Eric Bonabeau on Biological Computing

Eric Bonabeau is Chairman and Chief Scientific Officer at Icosystem. He apparently used to work for Microsoft Research. He's talking about social insects and what studying them can teach computer science. The chief question is "how do we shape emergence?" Less succinctly, "how do we define individual behavior and interactions to produce desired emergent patterns?"

His first lesson is "the whole is more than the sum of its parts." Eric spoke for a while about how Ants learn paths to food sources via pheromones and even find the shortest part in a robust way because pheromones evaporate. He's now showing how that can be used to solve the traveling salesman problem. This is biologically inspired since it makes several important departures from true ant behavior. The solution method compares well to other optimization techniques. I found a paper by Dorigo and Gambardella on this topic. Its being commercially applied by companies like Unilever.

The second lesson is "simple rules rule." Eric is describing how ants create bucket brigades organized by size with larger, more efficient ants closer to the nest. Simulation shows that this is the optimal way to organize labor. CVS Pharmacies was able to apply this rule to their distribution centers and saw a 34% increase in productivity.

The third lesson is "no one needs to be in control." He uses the example of collective transport where ants collectively try to carry something too large for any single one. They all push in random directions and make no progress until by chance more push in one direction and get them all going in a single direction. Another example is nest contruction in wasps and termites. He built a model based on a simple rule of brcks being deposited based on some stimulation in the environment. The only form of global communication is the structure itself.

The fourth lesson is "size matters." In small colonies, workers exhibit polyvalence, meaning all workers do all tasks. In larger colonies, specialization occurs. This is true, of course, in human organizations as well. Some organizations don't practice it well and that hurts their efficiency in significant ways. The State IT organization comes to mind.