Minds of their own

Font Size

Print

Feedback

Email

Discuss

: to crunch the data.

This is not just a clever party trick. Such experiments are essential for understanding how the brain stores specific pieces of data—a crucial first step for helping people with degenerative disorders such as Alzheimer’s and Parkinson’s diseases.

Throughout history, engineers have spent their lives inventing machines that were faster, stronger, more reliable or capable of greater precision than human beings. Whether they were Jacquard looms, combine harvesters or CAD-CAM gear, they were tools for amplifying some human skill or compensating for a weakness. But always they needed human intelligence to function.

That’s now changed. Neuroengineers build tools that think for themselves, making decisions the way humans do.

What if the machines acquired too much of a mind of their own? In the search for solutions, even a modest PC can manipulate data 10 million times faster than the human brain. Admittedly, humans can take certain heuristic short-cuts that save time, but it’s now over a decade since Deep Blue, an IBM supercomputer, defeated Garry Kasparov at chess. It did so, of course, by using its brute-force processing to predict, by trial and error, the course of a game up to 30 moves ahead; and then to compute which of the millions of possible moves would strengthen its own position best. What it could not do was devise its own strategies for playing a winning game.

There are machines around today, however, that can do just that. Over the past decade, a new technology known as ‘evolvable hardware’ has emerged. Like traditional brute-force methods, evolvable machines try billions of different possibilities. But the difference is they then continually crop and refine their search algorithm—the sequence of logical steps they take to find a solution.

To do so, they rely on so-called ‘genetic algorithms’, which use trial-and-error learning to mimic natural selection. With each run of the genetic algorithm, the highest-scoring solutions are retained as ‘parents’ for the next generation. Offspring solutions are created by swapping out portions of the parents’ blueprints, or by introducing some element of randomness to stir things up a bit—as happens in nature.