Collaborative Evolution: The Mona Lisa in Canvas (peterbraden.couchone.com)

I really like this because I can see the evolution taking place, including all the strange mutations. Reading synopses of genetic algorithms is often somewhat boring: "after 2000 generations, I had a fitness ranking of 54, but after another 3000 generations it had settled around 56!" The frustration boils down to the fact that they lack a story of how the algorithm arrived at its answer. It just has an answer, and it happens to work, and it's inscrutable. It's cool, because it works, but its opacity makes it a poor spectator sport.

Here, I get to see an entire story. I see innovation, progress, bold new ideas be attempted and fail. Of course, the story is all in my head, but I feel like I can see how it works, and what it's trying to do. My browser can't handle enough generations, but maybe after watching it more I could see what it's good at and what it struggles. Never underestimate the power of visualizations.

(aside: link got borked?)

oops, couchone has gone down. The source is on github at https://github.com/peterbraden/genetic-lisa

Surely there's a more interesting way to do this than random variations and picking the best. Why not constrain the color of the circle to the one that maximizes your objective function given its position? That isn't hard to compute.