June 17, 2003

Harold Cohen on artist programmers

by Andrew Stern · , 2:46 pm

As we have been discussing artist programmers and meaning machines on grandtextauto, I sent an email to Harold Cohen, creator of AARON, asking if he’d like to share his thoughts on the topic. To my delight, he wrote back with the following comments.

Harold Cohen:

I wrote my first program early in 1969, at which point, I’m sure you must realize, the option of using an existing package as opposed to writing your own program didn’t exist — there weren’t any packages. If there had been I suspect I’d never have thought computing had anything to offer me.

That reflection leads me to one rather obvious comment; I don’t see anyone saying why they got involved in computing, what they wanted from it. And in the absence of any driving personal need, questions about whether one needs to program or not seem very arbitrary.

In my own case — to make my point more clearly — I came to computing after a reasonably successful twenty-year career as a painter, at a stage in which I was feeling increasingly frustrated by my own lack of understanding of how painting “works;” what exactly is happening when I make some marks and other people, whom I’ve never met and know nothing about, claim that they know what the marks mean, what I intended by them. My need wasn’t to make art — I knew very well how to do that — but to understand what art is. You might guess how strongly-felt that need was if you know anything about what computing was like thirty-five years ago and how lunatic is was for an artist to become involved.

Given why I became involved, it seems clear that using a package, if any had existed, wouldn’t have done it for me, any more than using a paintbox helps you to understand painting. I had to write my own programs, and I had deliberately to make them non-interactive; because if I had allowed myself to intervene in the process I would be doing essentially what I’d been doing all along, and could expect no more understanding as a result.

It is nonsense (left-brain, right-brain bla-bla) to say that programing is too difficult for artists to learn . Anyone can learn anything if they really need to be able to do it. And if the need is powerful enough it really doesn’t matter how long it takes. It’s a sight harder to learn programming when you’re forty, as I was, than when you’re twenty — and it was a sight harder thirty-five years ago than it is today.

If there is no very powerful need, on the other hand, how can you know whether you should learn or you shouldn’t? I’ve taught for long enough — both painting and programming — to know that the desire to be where the action is — or where it appears to be — all too frequently overwhelms the newly-emerging glimmer of personal need students might otherwise be better able to recognise and to nurture. If their teachers can’t or don’t program — and few of them can or do — how likely is any serious discourse leading to informed choices about how to proceed?

Why programming? It seems to me that the only unequivocal argument for programming comes up when you know there’s something you need to find out, but you don’t know just what it is or what it should look like. But isn’t that the normal condition of art? Isn’t that what distinguishes art from academic art?

And isn’t that what makes the most significant distinction between programming languages — Lisp, C, even Java — and the various packages cited in your discussion that claim to be programming languages? Photoshop is a quite remarkable package (I use it a good deal, though never for making art) but a language allows near-infinite expressivity. Photoshop doesn’t, though it’s complex enough to create the illusion that anything is possible. Anything? Try asking it to decide what colors to use.

Packages are written by programmers who are obliged to make assumptions about what their target audience wants. When the target field is pretty stable and it’s pretty clear what the audience wants — 3D modelling for industrial designers, animation for film-makers, rendering for the academic artist — those assumptions may be reasonably well-informed. When it isn’t clear, either because the programmers still believe what they were told (about art?) in high school or because the audience and what it wants are undetermined, they’re not likely to be well-informed.

Not that it matters, fundamentally. Well-informed or not, all tools tend to do what they were designed to do, and they were all designed to do something. (You could figure out how to use a pistol as a flower-pot if you really wanted to, but mostly pistols tend to be used for shooting people; which is what they were designed for.) That isn’t to say that there isn’t space at the edges, in a package as complex as Photoshop, to do some things its programmers never imagined anyone ever wanting to do. But the space is comparatively small and being able to explore it — a need that only arises if you’ve already signed on — still rests upon what’s already in the package. If it’s not in the package, you’re out of luck.

Aren’t true programming languages equally limiting? No: they are variously limiting, but not equally limiting. Thus, for example, I tried for a couple of years to figure out how to get my own program, written at that time in C, to do its own coloring. I concluded finally that C simply lacked the expressivity to allow an adequate representation of anything as abstract as color. And it was only after re-writing the entire program in Common Lisp — CLOS — that I was able finally to see how I might solve the problem. That variability in expressivity is something we find in human languages also — there are good reasons why Eskimo languages have more ways of describing snow than English has, for example, and equally good reasons, probably, why they haven’t generated an Eskimo Shakespeare — but we’re a very long way here from the narrow confines set by any programming package.

From these perspectives, then, I find myself wondering, inevitably, what those of your discussants who advocate programming for artists are actually doing with the discipline. It would be too silly, wouldn’t it, to try replicating a small part of what Photoshop does so well, just to prove you can? That would be like bowing to a new version of the tiresome old anti-modernist argument that it’s ok to do Picassos so long as you could do the real stuff first. In the final analysis the proof of the (programming) pudding is in the eating (hm…) of the output. How come I don’t see any output offered as existence proof for or against the discussant’s positions?

In the final analysis, also, the argument for or against programming has to go deeper than the personal needs of the individual. Personal needs develop within the cultural context and the context is changing today with astonishing — and escalating — rapidity. There is no law that says the individual artist is obliged to buy into a view of the future largely dominated by commercial-technological interests. I don’t see anything absurdly passe about painting and I don’t see anything particularly compelling about using cellphones as an art medium. But if the shape of the future figures at all as a determinant to your personal needs, then you have a choice to make. If all you want is to be current, then the latest package will do fine. Just remember that there are solid reasons why oil-painting technology has served the artist for half a millenium, while most new technologies don’t make it through a single New York season. The staying power of the most popular current packages has yet to be demonstrated; I still don’t know of a single major artist who has adopted Photoshop as his/her medium of choice, for example. So if you aren’t committed to chasing the latest package — and most individuals seem to think the package they first acquired is the one-and-only — there’s a better-than-even chance that you will soon be more passe than the guys who stuck with painting.

If, on the other hand, you see yourself as part of the effort to define the future, then I don’t see that you have any choice but to program your way, in the dark, through an unknown and unknowable landscape. For what it’s worth, that pretty much describes my own work over the past thrty-five years. I began with the notion that a program could do some part of what a human artist could do. My program still functions more as a highly intelligent and talented assistant than as an independent, fully-autonomous artist. I don’t know how far I can go beyond that point — the landscape is still dark and unknown — yet I have very little doubt that the future will include autonomous programs that function as artists.

Existence proof? The two attached images, both large prints made by my program, AARON, hung in a print show last summer in the San Diego Art Museum. In case isn’t clear, let me explain that my regular habit is to let the program run overnight and decide which of thirty or forty images to print the following morning. Most of them will be emminently useable, but it’s too expensive and too time consuming to print them all.

The labels in the exhibition said the usual: “Digital Print by Harold Cohen.” I called the curator and said, no, they are digital prints by AARON, a computer program written by Harold Cohen. The label was duly changed. But the bottom line still read: “Courtesy of the Artist.”