Look at the shcoking beautiful snow fakes images as bellow, are you amazed by how can the nature produce such things?
Or maybe look at amazing plants as follow, can you think how God produce this from scratch?
But the answer may be simpler than you think. Because all these are some form of complexity system.
First of all, the complexity here is not the same as algorithm complexity in computer science. It normally accompanies with some form of emergence phenomena.
Emergence means interactions of interest are non-linear, which yields levels of organization and hierarchies. – John Henny Holland
If we pay attention our life, we can find a lot of emergence phenomena around us. Like the flocking behaviour of birds, schooling of fish etc. Consider the following periodic table as an example, have you ever wonder why there are so many materials with only very few basic elements?
Further,think about the carbon-based molecules like graphite and diamond, with the same basic element, only with a different structure, the diamond is one of the hardest materials on the earth while the graphite is one of the softest one.
Many scientists have done a lot of research of such things, and showed that emergence can be actually achieved by leting the individuals follow very simple rules. So in simple terms, emergence phenomena means the whole is not only the simple summation of every individuals.
Complexity system normally have several kind of behaviour in common. Here are just four normally observed behaviours in complexity system.
We will not explain every behaviour here in details, since it will be out of the scope of this article. But just to give some high level ideas of complexity system.
Take fractals as an example, consider the Koch Curve for partitular, we can make induction of the length of the curve as the following formula:
So we can make induction to show that as n tends to infinity, the length of the curve aslo tends to infinity, although all the curve is in a bounded area. This is a extremely efficient way to organize complexity structure to make use of the space. Hence we can find enormously example of self similar or fractal structure in our daily life, coach line, surface of mountain, tree branch, broccoli, lung lobe and more and more, you name it.
All these observation inspire me to look at problems in a new perspective. Just think about human neuros, they may be also follow the principle of emergence, every neuro play with very simple rules, but hundreds billion of neuros can produce extremely phenomena which we still can not understand.
Some other clues which indicate that human brain actually does not work in an artificial neural network style include that human brain is very energy saving, although brain is actually the most expensive organ in our body, it only cost about 400-500 calories. For comparison, Google Brain project, which is said to mimic the brain of human, but actually can only do very basic things of human like image classification, use 16,000 computers, and consume about 600KW. Can you image an brain cost so much? If every brain cost so much power, emission reduction should just be a joke, since everyone will cost much much more power than a car :).
I am always amazed by how clever and ingenious God create the world. He must be a extremely good complexity system expertise.
We have a long way to go on the way to the achievement of artificial intelligence. It is a hard way, which need the combination of a lot of different disciplines like biology, physics, math, complexity system, computer science and even philosophy. I don’t believe just to make machine learning more and more accurate itself is enough to lead us through this difficult path.
m00nlight 14 October 2015