstacles. "Driving is not about detecting, driving is a learned behavior," said Jen Hsun Huang, CEO of NVIDIA. The data collected by Drive PXes will be shared, allowing cars to learn the right and wrong reactions to different situations, essentially figuring out what to do from experience rather than a rigid set of pre-defined situations.
This discussion has been archived. No new comments can be posted.

NVIDIA To Install Computers In Cars To Teach Them How To Drive

Comments Filter:
  • by Crookdotter ( 1297179 ) on Wednesday March 18, 2015 @07:19AM (#49282177)
    Is that you?
    • If we teach computers how to drive by using US drivers....Skynet isn't going to be much of a threat...
      • by Anonymous Coward

        Hey, I'm not going to defend American drivers as good, but I will say as somebody who's been to Italy (Rome specifically) and India (Mumbai specifically), in the global perspective, Americans aren't that bad. They're still bad mind you, just not that bad.

        • Americans are oblivious (and I am one!), Italy/India you could make the argument it's intentional and bordering on malicious...that would be bad to teach Skynet...lol
        • I don't know... I've never been to either, but from what I've heard the traffic is total chaos, but the accident rates aren't correspondingly higher. That would suggest that the average driving skill is actually considerably higher.

  • by jeffb (2.718) ( 1189693 ) on Wednesday March 18, 2015 @07:20AM (#49282183)

    I recognize that analyzing lots of data across lots of cars, drivers, and routes might yield useful knowledge. I'll bet there are even insights that no single human driver could ever gain.

    But an awful lot of driving behavior comes from things that have nothing to do with anything this computer can monitor -- specifically, the driver's thought processes. If I slam on the brakes suddenly because I remember something I forgot at home, what will the computer make of that?/p?

    • by Anonymous Coward
      No, no, no. They'll learn from experience, just like we all do. The first time you make a hard right into that overpass support piling and your car crumples like a tin can, from that point forward you know not to ever do that again! And if in the vehicle in the next lane over has a lady's head bobbing up and down over the driver's lap, then you'd better give it a wide berth. These are all simple rules that can be learned.
    • Have you considered they might be studying human drivers to learn from their mistakes?
    • Re: (Score:2, Insightful)

      by Anonymous Coward

      How about: "I have to always be vigilant since at any time some idiot might slam on their brakes for no apparent reason, and sometimes that idiot might even be me!?

      • That is why we have filters ...

        * < 0.1% person doing something is just noise,
        * > 80% people doing it may be signal -- the context andconsequences needs to be considered.

    • by alen ( 225700 )

      that you are an idiot?

    • by ceoyoyo ( 59147 )

      It's called noise. Most people, most of the time will not slam on the brakes for no reason. Averaging over the behaviour of many drivers over long periods of time lets the algorithm discount your aberrant behaviour.

    • by Shoten ( 260439 ) on Wednesday March 18, 2015 @10:37AM (#49283625)

      If I slam on the brakes suddenly because I remember something I forgot at home, what will the computer make of that?

      That you live in Florida?

  • by Anonymous Coward

    ... in order to learn behavior you need to know where things are in space (aka detecting that they are there). What a dumb statement.

  • Only is this a success if the self driving car can appropriately extend a middle finger to other motorists who are giving them road rage.
  • by peragrin ( 659227 ) on Wednesday March 18, 2015 @07:23AM (#49282193)

    The computer isn't learning from experience it is being programmed by a different method. basically they are copying other drivers reactions to a set of obstacles so that the programmers don't have to create all those rules themselves.

    Think of it this way instead of manually programming a replacement robot arm on an assembly line they are copying the program code over directly to save time. This isn't a bad thing. However it is far from learning.

    • That sounds pretty close to the primary human learning method: monkey see monkey do
      • Only some people learn from monkey see monkey do. It is a valid method for some, but not everyone learns from that method. Monkey see monkey do results in users complaining about changing user interfaces. See windows 8 for examples.

        • I think you are being overly generous in your assessments of the frequency of monkey see monkey do learning. We do tend to be filled with hubris.
    • "Learning from experience" is a much better analogy for what they're talking about than your example of "copying the program code over."

    • by ceoyoyo ( 59147 )

      That's one of the most contrived explanations I've ever seen used to cover up "but my brain is magic!"

  • Comment removed (Score:4, Interesting)

    by account_deleted ( 4530225 ) on Wednesday March 18, 2015 @07:24AM (#49282195)
    Comment removed based on user account deletion
    • by dave420 ( 699308 )

      It would help if you didn't use the definition of animal behaviour from dictionary.com.

      Driving is a behaviour. If we look at the other definitions of behaviour at dictionary.com (on the same page you got your definition):

      • manner of behaving or acting.
      • Often, behaviors. a behavior pattern.
      • the action or reaction of any material under given circumstances

      we can see that you either made a mistake, don't know how a dictionary works, or decided to pervert the discussion by re-defining a word in order to tear apart

  • by rmdingler ( 1955220 ) on Wednesday March 18, 2015 @07:24AM (#49282197) Journal
    Even for the often flawed human drivers, this rings true. It seems one of the more common single vehicle highway accidents is the slight drift off the road followed by the panicked, aggressive over-correction... experience teaches us to gradually bring the vehicle back in line by fighting the gut-reaction to hurry.
    • by ledow ( 319597 ) on Wednesday March 18, 2015 @08:29AM (#49282599) Homepage

      I disagree.

      It's about being sensible. I have only ever been in a full skid situation once and I was able to apply WHAT I'D BEEN TOLD in the heat of the moment, without ever having done it before*.

      My father is car-mad and has run garages and serviced fleets of vehicles for decades. I just drive. I'm not interested in extreme situations or driving fast. But my father has for many years tried to get me on a skid-pan to "learn" how to control a skid. We never got around to it, and I'd never skidded.

      But when I *did* skid, I was able to refer back to what I was told and even those "things you have to do for yourself, because in the heat of the moment, son, you'll forget and do the instinctual things instead" - and applied them.

      I don't think driving is a learned behaviour at all, and I think it's EXACTLY the situations that are out-of-the-blue, unexpected, serious and panicky that you don't want to be interpreting the situation but "sticking to the rules you've been told, not what you 'feel' like doing".

      Driving while towing a caravan and the caravan starts to rock - you got there through stupidity ONLY, you're even more stupid if you're at that point and DON'T know what you're supposed to do if that happens. It's like getting into a strange car and not bothering to look for the brake before you start off (and that's in the same position for EVERY car!)

      If you follow the rule, you slow gently and stay straight. Sure, I bet a thousand drivers will tell you to spurt forward to bring the caravan back in line. But the rule won't hurt you, only inconvenience you.

      Skidding - you got there through stupidity PROBABLY (especially if such skid means you end in a collision because you're too close / fast), you should know how to handle it (it's in the Highway Code in the UK!).

      Emergency braking - instinct in all learner drivers is to stab the brake as hard as they can as fast as they can, which generates a skid that only ABS will save you from. Correct method is to apply brakes as normal, basically, but slightly quicker. Tell them that and it's what happens.

      Riding on your gut / learned reaction to a situation is a bad idea, especially if you've taken to playing games testing the limits of your driving/vehicle beforehand. You'll think you "know" when it's about to skid and how much you can turn before it will lose grip, etc. when in reality the surfaces are vastly different and determinant on every day and on every road.

      A computer has more than enough time to evaluate the problem, cause, and solution, and has no need to "guess" at the solution. It might not be able to avoid the collision - but then there's nothing it can do about that. Teaching it to work by an illogical application of arbitrary, self-formulated rules that can't be analysed or repeated reliably? That's just asking for trouble. Just program it to sit a few more feet back and follow the rules.

      I get told all the time that some things you can't "pre-teach", like clutch control - it's not true. It's just that your kids get bored with the theory when they first drive and just want to do it. If you tell them to expect loose pedal, slight contact, then dipping of bonnet as the gear engages, and slow, smooth actions from day one, then clutch control isn't hard at all. The problem is that we expect them to "jump in" and try it without knowing what to expect, and that's when you kangaroo and stall.

      But knowing what to expect is not about having done it several thousand times before, it can also just be about "this is what will happen, this is what you should do". It'll come swimming back to you when you need it.

      Emergency situations, you follow the rules. Getting clever "because you think you can go around him before he hits you" is exactly what causes the problems. Hell, from what I see of UK drivers, I bet a significant portion of accidents are people who DON'T want to get stuck behind the main accident and a split-second decision makes them pass him so they aren't stuck wai

      • It's about being sensible. I have only ever been in a full skid situation once and I was able to apply WHAT I'D BEEN TOLD in the heat of the moment, without ever having done it before*.

        The last time I was unexpectedly in a full-skid condition, I didn't have time to think. If I hadn't been in the habit of deliberately drifting that car, I'd have gone straight off the road. And I was going at a speed below the limit, and paying close attention to what I was doing. Must have hit an oily patch.

        On the other hand, when I've used ABS in the snow I had lots of time to think. Car in front of me (well in front, mind) tried to turn right and wound up turning left and sliding sideways instead. They m

      • > A computer has more than enough time to evaluate the problem, cause, and solution, and has no need to "guess" at the solution.

        Just to play Devil's Advocate. That is an underlying assumption there: That the computer has

        1) enough
        2) valid

        data to make to a solution that makes sense in that context?

        The problem with rules is that there is always exceptions. i.e. Sometimes accelerating will avoid the accident!

        Is the program smart enough to widen the search space and consider alternative solutions?

        The rest o

        • data to make to a solution that makes sense in that context?

          The problem with rules is that there is always exceptions. i.e. Sometimes accelerating will avoid the accident!

          Is the program smart enough to widen the search space and consider alternative solutions?

          The rest of your post is interesting.

          Assuming turbo-boost is inoperable there are only so many things we can do. Go faster, slower or same while going straight, left or right.

          For a computer doing some vector arithmetic brute force style across all possible reactions seems on its face to be quite trivial next to challenge of developing a valid model of the system/environment in the first place.

      • by GuB-42 ( 2483988 )

        On ABS-equipped cars, stabbing the brake *is* the correct action for emergency breaking. In normal conditions, no human can beat the efficiency of ABS breaking, even professional pilots. In unusual situations (gravel, snow, ...) a skilled driver can beat the ABS by basically doing a controlled skid, something that the system won't let you do anyways unless you turned on "suicide" mode.
        So much that many modern cars are able to detect emergency breaking situations and use 100% breaking power even if the pedal

        • This is one of the reasons I sometimes turn off ABS and other "traction" systems when driving on snow or ice. My sedan's sensors and algorithms clearly expect wet pavement, or at least some traction, or they don't let the tires turn, which means the car doesn't go if the snow/ice is too slick. Since I grew up driving in New England, I don't have a problem reverting to full manual drive, but I understand why some of the locals here in Virginia just abandon their cars when they stop moving, probably too afrai
  • by Anonymous Coward

    Accelerate and flip the obscenity bit.

  • He aint buying
  • by JeffOwl ( 2858633 ) on Wednesday March 18, 2015 @07:30AM (#49282247)

    What are they going to learn? How to not pay attention; how to not allow other vehicles to merge; how to force their way in when not allowed to merge; how to tailgate; how to brake check when others are tailgating; how to not use turn signals; what type of actions from other vehicles should cause them to rage; how to rage properly; how to ignore all the signs leading up to your exit and then cut across three lanes to take it at the last second; how to drive slow in the fast lane; how to pass when there isn't really room; how close they can get to a bicycle without actually hitting it; or hitting it, either way; ... etc..

    • by Wycliffe ( 116160 ) on Wednesday March 18, 2015 @07:45AM (#49282335) Homepage

      What are they going to learn?

      My guess is that they are going to learn what causes accidents and what doesn't cause accidents.
      We humans are really really bad at driving. You're basically stating all the bad things that we do.
      That is good input for a computer because it can see what causes accidents and remember not to
      do that and compare it to what doesn't cause accidents. I'm not sure I would trust a system like
      this to drive a car but it could easily be used to grade a computer (or a person) on their driving
      style.

    • Hopefully they plan on rolling this out in Finland or Germany where people actually know how to drive.
  • by koan ( 80826 )

    "Driving is not about detecting, driving is a learned behavior," said Jen Hsun Huang, CEO of NVIDIA

    It's about learning what to do with what you have detected, from position to collision detection is critical, then knowing what to do with that data is also critical.

  • Great, just great, where I live the rules of the road are treated as a "guide" line. A million ass hats experience of bad driving here will outweigh the few who actually driver responsibly, so I will end up with computers tailgating me and going through red lights.
  • Absolutely Necessary (Score:3, Interesting)

    by Atheraal ( 710104 ) on Wednesday March 18, 2015 @08:20AM (#49282551) Homepage
    This is an indispensable aspect of the road to automated driving, so to speak. Yes, learning human behaviour behind the wheel would be detrimental if it were applied directly to the AI's own driving algorithms, but consider that it's going to need to anticipate the actions of all the human drivers around it. Yes, I would like my self-driving car to have a very fine-tuned set of expectations for the idiocy it'll encounter.
  • Is this just a way to keep a robot-car manufacturer from specifically assigning weights to various bad outcomes and possibly avoiding lawsuits?
    Suppose a crash looks imminent. Whose life is more valuable? Instead of programming for this specifically, the manufacturer uses algorithms developed by obserations. Then the manufacturer could argue that it's not to blame when one person dies instead of another.

    In any case, this sounds like a great way to teach a computer how to drive badly.
    No one is a perfect dr

    • Suppose a crash looks imminent. Whose life is more valuable?

      The car will choose to follow the rules of the road as best it can in an imminent-crash situation. Your car won't be able to tell the difference between a bus full of old folks and a bus full of children any time soon. It's just going to stack into whatever is actually in its lane after bleeding off as much velocity as possible, it's not ever going to go onto the curb to hit the old woman to avoid smashing the day care minivan. Now can we stop asking this question, since it has a rather obvious answer? Pede

      • by Ichijo ( 607641 )
        Or the car wouldn't hit that pedestrian because it knows better than to drive faster than is reasonable and prudent [wikipedia.org] when its sight is obstructed by that row of parked cars. This is especially important in residential areas where children may be playing.
  • Ford has unveiled a car prototype designed to be parked inside a desktop computer. Its purpose is to teach cars how to compute.
  • When and where will these cars be tested on the roads, so I can know when to stay home? Sounds like a horrible plan.

    At least computers never eat or apply makeup while driving. But can they text? How many threads?

    I still don't trust the idea of giving 3000+ pound vehicles autonomy. Thankfully, they still need humans to pump gas.

  • Teach a bunch of chips to drive?

    They're Asian!

    PS. Do I get first stereotype at least?

  • Slow Down Cowboy!

    nVidia requires you to wait between each successful usage of the brakes to allow everyone a fair chance at avoiding accidents.

    It's been 4 minutes since you last successfully used your brakes.

  • If their experiment is successful, their computers should be able to give feedback fairly quickly as to what the driver should do before they actually do it. Advanced braking warning, correction if you get distracted/drowsy and start to drift out of your lane, speed corrections if the speed limit is changed or the road conditions changed etc etc. This idea is really cool and all, but they need to pack in a ton of sensors for any meaningful learning on the computer's part... once these sensors are there an
  • will the cars only control 3/4ths of the tires, with the last tire "slowly kicking in" when roads get slick?

Genius is ten percent inspiration and fifty percent capital gains.

Working...