Brains And Other Modular Systems Prove Efficient For Artificial Intelligence
Michael Harper for redOrbit.com – Your Universe Online
As most of us know, scientists are deep thinkers. Finding the answers to “how” or “why” are not always enough for these brainiacs, who often spend as much time asking more questions as they do answering them.
“Why does this work the way it does?” “How did this come into being?”: Questions like these are what drive scientists. And whatever the answer, there’s always another question to be asked, which is ultimately a fantastic asset to the human race.
Biologists, for example, have often pondered why many living things are arranged in similar patterns. Be it brains, blood vessels, nervous systems or even fungus, biologists like Richard Darwin, Günter P Wagner and the late Stephen Jay Gould have noted that these organisms are built in a modular shape.
Now, new research from Cornell University in Ithaca, New York has found that biological networks grow this way because it’s more efficient and, in a sense, more cost effective.
With this knowledge in hand, scientists say they’ll be able to make robot brains with the complexity of animal brains–either a fascinating or frightening concept, depending on who is asked.
According to Jeff Clune, formerly a scientist at Cornell but now an assistant professor of Computer Science at the University of Wyoming, these networks naturally form with a “cost of wiring” expense built in, thereby resulting in the popular modular shape.
This modular form is described as “dense clusters of interconnected parts within a complex network.” Brains, for example, consist of central modules with a series of nerves and pathways extending into other areas. Dawkins, Jay Gould and Wagner often referred to this phenomenon as the “evolution of complexity.”
For many years, scientists assumed living things grew in this way to help them move and grow more quickly. This shape could have also helped these entities evolve and grow more quickly than if they had been shaped in another way. While this explanation answers the question of “why” modular systems work the way they do, it hasn’t really answered the question of “how.”
That sent Clune and his team to find the answer once and for all.
To do so, these researchers simulated some 25,000 generations of evolution using computer models. As they watched, they noticed that these modules only appeared whenever they left an allowance for “cost of wiring,” or gave these organisms the opportunity to create networks composed of fewer and shorter connections. According to the research, when these entities are given this option, they chose the modular shape.
“Once you add a cost for network connections, modules immediately appear. Without a cost, modules never form. The effect is quite dramatic,” explained Clune in a statement.
Not only does this research explain why this shape occurs so often in both organic organisms and man-made organisms, (such as the Internet and subway systems) it also opens the door for more intelligently designed networks.
“Being able to evolve modularity will let us create more complex, sophisticated computational brains,” said Clune.
In conclusion, the researchers said this answer to the “how” of modular systems is “by far…the simplest and most elegant” solution they’ve found.