Slashdot Mirror

I just don't believe that every "big question" is at the extreme ends of the scale where cosmologists and particle accelerators seek their answers to the big "Why?"

Newton, Darwin and the others developed their great theories for things a lot closer to human scale and, while some may have felt they were reading the mind of God, they were also clearly focused on the problem at hand. (I say this as an admirer of the interdisciplinary efforts at the likes of Santa Fe Institute, as much more a generalist than a specialist.)

All I ask is a bit of balance so that enough weight is given to research at interplanetary and interstellar scales that we might give ourselves a chance to find their equivalents of plate tectonics, to my mind the most elegant piece of science since Darwin.

I am aware of this project quite a long time, since I'm working on a similar project called "Swarm-Bots" [web site www.swarm-bots.org]. According to me their use of the term "pheromone" is not more than a catchy adjective to label their work. The research, as displayed on their web site, does not take many ideas from the ethological studies of ant colonies. For instance the robots communicate directly with each other, NOT through the environment, which is what ants use pheromone for. To me, it is merely an integration of the dynamic programming technique with mobile robots coupled with VR display interface. For those new to the subject, there is a new approach, called "swarm intelligence" that aims to create intelligent systems from a group of distributed simple agents. An excellent description of this approach is available in the "Swarm Intelligence: From Natural to Artificial Systems By Eric Bonabeau, Marco Dorigo, and Guy Theraulaz" . In this approach, the agents communicate through the environment, called stigmergy, to achieve group level tasks. There is no centralized control, yet the whole system is very scalable and robust. I hope to report some news on the progress of the Swarm-Bots soon.

This paper came out last year. I found the analysis to be 10 times better than the conclusions that the Nature reporter jumped to (not to mention all the previous posts here, good lord!!).

To me, the key conclusion here is that the observed behavior IS rational self-interest. The previous ideas of self-interest I've encountered in economics always seemed to be distillations of raw stupid greed, without any allowance for intelligence on the part of the actor, and lacking any model whatsoever for social forces, e.g., shame, embarrassment, fame or notoriety.

This is the fundamental flaw in the idea of "the tragedy of the commons" - it assumes that the actors who overuse the commons have no social relationship to each other, and that their overuse will not be penalized in other arenas, which is of course totally ridiculous. This kind of theorizing has no use in the real world.

I normally don't reply to anonymous cowards, since they aren't very credible... ;} And if you'd follow the links I provided, you'd find plenty of citations and web links to "credible" sources of information.

However, in this case, I'll make an exception.

Check out:
Complexity International (a refereed journal) Santa Fe Institute (assoc. with Los Alamos Nat. Labs) CiteSeer ResearchIndex of Scientific Papers

"The Origins of Order" by Stuart Kauffman is an excellent book on this subject. It's pretty heavy reading, plenty of math and experimantal models, with a focus on biology and how the order we see in living systems arises out of chaos.

Since reading and doing an undergrad seminar on Symmetry in Chaos by Marty Golubitsky and Mike Field several years ago, I've been quite interested in this topic.

A more serious alternative to Emergence might be The Self-Made Tapestry: Pattern Formation in Nature.

I know very little about the politics of he "globalization" issue, and I hope I find a chance to learn more about it. Definition of concepts is always difficult and usually you just have to accept it. Context-sensitivity of words is very common.

One interesting point of view might be mathematics, or more exactly, studies of complex nonlinear systems (you know, the freaky chaos people).

There are numerous studies concerning the behaviour of complex nonlinear systems interconnected in different ways. The research of Stuart Kauffman (a theoretical biologist) is perhaps the most well known, as well as other research from the Santa Fe Institute.

One aspect is simply evolutionary - globally interconnected systems tend to converge fast, while sparsely interconnected systems (such as 2d-lattices) tend to converge slower, but they have higher diversity, which often results in better overall solutions.

Also, highly interconnected systems are rigid because each connection is also a constraint. I don't really know how to apply this to economical globalization. The problem is that the human culture is interconnected in so many ways and on so many levels. Globalization might force a radical self-organizational change in the connectivity structure of humanity, by reducing connectivity in many aspects, or in other words, reducing diversity.

One significant problem in many complex systems is that simple changes at a lower level of a system (in parameters or laws) can result in emergence of totally unpredictable and often undesirable effects in large scale.

Some call this "the invisible hand". It's a pretty well-known concept in many scientific fields, especially the science of finance and economy.

For example, globalization of economy forces countries to compete with their laws to get foreign investments and workers. The result is that companies control laws very effectively. Sometimes this may be good, such as for preventing wars, but quite often not. For example, countries that have stonger social balancing system may suffer in short-term economical competition, as their taxes can be forced to too low level.

Unfortunately, just like the watchmaker of biological evolution, the invisible hand of market economy is blind. Just like other nature, it doesn't have ethics nor does it care anything about humans, and is thus sometimes undesirable.

I mean, corporate life, it will find a way, and then comes the running and the screaming.

I'm not sure if this helps the terminology issue much, but hopefully it gives some directions.

Einstein did not help build the bomb. He wrote an influential letter to Roosevelt supporting the bomb effort. He made some contribution to gaseous diffusion, which is used to refine Uranium into weapons-grade material. That was the extent of Einstein's contribution. He did not work on the Manhattan Project.

Another misconception about the bomb is that relativity theory (E=mc^2) is somehown necessary for the design or conception of a nuclear device. This is simply untrue. Follow this link if you doubt the veracity of the previous sentence. Bomb design is based on basic nuclear physics, and the energy release can be calculated from electrostatic considerations.

I was surprised and pleased to see references to Stuart Koffman in this essay.

The research being done at the Sante Fe Institute with regards to complex adaptive systems, and the nature of complexity in general provide a number of insights to coders writing large software projects (and many other discplines...)

I would highly reccomend At Home in the Universe as a good introduction to the ideas behind research in CAS. ISBN: 0195111303

For those who like more thorough and academic texts the S.I. produces a number of conference and workshop transcripts which are chock full of great papers and enlightening discussion. ISBN: 0201626063 is a good one.

As software/hardware systems grow ever more complex, we will need to apply ever more powerfull methods to manage this complexity. Perhaps by learning from the experiences of millions of years of evolutionary biological computation to socio/economic progression and interaction we can begin to fashion methods of building software/hardware that can adapt and scale in ways we dream of...

While I do like Ruby, it doesn't have the support behind it that Python does. Thats why I use Python, and not Ruby.

This is a nice example of path dependence. You use it becuase more people use it, and so on and so on. Things that don't catch on sometimes don't catch on because of tiny, idiosyncratic, reasons, but then the competition snowballs. The canonical example of the QWERY keyborad is overused (and sometimes disputed), but you get the idea.

Besides, there are so many scripting languages. David Korn pointed out here on /. that ksh can do most anything perl can. Why not use ksh, then? Ad nausaeum.

While I do like Ruby, it doesn't have the support behind it that Python does. Thats why I use Python, and not Ruby.

This is a nice example of path dependence. You use it becuase more people use it, and so on and so on. Things that don't catch on sometimes don't catch on because of tiny, idiosyncratic, reasons, but then the competition snowballs. The canonical example of the QWERY keyborad is overused (and sometimes disputed), but you get the idea.

Besides, there are so many scripting languages. David Korn pointed out here on /. that ksh can do most anything perl can. Why not use ksh, then? Ad nausaeum.

(Story about Terry Gilliam finding his Usenet newsgroup, reading it, posting to it, and people not believing it was him...)

The very same thing happened with Douglas Adams. He visited the Santa Fe Institute when I was there (I think it was around 1993 or 1994). We showed him his newsgroup, which he got a kick out of. He posted replies to a couple of questions. Nobody believed it was him. (Adams was a really interesting guy in person, too.)

All the ad hominem attacks aside, Greenspun is a writer and thinker of significant clarity. Agree with him or disagree with him, but you always know where he stands and what he believes in. Even though I don't know anyone at ArsDigita and had no idea he'd left, I knew when I read this stupid piece of corporate bullshit that the real Philip Greenspun could never have had anything to do with it. Think I'll use it next time I teach George Orwell's "Politics and the English Language." So, I'm relieved. Relieved to know that someone like Philip Greenspun hadn't turned into just another corporate hack who can't even say what his company is about without confusing more than he clarified.

You can totally predict the evolution of Conway's game of Life, and other deterministic cellular automata, given the initial configuration. It's not necessary to solve the halting problem in order to predict the state in the future -- you just execute the completely deterministic rules. Simple. Conway's Life is awkwardly Turing complete, but it's inefficient for the purpose of general computation (much less efficient that a Turing machine). But at least it means that theoretically you could implement a higher level of Conway's Life (or any other computable function) in terms of a lower level Conways' Life implementation, but it would take a whole lot of time and space.

Andy Weunsche at the Santa Fe Institute has come up with a beautiful way to plot out the deterministic state map of any cellular automata rule: it's a colorful branching graphical fish-eye tree representation of the topology of every possible state and transition of a cellular automata rule (the basin of attraction fields).

You can see for yourself how a given cellular automata rule is completely deterministic, by viewing all the possible interconnected states at once. "Garden of Eden" states (that there was no possible way to arrive at through the rule, so they must be original conditions) are drawn at the extreme tips of the branches, that converge into cycles of the basins of attraction (repeating dead-ends where there is now way to break out). This is really wonderful stuff, well worth scrolling through the whole gallery:

http://www.santafe.edu/~wuensch/gallery/ddlab_gall ery.html

On the other hand, the halting problem has to do with one program's ability to predict if another program will halt (not to just simply simulate the program's execution at a higher level: because if the other program doesn't halt, the simulator will never halt either, therefore failing to give the result). It means that there are undecidable questions that a deterministic Turing complete program can't answer: even if the answers are out there somewhere, they just can't be reached by a Turing machine. It also depends on being able to represent any program as data (a number), that can be given to another program as input, which is essential to the Universal Turing Machine in "On Computable Numbers".

The paradox can be demonstrated by asking such a hypothetical program (called "HaltingProblem") to predict whether another subtly (yet insideously) modified version of itself, called "HaltingProblemNot", will halt.

Given a program "HaltingProblem" that attempts to predict if another program halts (taking as input data that program and its inputs), you can always construct another program "HaltingProblemNot" to give it as input, for which it will never be able to give you a correct answer.

"HaltingProblemNot" just has to call the first program "HaltingProblem" as a subroutine, and then it inverts the return value (not just logically, but by halting if it says it won't halt, and infinitely looping if it says it doesn't halt). An obnoxious trick (called diagonalization), but it's proven to work every time. The fatal Achilles' heel of logic -- Godel strikes again.

No matter how cleverly written, the original program "HaltingProblem" is doomed to fail given "HaltingProblemNot" and another program as input, by either looping infinitely or returning the incorrect result.

This does not mean the mind is any more powerful than a Turing machine, nor unpredictable. Nobody really knows for sure. The only thing we know for sure is that there are many things we'll never know.

Gilda Radner summed it up:

Years ago I implemented the rule described in the book, and a cell editor that I used to make a few circuits. Here's a circuit for a 2-d cellular automata rule that supports passing signals along wires and performing logical operations, with cross over, fan out, conjunction and negation -- all you need! Theoretically you could program anything -- but timing is everything.

http://catalog.com/hopkins/art/circuit.gif

The top left is an "or" gate, with a couple of looped inputs that endlessly repeat the same values, and a graphical "ground" that the output flows out of (not connected to anything). To its right is an xor gate, and after that are a couple loops and delay lines. Below those is a half adder, and a criss-crossing fan out that duplicates the pair of signals a couple of times. Below that are a couple of half adders composed together to make an adder with carry. Next to last is a part of a circuit that I can't remember what it does, and under that is a pulse widener, that duplicates a signal, delays it, and ors it back into itself to make it into a longer train of bits.

It was really fun making these, since the simulator and the graphics editor were running at the same time, so it was like soldering live logic, with signals flowing through it in real time!

I've lost track of the original Forth source code for the rule, which is based on a Margolis neighborhood using a lookup table, but it's described in the book.

More stuff on cellular automata:

http://www.catalog.com/hopkins/art/cell.html

Of course the granddaddy of cellular automata is John von Neumann himself, who designed a complex self reproducing cellular automata "universal constructor" on graph paper before it was ever practical to simulate them. It's reproduced in some historical ACM monographs on computer science.

Here's an actual implementation of John von Neumann's universal constructor, which is absolutely amazing to watch going about its business of reproducing itself:

http://alife.santafe.edu/alife/software/jvn.html

The paper computers would be great for implementing cellular automata, that you could draw on with a pencil while they ran in ferrociously real time!

-Don

The author of this article(in german) compared some of the available tools with the Windows program "Reference Manager". His conclusion: There are no absolutely recommendable programs, but Sixpack, Pybliographer (and XEmacs in the bibtex mode) are strong candidates.

She suspects that a process of top-down planning that "cascades" the process of form-changing will be needed to make the system change shape quickly.

They will not be able to make a robust system by trying to control these "building blocks" from the top down. The cells in our body are not being told by a controller that they are an arm, or a kidney, the information is stored in the DNA. Yet our bodies do have arms and kidneys.

Interesting research into complex systems has shown that robust systems are not controlled top-down, but are the emergent properties of lost of small agents that are reacting with each other based on a simple set of rules.

This type of research is the holy grail for scientists in this field, but we are still stumbling on much simpler problems right now.

Moto Mannequin

"With all appliances, and means to boot!" - William Shakespeare

...of exceptionally bright kids (no, I'm not bragging on my genes, they were others' biological children), I have a few insights to offer.

First of all, there is some risk of burnout. Don't concentrate too much on beating academic milestones. This is apparently where this kid excels, but grades and proficiencies may be an inappropriate set of milestones. They can give a combination of a false sense of success and invincibility and a learning style fairly inappropriate to the real world. (I've never had an employer who paid me to take tests. But the daily work I looked down upon as a student was a much better preparation for real life than any test I ever studied for.)

One of the most difficult things for the talented to learn is how to try hard. It's one of the most important lessons around, but the gifted (in sports, intellect, whatever) often have difficulty learning it.

Just think of Ralph Sampson and Slick Watts (sorry about the sports analogies). Sampson was born with a body and coordination which gave him extraordinary opportunities. Slick Watts had the wrong body for basketball, under six foot and then he got some rare disease at 13 and lost all his hair.

But Watts learned something Sampson never did: how to try harder than everyone else he ever met. It's not that the talented cannot learn it (Bill Russell and Michael Jordan spring immediately to mind). It's just a little harder for them.

How should this translate into "tutoring a prodigy"? Many ways: throwing that football around might help, if he's interested; but the key is taking his interests to the nth degree.

Suppose he's asking questions about assembler. Show him how Alan Turing conceived of a programmable computer from mathematical concepts put forward by Goedel. Show him how machine-language derived from the precepts of Principia Mathematica and David Hilbert's famous problems for the 20th century. (If he likes fiction, The Crytonomicon is a good introduction to how Turing conceived of computers long before the technology to build them existed.) Tell him why compiler theory is emphasized in CS programs, despite the fact that so few of us end up designing compilers. Show him how Turing invented computability theory before there were computers or even transistors or microchips. Show him a simple problem he can understand which is NP-complete.

Suppose he's interested in JAVA. Get him started with some good tutorials. Then tell him what object-oriented programming is. Show him the UML. Explain why somebody would want to invent a whole new way of thinking about programming (procedures versus objects). Ask him what thinks might come after OO. Then point out that some languages have a static view of object-oriented-ness, while others are built to change if the theory changes. Ask him if he wants to accept somebody else's paradigm (Bill Joy is a good choice if you want to copy) or if he wants to define the new paradigm. Then tell him to type "aspect-oriented programming" (including the quotes) into Google. Show him Ruby. Ask him to make up a new paradigm just for fun. Then help him try to implement it in Perl (which has a dynamic OO model which forces you to redefine what you mean by "object-oriented" every time you write a program).

Suppose he's interested in physics. Have him read Aristotle's "Physics" and Newton's Principia,. Then give him Feynmann and Einstein. When he thinks that's too easy, show him Aristotle's "Metaphysics." Tell him who the Vienna Circle was and how they sought to complete science. Then give him Ludwig Wittgenstein's Tractatus. When he decides that's the cat's meow, show him how Wittgenstein renounced all that in Philosophical Investigations.

Suppose he's interested in AI. There's plenty of material on the current state of the art which tries to make it easy to understand for the beginner. Show him the Santa Fe Institute's web site (www.SantaFe.edu). Get the NOVA video on chaos theory. Then tell him not all chaos theorists are fully accepted by most scientists. Get him Complexity: The Emerging Science on the Edge of Chaos and Dynamic Memory. Teach him neural nets, then point out how it failed to live up to its promise. Ask him if he thinks that's an inherent limit of the theory or that it's caused by an inadequately developed idea. Then show him genetic programming. Then take him back to Descartes and show him the mind-body problem.

Suppose he's interested in games. Teach him to program them. There are plenty of open-source game-design projects (my web site is www.FaerieMUD.org) where he can find any level of challenge in any kind of game he likes.

Suppose he's interested in the election or social problems or whatever....

It doesn't matter. Whatever the interest, show him that he can take it to some limit which will probably exceed his grasp. Let him fail, even if you have to show him problems which have baffled mankind for millennia.

There are two keys: start with his interests and take it to his limits. Then bring him back and show him that by trying very hard he can make real progress in places where he will make a difference.

Good luck, to you and to him.

A few minutes of looking hasn't turned up proof of the original intentions either way, but here is a quote from an article posted in 1991:

Ok. Here's the main question: We all know Life is universal, but has anyone given a manageable, understandable, _explicit_ construction that proves universality? The literature I've read (such as _The Recursive Universe_ and _Winning Ways_) talks about self-replicating machines, which is fascinating in its own right, but what I have in mind is a bit less ambitious. All I want is a universal Turing machine with one semi-infinite tape. Less exciting, perhaps, but at least something for which one could give an explicit construction that could be easily verified by hand.

So it appears that there have long been proofs that Life could host a universal Turing machine, but there had been no explicit construction.

This still doesn't answer the question of the purpose of Life. At this point, all I can say is that from what I vaguely recall reading somewhere, I am right and you are wrong. But I don't think a google search on "the purpose of Life" will turn up the answer.

there's a major problem there: the assumption that there's a reasonably straightforward mapping from gene to meme. but i think that current assumptions are probably highly simplistic, and in fact, perhaps in general the mapping is not discoverable.

not so long ago, i read most of a book by Stuart Kauffman called "At home in the universe: the search for the laws of self-organisation & complexity". It became a little less solid towards the end, but the early chapters contained what was, to me, a brilliantly illuminating discussion of the way that networks of genes function.

kauffman argues that the expression of a genome is not just just the simple reading of segments of the DNA which encode proteins which go away and build things, but that a set of genes really forms a boolean network, where the action of some gene can affect the expression of another gene, and vice versa.

what that means in computery terms is that the way your genes work is less like a shopping list (gene A implies obesity, gene B implies intelligence, etc), and more like a cellular automata. if you remember some of your computer theory, you might remember that many simple CA's (e.g. Conway's Life) are Turing complete.

so what we have is essentially an evolved computer program. and if you think that some people write bizarre code, wait till you've seen some that's generated by genetic algorithm. then multiply that by billions of year's worth of evolution, raise to the power of the Halting Problem, and that's the order of the difficulty of decoding the genome!

by way of illustration of the sort of complexity that can arise when even simple systems are evolved in the real world, check out Adrian Thompson's web page. In particular, this paper has a fascinating analysis of the properties of some genetically evolved FPGA hardware. now this stuff is really simple - we're talking digital components, 100 gates, evolved to perform a simple discrimination process.

the circuit worked, but they didn't really have the faintest clue of how! because it evolved, it pushed the physics of the FPGA as far as they would go. to quote from the paper:

There are numerous tactics that can be used to piece-together answers to analysis questions even for seemingly impenetrably circuits. We applied many of those techniques to the most advanced unconventional circuit yet produced. We still do not understand fully how it works: the core of the timing mechanism is a subtle property of the VLSI medium. We have ruled out most possibilities: circuit activity (including glitch-transients and beat frequencies), metastability, and thermal time-constants from self-heating. Whatever this small effect, we understand that it is amplified by alterations in bistable and transient dynamics of oscillatory loops, and in detail how this is used to derive an orderly near-optimal output. Certain peripheral cells fine-tune particularly sensitive time delays.

that's why i argue for caution in the use of genetic engineering technology. actually, i'm not sure i do. nature has thrown so many genes together for so long that i doubt we can come up with much that does anything really useful that isn't just a simple isolated gene-to-attribute mapping.

the claims that are made for genetic engineering are way overblown - genes might be the roadmap for life, but i bet they'll be an almost completely unreadable one.

there's a major problem there: the assumption that there's a reasonably straightforward mapping from gene to meme. but i think that current assumptions are probably highly simplistic, and in fact, perhaps in general the mapping is not discoverable.

not so long ago, i read most of a book by Stuart Kauffman called "At home in the universe: the search for the laws of self-organisation & complexity". It became a little less solid towards the end, but the early chapters contained what was, to me, a brilliantly illuminating discussion of the way that networks of genes function.

kauffman argues that the expression of a genome is not just just the simple reading of segments of the DNA which encode proteins which go away and build things, but that a set of genes really forms a boolean network, where the action of some gene can affect the expression of another gene, and vice versa.

what that means in computery terms is that the way your genes work is less like a shopping list (gene A implies obesity, gene B implies intelligence, etc), and more like a cellular automata. if you remember some of your computer theory, you might remember that many simple CA's (e.g. Conway's Life) are Turing complete.

so what we have is essentially an evolved computer program. and if you think that some people write bizarre code, wait till you've seen some that's generated by genetic algorithm. then multiply that by billions of year's worth of evolution, raise to the power of the Halting Problem, and that's the order of the difficulty of decoding the genome!

by way of illustration of the sort of complexity that can arise when even simple systems are evolved in the real world, check out Adrian Thompson's web page. In particular, this paper has a fascinating analysis of the properties of some genetically evolved FPGA hardware. now this stuff is really simple - we're talking digital components, 100 gates, evolved to perform a simple discrimination process.

the circuit worked, but they didn't really have the faintest clue of how! because it evolved, it pushed the physics of the FPGA as far as they would go. to quote from the paper:

There are numerous tactics that can be used to piece-together answers to analysis questions even for seemingly impenetrably circuits. We applied many of those techniques to the most advanced unconventional circuit yet produced. We still do not understand fully how it works: the core of the timing mechanism is a subtle property of the VLSI medium. We have ruled out most possibilities: circuit activity (including glitch-transients and beat frequencies), metastability, and thermal time-constants from self-heating. Whatever this small effect, we understand that it is amplified by alterations in bistable and transient dynamics of oscillatory loops, and in detail how this is used to derive an orderly near-optimal output. Certain peripheral cells fine-tune particularly sensitive time delays.

that's why i argue for caution in the use of genetic engineering technology. actually, i'm not sure i do. nature has thrown so many genes together for so long that i doubt we can come up with much that does anything really useful that isn't just a simple isolated gene-to-attribute mapping.

the claims that are made for genetic engineering are way overblown - genes might be the roadmap for life, but i bet they'll be an almost completely unreadable one.

Patently false.

Though using evolutionary computation to solve engineering problems have been around since Dr. Schwefel and his crowd used evolutionary strategies to optimize jet engine designs in the '60s, they are surprisingly not used that often by any of the engineering disciplines.

This is blatant flamebait.

First, evolutionary algorithms, of which GAs are a mere subset, can (and often) converge on very intelligent solutions. The article is a perfect example of an evolutionary algorithm coming up with a sound solution. Second, there are some problems that are very difficult to articulate using an evolutionary algorithm. For example it's not easy to map, say, a circuit design, which can be composed of arbitrary components linked in arbitrary ways, to a genome. What do the genes represent? Do you use variable length genomes? What effect will the genomic representation have on mutation and crossover operators? What trade-offs have you made by the operators you finally decide to use?

Again, false.

Evolutionary algorithms are indeed stochastically driven searches, but they leverage off information sharing between individuals. That is, they employ so called "building blocks" that Holland and Goldburg discuss using schema theory. Though schema theory is only directly relevent to GAs, since they deal with binary representations, the same notion holds for the other types of evolutionary algorithms. "Genes" that correspond to a good idea generally are propogated to their progeny.

Again, false.

There exist some test problems for which we know a priori the optimal solution and for which EAs have been exercised against. The most famous of which is the "De Jong test suite" which poses a number of problem spaces for EAs. Generally, the EAs find the optimal solution; the speed and efficacy being, naturally, dependent on certain run-time characteristics. E.g., mutation rate, cross-over points (if cross-over is used at all), etc.

Sure, there are other machine learning techniques for solving the same problems. But the point here is refuting your claim that "since they're randomly generated, you're not guaranteed an optimal solution" which is just plain wrong. If there exist optimal solutions, GAs can find them. Moreover, other machine learning techniques may have just as difficult a time as finding good solutions; for example the problem space may be infinite and/or overly complex so that regardless of the algorithm used, you're not guaranteed that you've gotten the best. This notion is called the "no such thing as a free lunch" theorem and was, if I recall correctly, proposed by Goldburg, et al. Though you may come up with an algorithm that beats the pants off of other algorithms for a specific problem, it likely won't be as good against other algorithms for other problems; and even for that specific problem, you're not guaranteed that somebody the next day won't dream up yet another algorithm that beats the snot out of yours.

Please get your damn "facts" straight before posting on subjects that you know damn little about. Here's an online resource. Go read it.

see this article

Nrrrt!

Groups of individuals optimizing to a very simple metric can create very complex systems (see Santa Fe Institute). There may be local "bad" results from greed, but is the greater results of groups of individuals motivated by greed really that bad? I don't know, but the answer is definitely not obvious.

machines have absolutely no reason to want the same things we do?
I am in an Artificial Intelligance division at a U.S. National Research Lab (can't say which, don't want anybody to know that I'm leaking this) we are working on models of intelligance networks that use, essentially, the necessities for biological function (eating, drinking, excreting, reproducing), as an intelligance model. The network runs on easy to produce microbots (bigger than nanobots, smaller than a penny) that use electricity flowing through the air (not flowing, but emitted by various things, toned down EMP) as water, metal as food and repair (they have tools to scrape shards of metal off of a metal block and high-heat fuse it onto damaged sectors of their body), and will collect bits of metal in a storage-bay type thing, in which they will construct other micro-bots. Our project is far from being completed, but rumor around here is that we may be getting military funding, so it might get done a bit faster.

Robotic Teenage Male Sex-Daemons roving the streets looking for tasty Human Teenage Girls to impregnate with their Metal/Carbon Hybrid CoDNA
Yes, but you might have Robotic-Teenage (developing its modular components) Asexual Reproduction-Microbots roving the streets looking for tasty PentiumIII-Linux-Boxes to impregnate with their Microbot-Larvae-esque things. Wasn't my idea.

that self-guiding code that learns from failures and suffers from overcompensation--in other words, code that can even evolve under feedback loops--is pretty rare, even among the best attack detection systems
All you need is one effective system that does all of the essential life functions. And we may be closer to making that system than anybody has known before.

what some *human* has programmed them to do. Tank or Pokemon, it's made by us
It was a great experiance when I realized that this wasn't true. Tierras are mutating bits of code that, in this case, fight it out to the death. Put one of these in a positive feedback loop, and.. well.. we're using a derivitive of this idea to actually program the microbots, along with a decentralized data bank via infrared packet TCP/IP to evolve a massive collection of response data that we can moniter. The microbots will fight, like Tierras, except they will be working with actual, physical robots, instead of bits of memory. The microbots will be able to reproduce, and if we put them in a plastic room filled with old computers, they should eventually fill it up. The project is exciting, although we haven't yet got official word on the military funding.

Peole who have treated it in Science Fiction include J. D. Bernal and Cordwainer Smith

I hope NASA make it work. The real problem is keeping the sail from tearing while it unfolds. This seems trivial, but it's not.

For a list of relevant web sites, see here

The Hitch-hiker's Guide to Evolutionary Computation contains links to some online software, most of which is free and open-source.

The comp.ai related FAQs.

The about.com AI page appears to be a good starting point for many AI related web sites.

Hope this helps.

I would add to this only a few of my own personal questions from having tried to get into AI:

• How can one approach AI from a non-academic point of view? I am not in college--I have a full-time job... How can I contribute? Is there shareware (or commercial software) is out there to help study AI?
• My experience, getting into the nitty-gritty of AI, is that it's laiden with mathematics--comeplex math, at that. Is there an approach to AI that isn't founded in mathematics? What can a non-mathematician contribute to the study of AI? Or do I have to bite the bullet?
• Friends of mine in CompSci (graduates) have all said "ALife is dead". They encourage me not to persue it, "if I want to work". Do you think the demise of ALife has adversely affected AI research? (For example, Santa Fe's ALife site is stale.)

I would also love to hear your comments on Cyc. Do you think they are going to 'win' the AI race?

Let me just add that I think the creation of AI is perhaps the most critical and meaningful pursuits available to mankind, as a whole. The universe exhibits an amazing capacity to create life--to share in that power as a race is about as ultimate a purpose as I can think of for humanity.

Best of luck to you.

As am I. (Disclosure, I studied math.)

Every logician who I have ever seen discuss the topic says that Penrose completely misunderstands the contents of Goedel's theorem. And furthermore his attempted application of it to AI is misguided. Here is a short explanation of Goedel.

Strong words?

Let me give you a quick sample of why Goedel does not apply to us. Goedel merely puts a limit on what absolutely consistent reasoning can determine. However our reasoning is inconsistent - we make mistakes. (Mistakes which historically have often taken years to discover.) And so Goedel says nothing about us at all!

Cheers,
Ben

But, back to Tierra. Tom Ray was the motivating force, if you want a contact point. The networked version hasn't been released, as far as I know; the other version is released under an open source license, copied below from the original location:

1) License Agreement

Tierra Simulator V5.0: Copyright (c) 1990 - 1998 Thomas S. Ray

Tom Ray, ray@udel.edu ray@santafe.edu ray@hip.atr.co.jp (the bulk of the code)
Joseph F. Hart, jhart@hip.atr.co.jp (general programming, Amiga support)
Matt Jones, mjones@condor.psych.ucsb.edu (Mac support)
Agnes Charrel, charrel@int-evry.fr, (tping code for network version)
Tsukasa Kimezawa, kim@hip.atr.co.jp (socket code for network version)
Kurt Thearling, kurt@think.com (CM5 adaptation, parallel creatures)
Dan Pirone, cocteau@life.slhs.udel.edu (frontend, crossover)
Tom Uffner, tom@genie.slhs.udel.edu (rework of genebanker & assembler)

If you purchased this program on disk, thank you for your support. If you obtained the source code through the net or friends, we invite you to contribute an amount that represents the program's worth to you. You may make a check in US dollars payable to Virtual Life, and mail the check to one of the two addresses listed below.

This is license agreement:

The source code, documentation, and executables can be freely distributed

The source code and documentation is copyrighted, all rights reserved. The source code, documentation, and the executable files may be freely copied and distributed without fees (contributions welcome), subject to the following restrictions:

This notice may not be removed or altered.

You may not try to make money by distributing the package or by using the process that the code creates.

You may not prevent others from copying it freely.

You may not distribute modified versions without clearly documenting your changes and notifying the principal author.

The origin of this software must not be misrepresented, either by explicit claim or by omission. Since few users ever read sources, credits must appear in the documentation.

Altered versions must be plainly marked as such, and must not be misrepresented as being the original software. Since few users ever read sources, credits must appear in the documentation.

The following provisions also apply:

Virtual Life and the authors are not responsible for the consequences of use of this software, no matter how awful, even if they arise from flaws in it.

Neither the name of Virtual Life, nor the authors of the code may be used to endorse or promote products derived from this software without specific prior written permission.

The provision of support and software updates is at our discretion.

Please contact Tom Ray (full address below) if you have questions or would like an exception to any of the above restrictions.

If you make changes to the code, or have suggestions for changes, let us know! If we use your suggestion, you will receive full credit of course.

THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.

has anyone ever used the Tierra artificial life software? I downloaded it years ago; hardly knew what to do with it, as I knew nothing about C.. got it to compile, but it always locked up my 386...

This article piqued my interest in the software again, and I found some info on it, for those interested...

links:

Web page
FTP Site
Documentation

A. Liu
surak_at_my-deja_dot_com

As someone who has tried to program adaptive systems based on the concept of artificial neurons that produce good results combining with other neurons that produce good results to form new neurons that have similarities with both parents, then must compete with those parents for the right to survive in the system and combine with other neurons, your point to me is an unavoidable conundrum. Using a number of neurons that permits reasonable computing time results in identical similarity between all neurons in a short amount of time, at which point the system can no longer adapt to a change in its environment. (entropy)

I hope that you will agree with me that creationism makes even less sense. Consider that computer simulations that use a central controller to determine how cellular automatons react with each other don't display any lifelike behaviours. In order to get non-linear simulations that display complex adaptive behaviour, the way is to define simple rules that entities act on, then let then react with each other. (chaos)

Another very basic scientific law states that living tissue can not spawn from non living matter. It was once very popular that when meat was left alone it turned into flys.... anyone?

I am not sure I know the law you are talking about, let alone that there is a scientific definition for "living matter." Not that I wouldn't like to be enlightened regarding the matter.

Good analytic scientists must question the validity of the theory of evolution, becuase there are problems with the theory. But there is some good to the part about survival of the fittest, in a large enought population, to produce entities that are better suited for the environment. It should be considered as a possible part of a larger puzzle that is not quite known. There are people who are considering these questions, and interested people could start with Stuart Kauffman's work on the subject.

In the meantime, a 13TB drive might help with some of my simulations :)

"if they can stop you from asking the right questions, you'll never come up with the right answers"

As someone who has tried to program adaptive systems based on the concept of artificial neurons that produce good results combining with other neurons that produce good results to form new neurons that have similarities with both parents, then must compete with those parents for the right to survive in the system and combine with other neurons, your point to me is an unavoidable conundrum. Using a number of neurons that permits reasonable computing time results in identical similarity between all neurons in a short amount of time, at which point the system can no longer adapt to a change in its environment. (entropy)

I hope that you will agree with me that creationism makes even less sense. Consider that computer simulations that use a central controller to determine how cellular automatons react with each other don't display any lifelike behaviours. In order to get non-linear simulations that display complex adaptive behaviour, the way is to define simple rules that entities act on, then let then react with each other. (chaos)

Another very basic scientific law states that living tissue can not spawn from non living matter. It was once very popular that when meat was left alone it turned into flys.... anyone?

I am not sure I know the law you are talking about, let alone that there is a scientific definition for "living matter." Not that I wouldn't like to be enlightened regarding the matter.

Good analytic scientists must question the validity of the theory of evolution, becuase there are problems with the theory. But there is some good to the part about survival of the fittest, in a large enought population, to produce entities that are better suited for the environment. It should be considered as a possible part of a larger puzzle that is not quite known. There are people who are considering these questions, and interested people could start with Stuart Kauffman's work on the subject.

In the meantime, a 13TB drive might help with some of my simulations :)

"if they can stop you from asking the right questions, you'll never come up with the right answers"

In particular, check The 2nd Law of Thermodynamics, Evolution, and Probability to see why the 2nd law applies to closed systems, not planets warmed by suns.

Charles Darwin's views on the evolution of the eye have been taken out of context. They were part of a thought experiment on evolution. See An Old, Out of Context Quotation on that and for some intermediate steps in eye evolution.

Another very basic scientific law states that living tissue can not spawn from non living matter.

I believe that you're thinking of Lamarck and spontaneous generation. Spontaneous generation is part of abiogenesis, not evolution. Evolution comes into play when you have living organisms to evolve. See Abiogenesis FAQ for details.

If you think that "evolution is a nice idea, but pure fiction", how do you explain the results that the A-Life folks get when the implement genetic algorithms on computers? ( Artificial Life Online )

Would you be implying that Evolution is taking place? If so I strongly disagree.

On the contrary, evolution is still going on around us today. See Observed Instances of Speciation for some examples. But, you don't have to go so far afield to look for evolution. My father was nearly killed twice by a newly evolved strain of strep that was immune to dern near all antibiotics. Remember the days when a little shot of penicillin would cure just about everything? No more. Resistant strains have evolved. Now penicillin is mostly useless.

The problem is that you probably have a faulty idea of biological evolution. The shortest and clearest definition I know is, "A change in allele frequency in a population of creatures over time." (An allele is an instance of a gene, say green peas vs. yellow peas in Mendel's experiment.) Who can doubt that that happens all around us, all the time? It's simply a fact. (See What is Evolution? for a better description than I can write.)

To get back on topic, if learning is formed by growing connections between neurons, then there should be an upper limit to it that can be roughly expressed in bits. I have no idea if 13 TB is close.

Emacs is good, but XEmacs is better. It even checks your html for errors.

Perhaps you should be more specific about "it". The default html-mode isn't too bright, but I've just discovered html-helper-mode, which is highly intelligent. I use it in conjuction with hitop, an html preprocessor.

The results may be found here. ;)

Iain.

"A society that will trade a little liberty for a little order will deserve neither and lose both."

Did we forget Objective C?