Slashdot Mirror
Emergence
The author makes a point that there are 3 main camps of scientific study.
- The study of simple systems - under a few dozen variables, like electromagnetism, or celestial mechanics.
- The study of stochastic systems - few million to few billion variables, like actuarial sciences and genetics.
- The study of disorganized complexity. Systems in the middle between a dozen and a few million variables, where the second order characteristics - how they interact, is of primary concern.
Deduction and induction work for the first two camps, but for the third, the interactions cause actions and reactions which are what scientists politely call counter intuitive, meaning your first thought is Huh? Or, in other words, it behaves quite differently from what your instincts and (so-called) common sense would tell you.
There are five basic principles for developing a system (or simulation of one) which can express emergent behavior:
- More is different. You get a very different behavior of the system when certain thresholds are reached.
- Ignorance is useful. Ants communicate with a vocabulary of around 20 words/ideas.
- Encourage random encounters. Much of the behavior of an ant colony comes from them just bumping into each other (or external things like food, or my foot).
- Look for patterns in the signs. Even with the limited vocabulary of ants, they can also express things based on the decay in the pheromones they deposit.
- Pay attention to your neighbors. Also described as "local information can lead to global wisdom."
One of the enduring myths we have, is that of the Ant Queen. The myth supposes that there is some central planning done in an ant colony. Instead, the queen exists only to pop eggs out. Male ants have such short lives, that in most species of ants, they have no mouths to eat with; they just don't live long enough to get hungry. The production of warriors and workers is stimulated by pheromones in the colony. Information on where to gather food is gathered through random acts of bumping into things. There is no ant which tells another to go lift that bale or tote that barge. It appears that our intelligence is a by-product of the neural interactions of our brains.
The economist Jane Jacobs had been studying things like this for years, and has been demonized by the majority of economists: they want to believe in some centralized controlling force, control that force, and you control the development of your economic system. People reading her books tend to think she worships sidewalks, instead, she values the communication that can only happen on sidewalks; people meeting each other and exchanging words. You can't say "hi" to your neighbors if you are each zipping past each other on the freeway.
One can experiment with emergent behavior with some software tools. The author explains a few, of which you are most likely to have experience with SimCity.
The main difference between chaos theory and emergent behavior theory lies in a couple important differences. A chaotic system has a number of determinable feedback loops, all of which are (usually critically) dependent upon the starting conditions. Emergent behavior has more to do with feedback loops causing totally different behavior, and when some threshold (usually population) is passed, the nature of the system drastically changes.
If you are looking for sample code to simulate things, you won't find it in this book. If, however, you want to get an overview of where this field is coming from, read this book.
You can purchase Emergence: the Connected lives of Ants, Brains, Cities and Software from bn.com. Slashdot welcomes readers' book reviews -- to see your own review here, carefully read the book review guidelines, then visit the submission page.
Groupthink
While A New Kind of Science may have lots of pretty pictures, and may be a decent survey of the field of cellular automata and its potential applications, and while Stephen Wolfram is no doubt a smart man, the quality of the book is overshadowed by his pathetically arrogant writing, wherein he pratically claims credit for CA, despite actually doing very little to even further the field. It's sad that people are beginning to think he really is a leader in the study. Please dissociate Wolfram and CA in your mind. Thanks...
The thing is, there's no reason to be scared of autonomous organization - it is literally everywhere around us.
Swarming/flocking/schooling algorithms are a great example of this. All it takes is a desire to be close to your neighbors but not too close, and the swarm/flock/school functions largely on its own - it can even go around obstacles and re-merge, it optimizes into aerodynamic shapes, etc.
I love complexity from simplicity. One of my favorites occurs from the standard predator/prey population equation. If you run it for a while, it switches into repeating cycles of population size. However, the positions and numbers of cycles are dependant on the equation conditions. If you plot the cycles vs. the starting conditions, you get this beautiful graph of the data starting at a single point, then branching, and again, faster and faster until it forms into pure chaos... and then from the chaos, emerges three clean branches, which then fall to chaos again.
This wizard will complete the installation of: AQP AA002! P O a @ P @1 Ae IoD'i
"...using cellular automata as a means of explaining higher order behavior (like Wolfram in A New Kind of Science)."
Well, not quite sure if it can explain as high order behavior as Wolfram yet...
This is, indeed, an interesting book and the reviewer fails to point out that Emergence goes into detail concerning the karma, moderation and meta-moderation system of this here web site.
:-) Towards the end where he's talking about emergent video games I got a little bored, but definitely a book that got me thinking. Worth reading even if you are aware of the way ants behave, because you probably don't know as much about slime mold as you should.
Author seems to think taco is a genius or something, but it's still a good read
John.
I read the first two chapters or so of that book. It's totally an essay strrrreeeetched into a book. Terribly boring (in a lite-on-content sort of way). On the topic of taking recommendations from Slashdot: A poster raved about "The Non-Designers Design Book," so I bought it. It's not completely worthless for total amateurs (like me), but it's pretty much written for the purpose of teaching secretaries how to make better-looking newsletters. Lesson learned.
Monstromart: Where shopping is a baffling ordeal
...but interesting, nonetheless. For two viewpoints that are more or less opposing, read Daniel Dennett and John Searle - the latter of whom is a latter-day dualist who talks a lot about emergence, aka emergent properties. Dennett thinks machines will be able to think, Searle doubts it.
I knew our collective hive mind would come in handy someday:
1. "I, for one, welcome our emergent overlords."
2. "???"
3. "Profit!"
4. "In Soviet Russia, our groupthink comes from emergent behvaior, or is it the other way around?"
5. "Who cares? Look, it's Natalie Portman!"
6. "Does Netcraft confirm it?"
7. "Yeah, but only in Korea."
8. "Netcraft does not confirm it. Old people are not quite dead yet."
9. "OK, that's the Monty Python reference out of the way. Has someone bashed China yet?"
10. "No, and we also haven't bashed Micro$oft yet, at least not until this line.
Crap. I'm only at #10 and the well's running dry. (What, you want me to yell "MEEPT!" or something?)
If you're a glass-half-empty type: we won't be as useful in the underground sugar mines as I'd previously thought. We're only capable of half as many thoughts.
If you're a glass-half-full type: or maybe we've achieved antlike emergent behavior using only ten words and ideas, making us twice as efficient as our formic emergent-behavior-exhibiting overlords!
but isn't this terrain Douglas Hofstadter covered about twenty-five years ago in Gödel, Escher, Bach? Does Johnson's book say much new? Has a quarter-century's "progress" in CA and AI brought us any closer to singularity? And will I ever stop posting this comment in rhetorical question form?
Welcome to the Panopticon. Used to be a prison, now it's your home.
This was also a theme from the Matrix. The machine world was not controlled by a single overlord but was instead made up of billions of different programs. All the way down to the "wind" or the "bird" programs. Taken individually they're all rather simple and pointless, but when taken as a whole they build something much more valuable.
No, no it's not. Now go find a sci-fi thread somewhere.
I quoted this excellent book and gave some future directions about using the bottom-up Emergence technique when dealing with Threat Modeling. Read the last chapter in my MSc paper Threat Modeling for Web Applications using the STRIDE model Comments welcome. Thanks
The essential concept is that each individual is a simple agent that operates autonomously, and makes very few very simple decisions as it goes about its work. The behaviour of one individual is unremarkable, but the behaviour that emerges from a large group of the same individuals is really quite amazing.
Because the concepts are really quite logically simple, this stuff is really simple to program too. Just fire up perl or java or any language that has a similar capability to OO concepts, write a simple object - your agent - that behaves according to a simple set of rules and responds in defined ways to certain stimuli. Make a wrapper program to create the playing field, instantiate as many 'agents' as you see fit, and let them loose. Tweak, rinse, repeat.
As an aside, when I was writing a simulation to emulate the behaviour of ants foraging (more to prove that perl and java were suitable languages for the task than to demonstrate anything new with ants per se), I went off and RTFM'ed quite a bit on ants. They're very interesting little critters in their own right. I picked the eyes out of the various behaviours of a bunch of different species of ants to come up with one that made a fun simulation (refer references below).
The bare mechanical simplicity with which some of these critters operate is really quite amazing. Take, for example, the concept of trail laying. I guess it's pretty widely known that many species of ants lay trails from food sources back to the nest to guide other ants to the food. (Try: find a line of ants climbing up the wall in the kitchen or somewhere, moisten your finger, wipe straight through the line (washing off the trail). They'll be all disoriented for a little while, but they'll quickly re-establish the trail, largely by random search). Anyhoo, what's really quite cool is how one species does it. The trail is just an emission from the back end of the ant that wipes along the ground as it walks. The mechanics are such that if the ant has a full crop, it puts pressure on the digestive tract, and forces stuff out the back. If its only lightly fed, it only forces a bit out the back, if its had a big feed, it forces a lot out the back and lays a denser trail. The outcome is that the ants lay stronger trails to the better food sources. Elegant, isn't it!
I could go on forever, but I won't. Some references below. Another behaviour that is probably even more interesting than trail laying is navigation. They're absolutely amazing. Various ants use various combinations of reference to the sun, counting the amount of ground that passes underneath them as they walk *AND* remembering turns!!!, and reference to major landmarks as they travel. Did I say amazing?
Anyhoo, here's a bunch of references on ant behaviour if anyone's interesting.
I find your ideas intriguing and I wish to subscribe to your newsletter.
Kind of like the continuum going from observing at the atomic level to observing at the macroscopic level. The physics of the atomic level is VERY different then the physics of the macroscopic level. Understandably, when you get to a point where you can't use one model over the other, things can get pretty hairy.
Can you say that the atomic level CAUSES the macroscopic level, i.e. one level emerges out of the other? My feeling is, it doesn't make much of a difference. The interactions you get depend upon your level of observation- they don't necessarily depend on what the interactions are at a different level. Observing at both levels is useful for different reasons. For example, for most low-speed aerodynamics, the model of air that you use is streamlines in the flow. For this situation and it's goals, it doesn't much matter to the airplane what is going on at the atomic level. The airplane is on a macroscopic level, so what matters is the physics of the macroscopic level. You fit the model to the same level and dementions of the thing you're observing. Remember calculus and limits? The limit works because it creates a fundmental building block of experience in a relavant dimension. Ex: dt is an infinitely small measure of a direction in time. But time is relevant in the demensions of the thing you're observing (actions the real world), so it's useful for the theory about the real world you want to create.
Due to my study on how people work, and that there are fundamental principles of human interaction that apply regardless of the individual person, I think it's probable that they're are fundmental principles of the overall interactions of an ant colony (we may not know them yet, but they are there). It's just that if you observe the colony at the microscopic level, you may not find them, since you're looking in the wrong place. Chaos Theory shows that even when behavior appears random, there are principles which create the randomness.
I guess it's nice these people have their "new" science to investigate- emergence. Hey, if it creates some new thought and gets people interested, I'm all for it. But I don't think that one thing is emerging out of another: it's just observing that for this particular level of observation, traits that were appearent at other levels have a bearing on the problem.
It's all the same thing, just different levels with different rules. (For example, duality is pretty much a law in the universe. You can't really equate the things that compose the duality, you can only recognize that the duality exists.)
"In our tactical decisions, we are operating contrary to our strategic interest."
Another good book on the subject of emergent systems is After Thought by James Baily. It is a quick and enjoyable read that takes a look at the evolution of mathematical and philosophical attempts at describing our universe from the ancient Greeks to modern day scientists. Specifically, he focuses on how we attempt to model the human brain electronically, and touches on parallel computing, cellular automata, genetic algorithms, and the techniques required to allow a machine to learn.
I only post comments when someone on the internet is wrong.
I found the following to be surprising and useful background for the glut of writing about complexity/emergence/universality etc. Lots of historical detail from J. S. Mill onwards about the use of emergence in philosophy. Good bibliography too of which I can recommend the Kaufman books as good fun:g ent/
http://plato.stanford.edu/entries/properties-emer
JESUS FUCK. You're telling me even Asimov couldn't distinguish between ITS and IT'S??
I have read the book, and that is not the impression that I, nor many other people who've read it, came away with. On the contrary, he repeatedly refers to it as his discovery- "...the new kind of science I [emphasis mine] have developed...", "...my discovery...", "...one of the more important single discoveries in the whole history of theoretical science..."
That's just in the first chapter, but it continues throughout. He makes little reference to others' work in the field, pretty much dismissing all work done prior to his becoming aware of the subject.
It is a good indicator of how lost in one's own world one is when none of the major peer-reviewed journals or scientific publishers is interested in "one of the more important single discoveries in the whole history of theoretical science". If a minor patent clerk can get his groundbreaking ideas out, one would hope an MacArthur award-winning, CalTech PhD-holding, Mathematica creator would be heard.
The whole book sounds to me like a guy who's been told how smart he was all his life, and is probably surrounded by sycophants, who has a lot of money and might have smoked a little too much weed.
It's okay, though, Stephen- the pictures are lovely.