Friday, January 22, 2010

Adaptive systems

Interesting post here:

In a Japanese laboratory, a group of scientists is encouraging a rapidly expanding amoeba-like blob to consume Tokyo. Thankfully, the blob in question is a "slime mould" just around 20cm wide, and "Tokyo" is represented by a series of oat flakes dotted about a large plastic dish. It's all part of a study on better network design through biological principles. Despite growing of its own accord with no plan in mind, the mould has rapidly produced a web of slimy tubes that look a lot like Tokyo's actual railway network.
The slimy tubes work independently of one another, it seems:
The mould's abilities are a wonder of self-optimisation. It has no sense of forward-planning, no overhead maps or intelligence to guide its moves. It creates an efficient network by laying down plasmodia indiscriminately, strengthening whatever works and cutting back on whatever doesn't. The approach seems as haphazard as a human planner putting railway tracks everywhere, and then removing the ones that aren't performing well. Nonetheless, the slime mould's methods (or lack thereof) produced a network with comparable cost, efficiency and tolerance for faults to the planned human attempt.
The lesson here is obvious: complex adaptive systems are as efficient as planning when it comes to relatively simple problems like a city's transport routes. When the complexity of the problem is increased, complex adaptive systems continue to function, as we know because we can buy cheap pencils. There's no example of similar success in the area of planning, that I can think of.

Why did I choose pencils as an example? Over to you, Uncle Milt:

1 comment:

Unknown said...

Are we are sure about this?

Could it not be that the present network is just such an example of adaptive networks?

We are looking at an organism that is adapting many times faster than we adapt, I would be interested in seeing the evolution of both networks adjusted for time - I have a hunch that they might look very similar.