|
Author
|
Topic: What is intelligence anyway
|
Terry Mullett
Member
Member # 1810
|
posted 04. November 2005 21:54
Hello all,
I've lurked around here for a couple of months and enjoyed it immensely. But now I have a question to ask. Please forgive if it sounds boneheaded.
There seem to be quite a few folks around here who are comfortable with Intelligent Design. I'm trying to get my head around it. Just to let you know where I'm coming from, I've assumed a naturalistic stance for decades. Still I think that most naturalistic attitudes I've encountered (my own included) tend to be naive toward complexity. I'm not really on the verge of abandoning naturalism or anything so dramatic, I just think that there's a lot to be learned and digested for most of us who hold that point of view. I see ID as a potentially very rich source for that kind of learning, and so that's why I'm eager to figure it out.
So I said I had a question....
The term "Intelligent Design" suggests to me that intelligence must be the key explanatory principle in that line of thought. Am I incorrect? Otherwise just "Design" would seem sufficient - it seems to me a distinction is being made. But I can't seem to see where there is very much time spent on saying just what we mean by "intelligence". I did see a definition in which intelligence is that which is necessary to produce design, but if the point is to charactarize some design as needing intelligence, that doesn't feel really satisfying. (Sorry about being sloppy but I don't have the reference handy at the moment ... I could find it if someone cares.) And besides that was just one little paragraph in one place. Certainly there's more.
One of my long-time interests is Artificial Intelligence. I know that researchers in this field have amassed a great deal of literature characterizing in great detail the features of the mind, the nature of thought, etc.. The same sort of thing can be said (more so I'd guess) for psychology, and it looks to me like AI has benefited a great deal in this from being intimately connected to psychology. I'm not supposing that the picture they have is a unified or complete picture, but gosh there's a lot of prior science here, and I think a sincere effort to be rigorous about it.
Am I missing something? Is there a great literature on just what ID means by "intelligence" that I've not yet found. If so, I'd appreciate anyone who could point me to it.
Design is another in the same vein. I'm somewhat familiar with the ideas of irreducible complexity and specified complexity, but it seems to me that these are the smoke while design is the fire. I'm looking the fire part. To my mind, design, creativity and intent are in a cluster of ideas that can prove to be very important for the future of AI, so I'm very anxious to see what thinkers in the ID sphere have to say about the nature of those things.
Thanks for making it to the end of my ramble.
IP: Logged
|
|
Micah Sparacio
Member
Member # 6
|
posted 05. November 2005 11:01
Terry, thanks for the very useful post. I apologize for the vaguery of my post.
I too am strongly inclined towards a naturalistic/physicalistic stance primarily because the universe and the things in it make the most sense to me as being unified in a single system. We can call that system physical.
At the moment, I depart from most physicalist views on the following points:
1. I'm not convinced of causal monism (that there is only one sort of causal type and it is at the base of physical reality).
2. I'm not convinced that reduction is the sole means of explanation for giving us a true picture of reality. Sometimes reductions fail, and that failure is beneficial to our scientific understanding (the failure of Alchemy)
3. I'm inclined to think that the cleanest division of reality is according to causal powers (and that causal powers are the best criterion for a thing's essential features - what makes it the kind of thing it is). If that is the case, it seems clear to me that one of the most obvious divisions occurs between the animate and inanimate (the classical distinction between active and passive things). Another, less obvious distinction is between those things that deliberate, plan and and act in a persistent, directed fashion.
So I have physicalist inclinations, though I'm not committed to the view that we must fit reality into reduction nor am I committed to the ultimate simplicity (monism) of the physical world.
My view might be called physicalist pluralism, where there are irreducible physical and causal types, just like at the level of chemistry there are irreducible chemical types (elements).
So that was my background. Here is the response to some of your questions. Short answer: there is no thorough scientific research project within the intelligent design community regarding the nature of intelligence. Though, I suppose that in some pre-theoretical way, most ID theorists probably think that intelligence is in some way causally irreducible (namely, its causal powers are unique and distinct, though dependent on its material base).
Long answer: the absence of such a research project makes sense, given the nature of the subject matter. The study of structural and functional properties is much more conducive to scientific research than the study of mental properties. However, if it is true, as I believe, that intelligence is to be understood in terms of causal powers (since all things are to be understood in this way) then the best way to study intelligence is to study the processes and effects of causal scenarios involving intelligence. Generalizing over such scenarios might help to characterize the nature of intelligence.
As a passing note, I should suggest that it was common for the ancient greeks to make distinctions between things based on causal powers (at the risk of being anachronistic). The active/passive distinction was pervasive. Plato even thought that the only source of motion was "soul" which we might characterize as intelligence. And, of course, many of the ancients thought that the most basic distinction to be made in the world was that between animate and inanimate things.
[ 05. November 2005, 11:06: Message edited by: Micah Sparacio ]
IP: Logged
|
|
Micah Sparacio
Member
Member # 6
|
posted 05. November 2005 11:27
One definition of intelligence that I've found particularly useful is one given by R.S. Peters (paraphrased because I don't have the reference)
Intelligence is the ability to vary movements relative to a goal in a way which is appropriate to changes in the environment that relate to the goal.
According to peters, most intelligent actions fall into what he calls the rule following purposive model. Thus, intelligent beings have to be capable of following rules in order to achieve some goal. In rare cases, the intelligent being encounters situations for which it has no rule to follow (the Frame Problem in Artificial Intelligence). In these cases, the intelligent agent has to be capable of inventing prospective rule-sets.
Another useful definition that I've encountered is that intelligence is "directed action." Things that have the ability (causal power) to cause things in a directed fashion are intelligent. Of course this definition seems too wide.
Finally, I wanted to say in my previous post that one of the troubling things about explaining phenomena in this universe is that there seem to be a hierarchy of levels, and we don't have good criteria for what constitutes a good explanation of phenomena at the higher-levels (human psychology) without mere reduction. Reduction, of course, has to end somewhere or else we have a recess problem. But how are we to understand things at the higher levels of reality. This might be called the deepness problem: how deep to phenomena go into reality. The question of course pesupposes that the more reduced, the more fundamentally real. I'm not sure that this position is strongly justified.
In any case, what I wanted to say was that physics might be able to give us the necessary conditions under which intelligence (or at least human intelligence) occurs. Necessary conditions reveal dependency conditions. But we are still left with the need of sufficient conditions: what are the causes of creativity, planning, directed action, etc.
IP: Logged
|
|
Terry Mullett
Member
Member # 1810
|
posted 05. November 2005 14:53
Hello, Micah, thanks for replying.
I have to ruminate some on the lack of such a research program making sense. I understand your point but it does give me some concern.
For instance, one of the ideas in AI research that has been gaining a lot of traction is that an effective approach to intelligence requires that it be situated. This may be related to the idea that, as you mentioned, its causal powers are possibly "...dependent on its material base". One thing I see as potentially a problem is the extrapolation, from that kind of understanding, to placing intelligence prior to the existence of a material world. Of course we could try to approach that problem by saying that there are other possible bases than material kinds which predate the material kinds, but I don't see where that gets us. It would basically make the materiality of the base irrelevant and give us a broader category of bases. Would we not then have to wonder whether intelligence is necessary to explain the broader category?
BTW I'm sure that previous line of thought was old hat for a lot of the people who frequent this board. I'm more trying to catch up here than break ground, so please no one think that I don't realize that. Just trying to show what my thought process is.
I guess what I really consider the big question is: how well-understood must something be before it can be effectively applied as an explanatory principle? Let's take another explanatory principle as an example: gravity. I'll pick this because I'm not so up on quantum mechanics or other more sophisticated physics. As I understand it, we are a little short of a complete account of gravity. However, we do have Newton's equation, among other things, and can say some rather specific positive things to characterize gravity even if we can't completely account for it. This kind of characterization allows us to explain or predict some details in the world with a high degree of confidence. In other words, it is from a well-characterized explanatory principle that we derive positive control over the subject matter. It seems to me that if intelligence is not similarly characterized, it can never have the same explanatory power. It would be as though physicists could only invoke the word "gravity", with some basic hueristic that it attracts bodies to one another... a much weaker stance.
Still, I'm sure people could have thought and spoken in terms of gravity well before Newton and others characterized it. And it could be argued that it was necessary to first catalog and analyze the observable effects before the characterization could be made. This is all well and good. But I do think that achieving the position of strong characterizations of fundamental principles is a breakthrough in any field which is necessary if we are to have much confidence in the conclusions drawn from those principles.
Of course I'm familiar with the idea that paradigms are not always commensurable between disciplines. However, if AI goes on to build successful real-world applications around a particular theory of intelligence (assuming it can converge on one), giving credibility through demonstration to that theory, and if that theory cannot support what ID needs intelligence to do, would that leave ID with a suspicious primary explanatory principle? I wonder if ID researchers don't really need a theory of intelligence that accounts for what AI demonstrates and still supports their own theoretical positions.
IP: Logged
|
|
Micah Sparacio
Member
Member # 6
|
posted 05. November 2005 17:40
Terry...I'll have more to say later but I just wanted to clarify something that I think I said poorly in the first place.
1. First, I wholeheartedly agree that there *should be* a research program in the intelligent design community on the nature of intelligence, though there is not
2. Second, what I meant is that it doesn't surprise me that there isn't a research program because the nature of intelligence is a much harder problem, than the mere identification and cataloguing of IC systems (which is no easy problem in itself)
IP: Logged
|
|
Terry Mullett
Member
Member # 1810
|
posted 05. November 2005 17:52
I'm also not quite on board with the definition of intelligence as simply directed action. Having seen (from Kitcher) how natural selection can also be seen in terms of directed action, I also think it would be too broad. Of course if I wanted to close my mind to ID right now, I could put those two ideas together and say that offering intelligence up as an alternative to selection makes no sense on the basis that selection has the charactaristics of intelligence in the first place. But that doesn't seem useful and I'm not suggesting it seriously.
What I got out of Kitcher is that having a goal is what distinguishes between selection and the purposive action of an agent. But it doesn't seem to me that the possession of a goal is itself causally irreducible. It brings to mind things like desires. It at least suggests the possession an internal representation of some state of affairs more advantageous than that which exists, and a way to get (representationally) between states of affairs in a coherrent way (rules, then). And these things would have their own explanations.
I hope no one minds that the last paragraph was an experiment on myself. After Micah had emphasized the word "ability" in the paraphrase of Peters, I wondered if I could get all the way through that thought without speaking of ability. It felt pretty artificial, and I don't believe it was entirely successful. Why would I want to try that? Here I'm taking ability as akin to effectiveness, as in potentially having an effect. As Micah said, causal powers. I wonder whether interacting with the outside world based on internal representations is a necessary condition for intelligence, or whether it is sufficient that the representations exist. The latter idea seems of no use at all to ID, which needs to explain phenomena, and some AI theorists question its value there, though in AI it does have interesting properties. A theorem prover program can start with a set of axioms and production rules and derive a set of theorems without moving the outside world. There could be some attribution of goal-seeking here without the need for that intent to change the state of affairs outside the program. Is this attribution sensible? (For a moment I'll ask anyone unwilling to attribute intent to a machine to imagine a physically idle philosopher instead). I'm not sure. One could say something is being moved with the production of theorems, but now I get fuzzy about what's internal and what's external. Does causality on purely internal states count? Some AI theorists, such as Brooks or Brickhard, claim that grounding in or interaction with a physical outside world is necessary component. A broader view may eliminate the word "physical", but I don't see that it changes the point.
All of this should relate to my original post, right? I'm wandering around. Anyway, what I seem to be heading for is this: AI has struggled long and hard with itself and others (such as Searle or Penrose) over what is sufficient for attributing something to intelligence, and continues in some ways struggle over it. That hasn't stopped research from moving forward, but it's apparently not a given. ID is interested in attributions to intelligence too. Is there really no one in the ID community engaged in a similar line of research?
IP: Logged
|
|
Terry Mullett
Member
Member # 1810
|
posted 05. November 2005 17:55
Micah, we seem to be posting pretty close to simultaneously. ![[Smile]](smile.gif) Yes, defining intelligence is a much harder problem.
IP: Logged
|
|
|
|
David L. Hagen
Member
Member # 323
|
posted 11. November 2005 23:37
Thanks Terry for raising the question of intelligence vs design.
As a mechanical engineer working with design, I have been exploring how to "reverse engineer" the universe and come up with design principles and to formulate a general theory of ID. So your question is a good one to distinguish intelligence from design in order to establish the high level goals.
A classic example of what is "Design" is the section "Seven Stages of Computer-Aided Engineering Design" in Ch 2 of Mechanism Design: Analysis and Synthesis Vol 1 by Arthur Erdman & Geoge N. Sandor, 3rd Ed. 1984. Prentice Hall ISBN 0-13-267782-2.
Briefly, these include steps of:
1A Confrontation: Encounter with a need to take action.
1B Sources of Information.
2A Formulation of problem. Clarify the problem to identify the real need.
2B Preparation of information and assumptions.
3 Generation and selection of design concepts.
4 Synthesis.
5 Analyzable model.
6 Experiment, analysis, optimization.
7. Presentation. To those who use & make it. Iterations.
The reverse engineering particularly focuses on colating and formulating the principles used in 1B and 2B etc.. To form a theory of ID, we need to be able to have at least minimal definitions of what is Intelligence, vs what is Design.
Some preliminary thoughts are:
What do human beings do in general apart from design?
How do we describe intelligence that is before/separate from design?
Does intelligence appreciate good applications of intelligence?
e.g. What do physicists mean by "elegant" solutions?
What do we mean when we comment on "Good design" from "poor design"?
Minimally, I think that Intelligence comprises the abilities:
To think and reason.
To communicate.
To ask and respond to questions.
To compare.
To evaluate.
These in turn require at minimum the capabilities to:
Store information.
Copy information.
Transform information from one form to another.
Process boolean transforms on information.
e.g. OR, AND, NOT.
Process algorithms.
I expect another aspect of adding "Intelligence" before "Design" is to contrast it to the public with the apparent goal oriented descriptions "natural selection" and "the selfish gene" in which neither involve any intelligence.
I would welcome feedback into what are minimal descriptions of Intelligence that in turn can be used to define minimal design requirements for an "Intelligent Designer." That will help in establishing a general theory of ID.
Keep exploring.
IP: Logged
|
|
Terry Mullett
Member
Member # 1810
|
posted 12. November 2005 22:39
Hi, David,
I'm glad to hear from an engineer. Frankly, one thing that has worried me a little was that the ID community may not include enough people who actually do design. That could result in the community's concept of design being theoretically interesting, even logically sound, but unrelated to anything real. Of course, that could just be an engineering bias
For instance, as someone who designs complex systems, I witness irreducible complexity cropping up in places where attention to design was lacking. This is distinct from the argument that IC indicates badly-done design, which I think is well-trodden ground, so I won't join that argument here. What I'm suggesting is that, in some practical contexts, IC can indicate an area in which design wasn't so much botched as it was missed entirely.
On the subject of elegance, I'd like to offer a viewpoint, as an aside. Everybody has an opinion, right? I'm neither a physicist nor a mathematician, but I think the most elegant thing I've ever seen is eπi+1=0. Schoolboy stuff maybe, but I said I'm not a mathematician. Anyway, I stand in much greater awe of Euler's little insight than of large exponents, generous significant digits, or specialized operations. In one concise statement, Euler related exponents and logarithms to trigonometry, and also related five of the most important fundamental constants in mathematics as well as all of the most fundamental operations. This little identity unifies the rational, the irrational, the complex and the trancendental, and sets the whole thing equal to zero. It makes the hair on the back of my neck stand up. What I find most remarkable is that all this (and, I suspect, more that hasn't been grasped) is accomplished with the simplest of statements. In comparison, big numbers and gratuitous precision seem crass, the intellectual equivalent of bludgeons, while mathematical arcana just seem so awkward that they must be missing some essential simplifying point. I guess for me, then, elegance is what distinguishes a deeply insightful formulation from one that is merely clever or merely forceful.
As a software architect, I find this familiar and comfortable:
quote:
Briefly, these include steps of:
1A Confrontation: Encounter with a need to take action.
1B Sources of Information.
2A Formulation of problem. Clarify the problem to identify the real need.
2B Preparation of information and assumptions.
3 Generation and selection of design concepts.
4 Synthesis.
5 Analyzable model.
6 Experiment, analysis, optimization.
7. Presentation. To those who use & make it. Iterations.
This is more or less what I recognize as real-world design. With a little tweaking of terminology, the science of software design also boils down to pretty much the same narrative, and that shouldn't be surprising. This is actually a general problem-solving approach, and so it can be applied to design activities in a broad sense, emphases modulated here and there.
The Software Engineering Institute has been developing a discipline around architecture reconstruction. This discipline addresses the problem of a software architect who inherits a complex system and must recover its design so that the system may be documented, enhanced or replaced without missing important functionality, rules or subtle interactions with related systems. I see some analogy between this and what ID attempts. I'm not sure whether it could add anything to your reverse engineering thoughts, but in case it could, you can find an entry point at http://www.sei.cmu.edu/architecture/ata_extraction.html .
I think if I were to suggest any addition at all to your list of minimum capabilities, it might have to do with interactivity. The theoretical problem with merely representing and manipulating information is that it requires an embedded interpreter, and then we start to see a regress forming. To make that bottom out somewhere ,we need an account that gives the stored (or, I think it's more accurate to say, represented) information and its processing an independent source of semantic content.
IP: Logged
|
|
|
|
Scott A. Richardson
Member
Member # 1855
|
posted 02. January 2006 11:32
It seems to me that:
Intelligence = unnatural selection.
Design = unnatural complexity.
IP: Logged
|
|
David L. Hagen
Member
Member # 323
|
posted 02. January 2006 15:50
Pithy point Scott
Terry thanks for input:
quote:
I think if I were to suggest any addition at all to your list of minimum capabilities, it might have to do with interactivity. The theoretical problem with merely representing and manipulating information is that it requires an embedded interpreter, and then we start to see a regress forming. To make that bottom out somewhere ,we need an account that gives the stored (or, I think it's more accurate to say, represented) information and its processing an independent source of semantic content.
I looked up these definitions for clarity:
quote:
intelligence [COMMUN] Data, information, or messages that are to be transmitted. [PSYCH] 1. The intellect or astuteness of the mind. 2. Ability to recognize and understand qualities and attributes of the physical world and of humankind. 3. Ability to solve problems and engage in abstract thought processes.
semantics [COMMUN] The branch of semiotics that deals with the relations between symbols and what they stand for, and defines the meaning that is prescribed for a statement by its originator.
design [SCI TECH] The act of conceiving and planning the structure and parameter values of a system, device, process, or work of art.
design engineering [ENG] A branch of engineering concerned with the creation of systems, devices, and processes useful and sought by society.
(c) 2003 McGraw-Hill Dictionary of Scientific and Technical Terms 6th Ed.
quote:
Intelligence General mental ability due to the integrative and adaptive functions of the brain that permit complex, unstereotyped, purposive responses to novel or changing situations, involving discrimination, generalization, learning, concept formation, inference, mental manipulation of memories, images, words and abstract symbols, education of relations and correlates, reasoning, and problem solving. . . .
semantic analysis [COMPUT SCI] A phase of natural language processing, following parsing, that involves extraction of context-independent aspects of a sentence's meaning, including the semantic roles of entities mentioned in the sentence and quantification information, such as cardinality, iteration, and dependency.
semiotics [COMMUN] The theory of signs and symbols, entities that represent some other thing; it includes syntactics, pragmatics, and semantics.
(c) 1984 McGraw-Hill Concise Encyclopedia of Science & Technology
Some Links to Semantics:
Semantic Web RDF Semantics
Any better definitions, links and references?
Per your suggestions, following are some possible high level requirements:
1) Intelligence: Semiotic activity including syntactic, pragmatic, and semantic processing.
2) Design: Creation of systems, devices and processes.
Possible particular requirements for Intelligence:
1.1) Establishing independent semantic content.
1.2) Converting the form of semantic content.
1.3) Processing represented information.
1.4) Interaction using semantic content.
Not being an expert in this field, I look forward to any clarification/correction and improvements.
Focus needs to be on summary evidence that can be used to establish goals and objectives for reverse engineering or software architecture discovery. In particular, on formation and use of biochemical codes and information processing including DNA, RNA, amino acids, self reproducing cells, glycoproteins in immune systems, chirality, and the central nervous system, etc. in complex systems, as pertinent to forming and distinguishing the explanatory power of various origin theories, particularly for macroevolution vs ID.
[ 02. January 2006, 16:13: Message edited by: David L. Hagen ]
IP: Logged
|
|
|
Printer-friendly view of this topic