Swarm Intelligence

elamdaly writes "Eric Bonabeau, Ph.D, a keynote speaker at the upcoming Emerging Technology conference, is a leader in the field of swarm intelligence and has focused on applying these concepts to real world problems such as factory scheduling and telecommunications routing. The concept itself is borrowed from nature; in this interview, that's where the conversation begins, with ants and other social insects. Dr. Bonabeau takes us from his childhood nightmares of carnivorous wasps to applying the theories of swarm intelligence to solving real problems in the business world."

We

We posted first!

We do this all the time

"....applying the theories of swarm intelligence to solving real problems in the business world."

We've found it works best when we all rush the intern at the same time. Down 'e goes! Ha! Whose nephew are you NOW?

Get his book

I recommend this book on Swarm Intelligence. It was written by experts in several different fields and is quite good.

From Amazon [amazon.com]

Michael Crichton - Prey

Or, for a lighter read, try Prey [amazon.com] by Michael Crichton [amazon.com]. Excellent novel, though not quite as good as some of his previous work (Timeline, anyone?).

Emergence

The arising of complex behaviour from simple agents is also known as Emergence. It is a subject I have recently begun to study and it deals with the amazing structures, methods of information storage etc which arise in complex networks. Very little of this emerging behaviour can actually be predicted if one is to only examine the behaviour of the individual agents. The reason I bring this up is to plug a book (a popular pastime here apparently!). Its called 'Emergence' by Steven Johnson and it is this book which introduced me to Slashdot. Did you know that slashdot's rating system allows it to act as a forum and knowledge repository on a large scale, without suffering from the needless wastes of spam which ruin other similar forums. Its a difficult topic to explain but the author devotes an entire chapter to Slashdot and how its design encorporates emergant factors to allow its success. Interesting stuff, and a fascinating book too! Will

in related news...

all sugar-processing plants saw a huge increase in network traffic

swarming behaviour

It's already being used in financial models. Explains everything from the dot-bomb crash to "tomorrow will, 2 out of 3 times, be like today"

Ever since Adam Smith.

It's already being used in financial models. Explains everything from the dot-bomb crash to "tomorrow will, 2 out of 3 times, be like today"

Not just that:

Adam Smith's "Invisible Hand" was exactly swarm intelligence emergent behavior: A large number of humans applying simple rules of self-interest organize large movements of capital goods into the production of more-desirable products by sending each other simple price signals.

Simiar comparisons might be made to the success of voluntary vs. totalitarian governmental systems, the free software movement, the explosion of network applications, and a number of other "less control gives better results" situations.

I think Dr. Bonabeau might find it useful, when trying to sell his ideas to administrators, to bring up these comparisons. Successful administrators and decision-makers already have a solid understanding of these concepts, so speaking in these terms should be immediately accessable.

Imagine going to a non-pointy-haired business exec or a conservative politician: Will he more quickly grasp an argument couched in terms of ants, or in terms of free markets?

Heck: Even central-control systems (such as bureaucracies and military staff command) work by giving only broad directions and letting the subordinates use their own intelligence (and local incentive structures) to work out the details. A fundamental lesson in officer training works as follows:

- Instructor gives the new second louies a platoon, a sergant, and a tent. Tells 'em to try to direct the men to pitch the tent.

- Each second louie tries to micro-manage the tent-pitching, with disastrous results.

- Then the instructor shows 'em how it's done:

"Sergant! (points to spot on ground) I want that tent pitched HERE!"

Then he goes away and lets the sergant handle it.

(Of course the sergant, in turn, passes the order on with only slight elaboration, maybe assigning labor division or providing feedback if somebody's slacking or screwing up. But mostly he lets the men, in turn, apply their own brains and brawn to doing their own pieces of the tent-raising.)

The man's a genius...

EB: My experience trying to "sell" the concepts of swarm intelligence to the commercial world is that managers would rather live with a problem they can't solve than with a solution they don't fully understand or control.

Guess he won't be giving the RIAA a call anytime soon, eh?

Hmm, insects.. a blueprint for a deadly worm?

Reading the article, I was just thinking how deadly a worm that was based on how a colony of wasps or ants would behave. Considering wasps and ant's don't have extremely complex brains, all someone with malicious intent would do is to give it a basic behavior and how to interact with other worms it might encounter and how to share information with one another. I have a feeling I'm completely out of date here, I'm not too up to date with the worms of today and even yesterday, and perhaps something like this has been set loose and/or killed.

Prey

Shouldn't Michael Crichton have given this address? His novel, Prey, did a better job explaining this.

ProfQuotes [profquotes.com]

Proverbs 6:6

"Go to the ant, thou sluggard; consider her ways, and be wise."

Considering how that's been around for thousands of years, interesting that no one's really done much about it until now. Maybe no one thinks they're a sluggard. ;-)

Re:Proverbs 6:6

Its not that people haven't been trying to emulate the behavior of insect swarms such as ants, beetles or bees for thousands of years its just that, like most current problems in science, the technology is just now matching up to the complexity of the problem.

The mathematical techniques are just being formed to handle these types of problems, based mainly on the numerical research that has been done in recent years.

So, I would say its more interesting that modern science is now capable to actually be wise from considering the ants ways, rather than someone conjecturing about being wise by thinking about the ants ways.

P.S. Proverbs havent been around for 'thousands of years', more like 16 to 17 hundred.

Prior Art

Hopefully this 3000+ year old prior art will prevent somebody from taking out a stupid patent!

Re:Proverbs 6:6

"Go to the ant, thou sluggard; consider her ways, and be wise."

Considering how that's been around for thousands of years, interesting that no one's really done much about it until now. Maybe no one thinks they're a sluggard. ;-)

Actually there is a long and fascinating history of research into swarm/colony intelligence in ants, from the groundbreaking work of EO Wilson [amazon.com] to the more recent work of Deborah Gordon [amazon.com] whose insights [paweekly.com] into the relationship between ant colonies as single organisms and the way that human intelligence emerges from the biology of the brain are startling. The study of ants and colony behavior is an exciting field that can inform many fields from weather systems to crowd behavior to artificial intelligence.

A more in-depth introduction...

by Tony White can be found here [carleton.ca].

Smart Dust

That makes me think of Smart Dust and the network intelligence of Smart Dust.

http://robotics.eecs.berkeley.edu/~pister/SmartDus t/ [berkeley.edu]
http://www-bsac.eecs.berkeley.edu/~warneke/SmartDu st/ [berkeley.edu]

Freenet

This certainly isn't the first attempt to apply these ideas to practical problems. Ian Clarke often describes swarm intelligence as one of the inspirations behind the Freenet [freenetproject.org] design, for example in this [choler.com] article he says:

"My motivation from the technical side was, firstly, really, I was fascinated by the idea of complex systems, which are formed from simple individual entities all cooperating. An example would be an ant's nest, whereby all of these ants are following relatively simple rules, yet they all work together to make this effectively a kind of meta-organism, which is the ant's nest, which can feed itself and reproduce and defend itself. So I was fascinated by that idea, and I was very interested in trying to apply that to a computer system. And by combining [this idea with my idealogical motivations], I essentially came up with Freenet."

business problems?

Hey Boss, we're not gonna make the deadline.

Boss: How about if I give you five thousand deadlines! ...I guess it would be a good motivational tool...

Hacker and the ants.

Reminds me of this post [slashdot.org] Once again Rudy Rucker proves prophetic. The protagonist of the story is a programmer named Jerzy Rugby who uses artifical life (ants!) to build and optimize robot source code.

Just like the ant analogy mentioned in the article, the ants were used for their collective ability to help build the smartest AI source. Again I recommend the read.

The Island Of Dr. Bonabeau

Dr. Bonabeau takes us from his childhood nightmares of carnivorous wasps to applying the theories of swarm intelligence to solving real problems in the business world."

For example, unleashing your army of carnivorous wasps to eat key performers at the competition.

Manager 1: "Where's Engineer Bob? He's supposed to finish project X-12 this week."

Manager 2: "He got eaten by carnivorous wasps."

Manager 1: "Wow. Sucks to have been him. Hey, that leaves us free for golf after lunch."

Manager 2: "Oh, right on, old boy!"

some cool links

be sure to check out:

The Swarm Development Group [swarm.org]

SFI [santafe.edu]

Complex Systems [umich.edu]

Mythical man month

I got to thinking about this real quick and, as i'm too lazy to read the actual interview and it's probably addressed there, what are the effects of diminishing returns?

The 'mythical man month' basically says that one programmer (or other worker) can produce more in one month than two workers each working half a month... who can do more than three workers all in 1/3 of a month. And further that just throwing more people at a problem doesn't really do much past a certain point. For some problems, it might be the case that one guy working for a month can do more than ten guys working for the same period of time.

How does swarm behavior overcome all of this great stuff?

I presume that it must be an essential part of the deal that the problem must be something very trivial for there to be great effects by swarming.

Re:Mythical man month

The mythical man month works as a concept because of the problems caused by the division of work.

For example:

• He can't optimize his task list over a long, flexible schedule.
• He has to explain the things he's done to the people after him.
• It takes time to be "in the zone," and the people after him have to spend that time all over again.

These things only occur to complex agents, like people. The idea behind swarm behavior is that the agents are simple, and need not individually perform complex tasks.

Boy, I don't like what that says about me as a Slashdot addict. :)

Re:Mythical man month

Context, context, context.

Programming right now is an activity that requires huge amounts of context to produce good output. Just being distrubed can cost big. Splitting the context in half will cost, it will not benefit.

Programming is an extreme problem, though. Some things, like "getting from here to there" requires much less context. You routinely set out on journeys with incrediblely incomplete amounts of knowlege regarding the conditions of your path. Sometimes you end up taking alternate routes because of obstructions. Compared to the amount of context maintained while programming, you set off to your destinations almost blind.

Only some problems can be "swarmed", mostly where there's some form of reinforcement that can be used. "Getting from here to there" is a great, obvious example of that, with the phermone trails reinforcement. On the other hand, the whole point of programming and its great attaction to me is the desire to never do the same thing twice, almost the exact opposite kind of problem. ("The number one sin of programming is code duplication." - me.)

Ah, yes... but...

... the mythical man moth can handle working in swarms and get more work out.

A business application

For a good picture of how this would apply to the business world, check out this article [business2.com] from Business 2.0 about agent-based supply chain solutions. Pretty interesting stuff if you've got a large, sophisticated network to manage.

class

There was a programming class at MIT [mit.edu] that used the swarm concept on AI for a game. The game was an RTS where each unit had it's own AI and could communicate to other units but not easily (short range, takes time). Each unit generally had a simple program, but your team had a fairly complex overall strategy. My team (Master Control Program) did pretty well in last years contest [mit.edu].

Dr. Evil

I remember with retrospective anguish my holidays in the south of France, when picnics turned into nightmarish fights against carnivorous wasps...

Ah, yes, those horrible days and nights writing poetry as child in the South of France. Sometimes Reginald wouldn't bring my water chilled, so I had to berate him. In the summer we made meat helmets! ~Peace out, Airrage.

Dirk Gently Navigation

This reminds me of the Dirk Gently Navigational techniques.

Follow someone who looks like he knows where he's going.. You may not end up where you want to be, but chances are you'll find your way somewhere interesting.

Me and my friend actually did that, arriving in Vancouver at 4:00AM. We followed a few random people to strange places (We stopped following the armed car when we figured that they might be getting a bit nervous). Befere long, we ended up in front of a Dennys. We stopped for breakfast/supper and then called Peter for directions to his place.

Tried it a few times since then -- as long as you've got a little time to spare, you can find some very interesting things about the place you're in.

Re:Dirk Gently Navigation

That technique works pretty well for eBay browsing. There are tools out there now that give you a GUI interface to enter eBay IDs. It presents a list of all items that person has bid on that are current. It has a 'favorites' feature so you can have 'favorite' people you track.

It finds the 'good' stuff, i.e. the things that anybody would actually bid on. By cultivating collections of people who buy the kinds of things I am interested in, I seldom anymore actually browse 'raw' ebay for items to buy.

Interestingly, when you pull up a query for an eBay account held by someone in Germany, eBay returns a message that they aren't allowed to gather and give out that information for German citizens.

Erm, is this thinking new?

What is open source after all? Don't think the system would quite work a business model, but for like minded volunteers it's already up & running.

I can't belive I'm quoting a shirt...

"Never underestimate the power of stupid people on large groups"

Fast, Cheap, and Out of Control

Rodney Brooks proposed something similar for space exploration in 1989. Did anyone else see the Errol Morris documentary that features him? His paper is here: Fast, Cheap and Out of Control: A Robot Invasion of the Solar System [mit.edu]

Been there Done That

This happens all the time in the business world. Any attempt at a new thought, in particular meetings, will be met with vile and a scorn, right before they they beat you to death with the conference room phone. It is a sort of a mindless action that is drilled into people in corporations in much the same manner as wasps or bees. Attack anything that threatens the stability of the nest!

So I guess this just proves, "The Future is already here!"

Swarm Intelligence: The book

There's a book on Amazon [amazon.com] by the same name (not the one mentioned earlier in this thread). It's a very good overview of artificial social intelligence models, very profound in places. Incidentally, Amazon offers it at a discount when bought together with Bonabeau's book of the same title.

People interested in intelligence and life as an emergent and evolving quality would probably also enjoy "Creation: Life and how to make it", "The Tipping Point", and "Figment of Reality". They should all be reasonably easy to find.

I think there is plenty of room for new inventions from those who understand both software technology and the emergence of intelligence from social models.

Read up! Enjoy!

Cheers, Jouni

Are we a swarm of cells?

I've always thought that the idea of multi-celluar organisms to be a misnomer - we are actually more of a tightly integrated colony of cells.

I would define a cell as the basic life form, and anything greater than a cell is not a single 'creature'. Humans, like the ant colony, are a giant collaborative effort.

Of course, there's something in our brain that gives us the sense of I, the individual, irreducible person. It's an illusion. But it helps us survive, I guess (By us I mean "we cells," not you).

genetic algorithms

This technique sounds suspiciously like genetic algorithms.

Put out a bunch of genes, see which ones survive. The ones that don't die, the ones that do are re-integrated.

Put out a bunch of ants, see which bring back food. The ones that do, copy, the ones that don't forget about.

Or how about neural networks. Put out a bunch of connections. The ones that work, strengthen, the ones that don't weaken.

Is it just me or is it all the same general idea.