Future of Computing: Inspiration from Nature

The intersection of biology and computer science is pushing computation beyond its traditional limits—forget algorithms think evolution. By Dennis Shasha DOI: 10.1145/2090276.2090289

Three years ago, Cathy Lazere and I began a study of computing on the edge of the possible [ 1]. We met scientists who described ways of controlling spacecraft from illions of miles away, embedding intelligence in smart bacteria, or building computers to run as fast as a million desktops combined. They work on the most
challenging applications in science, engineering, and even finance. We expected new computer
architectures and a variety of new software techniques. But we didn’t expect the common vision
that has emerged across all of these fields: The future of computing is a synthesis with nature.

Nature and biological thinking have inspired new ways to do computing. For example, computers control spacecraft throughout flight and after landing. Manual repair of onboard hardware, once in flight, is impractical to the point of impossibility; but innovative spacecraft engineers propose designing machines that will repair themselves. If you think back to your first programming course, you will appreciate what a change in thinking this represents. Instead of ensuring that every case is covered in your software design, you would design an intelligent machine that will adapt itself to possibilities that you don’t know about and sometimes can’t even imagine.

In the process the concept of “algorithm”—a recipe for arriving at a correct answer to a question in guaranteed time on well-performing hardware—loses its central place. For one thing, the question may not be known or may be vague (survive and do your job). For another, the hardware may be damaged. Instead of an algorithm, there is an approach evolutionary in inspiration (try, mutate, and evaluate, and repeat) and heuristic in guarantee (it will often work). The robots that carry out this approach will try to repair themselves and everything around them by what amounts to trial and error, but at electronic speed [ 2, 3].

Can such an approach work in the real world? Does nuclear reactor design sound real enough [ 4]? One of the people we interviewed was Louis Qualls who works at Oak Ridge National Laboratory. While much of what he does is classified, the approach and experiences are revealing. Qualls recalls designing a container for a spacecraft that also had to shield the spacecraft against gamma rays. The question was where to put the shielding material. There were effectively an infinite number of ways to make layers out of different materials. “The genetic algorithm said that putting the gamma shield in a thin layer in the middle yields the cheapest and lightest-weight solution,” Qualls told us. The shielding expert confirmed that this design could work. The genetic algorithm had discovered the solution by exploring thousands of possible designs.

Surprisingly, evolutionary algorithms contribute to the social health of human groups. Before he began using genetic algorithms, Qualls would take three weeks to come up with a preliminary design. Inevitably his design entailed compromises among design elements “and I made everybody real mad at me because nobody got what they wanted.” But when a genetic algorithm churns out 100 designs, the specialists begin to see a pattern of compromises.

People’s attitudes improve. “The algorithm helps different subsystem experts understand the constraints of others,” says Qualls. As a result, there’s more co-operation among team members.

Further, unlike human designers, the genetic algorithm is willing to try very different designs from previous ones it has tried. The algorithm has no loyalty to historical successes. If you doubt that humans preserve their biases when technology changes, just remember how many years passed in the early 1900s before cars no longer looked like “horseless carriages.” Consider also that many recent business and technological innovations— express mail, microprocessors, and the Web—arose from people outside of the establishment who found better ways to perform existing services. The establishment was stuck in its initial design frame of mind.

Other people we have spoken to, who work in fields as far away as finance and space-exploring robots, mirror these experiences. When you need to explore an enormous and sometimes unknown search space, genetic algorithms along with a variety of other heuristic techniques (such as simulated annealing or gradient descent) can often find great designs or enable machines to repair themselves.