Slashdot Mirror

← Back to Stories (view on slashdot.org)

Robot Hand Learns How To Learn From Babies

Posted by Zonk on from the everything-i-need-to-know-i-learned-in-the-comments dept.

jcasman writes "Wired's got a piece on building a better robotic hand at Stanford. The new robot is called Stair 1.0, and scientists are hoping to take a cue from human children for how to teach a robot to learn. 'When a computer fails at a task, it spouts an error message. Babies, on the other hand, just try again a different way, exploring the world by grabbing new objects -- shoving them into their mouths if possible -- to acquire additional data. This built-in drive to explore teaches us how to use our brains and bodies. Now a number of hand-focused roboticists are building machines with the same childlike motivation to explore, fail, and learn through their hands.'"

61 of 76 comments (clear)

  1. Sounds like the right direction by timmarhy · · Score: 1

    patients with existing limbs need to learn how to use them all over again. I'm really hopful we will see a major leap in artifical limbs in the next 50 years

    --
    If you mod me down, I will become more powerful than you can imagine....
    1. Re:Sounds like the right direction by snarkh · · Score: 1

      patients with existing limbs need to learn how to use them all over again.

      I sure hope patients with existing limbs will not have to learn to use them all over again.

      I'm really hopful we will see a major leap in artifical limbs in the next 50 years

      Or at least a hop.

    2. Re:Sounds like the right direction by somasynth · · Score: 1

      There are no major leaps, just small steps that eventually surpass the biggest of leaps.

    3. Re:Sounds like the right direction by timmarhy · · Score: 1

      Thats how it currently works. they must attach the sensors to other nerves and they have to train themselfs how to use the limb.

      --
      If you mod me down, I will become more powerful than you can imagine....
    4. Re:Sounds like the right direction by billcopc · · Score: 1

      That has less to do with the limb and more to do with the nervous system.

      If we could plug the artificial limbs into the right nerves (and have it all calibrated), theoretically the new limb would be a drop-in replacement and Kenny could run like a cyborg the next day.

      --
      -Billco, Fnarg.com
  2. Robot grabbing and shoving things in its mouth by JetlagMk2 · · Score: 5, Funny

    Am I the only one worried here?

    1. Re:Robot grabbing and shoving things in its mouth by ParaShoot · · Score: 3, Funny

      "Hello human, what are you?"

      *nom nom nom*

    2. Re:Robot grabbing and shoving things in its mouth by Necreia · · Score: 1

      Depends on what function you purchased the robot for.

    3. Re:Robot grabbing and shoving things in its mouth by Fx.Dr · · Score: 1

      Unless Real Doll plans on applying this to their products in the future, something tells me we'll be OK;

    4. Re:Robot grabbing and shoving things in its mouth by evil+agent · · Score: 1

      Ok, ok, keep your pants on...seriously.

      --
      End transmission.
    5. Re:Robot grabbing and shoving things in its mouth by pionzypher · · Score: 1

      And whether said model includes teeth.

      --
      I'll believe in corporations having personhood when Texas executes one... - advocate_one
    6. Re:Robot grabbing and shoving things in its mouth by Green+Salad · · Score: 1

      Lesson 1: "Don't put water in there. No! Water...Baaaaad!"

  3. About time! by sm62704 · · Score: 3, Funny

    'When a computer fails at a task, it spouts an error message. Babies, on the other hand, just try again a different way, exploring the world by grabbing new objects -- shoving them into their mouths if possible -- to acquire additional data

    Access gives me the most amusing error messages. "Error 3417: there is no message for this error" (the message is real, the number I pulled out of my ass).

    But thinking about it, a robot looking for better data might be a good idea, but a computer? That might worrry me.

    Don't forget that a computer, even one running a robot, is just an alectronic abacus, nothing like a human or any other animal's brain. The temptation is to anthropomorphise.

    -mcgrew

    --
    mcgrew's razor: Never attribute to stupidity that which can be explained by greedy self-interest
    1. Re:About time! by spun · · Score: 2, Insightful

      But a brain is more than likely, and to the best of our current scientific understanding, just a really, really complicated abacus.

      --
      - None can love freedom heartily, but good men; the rest love not freedom, but license. -- John Milton
    2. Re:About time! by merreborn · · Score: 1

      But thinking about it, a robot looking for better data might be a good idea, but a computer? That might worrry me.


      I don't know. I'd hate to see, say, an automobile assembly line robot accidentally skewer someone as part of its learning process. The most common application of robotics in the real world is for extremely dangerous industrial work, where learning by way of mistakes could very well be fatal.
    3. Re:About time! by pionzypher · · Score: 1

      Agreed, but of course this would never be used on a robot in such a situation. Those robots need no learning. They are repetitive things that should only vary when a human explicitly tells them to. This would be useful in creating a sort of base programming for a robot meant to walk, help elderly people, etc. You could do the base learning (i.e. putting things in your mouth) part in a lab to gain a nice template which you could copy to other robots. Follow this with a specialized learning period which would provide the refinements for its actual intended purpose. And of course you could always drop in directives of actions to never take, regardless of the circumstances.
      The luxury being you can always set concrete bounds for their learning. Where humans can rationalize the strangest things in the correct circumstances.

      --
      I'll believe in corporations having personhood when Texas executes one... - advocate_one
    4. Re:About time! by sm62704 · · Score: 1

      Even worse; how would the robot "learn" that when the old person stopped moving it was because it has stuck its appendage through the geezer's chest?

      --
      mcgrew's razor: Never attribute to stupidity that which can be explained by greedy self-interest
    5. Re:About time! by sm62704 · · Score: 1

      No it isn't, and there are no studies that would suggest such a thing.

      --
      mcgrew's razor: Never attribute to stupidity that which can be explained by greedy self-interest
    6. Re:About time! by spun · · Score: 1

      Yes, it is, and there are plenty of studies that don't just suggest that, they flat out state it. In fact, there are no studies that refute that staement, and I challenge you to find even one peer reviewed study that shows evidence that the brain is nondeterministic. There are no structures in the brain small enough for quantum effects to matter. The brain is deterministic. Sorry if that contradicts any dumb ideas you might have had about free will or a soul, but the universe doesn't care about your hurt feelings.

      --
      - None can love freedom heartily, but good men; the rest love not freedom, but license. -- John Milton
    7. Re:About time! by sm62704 · · Score: 1

      Actually I don't believe that a soul has anything to do with thought, and I think that free will is an illusion. That can neither be proven nor disproven, either. Thinking is a chemical peocess; MSG or alchohol as well as many other substances (maybe even all substances) can alter your mood, thought, and perception.

      A Computer is just a binary abacus. It can't think. That's not to say that something will never be invented in the future that can; if someone discovers Asimov's positronics maybe they can think. But a computer like the one you're using now will never think, although it can, of course, simulate thinking.

      --
      mcgrew's razor: Never attribute to stupidity that which can be explained by greedy self-interest
    8. Re:About time! by spun · · Score: 1

      Well we agree on the basics, I suppose. In fact, now I'm not sure what we disagree on. What qualitative differences do you see between a computer and a brain that would keep a computer from thinking?

      --
      - None can love freedom heartily, but good men; the rest love not freedom, but license. -- John Milton
    9. Re:About time! by sm62704 · · Score: 1

      Its archetecture. When someone starts talking about a "thinking machine" or about how computers should have rights or how computers will become powerful enough to think, I always ask "how many more beads to I have to string on my abacus before it becomes self-aware?"

      --
      mcgrew's razor: Never attribute to stupidity that which can be explained by greedy self-interest
    10. Re:About time! by spun · · Score: 1

      A computer is Turing complete. An abacus is not. Any problem that one Turing complete computational device can solve, any other can solve as well. Are you suggesting that the brain computes in a way that overturns sixty years of computability and information theory? Anything is possible, I suppose, but a premise like that needs more support than just, "Well that's what I think."

      --
      - None can love freedom heartily, but good men; the rest love not freedom, but license. -- John Milton
    11. Re:About time! by sm62704 · · Score: 1

      Thought and feeling are chemical processes. A computer can only simulate them.

      --
      mcgrew's razor: Never attribute to stupidity that which can be explained by greedy self-interest
    12. Re:About time! by spun · · Score: 1

      Ah, but if the simulation were accurate enough, would the internal qualia of the being so simulated be different from ours? How would we know? More importantly, how would it? If the physics of our world create the qualia of experience, then I think an accurate simulation would give rise to the same qualia.

      --
      - None can love freedom heartily, but good men; the rest love not freedom, but license. -- John Milton
    13. Re:About time! by sm62704 · · Score: 1

      Ah, but if the simulation were accurate enough, would the internal qualia of the being so simulated be different from ours?

      Does a simulation of a hydrogen bomb, no matter how accurate, produce radiation? Will a flight simulator take you to Paris? Simulation is not reality.

      How would we know?

      This is what I worry about. We surely will, sooner or later, come up with a devoce that can fool humans into thinking it is sentient, and you'll have nuts calling for "machine rights".

      More importantly, how would it?

      That is my point entirely - it wouldn't. It couldn't.

      If the physics of our world create the qualia of experience, then I think an accurate simulation would give rise to the same qualia.

      Again, no matter how accurate you simulate an atomic blast, there is no radiation and nothing in the real world is destroyed.

      --
      mcgrew's razor: Never attribute to stupidity that which can be explained by greedy self-interest
    14. Re:About time! by spun · · Score: 1

      That's a totally false analogy. The simulation of the atomic blast is not set up to interact with the real world. A simulation of a human mind would have to be. In fact it would have to be set up to interact in all the ways a human mind could, too be accurate. A mind's purpose is to interact with the world. With an atomic bomb simulation there is no feedback loop between sim and world. With an AI, there would have to be.

      The thing is, the questions we are asking are not only unanswered, I think they are unanswerable. But they are endlessly debated. The point is, the question doesn't just apply to simulations. There is no way for me to know if you even have internal qualia, or if you are a blank simulation lacking all internal awareness and experience. The question is moot until we have the technology to transfer our brains to a simulation, then people will really want to know if the simulation of them will have experiences or will merely says it that does.

      For a very, very good Sci Fi read on the subject, I highly recommend Greg Egan's Permutation City. It will blow your mind.

      --
      - None can love freedom heartily, but good men; the rest love not freedom, but license. -- John Milton
    15. Re:About time! by sm62704 · · Score: 1

      Or I could be in a straitjacket in a rubber room imaginbing all this.

      I think, therefore I am (I think). Am I?

      --
      mcgrew's razor: Never attribute to stupidity that which can be explained by greedy self-interest
  4. Related by UbuntuDupe · · Score: 4, Interesting

    This reminded me of this video that I favorited on youtube, in which a robot is "brought to life" and then "feels around" to model the world and itself, and then "figures out" how to walk.

    This seems really interesting and something I'd want to work on. Anyone know what I would need to learn and do in order to get involved on a theoretical or practical level?

    --
    Apology to Ubuntu forum.
    1. Re:Related by LexMortis · · Score: 1

      Am I the only one who is seriously freaked out by that thing?

      - The speed is fairly impressive for that thing
      - With walking there is a large thud with each step
      - It's making somewhat alien noises
      - It's all twitchy

      Perhaps it's a case of uncanny valley, except not with trying to be human but trying to be an animal/headcrab.

      I would probably use a bat/crowbar on that thing, just too damn creepy...

    2. Re:Related by Tacvek · · Score: 1

      Am I the only one who is seriously freaked out by that thing?

      - The speed is fairly impressive for that thing
      - With walking there is a large thud with each step
      - It's making somewhat alien noises
      - It's all twitchy

      Perhaps it's a case of uncanny valley, except not with trying to be human but trying to be an animal/headcrab.

      I would probably use a bat/crowbar on that thing, just too damn creepy... Bit too much HalfLife? On a more serious note, The self modeling system is highly interesting. However, all of the gaits shown in the video (especially those in the later part of the video) seem significantly less than ideal. I would find it much more interesting to see the same concept applied to a more normal leg layout, that is capable of more or less normal walking.
      --
      Stylish sheet to fix many problems in Slashdot's D3: https://gist.github.com/801524
    3. Re:Related by Stefanwulf · · Score: 2, Informative

      Anyone know what I would need to learn and do in order to get involved on a theoretical or practical level?
      The first thing I'd do is get a feel for the field. MIT's OpenCourseWare has some interesting stuff like Intro to Robotics. I've actually also found wikipedia to be helpful in determining the areas of specialization within a field, and some of the basic jargon that has developed. There will be certain levels of math, programming, and possibly physics or electrical engineering knowledge that you'll need to have, and I've found one of the best ways to get a feel for those is to go to a local university library and pull some papers/conference proceedings on the subject. Read a few of them (from different people, and preferably different conferences/journals) that have titles which interest you, and take note of things like the level of mathematics or engineering knowledge being applied. If you don't understand it, don't be discouraged...it's just a technical language used in papers, and it's not terribly hard to learn, especially when you're learning it within a directly applied framework. It's worth it too, being able to quickly and easily read the papers being published lets you benefit from a huge realm of work that other people are already doing.

      On a hobbyist level of involvement, my impression is that's it's far cheaper to get involved with the AI/control side of robotics than it is the hardware design. There are a number of freely/cheeply available robot simulators - some are listed at http://www.robotcafe.com/dir/Software/Simulators/ and http://www.google.com/Top/Computers/Robotics/Software/Simulation/. These let you play with the control systems without having to worry about constructing/purchasing the hardware. Alternatively, for less than $1000 US, you can set up a fairly cheap robotics lab with an AIBO, a wireless connection to your computer, and software like Tekkotsu or URBI.

      And of course, if some aspect of it really catches your interest and you want to pursue it professionally, your best bet is probably to start looking at studying with the academic departments which have been publishing the papers/materials you've enjoyed the most, or which have strong programs in that area of the field.
    4. Re:Related by UbuntuDupe · · Score: 1

      Thanks a lot for the links!

      I probably should have given a little background: I have a bachelor's in mechanical engineering and did some work toward a masters before dropping out. I remember the academic papers being very hard to follow even after all that schooling in in the subject area. (I was studying controls and the problem of closed-loop system identification, which seems to have a lot of overlap with what the robot in the video has to do: model a system while trying to manipulate it.)

      "When all you have is a hammer, everything looks like a nail." And likewise I'm tempted to do see where my M/E knowledge applies, but what I'm more interested in is not the mechanics of the robot itself, but understanding how it learns, which would make the "physics" of its world irrelevant until I want to make a practical demo.

      So far, I've been reading books aimed at broader audiences: Godel, Escher, Bach and The Emporer's New Mind, and well, everything by Eliezer Yukdowsky.

      Just for kicks, I sat down a week ago and tried to model how an intelligent being acts, and settled on a model: it compares sensory data (a sense-vector) to previous actions (action vector) to guess how the world works (the terms in a "world matrix" that it tries to imitate in its "internal model matrix"), and then chooses actions to optimize an unknown utility function.

      Looking at the site linked in the video's description, that's basically what the team did, except that they had a much more complicated model. (Mine had a 2x2 binary world matrix.) It seems they also helped it a lot by giving it good approximations of its environment. I'm unclear though how they made it "want" to walk.

      Again, thanks for taking the time to answer my quesiton.

      --
      Apology to Ubuntu forum.
  5. Re:Not too useful by Nos. · · Score: 5, Insightful

    Simple, you learn how they learn. Children learn alot in the first few years of their life, and the rate at which they learn can be quite astonishing.

  6. Re:Not too useful by physicsboy500 · · Score: 1

    The point of the article is not that we're learning specific tidbits of information from them, but instead we're learning how to learn from them. The mindset of a young child allows them to learn things at an immense pace and if this were copied as a computer algorithm it further advances AI technology.

    If you dig into the article it's a pretty interesting read about how the Stair robot is slowly learning small tasks. While it is currently not that impressive what tasks it is capable of, it is a staple of advancing AI.

    --
    The original generic sig.
  7. Couldn't resist. by eriks · · Score: 1

    I for one welcome out new robot baby overlords.

    Seriously though. If a robot could learn at even a tiny fraction of the rate that babies learn, that would be quite remarkable.

    1. Re:Couldn't resist. by Fission86 · · Score: 1

      I for one welcome out new robot baby overlords.
      I'm not worried what our new baby robot overlords will do, in fact i'm not even worried what our adult robot overlords (our deaths will be fast - see: skynet) will do.

      I'm worried about the angsty teen robot overlord that doesn't really know their place in the world yet but still thinks they're god. That and forcing everyone to listen to crappy emo music.
      --
      Coming to you live from another dimension.
  8. think of the potential by themushroom · · Score: 1

    This new device is fun on dates!

    --
    Laughter is the Spackle of the Soul.
    1. Re:think of the potential by UbuntuDupe · · Score: 1

      punchline: "Hah! You thought that was a robotic hand!"

      --
      Apology to Ubuntu forum.
  9. Re:With the whole... by Frequency+Domain · · Score: 2, Funny

    ...I sure hope they're not testing it out on babies.
    No, they use lawyers and politicians. They were using rats, but too many people protested.
  10. RobotCub - open source hardware by rich_from_shadow · · Score: 2, Informative

    RobotCub is particularly interesting because all the design is being posted on their website, so anyone with a spare 200kEuro can build one. It's an EU-funded project, and it's good to see government money widening the pool of Open Source stuff - see www.robotcub.org

    --
    We build robots - www.shadowrobot.com
  11. Old News by Agilus · · Score: 2, Informative

    Seven years ago, when I started grad school at UMass Amherst, they were talking about this stuff in the robotics lab. This is hardly new stuff.

    One of the ideas I got out of their talk about their research was that babies start using their muscles gradually, with only a few degrees of motion to start, and moving up from there. For instance, you'll notice that when a baby starts using his hands, he just grabs things for a while. There is no fine grained control of the fingers, and even the control necessary to consciously release whatever he grabbed doesn't come until later.

    --
    hackshop.com - My tech hobby project hub
    1. Re:Old News by StreetStealth · · Score: 1

      Still, I'm surprised we aren't reading more about robots (or software) that rely on procedural evaluations of actions and surroundings and "try" alternative paths to a goal. It seems to me that this should simply be the obvious way forward!

      --
      Your mind is clear / The things that you fear / Will fade with how much you / Believe what you hear
    2. Re:Old News by inKubus · · Score: 1

      I have a five month old and I've been observing his progress since birth. It is quite interesting how he learns. First of all, you have to understand that long before a baby is born it is moving around in the dark, learning muscle control (feedback loop). When it is born, the senses of touch, smell, hearing and sight are already there and developed. The baby just doesn't know what to make of the stimuli. Learning is when the baby starts associating one or more of these inputs with satisfying it's needs (hunger, changed, sleep, general attention). Babies "see" with their mouths, since the mouth is the most sensitive touch center, so they are constantly trying to bring things to their mouths, or their mouths to things. I heard about a project (Deb Roy at MIT) to capture literally every input a baby recieves while he/she is developing. He wired his house with cameras and microphones so all stimuli are being captured. They are also in tandem storing developmental milestones in the baby's life. From the resulting data warehouse they are hoping to find patterns that will help develop language learning models.

      Motion is quite a bit simpler, only with language do we truely understand why we want to grab something, and that influences everything about our grasp. We know to grab a rock with both hands, and an egg gently.

      --
      Cool! Amazing Toys.
  12. Building a better? by rich_from_shadow · · Score: 2, Interesting

    Not quite sure why it's a better robot Hand - Barrett have been around for some time, and their three-fingered system is good, but it's an interesting gripper. The people in the article are the customers, not the hand developers. Still, interesting research is interesting research ObDisclaimer - Shadow does five-fingered hands...

    --
    We build robots - www.shadowrobot.com
  13. Re:Not too useful by Jugalator · · Score: 1

    Especially how quickly they can get their parents attention by crying. :-S

    --
    Beware: In C++, your friends can see your privates!
  14. Re:Not too useful by moderatorrater · · Score: 2, Funny

    Too bad it doesn't work on Britney Spears...

  15. Let me be the umpteenth... by Nerdposeur · · Score: 2, Funny

    ...to welcome our drooling, pooping robot overlords.

  16. Misleading by Savione · · Score: 1

    I did not RTFA, but I read the magazine version. The title and summary are somewhat misleading. The robot does not learn from babies, but learns like them. I'm pretty sure no infantile humans were involved in its programming. Also, the Stair 2.0 is the new one the article talks about. The Stair 1.0 is at least two years old.

    --
    See it there, a white plume over the battle - A diamond in the ash of the ultimate combustion - My panache. --Cyrano
  17. baby programmers? by Essequemodeia · · Score: 1, Funny

    I heard that programmers are getting younger, but this is getting ridiculous.

    1. Re:baby programmers? by Oktober+Sunset · · Score: 2, Funny

      well at least they have an excuse for living with thier parents...

      --
      What if Tetris was invented by Nazis?
  18. Re:Obligatory Discworld reference by Abreu · · Score: 1

    +++?????++ Out of Cheese Error. Redo From Start.

    +++ Divide By Cucumber Error. Please Reinstall Universe And Reboot +++

    --
    No sig for the moment.
  19. Interesting about the shadowrobot.com hand... by tlambert · · Score: 1

    Interesting about the shadowrobot.com hand...

    I notice by the relative ratio of the length of the index finger to that of the ring finger that the hand is modelled on a male hand (the index digit is shorter than the ring digit). Have you done any studies on dexterity as to which configuration is better for manipulating objects?

    Thanks,
    -- Terry

    1. Re:Interesting about the shadowrobot.com hand... by rich_from_shadow · · Score: 1
      All four fingers on the Shadow hand are the same length - we mount them at different offsets to get the staggered tip effect.

      In general, we copied nature where possible, but sometimes engineering practicalities overruled us.

      So there's a curl on the little finger, but the palm itself is rigid. Making the palm flexible implies a lot more connectors on the electronics...

      --
      We build robots - www.shadowrobot.com
    2. Re:Interesting about the shadowrobot.com hand... by blincoln · · Score: 1

      I notice by the relative ratio of the length of the index finger to that of the ring finger that the hand is modelled on a male hand (the index digit is shorter than the ring digit).

      On my left hand, the index finger is shorter than the ring finger. On my right hand, the ring finger is shorter than the index finger. The index fingers of both hands are the same length. Does that make me a hermaphrodite?

      --
      "...always new atoms but always doing the same dance, remembering what the dance was yesterday." -Richard Feynman
  20. Re:Not too useful by physicsboy500 · · Score: 1

    Too bad it doesn't work on Britney Spears... or her fans
    --
    The original generic sig.
  21. Re:With the whole... by rich_from_shadow · · Score: 1

    I thought it was because there were things rats wouldn't do...

    --
    We build robots - www.shadowrobot.com
  22. This is not new by holywarrior21c · · Score: 1

    This is not just "better robot arm". Here is other example of this kind.http://www.iirobotics.com/bot-news/general/new-intuitive-ways-to-teach-a-bot!-20070605938.html
    This is just another way to teach robot by having direct interaction between the robot and the instructor. Robot yet still needs to be monitored closely by human and it does not do all that in its own. It still needs human's help. robot that learns like baby could be useful if it does something already well. maybe it can learn to create something better than it is originally programmed to do and play music more creatively and play sports such as golf and table tennis. but instructor himself is not perfect neither. lets say instructor teaches his robot to carry a fairly heavy box or large panel by simply putting(forcing) robotic hands on the sides of the box. then robot moves few feet and instructor realizes that it overstressed robotic arm. carrying it around with different position by finding center of weight by just lifting it up to knee high and experimenting comfortable position as humans do, could have prevented that. This is interesting progress anyway. I am bilingual who speaks English language as a second language. and during the beginning years of studying english, i thought a lot about how babies learn language. for instance, visualizing concepts and objects in your head, not just memorizing spellings and grammars. it is easy to ignore simple things. but we have yet to learn a lot from simpler ones(babies), and simple things that surrounds us.

  23. Geth? by DarthJohn · · Score: 1
    They start learning how to move hands, then lift things and put them in their proper place and that's useful enough. Eventually they ask "am I alive?" Then you're screwed, try to eradicate them and you end up forced into exile to wander the galaxy in a flotilla of old ships that are always on the verge of falling apart.

    Yea I know... nobody really cares, but that happens to be the first thing I thought of... been playing too much Mass Effect

  24. 1200 baud by xPsi · · Score: 1
    I for one welcome our new robo^H^H^H^Hbaby over^H^H^H^H.

    Oh, never fu^Hricking mind.

    --
    i\hbar\dot{\psi}=\hat{H}\psi
  25. Life With Limbs by swaza1 · · Score: 1

    Interesting digression, Science Friday just had a piece on rewiring nerves for amputees.

        http://www.sciencefriday.com/program/archives/200711301

    --

    "He that breaks a thing to find out what it is has left the path of wisdom."