Aiguy's Computer
by BilboOne of our recent, frequent participants here at Telicthoughts — aiguy — had what I consider to be a fascinating, intriguing comment here:
http://telicthoughts.com/are-dem-bunny-prints/#comment-193453, which I quote below:
Indeed - I can learn!
And so can my computer. Using genetic and other machine learning algorithms, it has learned to design complex machines that I didn't even understand until I saw them work. Not only that, but it decides all by itself what it is that it wants to design!
It really is so counter-intuitive to see how simple laws and a deterministic machine operating according to blind, natural processes gives rise to amazingly intelligent results without any intervention from me. I suppose it could be said to demonstrate "foresight" (it produces designs all at once without visible intermediate trial-and-error), but of course everything it does really is completely determined by its innate structure and its interactions with its environment. I have no reason to think it is aware of anything, so if it has beliefs and desires it doesn't know it. I really don't think it knows what it is doing - or why.
Yup, a completely unconscious, blind, unaware, deterministic physical mechanism operating according to nothing but fixed law and chance, incapable of doing anything but what it does, and there it is grinding out artifacts of complex form and function. Intelligence is so cool!
I thought this was be a good discussion topic. First, a few questions for aiguy:
1. I know next to nothing about computers. Could you explain for a layman how your computer goes about deciding what to design and how to design it?
2. This computer, you say, operates according "to nothing but fixed law and chance." The "genetic and other machine learning algorithms" that it uses — did it stumble upon them by fixed law and chance? Or were they programmed into it? And if they were programmed into it, was it by chance, or were they programmed by someone making an intelligent decision?
3. And these "genetic and other machine learning algorithms" — were they produced by chance, or were they intelligently designed?
I agree with you that your computer is intelligent, by the way. And I think your other comment,
This is why I say ID folks should be in the habit of qualifying what they think the evidence actually supports, rather than leaving it up to each reader's intuitive conception of mind.
is a very good one. If ID is true, it could be that the designer is a computer.
June 5th, 2008 at 7:03 pm
Let me add a warning: If I find comments that I consider non-topical, I will send them to the memory hole, regardless of who makes them.
Comment by Bilbo — June 5, 2008 @ 7:03 pm
June 5th, 2008 at 7:36 pm
Bilbo,
Say I write a program that selects some action. The choice could be fixed in the code, or it could vary with pseudo-random input, or it could even vary with "true random" input (say from radioactive transducers). I would say the computer is deciding which action to take in all three cases.
Now, I write another program which has the ability to select and run other programs. It selects one program or another based - again - on either fixed or random factors. This means this program is deciding which other deciding program to run. You can extend this as far as you'd like to.
The machine starts out with some basic abilities; humans start out with some built-in abilities too (you can recognize a face almost immediately after being born). I teach the machines some things by giving them software; your parents, teachers, and friends taught you some things too. The machines learn still other things by interacting with their environment and other machines - just like you learned from your environment and other people.
What do you mean, "intelligently designed" Do you mean by a person, or by a computer?
Excellent, Bilbo!
Comment by aiguy — June 5, 2008 @ 7:36 pm
June 5th, 2008 at 7:41 pm
Very nice.
In your opinion, is what you built an example of AI (artificial intelligence)?
Comment by Salvador T. Cordova — June 5, 2008 @ 7:41 pm
June 5th, 2008 at 7:45 pm
Hi Sal -
In my business, this is sort of a running joke. Example:
So that concludes the demonstration of HAL 21000. As you've seen HAL has written nine best-selling novels, proven all remaining mathematical conjectures, cured cancer, and designed faster-than-light transportation. Are there any questions? Yes, over in the corner…
That's great! But is it intelligent?
Comment by aiguy — June 5, 2008 @ 7:45 pm
June 5th, 2008 at 8:16 pm
I began studying computers partly to get involved in AI. I was particularly interested in designer substitutes that could:
1. build VLSI circuits
2. build spaceships
3. write software
4. build machines that could learn
What I realized was that there was no unified concept of what AI was, much less how it could be achieved
Indeed we do have expert systems and genetic algorithms which design things like VLSI circuits. In fact, once upon a time, spread sheets, word processors, compilers were viewed as AI…..but are these things AI?
This does relate to the ID debate, because if we have problems characterizing intelligence, then definition of ID is muddled by our definition or lack thereof of intelligence….
I think the "D" in ID is much more tractable and definable and can stand on its own. The "I" is another story…
In any case, I got by reluctantly through a CS curriculum, but it didn't fit me well. I went back and got an EE degree. EE and Physics better suited me….[it was professor of AI who lamented not pursuing physics, so his laments sort of motivated me out of the field of computer science]…
Comment by Salvador T. Cordova — June 5, 2008 @ 8:16 pm
June 5th, 2008 at 8:48 pm
Sal,
Correct. What's more, there is no unified concept of what natural intelligence is! This is a critical point that bears repeating in the context of ID: There is no theory of intelligence, and there is no unified concept of what intelligence is.
Your question is a matter of definition rather than discovery (a philosopher would say it was analytic rather than synthetic).
In other words, you are not asking me about any fact of the matter regarding what computers can or cannot do; you are asking me what the definition of AI is. My answer is: It doesn't matter what the definition of AI is. That was why my joke was (supposed to be) funny! Let's try it again:
Get it? What difference does it make whether we answer yes or no to this question? It doesn't change the fact that this computer has done all of these things. If you'd like to call those sorts of things intelligent, go ahead. If not, who cares?
Is Michael Jordan athletic? Well, yeah, of course. Is Tiger Woods athletic? How about a cheetah? What about a flea? I dunno - it just depends on what you mean by athletic I guess. You see?
Sal, I'd like to reiterate what you've just said, because in my view it is the single most important aspect of the ID debate. If you don't mind, I'll use bold font:
This does relate to the ID debate, because if we have problems characterizing intelligence, then definition of ID is muddled by our definition or lack thereof of intelligence
It's even worse than that, I'm afraid, because there are various different definitions of "design" floating around, and some of these build in ambiguous references to intelligence! For example:
1) design is "an artifact containing complex form and function (or CSI)"
2) design is "something created by an intelligent agent"
I have an undergrad degree in cognitive psych and graduate degrees in CS, and if I had to do it over again, I would major in philosophy (and be poor). Anyway, after 30+ years of thinking about these issues, I really do see ID being a non-starter as an explanation of anything until it begins to take a more rigorous look at what this concept of "intelligence" is actually supposed to mean.
So I'm really heartened by your response, and that of Bilbo here too!
Comment by aiguy — June 5, 2008 @ 8:48 pm
June 5th, 2008 at 8:57 pm
Well put. The effects (design) are what matter. We don't have to define human intelligence or any kind of intelligence to know what effects it produces. The analogy of cellular machinery to human engineering is enough to begin to investigate the rabbit hole.
(Edit: For those disturbed by the thought of investigating rabbit holes, feel free to substitute "looking for rabbit tracks", or something similar. :mrgreen:)
Comment by kornbelt888 — June 5, 2008 @ 8:57 pm
June 5th, 2008 at 9:22 pm
Salvador T. Cordova wrote:
Sal, the D without the I loses any meaning, at least in the canonical definition of ID:
Not that I find that definition particularly useful, but it certainly emphasizes the I as the source of the D.
Comment by olegt — June 5, 2008 @ 9:22 pm
June 5th, 2008 at 10:23 pm
The computer has no experience, and "intelligence" is something that relates to the experience of knowing. To project "intelligence" onto the computer only begs the question: Show me the computer that feels itself playing chess, or the toaster that feels itself baking bread?
Design is like the hammar, it is a tool, and can be picked up by an intelligence that uses the mere tool to build a house. It is the carpenter that feels himself building the house, the intelligently designed hammar feels nothing.
Comment by Stephen — June 5, 2008 @ 10:23 pm
June 5th, 2008 at 10:30 pm
I don't see why this would be counter-intuitive. Results are the product of human design even if indirectly so. Even the fact that human input is not needed beyond a point is not unique.
Comment by Bradford — June 5, 2008 @ 10:30 pm
June 5th, 2008 at 11:13 pm
Bradford,
Agreed - I think any attempt to use computers to prove that intelligence/design can occur without an intentional agent is doomed to fail. Even if a programmer designs a very complicated program that can run and accomplish many things flying by itself, you always not only have the spectre of the originally programmer in play, but the fact that the programmer can intervene at will with the program. It's like writing a play, sending the play off to the printing press, and recording the printing in action. 'Wow! Look at this! A machine is able to write and mass-produce a play all by itself!' It's a bit trickier than that.
Comment by nullasalus — June 5, 2008 @ 11:13 pm
June 5th, 2008 at 11:33 pm
Stephen:
This is why the computer [AI] analogy is so intriguing. If the artifacts are themselves intelligent, or even intelligent designers themselves, I don't think you can drop the "intelligent" qualifier for the theoretical designer of the artifacts. Even if the intelligent artifacts can't agree on what it means to be "intelligent," and even if they spend their time calling each other stupid. Or just 'Tards.
Comment by Joy — June 5, 2008 @ 11:33 pm
June 5th, 2008 at 11:46 pm
Right. There is a parallel to this in biology. If mainstream biologists are correct about everything pertaining to an evolutionary process and FLE never gets traction but the origin of life was the product of intelligent design then all life forms would be too. An intentional agent involved in setting up a self-perpetuating process would make such a process a product of intelligent design even without subsequent intervention.
Comment by Bradford — June 5, 2008 @ 11:46 pm
June 6th, 2008 at 12:35 am
That is a reasonable objection. As Mike also pointed out, there is not necessarily stable definition of ID.
However, let me point out that even the definition of "life" is confronted with similar issues as the definition of intelligence….
How does this relate to other issues. As I said, I see no urgency to change high school or college science curricula to include ID.
What about peer reviewed research? The hypothesis of ID is argued through various claims. Though the falsification or testability of ID directly or in totality might be difficult, I think certain claims used in support of ID can be empriically investigated.
The circumstantial evidence and various claims that are used to support ID can be validated. My personal interest has been in two areas that can be empirically explored:
Formally speaking, I suppose anyone could challenge whether the outcome of these two research areas has bearing on the question of ID — nonetheless these research areas will still return empirical measurments and observations. Whether one feels the measurements or observations justify an ID inference is another topic, but I think these are still observations and measurements that would be of benefit to all parties concerned….
With respect to genetic entropy, if the genome is decaying as Sanford postulates, this has serious medical implications. I will be working in a couple weeks with some other to outline what sort of research can be done along these lines. I think this specific question is worth exploring whether one believes in ID or not…..
Now back to the topic at hand….why do I support ID even if I can't define intelligence — well, I support the existence of love and life even though I can't define them adequately. I think there is good circumstantial evidence to support ID, but that is my opinion. Whether ID is science or not is less of my personal concern than whether it is ultimately true or not….
Even if for the sake of argument we classify ID as a crank philosophy, the question of whether and intelligence was responsible for life still remains….that is really the core of the search for ID.
Comment by Salvador T. Cordova — June 6, 2008 @ 12:35 am
June 6th, 2008 at 1:43 am
All -
You have merely transferred the ambiguity from the word "intelligence" to the word "design" here. Since there is no meaningful, unified concept of "intelligence", you decide that it doesn't matter, and ID will focus on "design" instead. When I ask what "design" is supposed to mean, I'm either told it is the product of "purpose and intent", or that it displays "apparent rationality and foresight", or that it is "something containing complex specified information". And so around and around we go, for lack of any consistent technical vocabulary.
If ID studies the designs and not the causes of those designs, then the "canonical definition" of ID (as it has been called here) seems to be in error: The theory of intelligent design holds that certain features of the universe and of living things are best explained by an intelligent cause….
Apparently we are back to requiring conscious awareness again, when I have been told repeatedly by several others here that consciousness was not at all critical to ID.
In any case, Stephen, how do you suggest we decide what has "experience of knowing" and what doesn't? Perhaps you can work this out with Kornbelt.
First, you have assumed humans are intentional agents and computers are not. Is this personal opinion, or have you solved the problem of intentionality?
Next, I don't see how the fact that the programmer can intervene has anything to do with it, since the programmer need not intervene at all as the system is behaving with intelligence.
Next, you haven't said why you are focussed on the question of "who designed the designer?" here. Either something is intelligent, or it is not. If a computer is not intelligent, then why can it solve novel problems, learn, generate functional designs, and so on, all by themselves with no intervention from anyone? I do not operate my computer systems from behind the scenes - once I flip the switch, they are entirely on their own. I do not know what problems they will be asked to solve, nor what answers they may give. I do not know what they will learn, nor what they will decide is important in any particular situation.
Next, you imagine that because computers have been designed by humans, this means that computers are merely the tools of their human creator, and have no intelligence of their own. And yet you also believe we humans have been designed - does that mean that humans are merely the tools of their Creator? But obviously you can't think that, because you just implied that humans are intentional agents (which I believe you think is the same as an intelligent agent, which I think is what is supposed to design designs, which are things that reflect purpose… sorry, this gets tricky when we have no technical vocabulary to work with).
Anyway, if the Designer can design human beings, and human beings can still be bona fide intelligent agents, then why can't computers also be bona fide intelligent agents, no matter who or what design them?
Even ID-friendly Bilbo believes that computers are intelligent - one wonders how can there be so much disagreement among ID proponents, even on these most fundamental of issues, and after all this time. Perhaps because ID has never committed to any technical definitions? You think?
It is nothing like this at all. If I want to know if something can fly, I build it, and then I see if it can fly or not. If I want to know if something act intelligently, I build it, and then I see if it can act intelligently or not.
Imagine you come across a strange entity in a field. You can communicate with it and it responds appropriately. You see it design novel machines, prove novel theorems, diagnose and repair complex equipment, and even write a new song. I ask you if you think this agent is intelligent, and you presumably would say yes, it appears that it is intelligent. I ask you if you think it was designed by an intelligent agent, and you say yes - it has all the signs of having been designed. Then I tell you it is a computer, and you say… Hey! You didn't tell me it was a computer! No way it's intelligent!
Saying Deep Blue does not think about chess is like saying an airplane can't fly because it doesn't flap its wings (Drew McDermott).
Until we have some clear, unified notion of what the word "intelligence" is supposed to actually mean in this context, don't you think it is premature to suggest what it might be responsible for?
And finally, regarding how counter-intuitive AI is, imagine having been told at the turn of the century that a machine which can only perform a few simple Boolean operations can be made to beat any human being on the planet in chess. Perhaps you think you would have immediately agreed it could be done. In that case you would have been the only person with such a clever intuition; many people kept denying the very possibility through the early 1980's.
Comment by aiguy — June 6, 2008 @ 1:43 am
June 6th, 2008 at 2:37 am
aiguy,
I don't have to in this case. If the computer is a bona fide intentional agent, then an agent can give rise to an agent (no surprise there), and the argument for all things 'natural' being part of or exhibiting hallmarks of bona fide agency strengthens. If the computer is not an intentional agent, then that which lacks intention, mind, and (by some definitions) true 'intelligence' can still be employed, instigated, and manipulated by an agent. Agency stays fundamental in this scenario.
It just fleshes out the problem with using computers as an illustration of 'no design'. I remember Kasparov being suspicious that Deep Blue's team was intervening in some way in DB's moves. Again, the agent always remains in a wide variety of ways with regards to computers. Even if there's no intervention ever, all that may mean is no intervention is needed - the computer can be trusted to fulfill the originating agent's desires.
Who designed the designer? I haven't gotten into that deeply. I'm basing this entirely on the validity of using computers to prove you can have design without intelligence. I take the opposite stance - computers show that you can have tremendous design that needs no obvious constant oversight, yet still originate with and act in accordance with an agent's wishes.
Well, no. Whether they're 'merely' tools, or are agents+tools, doesn't really matter all that much to me in this instance. An agent who creates agents which serve instrumental roles yet are themselves important as agents does not trouble me.
I'm in my own little category when it comes to ID, much as I have sympathy with a wide range of views, so it may not be fair to use me as an example - I'm the one who favors it in philosophy, but has strong reservations re: "science". On the other hand, the lack of precision in these definitions didn't stop a whole lot of people from railing AGAINST design and intelligence in nature prior to modern ID stepping up on the scene. No one derided them for being unscientific, loose with definitions, etc. For some reason, if an ID proponent does it, it's a sin, a virus that must be stamped out upon the instant. If an ID opponent does it, well, that's merely unfortunate - if that at all.
You seem to be basing a lot of this on my viewing a computer itself as intelligent/an agent or not. But as I've said - that actually doesn't matter to me in this case. If I'm faced with a computer, no matter how complicated or simple it is, I'm always dealing with 1 or more actual agents.
I remember John Searle giving a reply which I can't find now, but it was something along the lines of 'The amazing thing wasn't that Kasparov lost a game to Deep Blue. It was that an entire team of IBM engineers, working as an organized team with hundreds of thousands of dollars in funding to create a chess playing machine, and the human won against it.'
You may want to have a look at what Charles Babbage saw in the potential of programming re: a Designer. Impressive foresight there.
Comment by nullasalus — June 6, 2008 @ 2:37 am
June 6th, 2008 at 2:37 am
Bilbo,
You really have hit on the very idea that I have been attempting to convey here.
I don't think you are suggesting that a Dell PC with a flat-screen monitor was responsible for creating the universe. What you obviously means is that a computer operates according to deterministic law (or law+chance), and so for all we could ever know, the intelligence responsible for life might have been a blind, lawlike process.
Again: If ID is true, then there are two possibilities:
1) A blind, lawlike, intelligent process somehow existed and created the universe and life
2) A sighted, mindful, intelligent process somehow existed and created blind, lawlike processes which created the universe and life
The first hypothesis is a bit simpler, but I'd say either of these could be true. I'm certain, however, there is no way anyone could ever say which one was true.
(If you wonder what "blind" and "sighted" and "mindful" are supposed to mean here, look it up in the ID Glossary of Technical Terms. I think they are similar to "guided" and "unguided"…)
Comment by aiguy — June 6, 2008 @ 2:37 am
June 6th, 2008 at 2:48 am
Nullasalus,
I don't have much time, but let me respond to the central point:
Very well, I take it you concede that computers can indeed be bona fide intelligent agents.
In this case, then:
1) Computers operate according to deterministic law (or law + chance)
2) Computers are bona fide intelligent agents
3) Therefore ID can draw no distinction between law+chance and intelligent agency (aka "guided" vs. "unguided" processes)
And please read my previous post to Bilbo; it is an integral part of my arguments here.
Comment by aiguy — June 6, 2008 @ 2:48 am
June 6th, 2008 at 3:27 am
aiguy,
'I don't know. Maybe they can. For the purposes of establishing the strong possibility of agency in nature and reality, it's not important to me.' That should suffice for now.
We seem to go through this one over and over.
My replies:
1) It doesn't matter if computers are bona-fide agents or not. Not for my purposes. And oddly enough, not for yours. The fact that the question remains and has no (to my knowledge) definite resolution gets you where you need to go.
2) Your #1 is called into question either way, unless you're certain that humans or a human-like agent must operate according to deterministic law or law + chance (or, even if they do, that a human/human-like agent is indiscernible from a computer/computer-like agent). Because a computer can have a programmer, and the programmer can intervene (either directly, or via an operation that the programmer can guarantee will go off). I could picture ID considerations ranging to a program that was entirely front-loaded at the start of all creation, or an ID project attempting to find incidents in history that seem as if the programmer intervened to accomplish something the program wasn't going to accomplish on its own. I still wouldn't have faith in this being a falsifiable outlook.
3) From the philosophical end of things, the fact that humans as we know them have experienced so much personal success with design leads us to a whole lot of interesting realizations regarding the potential of 'agents like us'. We may never be able to completely resolve the question of human-like agency (or lack thereof) being responsible for nature, but.. well, I covered this with simulations.
As I said, I only popped in to mention that trying to take down the concept of an intentional agent working in nature by use of a computer example can't really succeed. By having a bona-fide intentional agent responsible for their creation, orchestration, etc, they are poison to a no-ID argument. They can't even really do good for a 'no way to decide' argument, at least in the philosophical end. You can fight over the agent's intentions, but doing away with the agent with computer examples is extremely problematic.
Comment by nullasalus — June 6, 2008 @ 3:27 am
June 6th, 2008 at 4:09 am
nullasalus,
But it does not suffice. I have presented a number of arguments that support this assertion, and your response has been that it doesn't matter if I'm right or not. But since I believe my argument depends on it, I must assume that your admission "maybe [computers] can [be bona fide intelligent agents]" is a concession in the affirmative.
Nonsense - we fully understand how computers operate, and they operate according to law + chance. The fact that they had an antecedent designer is irrelevant, and of course the fact that the antecedent designer is not known to operate according to law+chance is irrelevant as well.
Once again, I do not intervene in my program's operation. Furthermore, I can never guarantee what will happen in my programs. I can't predict their behavior - even in principle. They can also solve problems that I can't figure out how to solve. For example, they can prove theorems that I couldn't see how to prove, and they can design electrical components that are more efficient than I could figure out how to design, and they can induce regularities in data that I couldn't recognize. (I'm not actually personally involved with all of these sorts of projects - only some - but these are all factual statements about what programs can do).
Once they are activated, my computers' behaviors are completely independent of my will. Their behaviors change over time, and with their experience too. All of this can be said, in just the same way, about human beings: We are born with built-in abilities, then taught language and other skills by somebody else, and we learn from our interactions with the environment. There simply is no relevant difference you can point to (as my story about the entity in the field made clear, I think).
EDIT: Moreover, so what if I did intervene in my programs' behaviors? What's the point about this? Can't your Designer perhaps intervene in your behaviors?
And I like your whole idea about that as I've said. I should also tell my own philosophical counter-argument to the argument I've presented here: It's all about consciousness, whether or not it's epiphenomenal. The very fact of my own phenomenology, combined with my undeniable intuition that computers lack awareness and the vacuity of emergentist solutions, convinces me that something transcendental is going on. That's it - that's my counter argument. I think it is almost comically weak, but I think it's true - go figure.
I really couldn't disagree with you more on this one.
1) You are mistaken to say that computers do not operate according to well-understand principles of physical causation
2) You've failed to distinguish a) the supposed independence of us humans from our Designer's will from b) the supposed dependence of the computer on my will.
3) You won't try to argue that computers aren't bona-fide intelligent agents
So yes, I think it's clear that this argument holds up perfectly well:
1) Computers operate according to deterministic law (or law + chance)
2) Computers are bona-fide intelligent agents
3) Therefore ID can draw no distinction between law+chance and intelligent agency (aka "guided" vs. "unguided" processes)
Comment by aiguy — June 6, 2008 @ 4:09 am
June 6th, 2008 at 4:58 am
aiguy,
For the purposes of your argument, feel free to count me down as 'affirmative' if you like.
Some problems.
1) You 'do not' - even if I took this at face value, that doesn't mean that you can't. And if you can, it's fair game for me to imagine scenarios in which you would. And with that possibility being live, it's something that has to be addressed. Remember, I'm not mentioning anything here that's out of the ordinary with computers and programming.
2) You say you don't know what will happen with your programs. I could mince details on this, but it's enough to say - I don't care how complicated your program is. If you want it to, at the very moment it registers 4/1/2010, 13:52pm EST, spam the screen with 'HELLO WORLD' - you can. And that's just the beginning of what's possible. I object to your claim that you "can't predict their behavior - even in principle" - even you with your limitations can have some knowledge of what will take place, though it's imperfect. And the fact that you can have that much knowledge puts a dent in 'even in principle'.
So it's not enough to assert 'I know how a computer/program works (by deterministic laws, or laws + chance)'. You still have the possibility of the human or human-like, at the very least, to consider as meddlers in the program's operation.
I accept that there's definitely something 'special' going on with consciousness, so you get no argument from me there.
I never argued that they do not. I've argued that even if they do, you still have to cope with human and human-like agents (at the least) when considering them, because 'I create programs, and once they start running I don't intervene' just doesn't do the work you need it to do.
Are you trying to work free will into this question via a comparison between human minds+brains, and computers? The problem there is A) it's possible there are quantum effects present in human brains, though how is an open question (currently waiting on a paper about this supposedly happening in one testable way), and B) I'm not sure of what you mean by 'independence of humans from our Designer's will'.
I still don't believe I have to. You're trying to establish, to me, 'Imagine a program running which never had the programmer intervene', and my response is 'But I have no need to follow that thought, because programmers can and do do exactly that'. A question like this came up with Kasparov. Hell, it came up with Doug Hofstadter in a way - I remember reading about how someone sat him down in front of a computer (I believe at MIT) insisting they had created a new Turing Test style AI. He tested it for awhile, and then they let him in on a joke - it was someone in another room typing through a script. But what if it was a combination? A pre-scripted Eliza-style program, but one which a programmer could override at will?
Comment by nullasalus — June 6, 2008 @ 4:58 am
June 6th, 2008 at 6:06 am
Nullasalus,
I really don't understand your point here. Are you saying that I'm pulling a Wizard of Oz here, and secretly solving these problems for my computer from behind the curtain? It is certainly easy enough to control for, after all (you may run my computer in your room if you'd like). And as for "indirectly" guaranteeing its behavior; I can hardly guarantee behavior that I can't predict.
No, you're mistaken. There are two reasons that nobody could, even in principle, predict what a program will do, and neither reason has to do with how complicated the program is. First, it can incorporate random input, and second, it can learn from unpredictable environments, changing in unpredictable ways.
No, you really are mistaken completely. I am regularly suprised (and usually disappointed) by what my programs do.
Hmm, perhaps you think I'm pulling the old mechanical Turk trick here?
http://en.wikipedia.org/wiki/T...
So your "possibility of intervention" argument goes nowhere:
1) I can't predict what situations the programs will encounter, nor how random input might alter the program's behavior, and so I can't predict the program
2) The program can answer questions I can't answer
3) If you believe that AI programmers actually cheat when we run our systems (maybe we use secret radio-control devices?) then I'd mistaken you for someone far more rational
That's quite enough to defeat your argument, but on top of that, how do you avoid the possibility that your behaviors are simply instigated and manipulated by your designing agent?
Yes you did. You said: "Your #1 is called into question either way" (where #1 is "Computers operate according to deterministic law (or law + chance)") in post 193582).
First, I don't need it to do any work, because this particular premise is simply that computers operate according to law+chance, and has nothing to do with anything else.
Moreover, the fact that my programs run by themselves does every bit of the work that 'The Designer designs human beings, and once they are born He doesn't intervene" does.
No. What I meant was that you claim the computer is merely an extension of the programmer, but you can't explain why you don't consider that humans are merely extensions of their Designer.
Again - free will is not the question; I was referring to your issue with this "possibility of intervention". I really think that argument is dead though.
Wow, I really am missing your point here. The only thing I can think of is that you actually believe AI programmers hide inside their computers and manually operate their computers - ignoring the fact that my programs can do things I am incapable of doing myself. (Or perhaps I am only pretending that I can't solve the problems, in order to fool you into thinking my computer is figuring it out?)
Come on, Nullasalus - what in the world are you talking about? You absolutely must be joking.
Comment by aiguy — June 6, 2008 @ 6:06 am
June 6th, 2008 at 6:09 am
Oh… OK, I think I figured out what you mean. You mean that even if something looks like it's proceeding according to law+chance, it could always be the case that an agent which transcends law+chance is actually responsible? I recall a theological point like this - God moves the atoms, but He moves them as if they were moved by law? Is that it?
Comment by aiguy — June 6, 2008 @ 6:09 am
June 6th, 2008 at 8:45 am
Ah, a thread in my field of expertise (what little expertise I have anything, anyway).
aiguy, I simply do not accept any of your premises.
There is really no such thing as random input or unpredictable environments for a computer. The constraints that are put into the system to begin with are the ones that prevent this from being so.
Truly random input originates from outside the computer; nothing a computer generates is truly random.
You may get unexpected results from a computer, but certainly not unpredictable ones. A person merely needs to sit down and perform all of the calculations that a computer did to arrive at the exact same result the computer gave. Granted, the number of calculations involved may prohibit doing so in a timely manner, but it definitely is not a case of unpredictability. If unpredictability was defined so, we could call any algebraic equation unpredictable simply because it has variables.
There's also another problem on the opposite end you've mentioned only in passing - the output.
How does a computer know it is in the process of solving a problem? How does it know when it has been successful in solving a problem and stop processing? It doesn't. It requires constant intervention by human beings who interpret its output, define its successes, and define its failures.
Roger Penrose > Marvin Minsky.
Comment by angryoldfatman — June 6, 2008 @ 8:45 am
June 6th, 2008 at 9:46 am
You seem to be contradicting yourself. Certainly, inputs to a computer can be chaotic or random, e.g. weather data, or quantum randomness.
A measure of complexity is the inherent unpredictability of its behavior. Many systems are so complex that the only way to predict them is, as you point out, to run them with every possible input scenario"”and this is not always possible due to immensity of the possible inputs. We say that such a computer system is *deterministic*, but unpredictable.
It depends on what you mean by 'know'. It might churn numbers then print out an answer and halt. It could even say "working … working … done."
If we are talking about a complex simulation, there may never be a state of having solved a problem. There is no problem per se.
Comment by Zachriel — June 6, 2008 @ 9:46 am
June 6th, 2008 at 9:46 am
aiguy to nullasalus:
I do not see how an antecedent designer is 'irrelevant' to a computer-as-artifact argument. We understand how computers operate because WE designed and built them. Because they are our own intelligently designed artifacts, there just isn't a way to force anyone to pretend that they might be artifacts of an RM-NS style evolutionary process.
I understand (I think) why you wish to use this example, it's the one most well-known to you. But it doesn't suffice for what you want it to represent. A stronger argument in this vein, I think, could be generated using a duck-billed platypus.
IOW, it simply doesn't work to use a legitimate artifact of human intelligent design as an example of how a blind watchmaker mimics intelligent design. You aren't a blind watchmaker. The computer didn't evolve in the way you or a platypus presumably did. So what *IS* irrelevant here is your assertion that the computer operates entirely by "law and chance" once you've installed the software and turned it loose. Turning it loose does not negate the fact of physical and programming design. It's still an intelligently designed artifact - and we still know it's an intelligently designed artifact.
What relevance to the question of whether the computer is an intelligently designed artifact could it possibly have that the computer is not conscious (or consciously aware)? This again falls apart by the very existence of your example, which *IS* an intelligently designed artifact.
If both the intelligent designer and the intelligently designed artifact display the quality of "intelligence" and the ability to design, what difference does it make for one to be conscious and the other not? It's like you wish to do one of two things with your argument and counter-argument and aren't able to accomplish either…
1. Intelligence need not be conscious, and
2. Intelligent designs need not be conscious
…if we accept your postulation that both humans and computers are intelligent and focus on which is conscious, I don't see how it relates to the overall debate about whether life forms are intelligently designed artifacts. Because in your example, we already know that humans are conscious and computers are not.
Are you trying to somehow prove that because humans are conscious, if we're intelligently designed artifacts our designer cannot be conscious? Or if we're conscious we cannot be intelligently designed artifacts? The argumentative links here are tenuous and confusing. Can you clarify?
Comment by Joy — June 6, 2008 @ 9:46 am
June 6th, 2008 at 9:47 am
Regarding the following question from aiguy, this was why the Turing test was invented. But the question really belongs to philosophy. For example, Husserl will have us strip away our pregiven assumptions, ending in ontic meaning. The ontic meaning is grounded in consciousness (a transcendental subjectivity which has purified itself away from a cold objectivity), and that is as far as we can go which implies that consciousness is fundamental.
Comment by Stephen — June 6, 2008 @ 9:47 am
June 6th, 2008 at 11:03 am
If I may add some formalisms here to show why I think your objection, although reasonable, is misplaced.
I can say "the sky is blue" but "blue" is not a definition for what the sky is, nor even an necessary requirement for something being a sky….
The definition of design, as used in ID parlance is not associated with intelligence at all, it is simply saying something that is not describable by simple stochastic processes or regularity are designed:
I think this obscure paragraph in Dembski's work has be forgotten. Yes we can claim things that evidence design evidence intelligence, but that is a claim for design, not a definition of design….
"The sky is blue" is not a definition of sky.
Neither is "Design is what evidences intelligent agency" a definition of design. It is merely a claim about something we defined. Definitions are not the same thing as the claims we make about definitions. As Bill Clinton said, it depends on what you mean by the word "is"…..
So things with this mathematical property are called designed. It is a separate question whether intelligence is needed to generate such artifacts….
Comment by Salvador T. Cordova — June 6, 2008 @ 11:03 am
June 6th, 2008 at 11:25 am
That's because "The sky is blue" is parsed as "This sky is blue." (By the by, this is evidence in support of the statement that "Some skies are blue.")
Until now, I've never seen anyone cite that as an argument for clarity.
Excerpts from The Explanatory Filter, William Dembski:
Comment by Zachriel — June 6, 2008 @ 11:25 am
June 6th, 2008 at 11:40 am
Sal,
You're conflating definition with detection. EF isn't used to define what ID is. As the passage from The Design Inference shows, Dembski proposes EF as a means to detect design and from that infer that the object an intelligent cause.
Comment by olegt — June 6, 2008 @ 11:40 am
June 6th, 2008 at 11:57 am
I don't believe I conflated them, but I will grant it appeared to you that I did. And even supposing that I had conflated them, the issue is whether the definition of design in Dembski's work is formally decoupled from the existence of an intelligence, and I believe it is.
If the definition of design is decoupled from the definition of intelligence, at least the "D" part of ID is not subject to the concerns being raised about the definition of intelligence….
Furthermore, this definition of design, seems to embody the spirit of the word "design" when used in existing peer-reviewed literature and textbooks. It seems perfectly legitimate to speak of the design of the cell, and this seems to accord with Bill's more formal definition.
I'm not trying to defend ID as science in these discussion. If it all makes some of you feel better, you may call ID what you want for the sake of argument…
However, I like to set the record straight that at least Dembski's notion of Design is not committed to a doctrine of intelligent agency….it is a simple mathematical definition. That is all….
Comment by Salvador T. Cordova — June 6, 2008 @ 11:57 am
June 6th, 2008 at 12:06 pm
Third alternative might be that life is a sighted, mindful, (purposeful, responsive) intelligent process that elaborates itself, and is currently still in the process of limited self-modification. Some life forms have existed so long that they may appear to materialists to be blind and law like. However the fact that many living systems can obviously still modify themselves in purposeful, intelligent response to internal and environmental challenges can be viewed by non materialists as evidence against materialism. (There would be no way to detect whether or not some deity were involved in the process, since purposeful intent cannot be examined under a microscope. Its existence or non existence can only be inferred.)
Comment by Bert — June 6, 2008 @ 12:06 pm
June 6th, 2008 at 12:32 pm
Dembski: The filter is a criterion for distinguishing intelligent from unintelligent causes.
I'm not sure how much clearer it could be.
If so, then it merely says it is a pattern we have no ready explanation for.
Comment by Zachriel — June 6, 2008 @ 12:32 pm
June 6th, 2008 at 1:11 pm
Bert:
You know I agree, Bert! And this is what's ultimately at issue in this stupid 'culture war' of metaphysics being played out in the scientific, social and political arenas. Why is science attempting to limit by force what people may infer from evidence, so that only their 'approved' metaphysical implications can result?
Still seems mighty messed up to me, and it doesn't look to be getting any better as time goes on.
Comment by Joy — June 6, 2008 @ 1:11 pm
June 6th, 2008 at 1:39 pm
Joy,
That is simply not my argument at all. My argument is this:
1) For any detection criteria for "intelligence" that ID may want to adopt, computers are intelligent.
2) Computers operate according to "blind, unguided, natural processes"
3) Therefore ID can't distinguish between "intelligent cause" on one hand, and "blind unguided, natural process" on the other hand.
RM&NS and evolution has nothing to do with it.
It makes no difference to my argument at all. I was just sharing with Nullasalus why I don't happen to be a physicalist.
I'm not trying to say either of these things - as you see, my argument has nothing to do with consciousness either.
Hopefully my argument as stated above is sufficiently clear.
I'll repeat this in hopes it could clarify for some people why I think ID must admit that computers are intelligent in the same way people are:
Stephen,
I could not agree more. But this has nothing to do with my argument either. If ID wishes to say that "consciousness" is required in order for something to be considered intelligent, then ID turns into philosophical speculation about consciousness, period. Here at TT I've been told repeatedly that this is not the case - ID makes no particular claim about the consciousness of the Designer. I have been told the same thing by Dembski (personal communication). Instead, ID is about design, or about purpose and intent, or about intelligence.
Of course there is always this confusion because there are no technical definitions published for the concepts used in ID, which means we will always have these semantic problems when talking about ID Theory. While several here have indicated it really doesn't matter much how ID defines it, I think this discussion shows it matters a great deal.
Sal
I'm not trying to argue only that ID is not science; I think most people here (including MG, FMM, KB, and Nullasalus) agree that ID is not science. My point is a bit broader: For however ID would like to describe intelligence in terms of its effects, computers produce qualitatively the same kind of effects, and so ought to be considered "intelligent" in the same way humans are. The fact that computers are designed makes no difference, since (according to ID) humans are also designed, yet we consider humans intelligent.
If computers, which operate according to pure law+chance, are considered intelligent in the same way people are, then ID can't distinguish "intelligent cause" on one hand from "law+chance" on the other hand.
I really get tired of this semantic runaround. I would suggest that if (each version of) ID would adopt some technical definitions and stick to them, we would stand a better chance of discussing the issues with any clarity at all. If ID does not involve an intelligent cause for biological complexity, then I really don't know what we're all talking about here.
I don't believe anyone has countered my arguments here at all, and so I think we're left with the conclusion that there are no grounds to distinguish "intelligent cause" (or "telic cause") on one hand and "blind, unguided, natural process" on the other hand, unless one falls back on the standard metaphysical arguments for dualism. ID brings nothing new to those.
Comment by aiguy — June 6, 2008 @ 1:39 pm
June 6th, 2008 at 1:47 pm
Sal,
First, I've shown here that design cannot be distinguished from regularity and chance, and the reason is that computers (which operate by regularity and chance) can create novel designs.
But much, much more importantly, I want to point out the incredible equivocation that is ID! On one hand, people all over the world believe that ID somehow supports the idea of an Intelligent Designer of life and the universe. But in this "obscure paragraph", Dembski admits that it has nothing to do with intelligent agency at all!
Comment by aiguy — June 6, 2008 @ 1:47 pm
June 6th, 2008 at 1:55 pm
They are mechanistic extensions of human intellligence totally dependent on humans for their creation.
So do we. It's the origin of computers that defies a blind unguided process explanation.
Sure we can. This message was guided. The evidence is in the sequencing of the letters.
Comment by Bradford — June 6, 2008 @ 1:55 pm
June 6th, 2008 at 2:06 pm
Bradford,
1) I don't see why "mechanistic" has anything to do with the issue. What part of the Design Matrix criteria for design has to do with whether or not the designer is "mechanistic"
2) Computers are not necessarily totally dependent upon the programmer, because computers can learn from their environment.
3) We human beings are also dependent upon our Designer for our existence (I presume you agree with this). Yet you don't take this to mean that we humans are not truly intelligent.
Hmm. This makes your mention of "mechanistic" in your first point even more difficult to understand. Now you are saying that humans are also mechanistic? OK, if that's what you believe, that's fine with me.
Human beings are designed by a Designer, and computers are too. So what is your point?
Now you fall back on calling this "guided", without saying what "guided" is supposed to mean, but simply asserting that it is different from what mechanistic, algorithmic processes can do. My argument shows this is not the case.
If ID published a technical vocabulary with consistent definitions, we would not have these constant semantic runarounds.
Comment by aiguy — June 6, 2008 @ 2:06 pm
June 6th, 2008 at 2:19 pm
Hi Aiguy.
Very interesting thread.
And thank you Bilbo for posting it.
To the rest, excuse my absence, work finally had swamped me beyond the point of no return until this weekend. Wife and co-workers forced me to take this weekend off.
For what it is worth, it looks like Aiguy has forced an implied fundamental admission from even Nullasalus that there is nothing inherently different between an artificial intelligence and human intelligence.
Both might have required being front-loaded or continuiously manipulated by outside "meddlers"or "agents". It matters little if one calls these designers angels or ETs, the idea is the same. That is something higher is needed to qualify as intelligent.
Thus back to the old AI joke, "…but is it intellegent?"
I do suggest something more is needed for artificial intelligence to be truly non-deterministic. Psuedo random generators are deterministic. Massively interconnected and complex input is deterministic. While it can cause unexpected results, with enough time, energy and knowledge, the results could be explained.
What can't be explained with today's scientific knowledge is why a particular "random" quantum choice is made.
At best, the explaination is that a "random" quantum choice is forced to be consistant with all the other "random" quantum choices in the universe, past, present and future.
I suggest it is as simple and as complicated as that.
I am suggesting the "something more" is an interconnectedness with the entire universe.
I would be my opinion that if a computer isn't dependent on the only possible true source of randomness, it is nothing more than a complex turing machine, a fancy calculator.
Comment by Thought Provoker — June 6, 2008 @ 2:19 pm
June 6th, 2008 at 2:20 pm
aiguy wrote:
Asserted more like. Hardly shown.
". . . computers (which operate by regularity and chance) . . ." And initial design. And input. And specific parameters . . .
Comment by Eric Anderson — June 6, 2008 @ 2:20 pm
June 6th, 2008 at 2:21 pm
Not real hard to understand. Organisms and computers function in accordance with natural laws but more than references to laws of physics is needed to explain their causal origins.
Why not take a page out of Bill Clinton's book and say it depends on how "is" is defined? We can parse definitions all day long. Beats getting impeached.
Comment by Bradford — June 6, 2008 @ 2:21 pm
June 6th, 2008 at 2:36 pm
Hi Bradford,
Speaking of politics, we could always "Stay the course" regardless that the evidence and logic suggest otherwise.
What is inherently different between human intelligence being guided by an intelligent designer and an artifical intelligence being guided by the same said designer?
Comment by Thought Provoker — June 6, 2008 @ 2:36 pm
June 6th, 2008 at 2:40 pm
Hi TP,
More to the point, I think I've shown there is no observable difference between "intelligent cause" on one hand and "regularity and chance" on the other.
As has already been mentioned here, there are true random generators, for example http://www.random.org/
I think we both know that "random" is a hard concept to understand, but these generators are as random as random can be.
I disagree, but since there are true random sources available, this point is moot here.
Eric,
Actually I've articulated my arguments above. It really is you who have made an unsupported assertion.
Bradford,
So you say. How do you know? It can't be from the evidence, because both computer-like and human-like things can create designs.
This is absurd. As you can see here, ID proponents disagree even amongst themselves what these words are supposed to mean. I'm not the one introducing confusion here; I'm pointing out that the only way these arguments seem to make sense is if you don't try to say what they actually mean.
I don't think you can defeat the argument here, Bradford, but if you can it will require a principled counter-argument rather than the mere accusation of obscurantism.
EDIT: You failed to respond, for example, to my question regarding your point about "mechanism". What does "mechanism" have to do with your saying that the intelligence of computers is somehow different from that of humans? You are the one playing Bill Clinton here, I'm afraid, shuffling your definitions like 3-card Monte.
Comment by aiguy — June 6, 2008 @ 2:40 pm
June 6th, 2008 at 3:14 pm
Hey AIguy,
I think the problem is that it isn't the mechanistic part that is the problem but the fact that computers are an extension or augmentation of our own intelligence, and the requirements of that intelligence.
Our intelligence is able to identify problems and calculate that there is a possible solution to said problem (even or especially, if we cannot identify that solution on our own), and then we use computers to help us overcome our knowledge and precision gap. Whether that be tracking the weather, or designing a spacecraft to maximum efficiency and precision, computers are "designed" to fill some of those specific gaps in our own identifiable dis/abilities.
So in this regards I might say that while computers can and do mirror some of the aspects of human intelligence they are not representitive of it (Yet). Maybe when quantum computers are more prevelent in AI we will see that gap close.
Also in this regards I think we may be able to see a difference between a designer of us and a designer of computers. We do not seem to be needed by the designer in the same regards as we need computers. Our relationship with computers is more intimate like a designed symbiosis if you will.
Comment by Kuma — June 6, 2008 @ 3:14 pm
June 6th, 2008 at 3:14 pm
Hi Aiguy,
You wrote…
At least our disagreement is tangable and not simply semantics.
The point isn't moot if there is no such thing as randomness or chance.
I presume you are somewhat familar with quantum computing. The whole concept is based on the fact that qbits are interconnected and NOT random.
I am arguing that it make no sense to presume that "true random sources" exists in our known universe.
Radioactive sources are non-deterministic, but that doesn't make them random.
You might argue for all practical purposes (FAPP) but then for all practical purposes the interconnected quantum effects is the unknowable intelligent designer, agency, meddler, etc all rolled in one.
Regardless of what the definition of "is" is.
Comment by Thought Provoker — June 6, 2008 @ 3:14 pm
June 6th, 2008 at 3:25 pm
aiguy,
I'm saying that as a programmer, you're capable of intervening in your program. Your defense is 'Yes, but I don't'. Anything after 'yes' is unimportant to me.
'Can incorporate random input', 'can learn from unpredictable environments'. Can, meaning not not by necessity, but by possibility given your state as a programmer with finite mental capabilities (like all of us). But a program can also have code within it that will have a definite result at the right point.
Really, are you telling me that the result of..
10 PRINT "Hello World!"
20 GOTO 10
..is something that 'cannot even in principle be predicted'? Or an instantiation of 'after X amount of time has elapsed, run that program'?
Not all of the time. If you never had any idea what your program would do (other than 'fail', I suppose) each time you ran it, you wouldn't be aiguy. You'd be carwashguy.
Let me be clear here: I am not doubting that programming is complicated, or that you're capable of creating such a complicated program that you can't be 100% certain it will compile, or that you can't be 100% certain of what it will do totally. But you can still create a program you are near-100% certain will compile at a certain that at time 4:44pm will perform a given action. And you can still intervene if you like.
There's no defeat going on here, for all the reasons I've outlined. It really seems like you're arguing 'I can imagine never intervening in my program, or never coding something in which overrides other operations at a given point, therefore that's not a question you have to address.' Was Kasparov being irrational when he raised the possibility of DB receiving guidance? Was Hofstadter being irrational when he came to the conclusion that the 'AI' he was dealing with wasn't an AI after all?
Purely on the grounds that human or human-like intervention comes into play. Again, reference Hofstadter - when he was dealing with the "AI", was he dealing with a 'completely deterministic (law or law+chance)' system? To give an answer to that is to rule on human minds, despite there being a computer and programs between him and the fake.
Who says I don't consider that? Frankly, I'm open on that possibility. Doesn't seem to be pertinent here, though.
Here, let me try to package it concisely:
No matter what example you give with a computer, you're saddled with an agent right from the start. You don't even seem to be fighting this one, and that's fine. But the moment you're entertaining the possibility of an agent external to the computer (at any point, but particularly at the start), you're opening the door to the intervention of the agent at any point while the program is running. Arguing that you don't personally do this means nothing to the argument, because the possibility remains. Likewise, no matter how 'unpredictable' your programs may be to you, there do exist programs that are either perfectly or essentially perfectly predictable - and it's entirely possible (hell, at times it's common) to include code that will lead to an essentially/perfectly predictable result at a given point.
Put another way: It seems like you want to regard computers as examples of nature, as they're capable of puttering along according to exquisitely complicated programs, achieving intelligent results by nature of their program alone. My response is that computers demonstrate that you can have all that - and yet A) an agent can still intervene at will, and provide an on-the-spot personal touch within the program, and B) include within the program a point at which something definite will happen, no matter how apparently random the program is