Follow @ryderskull88
For the past few thousand years, engineers assumed that the best way to move water round a right angle was as you’d probably expect – just round a bend in the shape of a quarter of a circle. It makes sense; the water pressure is maintained and you’d expect it to flow perfectly round. The engineers did well. They were intuitive, and used the best configuration they knew of. When the first piped water flowed round the fist-ever bend in the first-ever pipe and nothing awful happened, I doubt anyone breathed a sigh of relief. It makes perfect sense it would flow around, and it did, and that’s the way the system stayed until very recently.
Turns out, a fraction of the circle is not the best way to do it.
In the 70s, work by Ingo Rechenberg and his colleagues at the Technical University of Berlin in the field of evolutionary computation showed that, in fact, water flows faster round a bend if allowed to “rest” slightly before moving through. For thousands of years, we’d been doing it wrong.
The work done at the Technical University developed into the science of evolutionary computation, where evolution strategy is used to optimise technical problems. To explain this, let’s use the water in the pipe example. In this model, various pipes of different shapes and sizes, all moving water at a right angle, are randomly modelled on a computer. We can imagine a vast plethora of differently shaped, weird pipes; some running up, down, and over each other, some tying a knot around themselves, and some widening and narrowing at various points around the bend.
The computer that is modelling all these weird and wonderful creations can now calculate the best few pipes out of the set – in evolutionary terms, the fittest – and save them for future use. The others are deleted. The surviving pipes are then “bred” by mixing the coding of the pipes together. The resulting new set of pipes is again analysed to find the most efficient water carriers, and again the insufficient pipes are discarded. This process is continued over some time until finally the best, most efficient pipe model is found.
This process of evolutionary computation can be applied to almost any mechanical problem you care to shake a stick at. Want the most energy-efficient way to fly 400 people around the word? Build an algorithm that takes into account their weight, air resistance and other problematic factors, and see what your evolutionary model builds. In all likelihood, it’ll look quite a lot like a very big albatross.
The limitations of this type of modelling stems from our lack of understanding and our technological limits.
There is no doubt in my mind that the aeroplanes of the future will be feathery. The only reason they aren’t already is because we lack the necessary technology to build a giant bird.
Similarly, the most efficient and powerful computer that scientists currently know of sits on top of your shoulders. Admittedly, you can’t square 18854648 correctly in a fraction of a second, but you can, for the most part, calculate the exact muscle movement necessary to catch a ball heading towards you at about 30 miles per hour. Which is something most computers have trouble with. Sadly, it’s logically impossible to expect a less powerful computer to model a more powerful one – it would be like asking a Nintendo to run Windows 8 without any problems. So we can’t prove that the brain is the best fit for the job. But evidence seems to suggest so.
Evolutionary algorithms are one of the most powerful tools available to the human race as far as engineering is concerned – it makes perfect sense that the method used to make the best computer in the world is will be equally good at making the best machines in the world.
Featured images:
License: Creative Commons image source
License: Image author owned
Sergeo Lee is the founder of the tech and media startup Edictive.com were he pioneered a technology platform for optimizing film production schedules. Now he is kindly sharing an article with us Evolutionary Algorithms and technology.
