Optimization by experimentation

A recent article (1) describes how honeybees optimized architectural aspects of their nests. By mixing five sided and seven sided shapes with hexagonal cells, they appear to have optimized the construction. This is further proof that repeated experiments are key to solving complex mathematical problems, those with no apparent closed form solutions.

Artificial Intelligence, thus, has significant potential to solve vexing optimization problems if computing constraints can be removed. It is likely a different way of thinking, away from the pure world of mathematics and equations. It is a much more flexible and adaptive way of solving problems. Applications are everywhere including medicine, engineering and economics. Optimization of systems with many non-linear parameters are mathematically challenging but in a regime that offers nearly infinite computing power at zero cost, there will be few problems that cannot be solved by pure random experiments. Guided search of the design space will accelerate solutions if the gathered intelligence is extensible. The trick here is not to fall prey to known heuristics but to let the machine wander where no human brain cell has ever been.

Humans are entering a world of zero cost experimentation, one that may allow them to go places they have not yet imagined.

(1) How geometry solves architectural problems for bees and wasps (sciencenews.org)