|
Author
|
Topic: Smallest Self Contained Unit of Intelligence
|
warren_bergerson
Member
Member # 262
|
posted 15. November 2003 07:58
Steve,
The mega-intelligence perspective starts by viewing intelligence as a property or phenomena of information processing performed by information processing units. Similar to your description, an information processing unit involves and is defined in terms of input, processing, and output. It is worth noting that ‘intelligence as a product or property of real world or materialistic information processing’ is not universally accepted.
The second key concept of the mega-intelligence concept is ‘specialized processing’. Specialized processing recognizing that just different cells in a multi-cellular organism perform specialized processing, so different information processing units perform different information processing. Different units may share a variety of logical properties but the processing performed or the functions performed are not identical. A particular act of intelligent behavior may require or depend on information processing performed by many different types or sub-classes of information processing unit.
The mega-intelligence perspective extends this specialized processing concept beyond the limits of an individual organism. The mega-intelligence concept recognizes that dispersed and specialized processing can involve processing performed in other units. This dispersed, ‘outside the human head’ processing is believed to be a key factor, even a defining factor in human intelligence.
The mega-intelligence perspective also recognizes what can be labeled ‘sequential progressive’ processing. In a simple form sequential progressive processing represents the step by step processing involved in many types of problem solving. On a much broader scale, sequential progress processing recognizes that 1)pre-genetic intelligent processing produced genetic processing, 2)genetic intelligent processing produced intelligent neuronal processing and 3)intelligent neuronal processing produced intelligent human processing.
Once you decide to pursue some form of the mega-intelligence perspective, or some alternative viewpoint, the next question as nobody asked is how do you apply it. As your comments suggest, analysis of intelligence as information processing ultimately requires looking at the processing performed by individual units. Part of the challenge for the mega-intelligence perspective is distinguishing between the common or universal features of all processing units and the features responsible for specialized processing.
My views on how to use the mega-intelligence perspective are outlined and discussed in the automatic learning thread referenced earlier. IMO, intelligent processing units are characterized by goal-seeking or teleological processing. This is quite different, I believe, than the process you described.
IP: Logged
|
|
Bruce Schuman
Member
Member # 1048
|
posted 02. January 2004 17:31
Responding to this initial idea of “the smallest unit of intelligence” – may I propose an approach that turns this discussion upside down? The approach offered by our moderator suggests that “intelligence” is too big to be fitted into a single nervous system – and in a related article http://groups.msn.com/LifeEngineering/unitofintelligence.msnw offers ideas regarding “all life forms”. This approach is described as “mega-intelligence”. I’d like to suggest an approach that defines “the smallest unit of intelligence” as possibly very small indeed – with the idea that, if defined well, we could then build out from this algebraic kernel or essential idea, into the most detailed and inclusive levels of complexity – including such composite objects as “the internet”.
One of my operating presumptions is that we should look for ways to remove any sort of stipulative values from our definitions – such that we aren’t trying to create universal solutions based on assumptions that essentially involve logically arbitrary boundary values (how would we distinguish “intelligent” from “non-intelligent”, except by stipulating a boundary value in some dimension?). My instinct is to look for irreducible algebraic or mathematical ideas that are of “absolute” generality, and then attempt to construct from these elements the kinds of constructs discussed in other messages here – “constructs” like “dogs” or “cockroaches” – or, indeed, “the internet”.
To offer an absolutely bottom-up approach -- what about taking the view that “intelligence is the ability to perceive a distinction”? I am persuaded that all conceptual structure is assembled in a recursive/hierarchical format across levels of abstraction – and that objects like “dogs” are best understood as “holons” – constructed units that are both parts and wholes – and defined in terms of a series of boundary values and attributes (this gets into the details of mathematical taxonomy).
My own research these days is concerned with what I am presently calling “the monad” – which, for me, is the “smallest conceivable unit of distinction”. It is clear to me that we can build any abstract concept by building composite assemblies of distinctions (building abstract objects like “dogs”). Way down at the bottom of the analytic hierarchy, where we find the smallest distinctions from which dogs are assembled (molecular, atomic, some number of decimal places in measurement, etc)
So, the question for me is – how do we make the lowest-level distinction – in color shading, for example – so that we can say “this is different from that”? This is my “lowest level unit of intelligence”. If I can answer that question, defining a solid algebraic unit that defines this distinction-making process so that it can say that .7847634 is different from .7847635 – then I think we can build any higher-level composite structure with absolute algebraic (and hence computer-compatible and real-world stable) precision.
So, I propose the definition that intelligence is that something that makes the distinction between two things that we formerly perceived as “the same thing”. This is the cellular division of intelligence at the lowest level – and if we fully understood this process, I am inclined to claim, we could then build from this irreducible unit of distinction any composite object, at any level of complexity – without introducing instability, or ambiguity.
What is needed for this “irreducible module”? Some capacity to sense, and some capacity to register a distinction. Build from there – and we can create an inductive definition chain that goes from the smallest detectable unit of meaning to the largest.
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 03. January 2004 08:36
Bruce,
If we take any intelligent behavior, we can always reduce it to information processing or input-output operations of the type associated with the operation of computer systems. Mathematician, can reduce any input-output operation to sets of simple elementary mathematical/logical operations. [I believe all such operations are ultimately reducible to sets of ‘input causes output’ or ‘A causes B’ type relationships.]
The fundamental problem with any such reductionist approach to defining intelligence is that you ultimately get down to tiny units which are indistinguishable from the logical, mathematical causal units which make up non-intelligent behavior. The same issue arises when you attempt to develop a definition which distinguishes living matter from non-living matter. Living matter, as far as is currently known, is made up of the same atoms and forces as non-living matter.
Elementary operations like ‘different reactions or responses to different stimuli’ (the behavioral equivalent of perceiving a distinction) are common to both intelligent and non-intelligent objects.
Again lots of people have tried to find elementary physical components or forces or processes which can differentiate living from non-living or intelligent from non-intelligent. None of them work or are likely to work.
The purpose of this thread was to explore the concept of ‘connected to other living things’. Life, at least on earth and in the absence of human interference, is in part defined by a connection with or dependence on other life forms. Intelligence, at least on earth and in the absence of human intervention, appears to depend on this same general type of connected to other life forms criteria.
The general observation that intelligence appears to be connected to life forms, suggests that it might be useful to explore the possibility of defining intelligence in terms of ‘what it is connected to and how it is connected’ rather than in terms of properties which can be isolated in small units.
As discussed in the definition of intelligence thread, the ‘connect to’ property does appear to be a useful concept in defining intelligence. The connect to property by itself is not adequate to distinguish between living and non-living or between intelligent and non-intelligent, but it seems to be a useful and possibly even essential component of an effective definition.
Reductionist approaches to defining intelligence might appear reasonable, but to date they have proved uniformly unsuccessful. This has, it appears, led to a wide spread belief that intelligence is undefinable.
IP: Logged
|
|
Bruce Schuman
Member
Member # 1048
|
posted 03. January 2004 15:47
Thanks for this comment (I haven't quite figured out how I can read what you wrote while posting a response -- maybe open a second browser window?)
How about -- I stick your entire message into this reply, and then respond in "dialogue format"....
*****
Just let me say -- that though I want to take a slightly different approach at the moment, and am working with "bottom up" models -- I want you to know that I am interested in these broader approaches, and think the kinds of questions and issues you are looking at are very interesting. So, if I seem to take a different tack, it's not necessarily because I "disagree" -- but because I am working with a slightly different approach...
And I might add that my own concern is not so much to try to "define intelligence" as it is to understand "how intelligence works". If I can create a process that does "intelligent-like things" -- we could debate whether or not it was "intelligent"....
Warren: If we take any intelligent behavior, we can always reduce it to information processing or input-output operations of the type associated with the operation of computer systems.
Bruce: By "reduce it to", I understand you to mean -- describe the behavior in these terms....
Warren: Mathematicians can reduce any input-output operation to sets of simple elementary mathematical/logical operations. [I believe all such operations are ultimately reducible to sets of ‘input causes output’ or ‘A causes B’ type relationships.]
Bruce: Ummmm -- I guess a lot depends on the details of these descriptors. If we have some highly complex unit and call it "A" -- we don't know much about "A". You remember the famous joke on this -- the lecturer is standing at the blackboard drawing a diagram, talking about some organic barnyard process, draws a circle, and says "Assume a spherical chicken...."
Warren: The fundamental problem with any such reductionist approach to defining intelligence is that you ultimately get down to tiny units which are indistinguishable from the logical, mathematical causal units which make up non-intelligent behavior.
Bruce: Hmmm. Now -- I am wondering about this -- because I think my own ideas on this subject -- what is happening down at "the lowest level of analysis" -- involves some degree or element of motivation. That's a theme that gets a lot of attention on this site -- and if we add some degree of motivation to our units of analysis -- it seems to me -- they take on a "life of their own" -- they become, in a sense, animated.
Last night as I was thinking about this, I was considering the idea that the general form of this motivation is "uncertainty reduction". That concept could be developed in to a general theory (or definition) of knowledge. "The objective of scientific inquiry is to reduce uncertainty". As we reduce uncertainty, we gain control, and we can relax. So, perhaps the desire to relax is the motivational drive behind uncertainty reduction -- and hence, from, perhaps, to all creation of knowledge.
Warren: The same issue arises when you attempt to develop a definition which distinguishes living matter from non-living matter. Living matter, as far as is currently known, is made up of the same atoms and forces as non-living matter.
Bruce: I am very sympathetic to that concern. Creating definitions of this sort always involve stipulating boundary values -- in ways that always seem to me to be logically arbitrary. So -- I don't want to try to distinguish living from non-living matter. I'll leave that to the experts.
Warren: Elementary operations like ‘different reactions or responses to different stimuli’ (the behavioral equivalent of perceiving a distinction) are common to both intelligent and non-intelligent objects.
Bruce: That's a good point. Maybe if we had an example to consider, we could get into more details.
Warren: Again lots of people have tried to find elementary physical components or forces or processes which can differentiate living from non-living or intelligent from non-intelligent. None of them work or are likely to work.
Bruce: I can see there could be problems with this effort...
Warren: The purpose of this thread was to explore the concept of ‘connected to other living things’. Life, at least on earth and in the absence of human interference, is in part defined by a connection with or dependence on other life forms.
Bruce: Just let me say -- that this is a theme that I find very interesting -- from a "spiritual" point of view. I do believe in "the interconnectedness of all life" -- and that this interconnection may be more profound than we realize, or take forms or exert influences of which we are still largely unaware....
Warren: Intelligence, at least on earth and in the absence of human intervention, appears to depend on this same general type of connected to other life forms criteria.
Bruce: Hmmm. It will take me a little time, and a little more reading, before I can see why this would be the case.... (In analogy to "if a tree falls in the forest, and no one hears it, does it make any sound" -- maybe you are saying "If a person does an intelligent thing, but nobody notices, are they intelligent"? )))))
Warren: The general observation that intelligence appears to be connected to life forms, suggests that it might be useful to explore the possibility of defining intelligence in terms of ‘what it is connected to and how it is connected’ rather than in terms of properties which can be isolated in small units.
Bruce: I will be happy to see how you develop that definition -- (thinking of such issues as what the structure is composed of, what are its "subunits" or elements, how they are related to one another, where they reside, how they produce or result in intelligent behavior, etc.)
Warren: As discussed in the definition of intelligence thread, the ‘connect to’ property does appear to be a useful concept in defining intelligence. The connect to property by itself is not adequate to distinguish between living and non-living or between intelligent and non-intelligent, but it seems to be a useful and possibly even essential component of an effective definition.
Bruce: I will look into that thread...
Warren: Reductionist approaches to defining intelligence might appear reasonable, but to date they have proved uniformly unsuccessful. This has, it appears, led to a wide spread belief that intelligence is undefinable.
Bruce: I appreciate the reply. These are complex subjects, with lots of details, and endless amounts of controversy. But the discussion is stimulating, and I appreciate the good company.
- Bruce
http://originresearch.com/sd/home.cfm
[ 03. January 2004, 16:22: Message edited by: Bruce Schuman ]
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 04. January 2004 10:47
Bruce,
One of the issues that appears to be central to the analysis of intelligence and intelligent processes such as evolution is the gap that exists between engineering/applied science/applied math/applied computer science and systems design and academic science/academic mathematics. A phenomena such as motivation provides a good illustration of the huge knowledge gap that exists.
Both academic and applied science/math are in general agreement that science involves the analysis of causal relationships and that in basic set theory, causal relationships are expressed in terms of or are reducible to mathematical functions. The applied/academic split seems to arise when you begin to consider complex causation involving or defined in terms of sets of interrelated causal relationships.
Applied science and mathematics use a complex type of causation that I have labeled DDP. In computer science such complex causal processes are expressed as computer programs. In various fields of engineering this type of complex causal process is expressed in other forms. Relevant to the discussion here, it is reasonably well established mathematical concept that going from elementary causal relationships to DDP’s involve only standard known logical/mathematical operations. From DDP complex causation to elementary set theoretic functions and simple ‘A always causes B’ causation does not involve any unknown phenomena such as motivation.
Designers of computer systems have succeeded in modeling, simulating, and creating testable predictive models of a variety of biological behaviors and brain functions. All of these successful simulations have been generated using a type or class of complex causal process that I label ‘goal-directed DDP’s’. As with DDP’s in general, goal-directed DDP’s are definable entirely in terms of known logical/mathematical operations.
While phenomena such as motivation are not a component of goal-directed DDP’s, motivation can be an emergent or apparent property of such a system. In other words, although motivation is not a logical property or process used to define and construct a goal directed DDP, a system build with a goal directed DDP can create the appearance of being motivated.
It appears that any type of biological behavior or process can be modeled and simulated by goal-directed DDP’s. As far as I know, no one has attempted to formally express and review this apparent principle, but at least some systems designers in attempting to simulate complex behavior appear use the principle.
To summarize, I am suggesting that many of the issues arising in the analysis of intelligence have already been addressed successfully by applied scientists, applied mathematicians and systems designers. Before we start to get creative in explaining intelligence we should look closely at the concepts actually used in successful simulations of complex behavior and in the successful redesign or reengineering of complex causal processes. Reviewing and analyzing successful applications, I suggest, is far more useful than reviewing the literature.
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 04. January 2004 11:21
warren_bergerson wrote quote:
To summarize, I am suggesting that many of the issues arising in the analysis of intelligence have already been addressed successfully by applied scientists, applied mathematicians and systems designers. Before we start to get creative in explaining intelligence we should look closely at the concepts actually used in successful simulations of complex behavior and in the successful redesign or reengineering of complex causal processes. Reviewing and analyzing successful applications, I suggest, is far more useful than reviewing the literature.
What are some specific examples of those successful simulations? Where can we read about them? How do those specific examples illustrate "goal directed DDP" processes? If "... at least some systems designers in attempting to simulate complex behavior appear use the principle," as he asserts, surely somewhere there must be publications describing that use. warren_bergerson says quote:
Reviewing and analyzing successful applications, I suggest, is far more useful than reviewing the literature.
That's real hard when warren_bergerson does not provide any specific examples of successful applications to review and analyze. (And where, pray tell, are those examples described if not in the professional literature of the field?)
RBH
[ 04. January 2004, 11:25: Message edited by: RBH ]
IP: Logged
|
|
|
Printer-friendly view of this topic