Why doesn't AI Falsify ID?

Hi - I'm new here, and I'd like to start a discussion in the Intelligent Design forum.The issue is this: It's clear that appropriately constructed computer systems can exhibit behaviors that qualify as "intelligent" (as described, say, by Bill Dembski). At first blush, then, it would seem that artificially intelligent computers represent an existence proof that deterministic (or stochastic) physical systems can generate complex specified information, contrary to the tenets of ID.I've generally heard two responses to this. The first is that computers don't actually exhibit intelligent behavior at all. I won't reply to this argument in anticipation here, except to say that finding a criterion for detecting intelligent agents that serves the needs of ID theory while excluding computers seems to be quite impossible.The more common response is that while computers might appear to be intelligent, they are only reflecting the intelligence of the real intelligent agent - the human programmer. But a moment's reflection should reveal the flaw in this line of reasoning: If computers are not truly intelligent because they are the product of another intelligent agent's design, then human beings - also the product of intelligent design according to ID - must not be truly intelligent either. If the IDist chooses to rebut this reductio ad absurdum by granting that computers are in fact intelligent in their own right, we are left with the conclusion that material processes must be capable of intelligent behavior after all.Any thoughts on this?Edited by aiguy, : No reason given.

---

Science is not simply reason - it is much less than that. It is reason constrained by empiricism.

Hi, GDR - There are are number of different problems with your response.First, while there is apparently no canonical definition of the term "intelligence" used by Intelligent Design Theory (an odd state of affairs, to say the least), none of the ID proponents I've read have suggested that "love", "joy", or "guilt" are requisite components. According to Dembski, "intelligence" is defined as "the ability to generate complex specified information". Computers most certainly have the ability to generate all sorts of designs, including those for irreducibly complex machines. One can also imagine a person who lacks the emotional responses we might call "love", or "guilt" but is still clearly intelligent (for example, a psychopath who happens to be a nuclear physicist).Second, obviously there is no scientific way for ID to infer that the Intelligent Designer expresses love, joy, or guilt either - but that doesn't stop IDists from claiming that the Designer is intelligent.Third, of course my computers can express love, joy, and guilt. In fact, the computer you are looking at right now will express love. Observe:You may not find this expression of love very convincing, but I'm not talking about subjective impressions here - I'm talking about science. Neither you nor anybody else has any scientific method to establish when something does or does not actually feel love, joy, or guilt.So, your complaint that computers don't currently seem to exhibit certain human emotions misses the mark. You've made a bald assertion with no explanation nor support, so it's difficult to respond to this. Also, nobody has said that humans are nothing more than intelligence, so you seem to be attacking a straw man here.

---

Science is not simply reason - it is much less than that. It is reason constrained by empiricism.

Hi, GDR - I understand that was your point. I made three arguments against that point, which I will summarize again:
1) Emotion (spontaneous or otherwise) has not been established (by ID theory or any other theory) as a necessary component of intelligence.
2) No ID theorist has even attempted to attribute emotion to the Intelligent Designer, yet ID theory holds that the intelligence of the Designer is still scientifically warranted.
3) There is no scientific basis for assessing whether or not a computer, or the cause of life, has emotions or not. You have not addressed my reductio argument: If the fact that engineers design computers means that computers are not really intelligent, then by the same reasoning, the fact that the Intelligent Design designed humans beings means that humans are not intelligent either. That is absurd, so something must be wrong with your assumption of design, or your reasoning about computer intelligence. How do you resolve this dilemma? In that case, we are in agreement. I'm new here, so let me lay my cards on the table: I happen to think there is probably something fundamental missing from our understanding of how life came to be, but I think "Intelligent Design Theory" is a vacuous semantic sleight-of-hand, utterly lacking in scientific merit. Too bad I'm fishing for ID proponents here... are there any? I basically agree with you here as well. More specifically, I think that "computers" will have to be built in a very different way before we might think they have mental experiences like we do. However, I disagree that emotion has anything to do with intelligence as the word is used in ID theory. (I think it has a lot to do intelligence in human beings, however, cf. the work of Antoni Damasio - Antonio Damasio - Wikipedia). Please review my reductio argument. Let's try this again. My argument is this:1) ID theory says "intelligent causation" is responsible for biological complexity.2) ID theorists define "intelligence" in various ways, but for any definition that has any hope of being scientifically useful in this context (e.g. "the ability to generate specified complex information" or "the ability to design irreducibly complex machinery" or "the ability to choose among different possible outcomes", etc), computers are clearly intelligent.3) If one denies that computers are intelligent on account of the fact that an intelligent agent (a human engineer) designed them, then we are forced to conclude one of two things. Either a) Human beings were not designed by an intelligent agent; OR b) Human beings are not themselves intelligent.4) Since (3b) is absurd, I believe I've shown that ID's claim that material processes cannot account for the generation of novel, complex form and function must be false.Edited by aiguy, : No reason given.

---

Science is not simply reason - it is much less than that. It is reason constrained by empiricism.

Hi nerd, The problem that intelligent computers poses for intelligent design is this: ID theorists attempt to demonstrate not only that evolutionary theory can't account for biological complexity, but that no combination of chance and necessity can account for it. All of Bill Dembski's claims about the "no free lunch" theorems, or that information is conserved, or that his explanatory filter can distinguish intelligent cause from chance and law-like causes... all of these arguments are intended to demonstrate that intelligent behavior can not arise from material mechanism. Otherwise, simply pointing to failures of evolutionary theory wouldn't amount to evidence for ID at all.But computers are in fact devices that operate according to material mechanism - deterministic physical law, optionally combined with random input. So, if you think that computers are intelligent, then you believe intelligent behavior can arise from purely material means. If the chance and necessity operating inside a computer can generate complex designs, then why couldn't the chance and necessity of evolutionary mechanisms do the same thing?Dembski sees this quite well, and has attempted to discredit AI (and the cognitive sciences in general) in order to preserve his arguments (see http://www.asa3.org/...yNeuroscience/PSCF12-1990Dembski.html). His arguments are embarassingly bad, but he does recognize that the reality of computer intelligence does in fact destroy all of his arguments for ID.

---

Science is not simply reason - it is much less than that. It is reason constrained by empiricism.

Hi TD, The basic problem I have with ID is that there is no attempt to actually provide a standard definition for "intelligence" at all. If you take (one of) Bill Dembski's definitions, i.e. "Intelligence is the ability to create complex specified information", then ID can be seen to be utterly vacuous:Q: What explains the specified complexity of biological systems?
A: Intelligence.Q: What is intelligence?
A: The ability to create specified complexity. It was really almost thirty years ago, in the AI heyday of the early 1980's, that we thought real AI was around the corner. Then "AI Winter" set in, many of the commercial ventures based on AI technology fizzled, and researchers discovered that the core problems were much more difficult that we believed. Since then, however, there has been steady progress, and some of us in the business are again (cautiously) optimistic that machines that give us the impression that there is really thinking going on inside may be a reality in the next decade or two. The big hurdle was the recognition that huge amounts of common-sense knowledge was needed for something to act intelligently, and the repositories that people have been laboriously creating for the past twenty years or so are just now beginning to be usable. The other change, of course, is that computers are now so incredibly big, fast, and cheap compared to previous decades.

---

Science is not simply reason - it is much less than that. It is reason constrained by empiricism.

Hi GDR, In that case, you and I basically agree. I just think we don't have a very good idea of what we mean by "intelligently designed". If you ask 10 people what they mean by "intelligence", you will get 15 or 20 different answers. No, you're mistaken about this. There are two ways computers might come up with different answers to the same question. First, computer systems (like people) are affected by their experiences and can learn. So the state of the computer the first time you ask the question may be different from the next time you ask, and you might get a different answer. The second way is that AI programs sometimes incorporate randomness. Simple little hacks like the "chatbots" on the net use random input to mix up the responses a bit, but sophisticated AI programs use it to explore different avenues for solving problems when the program is stumped. Unless someone attempts to position ID is science, then I have no problem with it. I'm partial to some pretty mystical hunches about the universe myself. This is always the problem in AI. I made your computer tell you it loves you a few posts back, but you didn't really think your computer loved you. Why not? What precisely would actually convince you that a computer could love - or have any other emotion? What scientific test would you suggest? Aha! That is certainly the heart of the matter. It is incredible to me that something called "Intelligent Design Theory" exists, and nobody has bothered to offer a technical definition of the term "intelligence"! What you'll find is that there is no definition of intelligence that suits ID's purposes. Either the definition is subjective and untestable, or the definition includes things that we really don't think of as intelligent.Example: "Intelligence is the ability to choose between different possible outcomes"
Counter example: "A thermostat chooses between different possible outcomes"Example: "Intelligence is the ability to freely choose an outcome without being determined by physical structure"
Problem: There is no scientific test to see if something is "freely" choosing or not. There is no evidence that anything "chooses" in a way that is not determined by its physical structure.Example: "Intelligence is behavior that is neither random nor fixed".
Counter example: "The behavior of a river is neither random nor fixed" I don't know what you mean by "intelligence". All fine with me, but I'd encourage you to think about just how different this intelligence of the Designer might be.Think about how much science has learned about the connection between brains/bodies and our emotions for example. Do we have good reason, in light of cognitive science, to believe that something without a brain, without an enteric neural plexus, without an endocrine system, etc would have anything like the same sort of emotions that we have, given He presumably lacks these essential components?

---

Science is not simply reason - it is much less than that. It is reason constrained by empiricism.

Hi GDR, Thank you, but it's just that I've been at for a long time. I've been doing commercial AI research since 1981, and I've been reading philosophy of mind since before then (gee I'm old). Funny, but I almost mentioned the third way AI systems might give a different answer to the same question: If there is a bug in the program, or a hardware problem. That might equate to a computer's bad day... The same argument could also be used to conclude that humans are manufactured on an assembly line out of silicon chips, so I don't think this sort of analogy does much for you.Again, this all comes down to what you mean by "designed". If you mean a human-like intelligence thought about building humans, wanted to do it, consciously made plans to do it, pictured humans in His mind's eye, and so on... no, I don't think there is any reason at all to believe that. Intelligent behavior can emerge from pure chance and necessity. Genetic algorithms have been popular in AI for the last ten or fifteen years, and many cognitive scientists have been suggesting that random variation/selection mechanisms might underlie our own cognitive processes. Of course a computer (a robot, really) could be made to sacrifice itself, without being explicitly pre-programmed to do so. AI systems can be given some general rules and knowledge, and enabled to learn on their own, and then they decide on their own what to do. Good point.

---

Science is not simply reason - it is much less than that. It is reason constrained by empiricism.

Hi Answers, Sure, computer systems confront problems not covered by their programming all the time. Think about the diagnostics programs that run on deep space probes. Nobody can anticipate everything that can go wrong with such complicated machines and extreme environments, so those programs are given a deep understanding of how everything works and general reasoning mechanisms to try and figure out what's wrong when something stops working. Random input is not required.I know it's difficult for people to see how deterministic computers could do anything they are not programmed to do, but they do. I can't possibly predict what my programs will do - I'm constantly surprised (and usually disappointed) by how my programs react in novel situations. Remember, in a sense, we humans are "programmed" too - we are born with some built-in mental abilities and basic knowledge, and then we are programmed by parents and teachers, and we learn by experience. Just like computers.Don't get me wrong - the very smartest computer systems are still complete idiots in many ways, and there is of course nothing that even begins to approach the science-fiction AI we see in movies and TV. But the basic elements of reasoning, learning, and flexible responses to novel environments are all there.

---

Science is not simply reason - it is much less than that. It is reason constrained by empiricism.

Hi Nerd, The point is absolutely central to all of Dembski's arguments, and he states it all the time. The entire paper of his that I cited above aims to discredit the idea that intelligent behavior can arise from material cause. His "explanatory filter" explicitly distinguishes "intelligent causation" from physical causation and chance. He argues for an "expanded ontology", which is another name for metaphysical dualism, because he believes mental causation is ontologically distinct from physical causality. What would you say the source of the intelligence is? The programmer? And what is the source of the programmer's intelligence? (And, again, Dembski disagrees with you, since he denies that everything in the universe operates under physical laws). If you believe that humans are natural, and that we operate under the laws of nature, then why do you think computers do not appear as the result of the laws of nature? Again: If you believe that you too have been designed, do you then believe that you function exactly as your Designer intended you to, and any intelligence that you exhibit is not possible without Him? Are you simply executing the program that the Designer loaded you with? Well, no. Genetic algorithms, for example, use random input, but there is no "design spec" for what they come up with. Nobody can predict what GAs will do, we just have to run them and find out. I don't think I've disproved the possibility of intelligent design. I think I've shown that ID's arguments against chance and necessity coming up with complex designs fail, and so we cannot possibly "detect intelligence" that could not be due to purely natural non-mental processes.

---

Science is not simply reason - it is much less than that. It is reason constrained by empiricism.

Hi nemesis, This, then, would be your own particular definition for "intelligence". I haven't seen anything in ID literature that lists this quality as a required component for something that is "intelligent". Can you provide a reference?Also, can you say how we can scientifically ascertain that the Intelligent Designer is cognizant of itself and others? This is an assertion, rather than an argument. I've addressed both of these points. Your first argument fails because of the reductio argument introduced in the OP (would you say you are only as intelligent as your Designer designed you to be?). Your second argument fails because it is not true - computers do things that nobody "assigns" to them, just as people do things that they've never been "assigned" to do. If you think you have a testable criterion for intelligence that includes people and the Designer of ID theory, but excludes computers, then tell us what it is. So you are saying that neither computers nor human beings are intelligent? I am offering no particular definition for "intelligence" at all. I believe the word is scientifically useless, in that while it can be used to loosely describe some sorts of behavior, it cannot be used to explain anything. I object to ID because it claims to be an explanatory theory that offers "intelligence" as its sole explanatory concept, but then fails to provide a standard technical definition for that term. I think it's a bit odd for us here on this forum to begin making up our own definition for this word, given there is supposed to be a scientific theory that explains biological complexity by invoking "intelligence".Aside from that, your attempt at a definition is not too good: First, there is no scientific test for "will" (but you may be interested to learn about scientific investigations into free will - read this: http://www.consciousentities.com/libet.htm). Second, you haven't supported your assertion that intelligence requires emotion. Third, why do you think learning is essential for intelligence. Do you believe that the Designer learns - or doesn't He perhaps know everything already?So, as your attempt at a definition illustrates, either the definition is untestable, or it does not manage to include the Designer and humans while excluding computers. Most AI researchers (including me) believe that human-like cognition requires much more complex physical interactions with the world that has been attempted to date. In other words, we won't really build a realistic aritificial mind until we also build a more realistic artificial body. Edited by aiguy, : typo

---

Science is not simply reason - it is much less than that. It is reason constrained by empiricism.

Hi TD, This old trick simply conflates "survival" of an individual, or set of individuals, with "survival" of a set of heritable traits. Evolutionary theory doesn't explain why any particular individual or species continues to reproduce or not - it explains how traits become fixed in a population. So it is really:Q: How does complex form and function appear in organisms?
A: Because random, heritable variation confers reproductive advantage to subsets of populations, and so these changes accumulate over generations.However, ID's attempt to explain biological complexity by "intelligence" (without actually describing what "intelligence" is or how it works) is in fact a vacuous tautology:Q: What explains the specified complexity of biological systems?
A: Intelligence.Q: What is intelligence?
A: The ability to create specified complexity.Edited by aiguy, : No reason given.

---

Science is not simply reason - it is much less than that. It is reason constrained by empiricism.

Hi Nerd, Funny, but I get this a lot. Dembski is the most visible and prolific "theorist" of "Intelligent Design Theory", but I have yet to meet somebody on any forum who is willing to defend his views. If you agree that Dembski's position is untenable, we can drop it... but then, can you refer me to some articulation of ID Theory that you do believe? You had said that a computer's intelligence does not arise from its mechanism, so I asked you what you thought it did arise from. Then I pointed out that if you suggest a computer's intelligence arose from the programmer's intelligence, you have simply offset the question, and needed to answer what you thought the programmer's intelligence arose from. I believe there is no scientific fact of the matter whether we say X is "intelligent" or not. Saying something is "intelligent" is just a loose way of speaking, conveying a general sense that something can reason in some way or other. It's like calling something "athletic" - it isn't an explanation of anything, nor is it a testable scientific category of things in the world. It's just a label that we use in informal conversation. You had said everything operates according to natural law, but then you said that computers did not appear as the result of natural law. So I questioned how that could be: Do you think that the actions of human beings (as when they create computers) somehow violates natural law? If not (and you had said not) then you must concede that whatever we do (including create computers) is "natural", and so computers do appear as the result of natural law. Of course people create computers, but as I explained again above, I think we've agreed that people are part of nature. I'm fine with that. (I assume we're leaving Darwinian evolution out of the discussion, which is ok with me). I meant that chance and necessity operating inside a computer comes up with complex designs. Yes (but not all AI design programs are GAs). Sure. Here's just one example I'm familiar with: http://www.mrc.uidaho.edu/~knoren/GAs/B-159_paper.PDF A selection criteria is not the same as a design spec. If I tell you to design an op amp, and tell you what it must do, and you go off and actually figure out how the amplifier should work, lay out the circuit, and build it, then which of us has actually designed the amp? I am not saying that any material thing is by definition not designed. I am saying that purely material cause - chance and necessity - has been demonstrated to be capable of designing. Again, if "sole catalysts" is your criterion, then human beings can't be shown to produce designs either. If you agree that humans can design things (even though they themselves were designed, and programmed by their parents and teachers), then you should agree that computers can design things (even though they themselves were designed, and programmed by their programmers). My point is that since we know that chance and necessity can produce designs, we cannot assume that chance and necessity did not produce biological designs - even if we do not understand everything about the material processes involved. ID theorists (like Dembski and Meyers and Johnson and Berlinski...) all attempt to show that chance and necessity cannot produce novel designs for just this reason.Edited by aiguy, : typo

---

Science is not simply reason - it is much less than that. It is reason constrained by empiricism.

Hi Nerd, The only answers I think science must leave out are the ones that independent researchers can't reliably test. I won't defend evolution theory as a being testable (I think parts of it are and parts are not), but I'm pretty sure that "the God option" is not something we can ground in empirical observation. I agree that there is no scientific definition of intelligence - but I don't think it's a philosophical problem either. I don't think there is a real referrent for the word at all. There is no unary property that some things have and some things don't have called "intelligence". It is very much like "life" in that regard. The vitalists believed that "life" was a property that some things had and some things didn't, and spent a long time trying to figure out what it was. In the end, they gave up - because there wasn't any. Again, I would have to say there is simply no fact of the matter either way - it is a matter of definition and not discovery. Does "athleticism" exist? Sure, you can something athletic if you want, but we'll never really agree on whether Tiger Woods is an athlete, or if a cheetah is althletic, or an elephant. Both of these words are just informal descriptive terms that have no scientific utility. That is why it can't serve as an explantion for things. Observe:Q: Why does a cheetah run so fast? A: Because it is athletic.
Q: How do you know it is athletic? A: Because that is what we tend to call things that run fast.In the same vacuous way:Q: Why do complex biological structures come to exist? A: Because of intelligence.
Q: How do you know the cause is intelligent? A: Because that is what we tend to call things that make complicated structures.In contrast, real explanations add knowledge:
Q: Why does the apple fall from the tree? A: Because of gravity.
Q: How do you know the cause is gravity? A: Because gravity acts in a well-defined way on all objects with mass, and independent researchers can measure mass in multiple different ways... I'm not sure if we agree on what "design" is or not. Can you provide an actual definition of the term that could be used in a scientific context? Hmm, I thought my case was pretty clear.... Can the same human engineer who designs an electronic circuit also design a biochemical enzyme? I can do the former, but not the latter, and my biologist friends can do the latter, but not the former. I'd still say we can both design, though. If you took all of the various AI design programs and ran them on the same computer, you'd have a system that could design all sorts of things... just like a person. And of course AI is getting better all the time. So what? You think because humans design computers, that means computers can't design things on their own? Do you think because an Intelligent Designer designed humans, then humans can't design things on their own? AI is artificial, meaning human-made, yes. I disagree. Let's define "intelligence" as the ability to create irreducibly complex structures. We test humans - yup, they can design IC structures. Then we test computers - yup, they can too. Now, some people think that human minds do not operate according to chance and necessity, but everybody agrees that computers do. So we have just scientifically demonstrated that chance and necessity can indeed give rise to intelligent behavior, QED.And so, ID cannot claim that IC structures in biology are indications of anything but chance and necessity working in nature.Edited by aiguy, : No reason given.

---

Science is not simply reason - it is much less than that. It is reason constrained by empiricism.

Hi Nerd, That's too bad. Perhaps some other ID-friendly sort here would like to try and show why AI doesn't falsify ID (or, more precisely, render ID moot - ID is of course always possible, but since chance and necessity can account for complex specified information, we have no reason to postulate anything else). I'm afraid you're quite mistaken. Many AI programs like genetic algorithms (as well as other types of programs, like statistical analyses) incorporate randomness. Usually pseudo-random generators are used, but some systems actually employ physical devices that transduce random physical events (like radio static or radioactive decay) into random input for the computer.So, many AI systems do use chance. And they all use "necessity" (which is simply referring to these deterministic laws of electricity, etc). Chance and necessity - that's how computers work. Just like evolution (and perhaps like brains too). It is not controversial how computers come into existence - we already know how that happens. The question is how does the complex form and function of biology come into existence.ID argues that only intelligent things can design complex machines, and that purely natural processes (i.e. those that consist of only chance and necessity) are incapable of designing complex machines.I have argued that since computers can design complex machines, and computers operate according to purely natural chance and necessity, then ID has no basis for claiming that natural processes couldn't give to rise to the complex machinery of life.Your only rebuttal is that computers are themselves designed by humans. However, this argument is a red herring. Nobody thinks that human beings designed life on Earth! So we eliminate human beings as candidates for the explanation of biology. What's left? Well, we know that pure chance and necessity can generate complex designs, so that's the best explanation. Why posit anything else?

---

Science is not simply reason - it is much less than that. It is reason constrained by empiricism.

Hi Quetzal, Please do. The odd thing is that I have yet to engage somebody on these any of these forums who is willing to defend Dembski-style ID. I've met out-and-out creationists, various flavors of vitalists and animists... and mainly those who just like God and don't like science and refuse to engage in actual debate.You made very good points; here are my thoughts:The enterprise of AI proper (the computer science discipline, as opposed to philosophy of mind or cognitive science) is quite explicitly aimed at reproducing human mental abilities. There is no pretense of establishing a general theory of intelligence, and that is not our goal. As you point out, Turing recognized "intelligence" is not a property that can be objectively defined and tested, so AI's operational definition is simply "those sorts of behaviors that people tend to call 'intelligent'". So while AI is not actually restricted to human intelligence (e.g. there is a lot of research into insect-like intelligence), we don't pretend that "intelligence" is anything other than a loose, subjective, informal way to categorize certain types of behaviors that strike people in general as being clever.The enterprise of ID, in contrast, is built upon a monumental equivocation. Clearly, if one takes Dembski's basic definition of "intelligence" seriously, ID is perfectly vacuous: ID explains specified complexity by "intelligence", which is scientifically defined as "that which can generate specified complexity".But by using the word "intelligence", which is laden with all sorts of folk-psychological connotations relating to human mentality, ID sneaks in any number of implicit and anthropomorphic assumptions that nobody could possibly support empirically. So while Dembski fills books with completely irrelevant mathematics to make it seem that we can scientifically detect something, that something is assumed by the unwashed masses to be a human-like mind. Free will, consciousness, emotion, empathy, intent... all sorts of these intuitive or philosophical concepts are what people typically associate with "intelligence".So the equivocation of the century is for ID to attempt to empirically support one meaning of the term (which is in fact tautologically, necessarily true of evolution or any process that can create complex designs) while pretending that all of these other distinctly non-scientific meanings of the word are somehow scientifically warranted as well.

---

Science is not simply reason - it is much less than that. It is reason constrained by empiricism.