Swarm Theory Makes National Geographic

g8orade writes "Swarm Behavior / Swarm Theory has made the pages of National Geographic. Brief but interesting article with several examples." Swarm theory has been discussed here a few times in recent years.

Comments

Nomenclature

[pubjames]

It seems to me that this whole field (what do I call it - complex systems? derived behaviour? emergent systems? swarm theory?) lacks a consistent language. It is a hugely important scientific field, but everyone calling it different names means it appears smaller than it really is!

Re:Nomenclature

[$RANDOMLUSER]

Agreed. And TFA itself is a little confused itself about the differences between the "Hive Mind" and swarming/schooling/flocking/herding behavior; which are really two completely different things.

Re:Nomenclature

[Gunark]

The correct term is Dynamical Systems, and its common, consistent language is the branch of mathematics dealing with dynamical systems (complete with its own vocabulary -- strange attractors, manifolds, emergence, chaos, etc.)

Re:Nomenclature

[AndersOSU]

When did dynamical become a cromulent word, and who decided that systems was too good a noun to be modified by an adjective like everyone else?

Insect swarms are smarter than insects

[idontgno]

Aunt Hillary would agree.

To the confused, Aunt Hillary is an ant hill, a character in Douglas Hofstadter's Gödel, Escher,Bach; an Eternal Golden Braid. The chapter she's featured in is subtitled "...Ant Fugue". (Which is the chapter following one subtitled "Prelude...")

Swarm Theory and Economics

[Attila Dimedici]

My fascination is with how similar this is to the theory of free market economics.

Re:Swarm Theory and Economics

[igny]

My fascination is with how similar this is to the theory of free market economics.

Or communism...

Re:Swarm Theory and Economics

[wvmarle]

you're mixing up communism and socialism. Communism, in the Marxist sense, has never been reached. The planned economy is a stage in Marxes theories on the way to communism, which is a utopia where everyone works for themselves, taking only what they need, giving what they can or think the group needs. Communes often come close to communism (and the words being almost the same is no coincidence).

Re:Swarm Theory and Economics

[xappax]

The idea that there is a net benefit for a group from the collective selfish actions of individual actors is closer to what this article is describing as swarm theory.

Actually, the article doesn't say anything about the collective selfish actions of anybody. In fact, in almost all the examples given, the actors are behaving unselfishly. The ants don't know exactly why they should go out and follow a given trail, the bees don't really understand why they should choose one nest over another - even a protester wasn't aware of how their movement to a particular street would help overwhelm police.

There is no apparent benefit to any of the individuals in doing any of that. In fact, I daresay that a "free market" ant wouldn't follow any trails, wouldn't bother to smell any pheromones, it would just chill in the nest and eat what the other ants brought, expending the minimum effort for the maximum gain. And free market ants certainly wouldn't automatically tell everyone else where the food-jackpot was that one of them had personally worked to find.

So I agree that swarms are unlike authoritarian communism. They're unlike authoritarian anything, simply because swarms are anti-authoritarian and non-hierarchial - any structure involving a boss or a "chain of command" cannot function as a swarm. However, they're definitely not behaving the way a free market does, either. The key thing to understand is that the actors in a swarm are voluntarily doing non-selfish things because those things, when done by a lot of actors all together, will result in a net benefit for all the actors.

So, swarms definitely have a sort of collectivist, socialist tinge to them, because they require all the actors to base their actions on what will benefit and sustain the group - not them personally. However, because of the lack of authority, swarms are sort of a more pure form of socialism that is inherently resistant to the corruption and oppression by things like governments or leaders.

I think swarms are one of the most important trends in society, because they're the one thing that terrifies all people in power - capitalist CEOs and communist dictators alike.

Re:Swarm Theory and Economics

[Attila Dimedici]

By that definition, there is no real difference between an ideal free market economy and communism. Except that in an ideal free market economy people have access to luxuries and in an ideal communist economy no one has any luxuries, since by definition a luxury is something you don't need. In an ideal free market economy, everyone's needs are met by the efficiencies of the economy. Of course, since this is not an ideal world, neither of these will ever happen. The question is which theory does a better job of meeting people's needs in the real world.

Re:Swarm Theory and Economics

[Elemenope]

No, in Communism there is access to luxuries, only they must be produced by the labor of the individual consuming them or appropriated by equal (i.e. no surplus value subtracted) labor from someone who can. Besides, it is narrow to assume that the economic definition of 'luxury' is equivalent to the practical definition of the same. Many here I'm sure can attest that access to sci-fi books and video games, while not strictly necessary for survival, are beneficial to their continued functioning as healthy individuals, and as such aren't really 'unnecessary'. There is such a thing as prioritized consumption.

Re:Swarm Theory and Economics

[tom_evil]

More like anarchism. Capitalism has corporate bosses, communism has party bosses.

One key to an ant colony, for example, is that no one's in charge. No generals command ant warriors. No managers boss ant workers. The queen plays no role except to lay eggs. Even with half a million ants, a colony functions just fine with no management at all--at least none that we would recognize. It relies instead upon countless interactions between individual ants, each of which is following simple rules of thumb. Scientists describe such a system as self-organizing.

Re:Swarm Theory and Economics

[Liberaltarian]

The principles, if one must cast a political/economic philosophy over them, would most closely resemble anarchism. Representative democracy and the modern business both have leaders and led, the rule-makers and the rule-followers: a disturbing amount of our economy and polity is top-down (also never meaningfully accountable) and couldn't be further from swarm theory. If we want to see a more rhizomic society, we'll have to think a lot further outside the box, that's for sure.

The closest the article's author comes to political applications is the Seattle '99 protests. However such tactics and structural sentiments abound in anti-neoliberal movements, from the Landless Workers Movement in Brazil to the Zapatistas in southern Mexico.

Re:Swarm Theory and Economics

[joh]

Communism's other name is "command economics". It's the idea that some wise and benevolent leader is better at allocating resources than a pack of ravenous self-interested capitalists.

No, it isn't. In real communism there's not even more a need for a state, a government or a leader. In theory at least.

Re:Swarm Theory and Economics

[cartman]

My fascination is with how similar this is to the theory of free market economics.

The theory of "swarm behavior" had already been elucidated in economics several decades ago, and its applicability to biology (and simultaneous co-discovery in that field) was described at that time.

Von Hayek described "swarm theory" and how it operates in the price system of a modern economy. Hayek elucidated how the price system coordinates the activities of millions of people, each of whom has extremely limited information, and without any kind of central authority. Each person follows simple rules based on his own local information and the result is an allocation similar to what perfect information of all participants would have suggested. Decision-making is purely distributed in a free market, and information travels by means of "price signals" to coordinate the activities of far-flung individuals who have never heard of each other. You can see this principle at work with regard to the extremely complicated interactions in the global economy. An example is the "ripple effect" of prices where you place a demand for something, which in turns affects the demand for what's necessary to manufacture that thing, which in turn affects the demand for something else, etc, until some shop in Taiwan manufacturing a part for a sub-sub-contractor in something seemingly unrelated shifts his production away from transistors because for some reason (in fact for millions of different reasons) the demand for that kind of transistor has gone down. The shop owner in Taiwan made the correct allocative decision that optimally satisfies the needs of an unbelievably complex interaction, but he didn't know why or how. All he must do is pay attention to local price changes--the price of this component has gone down, so produce less of it. The result of all this, even though it's difficult to believe at first, is that free prices produce the optimal allocation even though each individual participant was acting on limited information and was using only simple rules.

Von Hayek also made the additional claim that the price system itself was never devised by anyone, but was the product of an evolutionary development which nobody understood when it was occurring. Hayek speculated that the price system would never have been consciously devised, but resulted from people following "what worked" based on their own local information. People who followed the price system bred more quickly (in the malthusian sense--fewer of them died) and spread civilization. Of course on this point Hayek acknowledged that he was being very speculative.

The spontaneous order of the market was the basis for Hayek's claim that communism would quickly collapse--not because of a lack of incentives, but because of the need for conscious control. Conscious control was not a benefit (as claimed) but was actually a fatal flaw, because even the supremely intelligent leaders could never achieve the kind of coordination and distributed information that a market could easily achieve. As a result, Hayek believed that claims of capitalism being "anarchic" and "disorderly" were actually compliments to it, because only distributed, spontaneous order could ever hope to contend with the complexity of a modern economy.

Hayek immediately pointed out that the same principle was applicable to biology. He was then informed that a similar change was already underway in that field. He was told that ants and bees used something similar, etc.

Another example of "swarm theory" in economics is the efficient market hypothesis, which relates to stock markets, and which is mentioned frequently in the book about "the wisdom of crowds."

I have a sneaking suspicion

[Colin Smith]

That human consciousness is a swarm of neuronal interactions.

Re:I have a sneaking suspicion

[WormholeFiend]

And the rest of the body is a swarm of cooperating cells.

Re:I have a sneaking suspicion

[beyondkaoru]

and a computer is a swarm of cooperating transistors.

You want another example?

[ben0207]

See how many "I for one welcome our hivemind overlords" type posts we get with this story.

Alternatively

[z0idberg]

From TFA "Ants aren't smart," Gordon says. "Ant colonies are."

But apparently...

"A person is smart. People are dumb, panicky dangerous animals and you know it." - Kay

Re:Alternatively

[MarcoG42]

So, now you have a group of people with high IQ's trying to convince one another that *their* idea on the best way rid the valley of Count Bludsuckingfiend, which leads to much chart-drawing and counter-pointing. All the while the good Count is systematically draining their daughters' circulatory systems.

Now, you've got the low IQ guys, whose first thought may be "Der...maybe we should just kill him." The other two guys, not having any better idea, happily go along with the first.

"Practical Applications" of Swarm Theory

[chillax137]

They cite a practical application of Swarm Theory as optimizing the business operation of a gas producer. They say this technique was inspired by how ants learn to forage for food, but this technique is a standard (and pretty obvious) solution to numerical optimization. So while the idea is interesting and can definitely be applied to networks of robots, it is a retroactive explanation of something that has already been developed (for marketing purposes, I'm sure).

Unmentioned in the article

[Control Group]

It's not mentioned, but it seems an obvious sort of question to ask given the content they've got: is there anything to "real" (by which I mean, individual) intelligence other than swarm behavior at the neuron level? In fact, is the entire biology of any given animal (ourselves, obviously, included) anything more than swarm behavior at the cellular level? Or, if we accept the idea that cells are just a reproductive mechanism for DNA, is it just swarm behavior at the molecular level?

Which would have a fascination all its own, since I don't think anyone's ever argued that DNA has anything we'd call intelligence. If all of life arises out of swarm behavior at the molecular level, we've managed to take intelligence completely out of the equation.

Which, in turn, just makes this another facet of the belief that the entire universe is an emergent phenomenon of a vast set of simple items following simple rules.

The truly intriguing observation (from my point of view, anyway), though, is that this emergent phenomenon contains examples of exactly the same mechanism at so many levels of complexity. It wouldn't necessarily have to be true that simple interactions at the fundamental particle level would give rise to higher-order behaviors that can be macroscopically described as simple interactions at that higher level. It's the fractal nature of the mechanism that is most intriguing, I think.

Re:Unmentioned in the article

[ThosLives]

Yeah, I don't really see this as "swarm intelligence" so much as a system with self optimizing behavior.

Take the example of the gas producer / distributor. They have a system of equations linking variables (the routes that trucks can take, cost to operate the trucks, and price of the product at various plants) which is solved for an optimal solution. The optimization is simply to find the maximum profit - it's a very simple optimization problem. (For mathematical definitions of "simple".) The fact that it's not intuitive doesn't really mean anything other than intuition isn't a good method for optimizing systems.

The interesting thing is that the biological system of an ant hive developed to be an "optimization solver" - which isn't really that surprising considering the whole point of a biological system is to minimize some potential (as happens with all physical systems). It just so happens that with biological systems, minimizing that potential also increases the probability of the system existing for longer periods of time (in other words, perpetuating the species). I think ant hive behavior is kind of anthropic - if hives were not optimized that way, ant hives wouldn't exist because all the ants would be dead.

So, yes, this is nifty stuff, but I don't see it as "intelligence" so much as an optimization problem.

Of course, it may be the case that "intelligence" is the result of an optimization, but it may also be the case that "intelligence" falls into "Godel space" (i.e. that space where something exists but can't be proven because logic, being sufficiently powerful, is incomplete).

Re:Unmentioned in the article

[iangoldby]

There is an interesting question of reductionism here. There a good form of reductionism, were a complex idea is described in terms of an aggregation of simpler parts. But there is also a bad form of reductionism in which the complex idea is claimed to be 'nothing but' the aggregation of simpler parts.

This bad reductionism has been called alternately 'nothing buttery' and 'Greedy Reductionism'. (Greedy Reductionism.)

In this case we just need to be careful not to suppose that if intelligence might perhaps be well-described in terms of swarm-theory of neurons then intelligence is not "real".

(Incidently, there is a difference that might be relevant here between describing something and explaining it.)

Re:Unmentioned in the article

[Hal_Porter]

Which would have a fascination all its own, since I don't think anyone's ever argued that DNA has anything we'd call intelligence. If all of life arises out of swarm behavior at the molecular level, we've managed to take intelligence completely out of the equation.

DNA doesn't but the process of evolution manages to make perfect designs from swarm like rules I think. I'm pretty comfortable with the idea that all intelligence is emergent behaviour from swarms actually.

The truly intriguing observation (from my point of view, anyway), though, is that this emergent phenomenon contains examples of exactly the same mechanism at so many levels of complexity. It wouldn't necessarily have to be true that simple interactions at the fundamental particle level would give rise to higher-order behaviors that can be macroscopically described as simple interactions at that higher level. It's the fractal nature of the mechanism that is most intriguing, I think

You know there's a hole in all this that I think means we don't understand it. Once DNA and proteins are there, evolution has booted up and can do anything. But DNA and proteins are two complex to appear by chance. My hunch is that some evolution like process which was not dependent on DNA and proteins must have been running first, and it made them as tools before they took over. But no one to my knowledge has explained how this could work, though some people have tried.

But life appeared pretty soon on earth, almost instantly, and I don't believe in luck or supernatural intervention, so I'd expect someone to figure it out sooner or later.

Re:Unmentioned in the article

[Vireo]

I think ant hive behavior is kind of anthropic - if hives were not optimized that way, ant hives wouldn't exist because all the ants would be dead. This is clearly the antropic principle at work.

Re:Unmentioned in the article

[lawpoop]

You're you're own parent? Woah... ;)

I don't fault Wolfram's idea in general, but I think the main criticism is that he makes associations that are unwarranted. Combine that with the tome he wrote over a number of years in almost complete isolation, claiming that it would totally revolutionize science, it makes him come off as a little crazy. If it weren't for the fact that he is a genius, and he has contributed immensely to various fields, I think people would dismiss him as a total nut. But even productive geniuses can go insane, or have an 'episode' where they go off on the wrong track for a while.

Basically he takes the patterns you can get from the game of life mathematics, and says it can describe almost anything, and then shows how, as one example, patterns in sea shells look similar to output from a game of life checkerboard.

Not that game-of-life examples can't be used to describe things, but to jump to the conclusion that it can describe everything, and the implication that it *is* the fundamental function of the universe... well, time will tell! ;)

Re:Unmentioned in the article

[lawpoop]

Yeah, I don't really see this as "swarm intelligence" so much as a system with self optimizing behavior. I think the reason people are talking about this is that it goes against the sort of inborn intuition about where intelligence lies in living organism.

Without critical study, we seem to have the inborn idea that the individual mull-cellular organism is intelligent. Humans are intelligent, dogs less so, plants, not really at all. If a group of organisms is acting intelligently, we assume that each one of them has to be pretty smart, or else the whole group couldn't be smart. In the case of swarms that exhibit intelligence, none of the organisms seem to be that smart -- or at least, they don't have the complete set of smarts that is shown in the group behavior. In fact, they are pretty simple when it comes to interacting with groups.

So when studying ant colony behavior, there was kind of a conundrum in the field for a while. If individual ants are dumb, why does the colony behave so intelligently? People where then looking for the hidden smarts inside each individual ant. Or, another possibility is that colony behavior really isn't that smart, despite it seeming so to us.

But it turns out colonies really are smart, *but* there are no hidden smarts in the ant. The ant really is dumb. It's only when you combine their simple behavior in the swarm that you find intelligence. It's not in the ant; it's in the colony.

This is a paradigm shift in the understanding of complex behavior of multicellular organisms. We have had good evidence of individual organisms acting smart, that was never in question. But until now, we have never had good scientific, mathematical evidence of intelligence at the group level. People may have suspected it, but now they have evidence to convince skeptics.

Re:Unmentioned in the article

[jc42]

This sort of reminds me of the the Chinese Room Argument. The gist is that a person is isolated in a room with a complex instruction manual, and that person receives cards with Chinese characters. Using the instruction manual, the person translates the characters into English. The argument is that the person in the room doesn't really understand Chinese. He's executing instructions that lead to a Chinese translation.

And just about anyone who knows more than one language understands the fallacy behind this scenario. For a lot of fun examples of the results, visit engrish.com and wander around a bit. You'll quickly find a lot of example of the results of word-by-word translation. In the case of east Asian writing, translating individual characters without understanding that there are multi-character words can lead to especially humorous results. But even some of the "correct" translations are hilarious. And sometimes, of course, you just can't tell what was meant unless you can read the original.

The reason that computerized translation has been "5 years away" for so many decades is that the job requires intelligence and understanding. Doing it mechanically as a text-substitution process simply doesn't work very well. But it's a good source of humor.

(Reports are that the Chinese government is preparing for their upcoming Olympics by replacing a lot of multilingual signs with better translations. This has led to complaints that they are taking a lot of the fun out of trying to find your way around in Beijing. ;-)

Sentient Groups & Algorithm Difficulties

[aldheorte]

A Fire Upon The Deep novelizes the potential of sentient consisting of several physically individual members who do not have sentience as individuals, although this runs tangential to the plot.

Everyone with some algorithm design experience knows that you can get complex behaviors (often known as bugs) with a set of simple rules. Unfortunately, the wide range of problems to which we apply computers, generally by business demands, require rigorous certainty. We want to know exactly how many beans were shipped, not an estimate. Individual instances of an algorithm cooperating via simple rules inherently introduces uncertainty or reflects a very inefficient approach to solving a certain problem. This goes against the grain of classical training and thinking about computing.

Collective intelligence may also depend on all individuals having some level of variation, yet cooperating through simple rules. In this case, the emphasis goes to the protocol and not the algorithm. I believe that further research will find that some level of individual variation will become recognized as an essential element of perceived group intelligence, important to breaking recursive feedback loops and deadlocks. Unfortunately, attempts to emulate this in computing will run into the issue that group perceived intelligence may not be determined so much by design, but by fitness for a particular, narrow purpose, with truly remarkable group intelligence requiring many iterations exposed to actual operating conditions or good simulations thereof.

Antsdot

[mhannibal]

Ant colonies sound a lot like slashdot it seems...

Re:Antsdot

[TrappedByMyself]

Ant colonies sound a lot like slashdot it seems...

I disagree. Ants get smarter when put into large groups...

Comment removed

[account_deleted]

Comment removed based on user account deletion

Re:The brain is a swarm

[SatanicPuppy]

Would your body classify as a swarm of atoms?

Unlikely. A swarm is composed of units that are functioning individuals as well, with their own individual complex behavior patterns.

That's what makes swarm theory so interesting. if they were all working together because they were effectively cogs in the swarm "machine" then the fact that the sum is greater than the parts wouldn't be interesting at all.

What a great way to stop an ant population!

[erroneus]

Perhaps some genius chemist will come up with a way to infect or affect an ant's sense of smell/touch/taste in such a way that foragers never go out and thereby starve the colony? It wouldn't be poison in the direct sense and would hopefully be safe for plants, animals and children. It would be like boric acid but better.

Article Missed a Major Point

[CyberGarp]

It said that "Swarm Theory" was being applied to business operations. I call bullshit. A computer model was run at night that provides the orders to all the drivers each morning. This flies in the face of the premise of swarm theory. If each driver were given a simple set of rules to follow for driving then it would be a direct application of swarm theory to operations. However, it's not swarm theory applied to operations, because each driver gets an order from corporate each morning. No local decision are made. It's just another algorithmic approach to combinatorial optimization with centralized management, which till I see a Big O notation, and some papers, I withhold comment on the computer model.

How to Kill the Bad Aliens Now?

[J. L. Nelson]

In Hollywood movies such as "I, Robot" and "Independence Day" a non-swarm organisational structure is assumed and actions by the hero, such as destroying the central processing core or blowing up the mothership, generally puts an immediate and dramatic end to the world's invasive trouble. What would happen in a movie where the invasive enemy had a swarm organisation? I'm not a movie buff at all, so can anyone point out any examples of this? Perhaps Hitchcock's "The Birds" (which I haven't seen) or some killer bee-type movie? On the other hand, movies like "I, Robot" and "Independence Day" also glorify the worth of individual thinking to benefit the whole, but more in terms of exceptional individuals who stand out over the rest, rather than to point out how a group of independent-minded individuals can overcome obstacles. (Again, any examples to the contrary?) Yet, it seems to me that it would be more beneficial to society as a whole to propaganise the swarm organisation. Will Hollywood catch on--with effect--or is Western society too enamoured with the cult of the individual for it to make a difference?

Re:Swarm Theory and Free Market Economics.

[xappax]

Actualy they are very much more like a free market.

Communism is a tightly hierarchical system in which all decisions are made at the top and everyone has to do what they are told by the chain of command.

I don't want to seem snotty or disrespectful, but please read what someone's written before disagreeing with them. You're right. As I wrote above, authoritarian systems - including communism - are not swarms, and in fact are usually set up to deliberately suppress swarm behavior (which undermine centralized power). So swarms are not a good example of communism.

I think, however, you've fallen into the classic trap of thinking that there are only two economic models: communism and free market capitalism. It reminds me of when I was a young kid and thought that if you weren't Christian, it meant you were Jewish :)

There are a ton of socio-economic models which critique and are sometimes opposed to free market capitalism - and only one of them is communism. The rest are things like participatory economics, anarchism, gift economies...I would say that swarms are more closely related to some of these models.

Human beings in a free market make decisions based on the information we get from our interactions with others in society

That's true, but irrelevant. All life forms make decisions based on the information they receive, that has nothing to do with swarms. The interesting thing about swarms is that when you get a bunch of actors together, and each one of them follows a pattern of behavior that has no benefit to the individual, you get an overall emergent result which benefits the whole group. Individual humans in a free market environment base their decisions on what will help their personal interests to the exclusion of anyone else's - that's the hallmark of the system.

Swarms are like a proof-of-concept that when people are able to stop being myopically selfish and participate in a collective "organ" that's larger than them, rewards return to them which couldn't have been anticipated with a free market perspective. In one way, this is a kind of creepy realization, since it suggests that the most efficient mode of socio-economic organization would be some kind of Borg-like hive-mind. Obviously, I don't think that'd be a good thing, but I do think there's room for individuals participating in collective swarms when it comes to important matters (like food,clothes,shelter), and going their own ways when it's not.

Comment Missed a Major Point

[theefer]

The swarm intelligence algorithm is ran offline to determine a solution to the global problem. Indeed, ants "run" the "algorithm" inline as they don't leave the nest with a full plan of action, but the method used is still swarm intelligence, as opposed to, say, standard heuristic-based TSP solvers. The reason why it's not ran inline is that the cost of doing so is larger than the benefit, since the conditions are not very dynamic.

By the way there are many papers on the topic, although it's quite recent, just citeseer for "swarm intelligence".