Researchers solve biological mystery and boost artificial intelligence

January 29, 2013

By simulating 25,000 generations of evolution within computers, Cornell University engineering and robotics researchers have discovered why biological networks tend to be organized as modules – a finding that will lead to a deeper understanding of the evolution of complexity.

The new insight also will help evolve artificial intelligence, so robot brains can acquire the grace and cunning of animals.

From brains to gene regulatory networks, many biological entities are organized into modules – dense clusters of interconnected parts within a complex . For decades biologists have wanted to know why humans, bacteria and other organisms evolved in a modular fashion. Like engineers, nature builds things modularly by building and combining distinct parts, but that does not explain how such modularity evolved in the first place. Renowned biologists Richard Dawkins, Günter P. Wagner, and the late Stephen Jay Gould identified the question of modularity as central to the debate over "the evolution of complexity."

For years, the prevailing assumption was simply that modules evolved because entities that were modular could respond to change more quickly, and therefore had an adaptive advantage over their non-modular competitors. But that may not be enough to explain the origin of the phenomena.

The team discovered that evolution produces modules not because they produce more adaptable designs, but because modular designs have fewer and shorter network connections, which are costly to build and maintain. As it turned out, it was enough to include a "cost of wiring" to make evolution favor modular architectures.

This theory is detailed in "The of Modularity," published today in the by Hod Lipson, Cornell associate professor of mechanical and aerospace engineering; Jean-Baptiste Mouret, a robotics and computer science professor at Université Pierre et Marie Curie in Paris; and by Jeff Clune, a former visiting scientist at Cornell and currently an assistant professor of computer science at the University of Wyoming.

To test the theory, the researchers simulated the of networks with and without a cost for network connections.

"Once you add a cost for network connections, modules immediately appear. Without a cost, modules never form. The effect is quite dramatic," says Clune.

The results may help explain the near-universal presence of modularity in as diverse as neural networks – such as animal brains – and vascular networks, gene regulatory networks, protein-protein interaction networks, metabolic networks and even human-constructed networks such as the Internet.

"Being able to evolve modularity will let us create more complex, sophisticated computational brains," says Clune.

Says Lipson: "We've had various attempts to try to crack the modularity question in lots of different ways. This one by far is the simplest and most elegant."

Explore further: In search of the biological significance of modular structures in protein networks

More information: "The Evolutionary Origins of Modularity," Proceedings of the Royal Society B, Jan. 30, 2013.

Related Stories

'Endless Forms' uses the Web to breed 3-D printable objects

August 18, 2011

Just like generations of plants and animals evolve in nature, Cornell engineers are allowing anyone online to guide the evolution of printable, three-dimensional objects, aiming to revolutionize the design of art, architecture ...

Globalized economy more sensitive to recessions: physicists

October 18, 2010

By applying the same rules that explain how genomes evolve, Rice University physicists have shown that the world economy is more sensitive to recessionary shocks and recovers more slowly from recessions now than it did 40 ...

Learning to evolve: With a little help from my ancestors

July 30, 2007

Learning to fly is easy, if you are a bird. But why is it that birds learn so easily how to fly" It is well known that birds learn through practice, and that they gradually refine their innate ability into a finely tuned ...

The evolution of division of labor

January 24, 2012

Division of labour is not only a defining feature of human societies but is also omnipresent among the building blocks of biological organisms and is considered a major theme of evolution. Theoretical Biologists Claus Rueffler ...

Recommended for you

Mammal long thought extinct in Australia resurfaces

December 15, 2017

A crest-tailed mulgara, a small carnivorous marsupial known only from fossilised bone fragments and presumed extinct in NSW for more than century, has been discovered in Sturt National Park north-west of Tibooburra.

Finding a lethal parasite's vulnerabilities

December 15, 2017

An estimated 100 million people around the world are infected with Strongyloides stercoralis, a parasitic nematode, yet it's likely that many don't know it. The infection can persist for years, usually only causing mild symptoms. ...


Adjust slider to filter visible comments by rank

Display comments: newest first

Whydening Gyre
1 / 5 (5) Jan 29, 2013
Jan 29, 2013
This comment has been removed by a moderator.
1.7 / 5 (6) Jan 29, 2013
"simulating generations of evolution"....this is so unscientific it makes me sick. Any clues as to how they achieved these most improbable events? Did they take into consideration degration as a part of evolution?

It's hogwash.

not rated yet Jan 29, 2013
Great, just what we need, cunning unkillable robots who have no use for humanity. Singularity here we come! :)
1 / 5 (6) Jan 29, 2013
Once you add a cost for network connections, modules immediately appear. Without a cost, modules never form. The effect is quite dramatic..
IMO it's just a rather trivial example of spontaneous symmetry breaking, which happens during cooling of system of hot atoms for example. At the moment, when the exchange of energy between particles takes more time, than the atoms can afford, the formation of nested structures (molecules, density fluctuations, droplets and crystals or even more complex structures) will occur. It means, in poor energy conditions the evolution of life goes through speciation, organels and sexual dimorphism. The gradual evolution of these structures is the consequence of situation, when the conditions for terrestrial life are getting worse - usually as a consequence of the gradual depletion of resources, but cataclysmic changes may play its role too. Every sudden change makes the situation for existing organisms (which are adopted to current situation) worse.
1 / 5 (5) Jan 29, 2013
In analogy with well known situation, until the Internet connection remains good, the people have no needs to form temporary file caches and storage servers. But when the connectivity gets worse, the people will build the file caches, which will reduce the network traffic between remote servers at distance. Analogously the level 1 and level 2 caches were implemented for computer processors, when their increasing computational power made the throughput of bus a bottleneck of further increasing of computer speed.
1 / 5 (4) Jan 29, 2013
As another analogy may serve the monetary evolution. At the very beginning no money existed at all - just the exchange of various kinds of goods. The goods in primitive communism society served both like subject of trade, both like subject of exchange. During evolution of society in Sumerian empires 3500 - 3000 B.C. the low speed of this exchange did become a brake of the further evolution, so that new independent structure: money were introduced. Another degree of condensation has lead into evolution of more complex financial structures (stock exchange derivatives, swaps, forward rate agreements, etc.). From this perspective, the monetary economy evolution is just process of increasing modularization of value exchange. Another example is the modularization of programming languages at the moment, when the laboriousness and complexity of their programming did become a brake of the software development.
4.6 / 5 (9) Jan 29, 2013
"simulating generations of evolution"....this is so unscientific it makes me sick. Any clues as to how they achieved these most improbable events? Did they take into consideration degration as a part of evolution?

It's hogwash.

It is hogwash because you are ignorant. Evolution inside numerical simulations has been a standard stock in trade for computer scientists for decades. The software code is quite simple, capable of being created by many high school students. In fact any college worth its salt teaches courses in simulated evolution in their software departments.
4.2 / 5 (10) Jan 30, 2013
It's hogwash.

They're using a genetic algorithm, which is exactly what they say:
"simulating generations of evolution"

It takes the current state, mutates it in various ways (for a NN probably by adjusting signal strengths and interconnections between neurons with a pseudo-random function) and selects from the resulting population the one that is best at solving a target function.
That one is then used as the seed state for the next round of mutation.
Rinse. Repeat.

It's a rather elegant (and simple) way of getting at a solution without hard-coding it.

If you want an example of what this can look like go here (where a genetic algorithm slowly but surely designs a car to go over a predefined terrain):

You'll notice that the 'cars' look pretty sucky at first. But if you let it run a couple of minuteshours it comes up with pretty good (and sometimes surprising) solutions.
5 / 5 (3) Jan 30, 2013
@AP - nice link, bookmarked cheers!
1 / 5 (1) Jan 30, 2013
This is amazing. Before this work, neither I nor my peers ever thought about making things modular! This will clearly solve the problem of strong AI.

I shall prepare to change fields forthwith.

Thank you evolutionary simulator dudes who put in a cost for wiring ... and got out ... less wiring. Is there no limit to what these dudes can do??
not rated yet Jan 30, 2013
So simple, nobody ever did it before
not rated yet Feb 03, 2013
OK but in nature I guess network connections indeed have a cost. But in a computer, do they? I mean, the modular structure might be optimal in nature but not in artificial system, in which case there is no need to try to reproduce it, is there?
1 / 5 (2) Feb 04, 2013
I mean, the modular structure might be optimal in nature but not in artificial system, in which case there is no need to try to reproduce it, is there?

Since I program NNs at home I might give an answer to this: There is a cost. If you want even a semi-serious NN implementation you have got to go (massively) parallel. And it costs a lot more (in terms of time) to send information accross processes (threads, multi-processors, and most if you have linked several, physical machines which have to communicate via LAN).

So lumping functional stuff together makes sense.

My implementation runs on a single multi-processor machine with hyperthreading (4 processors, 8 hyperthreads). On loading the net I distribute the neurons accross the threads so that the number of cross-thread connections is mimimized. This implicitly gives me local clustering of closely connected neurons.

Please sign in to add a comment. Registration is free, and takes less than a minute. Read more

Click here to reset your password.
Sign in to get notified via email when new comments are made.