Using Swarm Intelligence with AI

In the same vein as the artificial life story I discussed on yesterday’s podcast, The Economist has an article discussing the application of swarming algorithms developed by observing ants. It is actually a pretty good primer on a field that has been around for almost two decades, focusing specifically on Dr. Marco Dorigo who was instrumental in launching the development of swarm intelligence.

In 1992 Dr Dorigo and his group began developing Ant Colony Optimisation (ACO), an algorithm that looks for solutions to a problem by simulating a group of ants wandering over an area and laying down pheromones. ACO proved good at solving travelling-salesman-type problems. Since then it has grown into a whole family of algorithms, which have been applied to many practical questions.

The idea overlaps directly with the bottom up approach to artificial intelligence. Trying to understand, model and execute all the complexity of even a simple mind ab initio is cost prohibitive. Tracking masses of simple states and equally basic rules has become much more tractable, especially with increasingly parallel capable computers. I think solving NP hard problems may be overstating things. The way swarms explore problem spaces though is clearly fruitful in terms of fast, high quality optimization and approximation. Mind as an emergent phenomenon of swarm-like, bottom up systems undoubtedly relies quite a bit on the resilience to noise and fuzziness that also characterize these systems.

If you want an excellent fiction treatment of this idea, read Cory Doctorow’s “Human Readable“. If you think trying to hash through the problems of network neutrality is hard with our traditional computers and networks, think of the same questions applied to swarm driven route optimization.

Artificial intelligence: Riders on a swarm, The Economist via Slashdot