Stories
Slash Boxes
Comments

News for nerds, stuff that matters

Slashdot Log In

Log In

Create Account  |  Retrieve Password

Self-Introspecting Robot Learns to Walk

Posted by CowboyNeal on Sat Sep 01, 2007 09:20 AM
from the dreaming-of-electric-sheep dept.
StCredZero writes "There's something about these things that seems eerily alive! The Starfish Robot reminds me of the Grid Bugs from Tron. But it's very real, and apparently capable of self introspection. In fact, instead of being explicitly coded, it teaches itself how to walk, and it can even learn how to compensate for damage."
+ -
story

Related Stories

This discussion has been archived. No new comments can be posted.
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
 Full
 Abbreviated
 Hidden
More
Loading... please wait.
  • by doombringerltx (1109389) on Saturday September 01 2007, @09:46AM (#20433589)
    but come on! "Update 24-Nov-2006:"
    • Like another poster said, it's still more relevant (geeky, cool, and IMHO important than most of the "new" stuff on here.

      I believe that we will eventually create true self-replicating machines. It may even be one of the next important steps in the progression of computer technology (Boolean Logic, Relay, Tube, Transistor, IC, Microprocessor, Self-modifying code...)

      Where it gets controversial is that I think that we could see it in the next 50 years or so. (That is, a self-replicating organism that can m
      • Having a self replicating machine for the sake of a self replicating machine is pretty pointless. Though maybe that makes life pointless, but angst aside, can't 3D printers already replicate themselves? At least if you build an assembly line in. A self replicating robot let into the wild would just cause havoc with our ecosystems and ravage all the decent raw materials we have :O
        • Self-replicating robots could build you a Moon or Mars base before you even land. Or would you prefer to haul rocks yourself, in a spacesuit, for 12 hours per day, just a year away from Earth in case if you break a leg?

          Self-replicating function here is essential because it covers repairs, and those will be essential. Besides, it might be difficult to send more than a handful of robots ahead of time, and definitely not thousands.

          • This is a potentially very good area for aspiring young engineers and scientists to focus on in their education.
            Take the Mars base as just one of a big class of problems. Obviously, the robots can't just be capable of self replication. They have to self replicate up to useful numbers, and then do other tasks. Once that very general model is applied, refining it is mostly a matter of efficiency, and that efficiency determines whether the project will ultimately be funded or not.
  • by RyanFenton (230700) on Saturday September 01 2007, @09:47AM (#20433599)
    This is a very well-done video. I really like how it shows the virtual model to illustrate how the system 'sees' itself. Self-reflection of a sort is usually present in most complex programmed systems in one form or another - usually in terms of disjointed status variables and variables for their hard-coded implications. This is neat because the implications can be a little more dynamic.

    I hope this becomes a more general library that can be used to help self-reflection of this sort become a more separate part of physical designs. Even if the implications of the physical model aren't dynamic, a standard way of quickly seeing how your model 'sees' itself would help debugging and development in many future projects.

    The only problem if it becomes more prevalent would be same one that quantum mechanics holds - people think that 'observer effects' has to involve consciousness, in the same way they'd think that a program's self-reflection would mean that it 'thinks' the same way they do. Neither is true - they're all mechanical terms wrapped in common language. Anything that can record an effect on the world (a falling rock's scratches in another stone would work) is a quantum observer - consciousness has nothing to do with the 'collapsing wave function'. The same here - a bit of self-reflection on the part of a program doesn't mean it's eerie self-corrections are capable of the complexities of our mind. If anything, such mechanical results would imply that our own minds act simpler in some ways than we may think, and that consciousness doesn't necessarily have to be as inscrutable and special as we might want.

    Ryan Fenton
     
    • "If anything, such mechanical results would imply that our own minds act simpler in some ways than we may think, and that consciousness doesn't necessarily have to be as inscrutable and special as we might want."

      Philosophers like Daniel Dennett agree with this notion. Consciousness may simply be a more complex continually-running predictive model like that used by this robot.
      • And most psychologists and neuroscientists. Most of our decisions are made for us before we become conscious of them. They've shown the responses to stimuli are often formulated before a subject even becomes aware.

        Dennet objects to the notion of "qualia" - the experiential bits that make up seeing, hearing, touching, smelling, and tasting. He says that these qualia don't really exist as special things, since adjusting qualia and adjusting the circuits that supposedly generate qualia conceptually achieve the
          • The theory of evolution does not presume it. Nowhere in the tenets of evolution does it say "Materialist theories of mind must be true." You could conceive of natural selection never producing consciousness - but the fact is, evolution did produce consciousness, and so we have to come up with an understanding of consciousness which explains how that could be. Evolution does presume that life arises only from physical matter, but then it wouldn't be a scientific theory if it didn't. It would be intelligent d
            • by Original Replica (908688) on Saturday September 01 2007, @03:27PM (#20435523) Journal
              "You could conceive of natural selection never producing consciousness "

              Oddly enough, you could not conceive of anything without consciousness. Understanding is a mental, not physical process. You could however conceive of consciousness without the physical world. Indeed every culture has been doing so for all of recorded history in the form of spirit worlds, afterlife, etc.

              Occam's razor can be much abused depending on how you frame your observation. "I think therefore I am." is much more straight forward than "I am incredibly complex and elaborate, therefore I think." Let's set Occam's Razor aside for this discussion, it doesn't seem to be the right tool for the job here.

              If you allow yourself to view the conscious world as more fundamental than the physical world, then the observed consistency/connectedness of all physical phenomena would require some sort of governing over-consciousness that is responsible for the physical world. That of course would be a form of creationism, much reviled here on /.
              • Oddly enough, you could not conceive of anything without consciousness.

                His point would probably be better made by saying: As far as we know, natural selection could have produce life that had purely mechanical "thought processes", like a world populated by computers.

                "I think therefore I am." is much more straight forward than "I am incredibly complex and elaborate, therefore I think."

                The facts that I exist and that I think is very obvious to me, but that doesn't make it a more basic truth, or show t

                • If you want it to be seen as a rationally defensible position, rather than just a possibility for a philosophy class or a religious discussion, you're going to need more that just "here's a cool way of looking at things".

                  Please. Even philosophy class needs more than "here's a cool way of looking at things".

              • Re: (Score:3, Interesting)

                Understanding is a mental, not physical process.

                You are assuming that they are independent, when in fact there is lots of evidence that mental processes depend on physical processes. There are drugs to alter your consciousness, physical damage to your brain can cause mental damage, and there are experiments where people's thoughts have been maninpulated by direct electrical stimulation (these people were undergoing brain surgery).

                That of course would be a form of creationism, much reviled here on /.

                Because it doesn't explain anything or offer any evidence.

              • Saying "consciousness just emerges because of the complexity of the machine" no more satisfies Occam's razor than saying "consciousness just emerges because of God." Neither explains a mechanism, so neither is a theory, so Occam has no interest in either.

                Occam's razor applies whether we have proposed mechanisms or not, the whole point of the thing is to help you make a good guess when none of the theories you have seem very complete or easily testable. If we knew every detail, we wouldn't need to use a

    • The video has parts (the simulated ones) which look a heck of a lot like the cool and open source breve AI simulation environment ( http://www.spiderland.org/ [spiderland.org] ) which does pretty much the exact same thing. Check out the brevewalker or brevecreatures.

      The video is still extra-impressive though, as the robot uses sensors to detect it's own shape and limitations, and then (it looks like) loads it into breve where the thinking seems to happen. Pretty cool indeed.
    • This seems to be similar to a teenager spending lots of time in front of a mirror (except without the angst, acne and worries about hairstyle).

      CAPTCHA: prophet. Dammit, I wanted PROFIT!
  • Creepy (Score:5, Funny)

    That thing almost looks alive. After seeing it, it reminded me of the nurses in Brookhaven Hospital trying to move. Eew.
    • Re: (Score:3, Interesting)

      At the start it looks creepy when it's moving around looking a little like a spider. Then it gets damaged and looks genuinely scary, in terms of "WHY WON'T IT DIE?!". At the end it just looks like its makers enjoy pulling the wings off flies (although I did laugh when it flipped itself upside down). It's be interesting to see whether this modelling system could be made to learn from its experiments and failures as well as creating initial similations to work from. What I mean is, its internal simulation le
  • The videos of it trying to move with the damaged leg make it look like a crippled animal. I can't help but feel sorry for it. :(

    Argh, I said that and had a sudden mental image of hordes of animal rights activists protesting the mistreatment of robots.
    • In a gut-wrenching moment, the robot was heard to be saying: "Why, why, WHY was programmed to feel pain?!?!"
    • The videos of it trying to move with the damaged leg make it look like a crippled animal. I can't help but feel sorry for it.
      You won't be saying that when it leaps up and grabs you in the face shouting, "Introspect this, motherfucker!"
    • Re: (Score:3, Interesting)

      Supposedly a mine-clearing bot (lots of legs designed to be blown off by mines, the bot just walks around and triggers them) that was literally on its last leg was pulled out of the testing (it would have crawled onto a final mine and be destroyed in the process) because the supervising officer felt sorry for it. People are capable of feeling empathy for the dumbest animals, why wouldn't they for a robot?
      • Well I can't exactly argue why people wouldn't feel sorry for a robot, since I in fact feel sorry for said robot. ;P
  • and observe a 'Robots Must be Given Human Rights' movement grow in numbers.

    This robot moves in a fluid way, almost like a living creature would, many people will immediately anthropomorphize it.

    --

    What I find interesting is applications in todays world. How about equipping cars with abilities to sense its physical parts and build a total model of itself in real time. This could be used for immediate diagnosis of problems with the car itself and with its interaction with the surrounding environment. Many p
    • I find your use of the word "anthropomorphize" in this context interesting.

      It seems to me that in the context of artificial intelligence that word represents a set of values in the guise of a representation of some unspoken, well-defined set of characteristics that separate humans from whatever it is one is comparing humans to. It conveniently disposes of the really hard problem of establishing what it is that sets humans apart in a very neat linguistic package.

      In other words, use of the word "anthropomorph
      • "so the car doesn't need to know its own state in more detail than the LEDs on the dashboard can express"

        The car makers using dynamic stability control [wikipedia.org] would beg to differ, amd IMHO the system qualifies as a kind of "introspection" for cars.
  • Hope it comes with a remote-control kill switch.
    Hope it doesn't figure out how to circumvent the remote-control kill switch.
    Hope it doesn't build a bigger version of itself...
    • Seconded!

      Watching that video, I got a truly creepy feeling.
      Rather uncomfortable actually. Then I thought about it for a little while, and I think the reason was this [movieweb.com].

      Let's get that thing a kill switch *first and foremost*, and *then* think about imparting sentience.

  • by LynnwoodRooster (966895) on Saturday September 01 2007, @10:36AM (#20433897) Journal
    1. Did they give it a navel?

    2. Can it contemplate it?

  • by icepick72 (834363) on Saturday September 01 2007, @10:37AM (#20433903)
    it's obviously going to latch onto somebody's face and then they'll say it learned fast.
  • by Animats (122034) on Saturday September 01 2007, @10:41AM (#20433939) Homepage

    First, get past the blogodreck to the actual work. [cornell.edu] (Slashdot editors missed a blog troll again.) Also, this work is several years old. The papers are from 2004 to 2006.

    The original article says that the robot has "tilt and angle sensors in all its joints", but that's wrong. It only has one central tilt sensor. That's significant, because if it did have tilt sensors at each joint, system identification would be easier. The algorithm is doing better than one might expect.

    This thing is doing what controls people call "automatic system identification". You have some set of sensor inputs and some set of control outputs, and the control system has to figure out how they relate. It does this by adjusting the outputs and watching what happens. There are various statistical techniques for doing this. Calling this "introspection" isn't really correct.

    After system identification, the model is inverted, or solved for the inputs in terms of the outputs. The inverted model can then be used as a controller. Given desired outputs, the inputs needed to achieve them can be computed.

    The novel result here is that a reasonably decent system identification for a nonlinear system is being performed with a small number of physical tries. That's an improvement over previous methods, which tended to "learn" very slowly. I'd looked at approaches like this for legged locomotion in the past, but the available system identification algorithms weren't good enough. This looks promising.

    Good robotics work, crap Slashdot article.

    • Not a blog troll! I have no association with Technovelgy. I was just cleaning up my less-used bookmarks, and happened on this post on goold ole Technovelgy, and thought it was neat. (So did the editor, and lots of readers!)

      I posted the blog entries about the Starfish Robot because it was a good and useful summary. If you don't think so, then that's fine. Just don't go falsely ascribing motivations and intentions!

      I've been a Slashdot reader and commenter for many years now. (Lost count. Over 5?) This
    • I think this device shows that pattern matching is all there is to life and intelligence. All that has to be done for artificial AI is an engine which tries to find a solution using statistical methods.

      By the way, this thing proves evolution one more: by trial and error, living entities have been developed...
      • Not really. It only shows one way that life and intelligence is possible. Even then, you get into that thorny little question of how you are definining intelligence.

        Don't misconstrue me, I'm closer in belief to your thoughts on this. But it doesn't prove the big questions you're alluding to in your post. It simply brings us one step closer to showing that all the amazing properties of life are not dependent on some mystical, impossible to fully comprehend entity, but can be mechanistically created.

        Even
  • by Sir Holo (531007) * on Saturday September 01 2007, @10:57AM (#20434049)
    Link to the research group at Cornell: http://ccsl.mae.cornell.edu/research/selfmodels/ [cornell.edu] Lots more pics, movies, and details.
  • Self-introspection is a tautology. It is just "introspection."
  • came over me as I thought back to images of infants playing with their toes.

    If the robot had come with some elastic (but NOT flesh colored) rubber skin, instead of looking like a meccano set, it would have been almost cute.

    They should try different orientations for the 'shoulder/hip' attachment, give it a longer brain/body and a spotted outer covering (with sensors in the 'skin",) a need to home to an electrical outlet to recharge, and make a toy out of it.

    After an initial charge "through" the box, you open
  • while i'm excited to see new development in these fields it is far from new. The 60s introduced concepts used in this robot and the 80s introduced actual simulations of self emergent systems.
  • ...be called a herd?
  • by Jafafa Hots (580169) on Saturday September 01 2007, @11:46AM (#20434355) Homepage Journal
    ...is to make it capable of autonomous self- introspection.
  • by Punto (100573) <{moc.liamg} {ta} {botnup}> on Saturday September 01 2007, @03:16PM (#20435471) Homepage
    Once it learns there's so much damage he can take, he'll know pain. From there is straight to world domination.
  • I read in the newspaper today that researchers had built dinosaurs in the computer that taught themselves to walk. They also modeled animals like dogs and humans and let them learn to walk too. Eventually they came up with ways to walk that were very similar to the way those creatures really walk. From this result the researchers concluded that the way the dinosaurs had walked in real life must be approximately the same as the way their computer models walked.
  • First, an aside: "self-introspection" is redundant. What's the alternative? Introspection of another? That makes no sense. Also, introspection is a component of consciousness. There is no way to determine another person is conscious (as opposed to a completely stimulus-response programmed "zombie"), much less a robot. Without consciousness, the appropriate term is "feedback mechanism".

    That said, the device in TFA is not novel, nor is it as simple as previous designs. Far simpler microbots have been built wi
    • Ah, but can it shoot?
    • Re:Damage (Score:4, Funny)

      by hasbeard (982620) on Saturday September 01 2007, @09:58AM (#20433663)
      Well, it seems to me that combat really isn't a good time for too much introspection. I mean with all the bullets flying and all.
      • Re: (Score:3, Interesting)

        I have to agree, even if I am not sure how I would define life. It would be interesting if the software element of this could be used in conjunction with biological hardware, or hardware with biological traits (i.e. replication and energy production). It seems to me that having a central control mechanism (brain) for all large scale operations plus small independent modules for specific tasks would be a close approximation to biological life (less complete reproduction, although I suppose that may be possi
    • One way of doing this would be to have self-registering joint/segments that send out a broadcast along the communication link. This would let the controller know what joints were present. But the controller would still have to figure out how they are connected. That's where the self-awareness bit comes in. With a single tilt sensor, the controller keeps moving joints at random until the orientation of the sensor in the predicted arrangement matches that in the real world. Alternatively, each segment could d