Sunday, February 26, 2006

How making a work of art is like hill-climbing in AI

I was at a meeting of Working Title Playwrights, a playwright club I am associated with in Atlanta, and the group was giving some playwright comments about the play we'd just heard. One of the comments suggested major changes to the play. Evan said that the comment referred to a different play-- that incorporating the comment would destroy what the play was and be a completely different play. I found this idea interesting. As we make comments we need to decide whether our comment is a suggested change to the current play, or so radical a change that it's something else entirely.

While I was working on Medea: The Fury, my co-author Xiao Gong Ji asked me if it was pulitzer-prize winning material. I said no. Xiao thought that the Jason and Medea story was just fine as the basis of a pulitzer-prize-winning play, and suggested we think about how to change our play so that it was deserving of a pulitzer. I found the idea disturbing but also inspiring.

I just wanted to draw a parallel between thinking about improving a work of art as you develop it and the notion of hill-climbing in AI (See for a technical description). Basically it means that you are trying to get as high up as you can, but you can only see a few feet in front of you. So you walk upwards. The only problem is that when you can't go up any more (because you're at the top of a hill) you can not tell by this method whether or not you're at the top of the highest hill. You might be stuck at the top of a smaller hill. This is called a local maximum.

One can think of a local maximum like the best version of a particular work of art. That is, if you improve a painting or a script as much as possible, then the local maximum will be the best that it can be, while still being what it is.

I'll try to make it clearer with an example. Medea: The Fury is a play I'm pretty proud of, but I'm sure it can be improved. For example, changing every word so that the play is identical to Romeo and Juliet would make it a better play, because R&J is a better play than Medea: The Fury. See something wrong with this? I'm at another maximum, not Medea's maximum.

In hill-climbing, you sometimes have to go downhill before you can go uphill again, up another hill which might be higher. So it is with art. You sometimes have to make the painting or play or piece of music worse before it gets even better. These are choices I make constantly with my art, and there seems to be no easy way to make these decisions.

No comments: