For those of you who already understand the artificial intelligence concept of hill climbing, skip to the next paragraph. Imagine you're in a very foggy land where you can only see about ten feet an any direction, and your job is to get to the high point of the land. You can keep walking uphill until you can't anymore. But how do you know you're on the highest hill? You might have to walk downhill to get higher on another hill. Computer programs in many instances face a similar problem. It's called Hill Climbing. Take, for example, a computer program that is route planning for a trip from Atlanta to New York. It might plan to take one step closer to New York, when the actual best thing to do is to go away from New York toward the airport.
I've been giving a lot of art critisizm over the years, as well as creating art, and I find that improving on a piece of art is to face the hill climbing process. Once I was talking with some playwrights, and somebody got a suggestion to change the play in a very fundamental way. Another playwright commented that taking the suggestion would make it a different play, and we are supposed to be making comments on improving THIS play. In other words, tell her how to get to the top of THIS hill, and don't tell her to climb another.
When I work on a script, I can sometimes think of changes that are very bold. Such changes make the play I'm working on worse, but I have a feeling that eventually I can work through and climb a different hill and make a better, different play. I've got a play right now that I know is flawed, but I don't know what changes to make that wouldn't make the play worse. Well, maybe I need to remember that I'm at a local maxima (the top of a small hill) and that making the play worse might be a necessary step to making it eventually better.