Really interesting interview with Will Wright at The Onion’s AV Club (no, it’s not a parody.)
I realized at the end of SimAnt that the simulation we built for the ants was almost more intelligent than for the guy, because the guy was being done using traditional programming, whereas the ants were done using this distributed environmental intelligence of the pheromone trails. I began wondering, could we build a more robust simulation of human behavior if we adopted this ant model, where we distribute the intelligence not through the agents, but through the environment?
(Via Robin.)