91av

Read more: “Seven technologies to disrupt the next decade“

We are on the cusp of a new era in the history of invention. That’s the implication of software that can automatically “evolve” technology, and create designs that often no human would come up with. It’s already transforming fields as diverse as robot locomotion, computer security and drug design.

Genetic algorithms mimic natural selection by describing a design as if it were a genome constructed from segments. Each segment describes a parameter of the invention, varying from its shape, say, to much finer grained aspects, such as electrical resistance or a chemical’s molecular affinities. By randomly changing some segments – or “mutating” them – the algorithm improves the design. The best results are then bred together to improve things further.

Until now your average desktop computer didn’t have the processing power to crunch through millions of generations and chuck out the undesirable mutants. That’s now changed, so genetic algorithms are poised to have a profound effect on R&D, says John Koza at Stanford University in California, who has pioneered their use in engineering design. He has “bred” designs for efficient radio antennas this way. What’s really interesting, he says, is that it is not always clear why the evolved invention works: no human would have come up with his antenna’s weird, zigzag shape. In addition, software can be set to design around existing patents.

Evolved invention is catching on in all sorts of fields, says Robert Plotkin, whose 2009 book Genie in the Machine tracked the rise of the technique. Drug discovery firms are becoming big users, for instance, evolving new molecular mechanisms to reach receptors no human would have thought of. The technique has also been used to improve the walking gaits of robots. “Most of the evolved inventions are not necessarily dramatic – but they are producing a steady stream of improvements,” says Hod Lipson, a roboticist at Carnegie Mellon University in Pittsburgh, Pennsylvania. “The big picture is that they are having a profound cumulative effect in accelerating innovation.”

Still, don’t expect to hear inventors admitting they use genetic algorithms anytime soon: they will likely still take the credit for the work, says Lipson.