Aaron, Harold Cohen's computer artist.

April 26, 2003 1:40 PM

"If what AARON is makingis not art, what is it exactly, and in what ways, other than its origin, does it differ from the "real thing?" If it is not thinking, what exactly is it doing?" Asks Harold Cohen, the inventor of AARON, the computer artist. Now, if you own a PC, you can download an Aaron Screensaver (9Mb, asks for details but worth it) and have it produce original works of art on your desktop.
posted by jamespake (49 comments total)

art, not art.

Whatever it is, it's butt ugly.
posted by delmoi at 1:50 PM on April 26, 2003

The art there is the program, not the output, if you ask me. Fascinating though, it's sort of concrete evidence of a lot of the thigns i've thought about when it comes to art.

Art school rots your brain, I'm proof.

What is it about art that makes it move people? I've done so many drawings and investigations into trying to pick visual language to pieces and I'll ive come to is that it doesn't matter, if you don't have a trust fund noones ever gonna see it anyway.

I hate you, Marcel Duchamp, I hate you. You ruined everything.
posted by atom128 at 2:09 PM on April 26, 2003

I work for Ray Kurzweil (not for KCAT, but another company in the same offices). There's a big plasma screen outside the conference room that shows this program running all the time. It can be hypnotic... especially during a boring meeting. As far as the output -- I agree that it's not the most beautiful thing, and it's missing any kind of evocative element (which, to me, is fairly important in the art/not debate).
posted by Turd Ferguson at 2:15 PM on April 26, 2003

That last link is evil Java hell on my browser, FWIW.
posted by jeremias at 2:33 PM on April 26, 2003

Come on, Mr. Cohen. You created Aaron. You gave Aaron the ability to make pictures and probably defined the parameters that Aaron uses to "independently" make art. If there is an artist here, Mr. Cohen, it is you. No amount of saying "but I don't tell it what to make" changes the fact that you made a program which makes a certain kind of art.

When my copy of MS Word decides, one morning and of its own volition, to stop processing my words and to start composing concertos, call me back and we'll talk about machines making art.
posted by Joey Michaels at 3:11 PM on April 26, 2003

I thought I'd post this because I installed the screensaver a few weeks back and I really like it - it keeps surprising me when I come back from a break. But I was also really interested in the papers by Harold Cohen (on his home page, above). He seems to be trying to get the computer to do what he does when he paints and his description of that process is fascinating from both an artistic and a programming point of view. I also think it's cool that Kurzweil has made this available - I liked their Cybernetic Poet as well. And, sorry for not warning about the Java link - sorry if it crashed your computer.
posted by jamespake at 3:15 PM on April 26, 2003

Aaron ate my 266 a couple of years ago, maybe I'll give himit another chance to behave on my 1.5GHz.
posted by hairyeyeball at 3:35 PM on April 26, 2003

> If there is an artist here, Mr. Cohen, it is you.

Exactly. Tossing in a few random functions and wacky parameters does not AI make.

Its like saying a photoshop filter is the artist when really its just an app.
posted by skallas at 3:39 PM on April 26, 2003

Come on, Mr. Joey Michaels. Your mother and farther created you. They gave you the ability to make pictures and probably defined the parameters that you use to "independently" make art. If there is an artist here, Mr Joey Michaels, it is your parents. No amount of them saying "but I don't tell him what to do with his life" changes the fact that you made a child which makes a certain kind of art.

When a rock decides, one morning and of its own volition, to stop laying around on the ground and to start composing concertos, call me back and we'll talk about objects making art.
posted by fvw at 3:46 PM on April 26, 2003

atom128 & Joey Michaels, I agree completely. AARON is art in itself in the way it makes us reflect on what constitutes art form. AARON's output may be art, although extremely mediocre (seems like a contradiction in terms), IMHO, but AARON is not the artist; the programmer is.
posted by cx at 4:08 PM on April 26, 2003

Aaron is cute. I think the question is not whether Aaron produces art. Art is a very vague term; I would just submit that if Aaron's output is art, then so is that of any fractal generator. Those have been around for ages, so the only really new thing about Aaron is the type of art it produces. And common sense tells me that at some point, somewhere long ago someone probably wrote something very similar in his basement on an 8-bit computer. One cool thing about Aaron is that he draws the images in real time, object by object; I think it would be much less impressive if it just rendered the finished images onto the screen.

The real question is whether Aaron is creative, and this, IMHO, can be answered clearly with "No". (Of course, whether all human artists are creative is another question ..) To be creative, Aaron would not only have to vary the parameters passed to his functions, but the functions themselves. And to do this effectively, it would have to know what it's doing -- awareness, knowledge, a little emotion perhaps. And Aaron is not really playing in that league.

Another approach is to rely on a human to select the functions that produce artistic output, but to have a computer produce the functions. This is what Qbist does, and it's pretty cool -- "evolutionary art".
posted by Eloquence at 4:09 PM on April 26, 2003

When a machine decides, one evening of its own volition, to stop looping through a set of movements and to start cutting off its ear taking it to a brothel and asking a prostitute to "keep this object carefully", call me back and we'll talk about machines making art.
posted by none at 4:13 PM on April 26, 2003

fvw: I call false analogy on you.
posted by Joey Michaels at 4:25 PM on April 26, 2003

I collect "bad art" at thrift stores. This is similar (as delmoi said, it's "butt ugly.") However it lacks (obviously) human soul. If I could say "repulsive" in binary code, I would.
posted by kozad at 5:02 PM on April 26, 2003

I've been working on a piece of software that generates images without human input for about seven or eight years. It creates abstract textures rather than pseudo-impressionist portraits, and I suspect its algorithms are rather simpler than the ones Aaron uses, but it has inspired some of the same questions.

I didn't go to art school, so instead of hooking my project up to a giant pen plotter and selling off its work, I built it into a little application that creates desktop wallpaper. Still, it's a computer program that designs and creates colourful, striking images that are usually pleasant and sometimes beautiful. Is it an artist? Are the images it creates pieces of art?

I've decided that the answer is no. If it were yes, then I would have to conclude that the entire planet earth is one big machine artist. After all, its geological processes churn out beautiful sweeping seashore panoramas, brilliantly coloured eroded sandstone canyons, and crisp, harsh, pristine mountain peaks. All very nice to look at, but accidents of nature nonetheless. It takes a human being looking on to say: that is beautiful, or that is interesting, or that moves me.

So the output of my Starfish program is something like a natural artifact: an inevitable - if unpredictable - result of ordinary mechanical processes. It still takes a human being looking on to say that its output is beautiful, or interesting, or evocative. It takes a creative intent to produce a creative work, and until computers achieve sentience they can have no such motivation.

Aaron has been tailored toward a different sort of graphical output, one that approximates images we are used to seeing as products of art. But it is Aaron itself that is the piece of art, and Harold Cohen the artist. I'm sure he has spent countless hours running his program, thinking about its output, coming up with ways to channel it towards the sort of work he wants it to produce while leaving it latitude to come up with ideas that surprise him. That is where the creative endeavour lies.
posted by Mars Saxman at 5:08 PM on April 26, 2003

Joey Michaels: You claim anything which merely does what it is programmed to do cannot produce art. Everything you do is just what you are programmed to do in combination with inputs, just like the computer program is, albeit slightly more complex.
posted by fvw at 5:43 PM on April 26, 2003

leave art alone you monkies.

go back to bush.
posted by Satapher at 5:44 PM on April 26, 2003

Machines think in the same way that submarines swim, fvw. Don't confuse the two.
posted by Hildago at 7:02 PM on April 26, 2003

fvw - I agree quite wholeheartedly with you. I think most people try to determine how human-like computers can be, but fail to realize how computer like humans are. Recent scientific advances in genetics seem to support the idea that we are simply a product of our programming (genetics) and our inputs (environment). Our programming might be significantly more complex than any presently available computer, but that doesn't mean it will always be that way.
Of course people are disinclined to believe this because it's somewhat unpleasant to believe that you have no real control over your life. (And it doesn't agree very easily with religion either.)
posted by Wingy at 7:03 PM on April 26, 2003

Oh, and by the way, people interested in this subject area should consider reading Kurzweil's The Age Of Spiritual Machines. It is an excellent analysis of how computers have advanced, how they are likely to continue to advance, and what distinguishes humans from computers.
posted by Wingy at 7:09 PM on April 26, 2003

I think most people try to determine how human-like computers can be, but fail to realize how computer like humans are.

Okay, I'll fall for it: humans have freedom of choice, computers don't, or if they do, it's only the freedom to choose a pre-existing option as given to them by a programmer.

This leads us into a huge can of worms, since it's easy to argue that humans beings have possibly been given a range of choices by Someone/Something/Environment/DNA/HigherPower/AddYourOwnEtcEtc that is also relatively narrow, just as computers have. While we have an awareness of what choices we make as well as their outcomes, we have also assumed that computers do not, unless programmed to, and possibly even at this point their ability to understand the outcome of their decisions can only be very limited.

...

I was going to say more but I really don't have a clue as to what I'm talking about. It's a damn complex issue, one that philosophers and scientists have been debating for years. Personally, I feel that AI will happen, but that's a genie that I can't even imagine.
posted by ashbury at 8:17 PM on April 26, 2003

Oh, and my real name is Aaron, and I would appreciate it if you would stop talking about me.
posted by ashbury at 8:18 PM on April 26, 2003

wingy, I don't think people are disagreeing with the idea that the complex organization of matter can produce intelligence and willful behavior. The point is that this program is not self motivated. It allows for a certain level of randomness, but artists are conscious of their intent and are expressing their reactions to their experiences and memories etc. This program has no experiences or memories or intentions.

Kurzweil believes we'll get to conscious and then spiritual machines within a couple decades, but he doesn't consider pre-programmed random generators to be those machines.
posted by mdn at 8:22 PM on April 26, 2003

we are simply a product of our programming (genetics) and our inputs (environment)

He he he... yes, it sounds simple, doesn't it?

Trouble is, though, that not only is the environment itself complex beyond human understanding - or so theoretical physics suggests - but how we interact with the environment and what input we derive from it is even more complex - and even though our genes are mapped, our programming changes constantly as our brain adapts to damage and wear.

The last unknown in this equation of interactive variables are perception it self. Even though we know the mechanics behind it is most likely impossible for us to perceive our perception - how the chemicals and currents in our body are transformed into what is our mind - and what poets would describe as our soul.
posted by cx at 8:26 PM on April 26, 2003

Joey Michaels, it might interest you to know that Cohen would actually agree with you. a couple of comments from the creator of Aaron(via wired):

"Most everybody else does consider it to be creative," Cohen said. "I personally do not, because I have rather stringent views on what creativity would demand. But it's considered creative enough that the president of the American Association for Artificial Intelligence cited it in his inaugural address last year as one of the only creative programs in existence."

"If the program did a drawing in August that it couldn't have done when I stopped programming it in January," he said, "then I'll consider it creative."
posted by poopy at 8:51 PM on April 26, 2003

Thank you for the link, poopy. I am sorry now that I was flippant towards Dr. Cohen on the basis of the original link.

fvw: when Aaron is able to choose to come to Metafilter and argue this for itself, I'll grudgingly agree that you have not committed a logical fallacy.
posted by Joey Michaels at 10:01 PM on April 26, 2003

More interesting is perhaps the dilemma of consciousness:
If we submit that machines can be conscious, how do we tell a conscious machine from a machine that is not conscious?
posted by spazzm at 10:02 PM on April 26, 2003

Hildago: Nice quote, but I kind of think you missed the point. Edsger Dijkstra's point was that the entire question is moot unless you strictly define thinking, at which point you either create a definition which does or doesn't include computers. However, if you don't include computers you're going to end up excluding them on some artificial principle, not on the fact that they're truly different.

Wingy: Exactly. I study AI and it's amazing how often you run into this kind of mentality, even amongst the professors. My current understanding of what most people consider intelligence includes that we should not be able to understand it :)

Ashbury: What makes you think that you have more freedom of choice than an (admittedly very complex) computer?

Mdn: Motivation? That's subjective, I could say word is motivated to making sure your documents are free of spelling errors. Motivation is a highlevel, imprecise description of an underlying process, just like "the gas wants to fill the entire space of its container".
If the program saves earlier interactions with humans and other programs and uses that in its formulas, why isn't that memory?

cx: just because we don't understand how it works on the low level doesn't mean it works any different. More than that, occam's razor suggests that unless we have any reason to believe it isn't so (which I haven't seen yet), the sanest assumption is that we are just complex computers.

Poopy: That'd be easy to make, just make the computer save a random state between runs, and all output wil have been based on all previous outputs.

For everyone who hasn't done so yet, read Gödel Escher Bach by Douglas Hofstadter. It's about a lot more than just this, but it's a terrific read, and well worth the price (which is rather high, but it's a very thick, very thoughtprovoking book)
posted by fvw at 10:04 PM on April 26, 2003

Joey Michaels: So anything/anybody incapable of defending itself on MeFi is incapable of intelligence? Congratulations, you've just put yourself in that category.
posted by fvw at 10:39 PM on April 26, 2003

hey kozad, 011100100110010101110000011101010110110001110011011010010111011001100101
posted by RubiX^3 at 10:50 PM on April 26, 2003

We've secretly replaced MetaFilter's community consciousness with Folgers Crystals. Do you think anyone will be able to tell the difference?
posted by ZachsMind at 10:50 PM on April 26, 2003

fvw: Congratulations, you've just made an obnoxiously unnecessary personal attack.

Actually.. some consider the ability to act illogically to be the mark of humanity that differentiates us from animals and other non-human things like computers.
posted by Wingy at 11:10 PM on April 26, 2003

Oh, and as for this program itself, I don't think it really produces art, because, as poopy's quote suggests, it does not progress over time. It is a static object (even if it contains random aspects.. it always obeys the same boundaries and standards). Creativity (and perhaps intelligence) require constant reassessment of boundaries and standards, and change according to those assessments.

As for AI: The day we create a program that writes programs, and it writes a program-writing program that is better than itself, we will have true AI.

Or at least that's how it seems to me.
posted by Wingy at 11:18 PM on April 26, 2003

As for AI: The day we create a program that writes programs, and it writes a program-writing program that is better than itself, we will have true AI.

But I can't create a human-creating human that creates humans that are better than myself, and I'm intelligen--

Oh.

Damn.
posted by Slithy_Tove at 11:24 PM on April 26, 2003

Wingy: By reassessment of boundaries, do you mean it should be possible to overcome any boundary? If so, humans don't count either. If not, who decides what boundaries count and which don't?

First of all, what's your definition of better in that context? Not that it really matters, because the definition doesn't really work:

1. Create a program-writing program (doesn't matter how bad)
2. create a program that'll create the program in step 1.
3. Mutilate the program in step 2 in such a way that it's worse than the one in step 1.

The problem you're hinting at is ofcourse the fact that you feel anything this program creates is already locked inside it, yet for humans this isn't so, which brings us back to the original "It can only be intelligent if I don't understand how it works" viewpoint.
posted by fvw at 11:32 PM on April 26, 2003

You claim anything which merely does what it is programmed to do cannot produce art.

Great, you've proven his claim false. What you have not done, is prove that AARON is creative. I'm interested in what your opinion is on the subject.
posted by cohappy at 11:46 PM on April 26, 2003

Edsger Dijkstra's point was that the entire question is moot unless you strictly define thinking, at which point you either create a definition which does or doesn't include computers. However, if you don't include computers you're going to end up excluding them on some artificial principle, not on the fact that they're truly different.

I figured the point was that what computers did was similar in effect but not in process to what humans do, in the same way that submarines and swimmers perform the same function in very different ways. In regards to thinking, call both methods of solving a problem intelligence if you wish (I don't mind), but clearly they work differently, so why must the limitations of one be the limitation of the other?

We know that computers cannot make decisions on their own, but I'm not aware of any test that shows that humans are not capable of doing something they have not been prompted to do. In fact, given how complicated the factors involved are (see cx's comment), I sort of doubt it's possible to devise a test that shows how what you and I do is the result of a sort of programming. In any test of that sort, you'd always get some people who didn't perform as expected -- something that does not happen in a computer program. Therefore, if the so-called programming is so complicated that it can't even be verified, what evidence is there to even say it exists?

I've got a quote around here somewhere (I've got a lot of quotes) that says what I want to say:

"It's interesting to look back through history on this one. Each age has its pinnacle of technology, and each age uses that technology as a metaphor for nature, for the universe. In ancient Greece, the technological marvels were musical instruments and the ruler and compass. The Greek philosophers tried to build an entire cosmology from number, harmony, proportion, form, and so on ­ from mathematics, basically. Remember the music of the spheres? The Pythagoreans believed that nature was a manifestation of rational mathematics. Later on the pinnacle of technology was the clockwork. Newton wanted a clockwork universe, the entire universe as a gigantic clockwork mechanism, with all the parts interlocking and ticking over with infinite precision. Then in the 19th century along came steam power, and the universe was then depicted as an enormous heat engine, or thermodynamic machine, running down toward its heat death. Today the computer is the pinnacle of technology, so it's now fashionable to talk about nature as a computational process. All of these ways of describing the world capture to a certain extent the way it is, but I would say that the universe is a universe, not merely a clockwork or a computer or whatever." -- Paul Davies

posted by Hildago at 12:11 AM on April 27, 2003

cohapy: Hey, I can only debate with one person at the time :)

I'm not claiming AARON is creative. What I'm claiming is that the answer to that question depends entirely on your definition of creative. There is however no intrinsic difference between what AARON (or a program that prints "Hello World!") does and what a human does, apart from its complexity.
posted by fvw at 12:15 AM on April 27, 2003

fvw - Ok.. I guess I was getting ahead of myself. I started arguing points related to kurzweil that haven't really come up in this discussion. Kind of a bad reflex I suppose.

What I'm really getting at is not so much the argument of intelligence but the argument of whether computers can exceed human capabilities, whether it be in art or many other fields. In order to go beyond reliance on humans for programming (and therefore serving as an extension of human capabilities) computers would need to be able to program versions of themselves that would improve over time. Otherwise, they can never function as more than an extention of human knowledge and methods. They can't develops new methods, only use human methods more effeciently and effectively.

I think that will be the biggest hump that progress of computer intelligence will need to get over.
So I guess you're somewhat right about the "understand how it works" viewpoint, but I would phrase it differently. I'd say, as long as a machines are restricted to operating in ways that we understand, they will be unable to break free of our capabilities.
But I'm not really discussing intelligence here, because it really comes down to how you define it then. And in any case, I think we'll someday understand pretty fully how human beings work, so by that standard I'd have to maintain that humans are not intelligent either.

As for boundaries, it's basicly the same question. Computer programs run based on human-defined boundaries. Even an evolving program, say an evolving chess program, judges it's improvement on factors defined by the original human programer. It only considers a move that exchanges a pawn for a queen to be a bad move because that's what it's predefined boundaries tell it, even if they do not tell it what specific moves to make. When that program has the ability to determine what boundaries will be best used for evaluating the success of a move/game/program, it will have reached a completely new level of capability.

Don't get me wrong, I don't have a programming background and I'm well aware that I'm not an expert on this topic (drawing mostly from Kurzweil's book and my own twisted brain). It's merely something I enjoy trying to wrap my head around, and I think I have a decent grasp of it at times, but it's often harder to make sense in words than thoughts. I do enjoy your well thought out arguments though.
posted by Wingy at 12:19 AM on April 27, 2003

Hidalgo - excelent quote. The fact is that the only way we can define things is through comparison to other things. Defining something as itself is universally accepted as inadequate, yet in all truthfulness, it's the only completely accurate defination possible.
posted by Wingy at 12:24 AM on April 27, 2003

Food for thought:

Inanimate objects cannot become artists. However, artists can become inanimate objects.
posted by son_of_minya at 1:34 AM on April 27, 2003

Based on what I know of recent AI research (look up pfeifer's book Understanding Intelligence for a wicked cool overview), it's looking more and more like we are able to create truly intelligent machines, but that this intelligence is limited by the simplicity of the inputs we are able to give to them, and the amount of processing power we can allot to simulating neural structures. I imagine that with some work, AARON could be transformed into a program that learns over time, and whose artwork changes accordingly. What it learns from would be the choice of the programmer, but the direction it goes with its knowledge would be, except in some trivial cases, unpredictable. Of course, our artwork is produced by the full force of a human level intelligence, and we can in no way expect such an intelligence to fit in a 9meg program running on a 1.5ghz processor with even a 200gig hard-disk. We, as people, are simply a few orders of magnitude more complex (intelligent?) than that. But in its niche, making pictures of people in art galleries, AARON is potentially capable of being an evolving artist, with a conceptual core, favorite shapes, what-have-you. I imagine his concepts would be fairly alien to us, though, as his mind and world are so much different from our own. What is profound to him may be simple to us... But he has even in his present incarnation shown that what is simple to him - the act of making these pictures - can be profound for us.
posted by kaibutsu at 5:09 AM on April 27, 2003

"Am I the artist or is Aaron the artist? I don't know how to answer that." - Harold Cohen.
posted by jamespake at 5:58 AM on April 27, 2003

10 PRINT "I THINK THEREFORE I AM"
posted by inpHilltr8r at 12:15 PM on April 27, 2003

What strikes me looking at this this is that it's much like a spider's web. Many, many people enjoy the beauty of a spider's web. Just look at many pictures of dew laced webs are out there.

The spider is working a very similar level as AARON here. It uses fairly simple rules to create it's web, as AARON does with it's "art". It's not motivated to create beauty, that's just what it does.

I don't know if you'd call a spider web art, or a termite mound architecture, but there's definitely beauty there.

Also think about artist like Escher who created art using mostly mathematical concepts and subtly changing patterns. You can't tell me there's some deep human truth to a fish turning into a bird. Sure you can give it personal meaning for you, but there's no real message being communicated. However, it's definitely a thing of beauty, and most consider it art. I'm not sure what separates this from AARON's work.
posted by betaray at 12:24 PM on April 27, 2003

Another thing to consider: the artistic savant. Adrian has an talent for art that's wasn't developed, it's just there. The author of the article claims that the art is a way of communication, but is it really more than an output given his current inputs at the time and the history of inputs he's received? He creates because it makes him happy, and what he creates is based on his environment. Just because we are naturally equipped to see the output and understand the inputs doesn't mean it's a message. I would bet that Mr. Coen could look at AARON's art and give us some understanding of the inputs, but again AARON's not trying to communicate with us.

Ahhh... all the beautiful shades of gray.
posted by betaray at 12:49 PM on April 27, 2003

What's the big argument here? All AARON does is randomly place X number of androgynous humanoids next to Y number of prickly fern plants behind Z number of bland walls, using primary and secondary colors. Hell. A kindegarten class can do that.
posted by ZachsMind at 1:35 PM on April 27, 2003

fvw: So anything/anybody incapable of defending itself on MeFi is incapable of intelligence? Congratulations, you've just put yourself in that category.

I call argumentum ad hominem on you.

My point was that AARON does not choose to make art. Dr. Cohen, on the other hand, chose to make a program that makes art. My parents chose to raise a child who would go into business. As it turns out, I went into art. Could AAROn be programmed to do one thing, but choose to do another? Absolutely not under under current circumstances. Dr. Cohen might not have any control over what AARON creates, but AARON won't suddenly "wake up" one morning and think, "hey, today I am going to be a spreadsheet instead of an artist."

Similarly, AARON cannot choose to come to Metafilter (instead of making art), read what I've written and then personally choose attack my intelligence. Dr Cohen could. You could, too. (I confess that I was following up on the wording of my original reply in order to make this point in brief instead of explaining it in lengthy detail.)

So, in my opinion and based on everything I've read (including Godel - Escher - Bach), what AARON does is intrinsically different from what a human being does. I hope that this is more clear and apologize for upsetting you.
posted by Joey Michaels at 4:49 PM on April 27, 2003

Joey Michaels: fvw's argument was not strictly logical, but it did bring up a good point. "Food for thought" and "logic" do not always have to agree. The history of philosophy, IMHO, can be cut in half by discarding such arguments over "logic." Logic, taken to the extreme, can lead to things like arguing over how many angels can dance on the head of a pin.

You were right about fvw making an illogical statement, the first time. Parents are not programmers. Programs are not human beings. The analogy was ridiculous.

From my POV, however, fvw was merely making a reactionary statement to your "rude argument" against Cohen. This whole exchange would have been better, if it wasn't personally motivated.
posted by son_of_minya at 6:59 PM on April 27, 2003