robot artist draws portraits

Posted by Lee Wells | Sun Dec 30th 2007 2:37 a.m.

This humanoid robot holds a pen in its hand and can draw an image of any person who stands in front of it.

Artist Robot by Sylvain Calinon

Created by robotic researcher Sylvain Calinon, the robot recognizes when there’s a face in its field of view, then snaps a digital photo and extracts the major characteristics of their visage. Once that’s done, the robot turns itself into an X/Y plotter, picking up an old-fashioned quill pen and gradually filling in the details of a portrait.

I think it would be really cool to set up a bunch of these in a row, creating different styles of art using ink, pastel, oils and acrylics. Sylvain, are you listening?

<object width="425" height="355"><param name="movie" value="http://www.youtube.com/v/aosl53ie3mw&rel=1"></param><param name="wmode" value="transparent"></param><embed src="http://www.youtube.com/v/aosl53ie3mw&rel=1" type="application/x-shockwave-flash" wmode="transparent" width="425" height="355"></embed></object>
  • Pall Thayer | Sun Dec 30th 2007 7:54 a.m.
    We did some similar stuff (albeit, minus the robot) at The Institute of Everyday Life (http://iel.hexagram.ca), the lab I worked at in Montreal. We used a simple camera and an old HP Plotter. Jpegs or pngs taken by the camera were autotraced into vectorized svg's and then plotted by translating the svg xml into hp-gl, the ancient plotting language understood by the plotter. The whole process was automated with a couple of perl scripts. One of the things I feel is conceptually important for this sort of thing in an art context is to avoid pretending that the machine/robot/whatever actually knows what it is drawing. Because that's exactly what's missing from this type of imagery. A true recognition and familiarity with the subject matter. The result is invariably a collection of outlined areas, eventually forming an image but void of any expression. Making machines that draw is nothing new, but how do we make those machines expressive? And do we really want to?

    Pall
  • Philip Galanter | Sun Dec 30th 2007 5:25 p.m.
    If someone can design a machine that "has" something to express, making the machine expressive will be easy.
  • Steve OR Steven Read | Sun Dec 30th 2007 6:07 p.m.
    I like that its a cute little robot with a little robot hand and fingers. I like that he asks for the tool to be placed in his hand. Yeah I suppose the capture/drawing portion itself is samey, but because its a little bot sitting at a little desk, facing with the subject directly, I like that. Reminds me of the renaissance automatons. But I agree perhaps something more expressive or unpredictable would be nice in the paper drawing artifact. Not sure if that was the intent though, looks like alot of the craft went into the look and behavior of the robot itself.
  • Pall Thayer | Sun Dec 30th 2007 6:17 p.m.
    That's a wonderful point, Philip. I don't believe we'll ever get there but if we do, will whatever-it-is still be a "machine"?
  • Philip Galanter | Mon Dec 31st 2007 1:02 p.m.
    Well, we are teetering on the brink of a very long and complex topic. We could branch off and try to define life (i.e. are humans machines?), and sooner or later will come the issues of qualia and consciousness and the Turing test and whether or not you can detect (from the outside) whether an agent is a (philosophical) zombie or not...etc...

    But avoiding all that...

    I'd say the question comes down to what you mean by "machine." In terms of everyday use most people today would probably say such a thing would no longer be a machine, because machines have no consciousness, and the everyday notion of expression requires consciousness.

    • Pall Thayer | Wed Jan 2nd 2008 9:15 a.m.
      If we extend the meaning of "machine" to include machines that have some form of consciousness and that are therefore capable of expression, then we sort of defeat the purpose, don't we? Isn't that one of the big benefits of machines? That they work for us tirelessly without complaining or injecting any personal subjectivity or opinions into their task?

      I'm not saying that this type of work can't be expressive but sometimes people make the mistake of projecting the expression onto the machine rather than the human behind the machine's creation. Another mistake I see sometimes is when people refer to the images created by these machines as the "artwork" as opposed to the whole being the "artwork".
  • Lee Wells | Mon Dec 31st 2007 2:47 p.m.
    image

    I put together this little image to show how through simple telepresence learning a robotic machine could be tought to paint naturally. The robot is from the Greenman project from the 1980's.
  • Lee Wells | Mon Dec 31st 2007 2:48 p.m.
    image

    I put together this little image to show how through simple telepresence learning a robotic machine could be tought to paint naturally. The robot is from the Greenman project from the 1980's.
  • Erika Lincoln | Wed Jan 2nd 2008 4:51 p.m.
    To move horizontally from the discussion, the question I have is does the machine draw or does it render or plot?
  • Pall Thayer | Wed Jan 2nd 2008 8:10 p.m.
    That's a good point, Erika. I started writing a reply but decided to take a little time to mull it over a bit but have now ended up pretty much where I started. I would feel inclined to say that "drawing" is a human act involving interpretation, expression and a level of consciousness that machines are incapable of. Plotting is based on perfectly defined x and y (and sometimes z) coordinates. What the machines of today need, at the very least, to be able to make a mark. It's a bit like the difference between speaking and communicating. We can make machines that speak but what happens if a person that the machine is speaking to doesn't understand. Will the machine be capable of re-interpreting what it said and speaking it in another manner? Will the machine be capable of guaging the person's level of understanding in an appropriate manner? Usually if someone doesn't understand, we scale downwards but there are occasions where you have to scale upwards. Say things in a more complex manner, to avoid confusion and usually, our conscious assessment of the situation, the subject matter and the person we're communicating with, tells us what's appropriate. So I guess you could say that drawing is communicating whereas plotting is simply marking. The machine doesn't really care what the end result is or looks like. It just goes from one point to another, obediently and without any conscious idea or care as to why. That's something we humans really can't do, even if we try.
  • Lee Wells | Wed Jan 2nd 2008 9:33 p.m.
    image
    Hi Pall:
    Does your use of Human include other mammals too?
    Does art need to be understood by everyone?

    Humans still need to be trained to think.
    image
  • Philip Galanter | Wed Jan 2nd 2008 11:27 p.m.
    This all points to one of the many reasons that I think generative art is such an important topic for theory. It's something in between "expression" and "plotting." Expression implies some form of consciousness. Plotting implies a simple mapping from input to output, i.e. something that is more like transcription than writing, and thus not an act of creation per se.

    I'll avoid the term "creativity" here because it's ambiguous enough to confuse. But I think it's pretty clear that *creation* does not require consciousness. But it's more than mere plotting. The are all kinds of creation going on in the universe, and almost all of it is unconscious and yet self-organizing. (Leaving aside various theological counter-proposals.)

    In generative art the artist harnesses and cedes control to these same, or similar, creative forces. The artist, of course, makes the indirect decision to use the system. But the significant form emerges directly from the generative system even though the system itself is (so far) unaware of it.

  • komninos zervos | Thu Jan 3rd 2008 2:35 a.m.
    a very effective visual argument
    komninos
    komninos.com.au
  • Pall Thayer | Thu Jan 3rd 2008 12:19 p.m.
    I'm not so sure that we can avoid the term "creativity" as it's pretty central to the issues. I think it's ambiguous only if you can't determine the context in which to use it. Within an artistic context I would define it loosely as meaning something along the lines of being able to break "rules" (or perhaps it would be better to say "not follow 'rules'"), break from convention and come up with new ideas as a result. But breaking rules and breaking from convention isn't an easy task. Even when we make a conscious effort to break from the norm, it's very likely that while breaking from some conventions, we'll still hold on to others. What I find exciting about generative and machine-based creation systems is the ability to break from convention and be entirely non-selective in the outcome which can lead to chance discoveries we may otherwise not have come up with on our own.

    Lee:
    Obviously, my use of the term "human" doesn't refer to other mammals (as they aren't human). But of course animals have consciousness and are capable of expression. They are a far cry from our "machines".
  • Steve OR Steven Read | Thu Jan 3rd 2008 3:52 p.m.
    Yeah Pall, I'm with you. I think chance when used in a generative system is a form of creativity. The machine is creative when it openly asks the universe for a random input. I guess the human tells the machine when and how to ask for the random, but a chance outcome occurs nonetheless. This robot's drawings might benefit from more chance. For instance one simple way to do that, without even having to re-engineer it, would be to add chance input to the siting procedure. Instead of of just sitting so centrally and carefully, one could do any number of things when the picture is being snapped. I guess both the human and the robot would be creative in that case. And obviously the image processing or movement algorithms could use utilize additional chance operations too... thereby increasing creativity of the robot? I am going to avoid the topics of consciousness or personality of the robot, because I don't want to hurt the robot's feelings.
    Stephe
  • Steve OR Steven Read | Thu Jan 3rd 2008 4:01 p.m.
    Lee I love animal paintings. Late DeKooning == Elephant Painting. I have friends who say 'art' and creativity only refers to the human. I guess define the terms however you like, but my own definition includes machines and animals too. I like those male birds (forgot the name) that woo the women with their personal style of building and displaying these weird piles. And I always thought pack-rat and bird nests were creative art. I've personally seen some amazing pack-rat nests!
  • Lee Wells | Fri Jan 4th 2008 1:01 a.m.
    <object style="width: 100%" type="application/x-shockwave-flash" data="http://www.reuters.com/resources/flash/includevideo.swf?edition=US&videoId=72248" width="344" height="320"><param name="wmode" value="transparent" /><param name="movie" value="http://www.reuters.com/resources/flash/includevideo.swf?edition=US&videoId=72248" /><embed src="http://www.reuters.com/resources/flash/includevideo.swf?edition=US&videoId=72248" type="application/x-shockwave-flash" wmode="transparent" width="344" height="320"></embed></object>
  • Rob Myers | Fri Jan 4th 2008 4:08 a.m.
    Pall Thayer wrote:
    > Within an artistic context I
    > would define it loosely as meaning something along the lines of being able to break "rules"
    > (or perhaps it would be better to say "not follow 'rules'"), break from convention and come
    > up with new ideas as a result.

    Margaret Boden describes this as "H-Creativity", historically unprecedented creativity. She characterizes it as adding new axioms to the state space of an activity.

    Most artists and many perfectly good works of art don't do this so it seems a little unfair to demand it of computers.

    In Artificial Intelligence research it is a running joke that as soon as a computer can perform any given activity the common sense definition of "(true) intelligence" will become modified to exclude that activity. Solving algorithms? Playing chess? If a computer can do them they can't be true tests of intelligence.

    It's the same with "creativity". Creativity is a gestalt or emergent "dunno" the same as "intelligence". And it's worse because people tend to demand H-creativity rather than P-creativity, which would just be personally unprecedented creativity.

    Can a computer draw? Not creative. Colour? Not creative. Solve puzzles? Not creative. Create drama and meaning? Not creative. Map inner states to outer compositions like a good expressionist? Not creative. Cut off its ear? Er...

    Good reading:

    "The Creative Mind" - Margaret Boden.
    "Fluid Concepts And Creative Analogies" - Douglas Hofstadter.
    "AARON's Code" - Pamela McCorduck.
    "Affective Computing" - Rosalind Picard.

    - Rob.

Your Reply