Will Evolutionary algorithms be used in next-gen electronics?

Evolutionary Algorithms (EA) mimic the processes of natural selection and random mutation by breeding, selecting the best and then rebreeding. This process continues till the system finds the best offspring and then uses it as the final design.

evolutionary algorithms

To further explain the technique take an example where designers have two designs of an antenna, they will use these designs to develop more designs that borrow characteristics from both. The offsprings are tested. Most of these offspring fail to work, but the survivors are kept for further analysis, which include testing thousands or sometimes even more designs. The final result is an antenna that is better than any of its ancestors.

The technique is useful in technologies in which the scientific foundations are not very firm and detailed. For this technique researchers have to make use of ultra-fast computers that can process trillions of instructions per second since getting to the best offspring requires the system to simulate billions of results and then give the best one out of them. This technique is now being helped by some ultrafast chips by Intel and AMD. These chips contain more than one core that allows programmers to use each core for unique tasks.

Proponents of EAs feel that they can use this technique to replace traditional methods in many fields from designing new types of optical fiber and USB memory sticks, but critics feel the other way. They are of the opinion that designers will not be able to know much about the final offspring and won't know its trade offs or the places which are critical and need more attention in the design phase. These critics also argue that since no human understands the designs completely its failure recovery can lead to major problems.

The research is finally stuck in two stages, one wants the design to work and other wants to know how it works. Getting an answer to the first question is no problem, but the second question means that the electronics devised with the process can become too vulnerable to repair.

Image

Via: New Scientist Tech