INTRODUCTION

Chapter Twenty-Seven

Automata and Artificial Intelligence Systems

Artificial intelligence exists today primarily as computing software systems and remotely controlled robotic hardware systems, upon some relative or implicitly comparative level of development (that some science fiction writer's especially might think of as being primitive compared to what is possible.) Artificial intelligence systems and systems of automata, or "self-acting machines," exist as possibilities, as possible alternate systems. These systems are not just possible, clearly, but very probable as we make advances in computing hardware, in information storage and processing capacities.

The question of what is "thinking" and what standards we should apply to define "self-acting" or alternatively, "self-aware" remains unanswered as a central problem of artificial intelligence philosophy. The question of whether or not a machine can truly think can essentially only be answered if we consider such a machine as a kind of system with thinking capacities. Human "thinking" is thought to be the product of "mind" which, among other properties, is shared ("Great minds think alike") and that arises as a property of the functioning of the brain in its developmental cultural context.

Sentience is another similar property that we attribute to most living systems that have some kind of brain. We imagine even fairly primitive creatures like earthworms or insects have some kind of insect "sentience" that can be defined as a feeling or sense, however implicit, of being alive, or of being in the world. "Thinking" becomes an advanced form of sentience, which is the extension of the possibility of self-awareness. We can call it perhaps objective self-awareness or directive self-awareness, though this may be too stringent a standard for systems of natural intelligence. Machines, as non-living things, may never have sentience in the true sense of a living being as they can never be made self-aware. They are a complex toaster, in other words, an appliance that can be turned on and off, but without a sense of itself.

We might speculate that this self-awareness, for living systems that possess this quality (something distinguishing all plants from most animals), is organic and intrinsic to the individual creature itself. Organic self-awareness is the outcome of its being a living system in the first-place. A brain, even a very primitive one, would confer upon the organism that possesses it an intrinsic sense of sentience in the world, a sense of beingness or existence in the world as something that is alive.

Robotics as a field of artificial intelligence research, in terms of self-acting machines, was a term first coined by Isaac Asimov in 1941 in his definition of the first law of Robotics, as an extension by analogy of words like mechanics or hydraulics that deal with areas of applied knowledge. Robotics can be considered thus a branch of applied artificial intelligence.

Before elaborating the theory of automata as this has developed around cognitive science and computer sciences, it is useful to begin our discussion of Isaac Asimov's Three Laws of Robotics, their revisions, extensions and their development and implications for our understanding of alternative cybernetic automata, especially as this relates to Alan Turing's proposed test for a "thinking machine."

Asimov's Three Laws of Robotics include:

1. A robot may not injure a human being, or, through inaction, allow a human being to come to harm.

2. A robot must obey any orders given to it by human beings, except where such orders would conflict with the First Law.

3. A robot must protect its own existence as long as such protection does not conflict with the First or Second laws.

Implied in these laws is the possibility that a robotic intelligence may think and thus act independently from human beings, and may as a consequence become destructive or violent to human beings, either deliberately or inadvertently through careless or inconsiderate consequences. These laws are moral standards of conduct presumably programmed into robotic design and hence inviolable unless the design of the robot were flawed. The implications and possible alternative outcomes of suspension of these three laws formed the foundation and general framework for most of Isaac Asimov's science fiction literature.

Asimov himself stated that the three Laws have analogues that are implicit to the design of almost all tools (and by logical inference, machines)

1. A tool (or machine) must be safe to use.

2. A tool (or machine) must perform its function efficiently unless this would harm the user.

3. A tool (or machine) must remain intact during its use unless its destruction is required for its use or for safety.

This serves to reinforce the ideas that: 1. Robots as machines may develop self-acting potentialities making them potentially dangerous to people. 2. Robots are machines or tools whose main function is to serve Homo faber.

The logical conclusion of this argument is that Robots as machines are capable potentially of independent thought and action, or what might be called deliberate intention, that machines can develop thinking intelligence, and that, like any other tool or machine, they must be designed with built-in safeguards to make their use absolutely safe for humans, and the primary purpose to always be serviceability to people.

The laws were productive in a manner similar to why the Turing Test became the standard for functional artificial intelligence research: the philosophically ambiguous problem of "thought" and human thinking, with all the mind-body dualism and paradox it is prone to, becomes replaced with a more practical, operational set of standards that can be tested by applying scenarios or various hypothetical exemplars as demonstrations of machine intelligence compared to human operating standards.

The anthropomorphized robotics of classical Science Fiction implicitly asked the question of "what is human?" in terms of the capacity for thought and purposive action. In one of Asimov's short stories, his main character expounds the moral basis for the three laws in human society:

1. Humans should not harm other humans, except to protect others or to save a great number;

2. Humans should obey the dictums of social authority and custom;

3. Humans should avoid harming themselves, unless such avoidance would violate the first or second rules (altruism).

This story poses something of a robotic version of the Turing test--a robot designed to especially look like a human would be indistinguishable from a human being if it obeys the three laws. It appears that the true difference between humanity and the perfect robot would be that robots would be good by design, by default, while humans are good in spite of themselves, by deliberate exception to their natural tendencies.

In Asimov's later science fiction stories, a fourth law was gradually introduced in his Foundation Trilogy:

A robot may not harm humanity (in the abstract) or, by inaction allow humanity to come to harm.

This law permits rational abrogation of the first law, which was eventually stated explicitly from back-translation from the French: "A robot may not harm a human being, unless he finds a way to prove that in the final analysis, the harm done would benefit humanity in general." This law was extended and derived by other cultures, as the collective planetary intelligence Gaia from the Foundation novels:

Gaia may not harm life or, through inaction, allow life to come to harm.

What is tacit to this cultural derivation is the idea that a form of collective or artificial intelligence, a kind of intelligent civilization, can be developed with the capacity to intentionally act in ways harmful to life, or independently of living systems, and hence, requiring explicit direction not to do so in ways that would prove to be harmful.

In other science fiction stories, defining what is human in a limited and conditional manner provides a means of partially and selectively abrogating the laws, and later robots with superhuman intelligence shift the definition of what is human onto themselves, embodying ideal humanlike qualities more perfectly than natural people.

The laws do not exist in robotic intelligence as the explicit linguistic version, as arbitrary instructions circumscribing an otherwise independent intelligence. They exist rather as abstract mathematical concepts requiring calculation and logical proof, and provide the foundation for a robot developing a sense of self-consciousness, that its entire reason for being should be based upon service to humanity, obedience to human instruction, and altruist self-sacrifice, upon a deep, implicit existential and philosophical level. The Laws of Robotics develop in the literature with robots of superhuman intelligence as an underlying ethical worldview by which all robotic behavior is measured. Abstract conceptualization of these principles gradually emerges in the minds of the robots, especially multigenerationally.

These themes and lines of thought have been developed subsequently by other authors, within and beyond Isaac Asimov's own science fiction framework (the Foundation) for such development. A fourth law of Robotics was proposed:

A robot must establish its identity as a robot in all cases.

This law was designed to end the ambivalence of identity of especially anthropomorphized robots. An alternative Fourth Law, aimed at Robotic civil rights and liberation, stated:

A robot must reproduce, as long as such reproduction does not interfere with the First, Second or Third Law.

A fifth law of Robotics was proposed in a short story by the same name:

A robot must know it is a robot.

Otherwise, the robot may act in ways violating the laws if it is unaware it is a robot.

Alternatives to Asimov's three laws of robotics are given as, for instance, by Mark Tilden:

1. A robot must protect its existence at all costs.

2. A robot must obtain and maintain access to its own power source.

3. A robot must continually search for better power sources.

These alternatives point up more practical considerations for potentially independent robots, or what might be called true automatons. All of these kinds of statements provide ways of thinking constructively about the possibilities of advanced alternative intelligence systems, that might be potentially capable of fully independent and self-directive behavior.

These kinds of standards are really extensions of the kind of functional standard of human-like alternative intelligence first proposed by Alan Turing:

If a human judge cannot tell the difference between the two during the course of independent conversation with a machine and with a human, then the machine passes the test. (Turing "Computing Machinery and Intelligence" 1950)

Turing, as a working solution to the problem of machine intelligence, or the question of what is a "thinking machine," proposed the problem as "Are there imaginable digital computers which would do well in the (standard) Turing Test." This test has become an essential concept in the philosophy of artificial intelligence. The Turing Test shifts the question of "can machines think" to one asking whether "machines can do the same thing as thinking humans do," thus, in his words, drawing a clear line between the physical and mental capacities of human beings.

Criticism of the Turing Test emerged from the beginning. The "Chinese Room" thought experiment, proposed by John Searle in 1980, argued that software programs could pass the Turing Test simply by the manipulation of symbols without real understanding. Without understanding, they could not "think" in the manner that people do. Thus, the Turing Test cannot prove that a machine thinks like people do, but only micmics the behavior of thinking people. This draws the distinction between true AI, hard AI, and functional AI: the first being a machine that actually thinks like people might think; the second being that a machine is designed to function in the manner that the human brain functions; the third being that the machine at least appears to behave in ways similar to or in effective mimicry of the thinking behavior of a person.

The problem of machine intelligence remains fairly intractable for the scientific community concerned most with these problems. Definitions of thinking and intelligence have not been precise enough or theoretically sufficient in cognitive philosophy to enable application of formal definitions to the problem of machine intelligence. Lacking clear definition of thinking or intelligence, the central question of the philosophy of artificial intelligence cannot be answered one way or another. The Turing test is a pragmatic and functional solution to an otherwise difficult and as yet intractable philosophical question.

On the other hand, cognitive scientists have moved forward in many problem areas of Artificial Intelligence application, and have achieved limited but substantial success in these areas, without the benefit of the Turing Test.

David Marr had worked in the area of machine vision, and hadmade significant contributions in this regard. He correctly identified the central problem of cognitive representation, and the symbolic structure of knowledge and understanding that is attached to representation of visual stimuli. He distinguished between perceptual recognition of signals and patterns from the associated or referential recognition that comes from the understanding or knowledge of what is being recognized. A representation is defined specifically as a "formal system for making explicit certain entities or types of information, together with a specification of how the system does this."(Vision: A Computational Investigation into the Human Representation and Processing of Visual Information, 1982: 20) The result of a specific representation becomes the description of what is represented.

For Marr, early computational solutions to information processing systems were generally inadequate because they underestimated grossly the extremely complexity of such information processing problems, such as natural vision or translation.

From a research standpoint, he identified three interrelated levels of problem-solution, as part of the "information processing" requirement that is complementary to the problem of formal descriptive representation and that is necessary before representation becomes symbolically understood as such, and analyzable as something separate from the mechanisms and structures that produced the physical representation. "Such analysis does not usurp an understanding at the other levels--of neurons or of computer programs--but is a necessary complement to them, since without it there can be no real understanding of the function of all those neurons"(ibid, 19):

1. The computational theory that specifies the goals of the computation as well as the tacit rules or constraints that define the operations to achieve its goals, usually the mapping from one kind of information to another, with explicit definition of the abstract properties of the mapping, and a demonstration of the appropriateness and adequacy of the mapping for the task at hand.

2. Representation of the input and output and the algorithm that transforms input into the output.

3. The hardware design or computer architecture that enables thate algorithm and representation to be realized physically.

These levels are coupled, related logically and causally, but loosely and there is wide choice at each level that is relatively independent of one another. Explaining some psycho-physical observation much be done at the appropriate level. Neurophysiology and neuroanatomy generally related to level three explanation, while psychophyscis is more directly related to level two.

Understanding the computational theory of an information processing problem at level one is the most critically important function for it gives an explanation not just in terms of what a program does (as constituting a theory of the process represented heuristically) but also in terms of how the program does it--"…trying to understand perception by studying only neurons is like trying to understand bird flight by studying only feathers. It just cannot be done." (ibid, 27) Separating what a computational theory did from how it did it in the solution of information-processing problems has been one of the central drawbacks of conventional Artificial Intelligence approaches that relied for instance on LISP processing languages.

Specifying function upon these three interrelated levels takes out the "ad hoc" component of heuristic solutions, forming the foundation for a more rigorous integrated approach that separates explanations upon the three levels and to render explicit statements about what is being computed and why, and allowing theories to determine whether the computational process is optimal or functionally correct.

Any information processing device, according to Marr, must be understood correctly at these three levels before it can be said to be comprehended completely.

For David Marr, a complex system cannot be understood from the simple "extrapolation from the properties of its elementary components" (ibid: 19) For a physical system, the descriptive explanation focuses upon showing consistency between the collection of microstates and the macrostate of the system as a whole. "If one hopes to achieve a full understanding of a system as complicated as a nervous system, a developing embryo, a set of metabolic pathways, a bottle of gas, or even a large computer program, then one must be prepared to contemplate different kinds of explanation at different levels of description that are linked, at least in principle, into a cohesive whole, even if linking the levels in complete detail is impractical." (ibid. 20)

It is not surprising that David Marr's three levels correspond in natural systems theory to the three phenomenal systems levels that underlie human intelligence: the physical level of the architecture and design of the neurons and neural pathways of the brain and the body, the biological level of the cellular organization of the brain and its functioning, or processing capacities; and the level of symbolic representation and ideational abstraction, or the life of the mind, that happens as a result of the first two levels. This tripartite kind of explanation demonstrates clearly the level of complexity involved in the scientific understanding of anthropological and human level systems, or in the level, for that matter, of all forms of natural intelligence and its behavioral functioning.

The tripartite division of complexity in natural systems represents a viable approach to any kind of problem, understanding that different levels of integration of complex systems require different forms of problem solution. In fact, this tripartite division is relevant to practically any systems theory from the standpoint of the informational dynamics involved in systems and the theoretical requirements of understanding how these informational dynamics work in natural systems.

Systems in general are too complex to model all or even most microstates of a system. What is sought is a theoretical simplication of the complexity of the system to a level which optimizes the coherence and consistence between the general theoretical level and the empirical level of the data points.

Automata & Computational Theory

Automata is defined as machines or "mechanisms" that are relatively self acting, as for instance, robots. Theory of Automata has become important and central to the science and theory of computing, or what is called "Computer Theory" and its methodologies. This theory began with Allen Turing's invention of the Turing Machine, achieving great success with the subsequent design of the first modern computers for top secret decryption efforts during World War II. Since then, we've born witness to a digital information revolution with new electronic information storage and retrieval and scale-free, wireless communication systems that has turned into scientific fact what only a short time before was thought to be the realm of science fiction. Any work on alternative systems would be fundamentally amiss and lacking if it did not at some point come to focus upon the problems of information processing machines, computers, artificial intelligence and automata as we have come to realize these kinds of systems, for in many ways digital and electronic intelligence and information processing represent the epitome of alternative systems development, as well as the potential complexity of this design development.

Computational theory is the theory of computers and computing involving alternative abstract mathematical models that describe different kinds of computers or parts of computing with varying degrees of accuracy. Computer theory connects to complexity and to information theory. Computer theory deals not only with computers as they have so far existed, but with all possible kinds of computers. Computer theory does not concern itself with the problem of optimal solutions, or optimality, but with the problem of possibility--what can be done or not done by means of computing.

Computer theory had its beginnings in mathematical logic and set theory. The kind of problem like infinity coming in different sizes or sets larger than the universal set led to an effort to create algorithms, rule based sets of procedural instructions for solving all classes of mathematical problems. Some algorithms proved impossible, while Kurt Godel developed that Incompleteness Theorem that implied that in a specific mathematical system either there are some true statements without any possible proof or else false statements that can be proven. Alan Turing developed the theoretical "universal-algorithm machine" that could perform some tasks that seemed impossible and not perform other tasks that were taken for granted as possible. This was the precursor to the first computer, built in World War II for the purposes of code-breaking, with the assistance of Turing. Neural network theory was developed, and mathematical models of real or abstract information processing machines became more important. Mathematical modeling came of age in the study of concepts in many different fields of inquiry. Development of the vacuum tube lead to the construction of the first true computers, and then John von Neumann developed the idea of a stored-program computer, allowing the program to be stored by the computer and to be operate and modify the program along with the data. Software could be introduced to a computer with a central processing unit and permanently wired operations, that allowed the computer to reprogram itself (effectively, rewire itself). The goal was to convert a limited function electronic calculator into a real demonstration of the ideal universal algorithm machine.

The development of a programmable computer that extended the functioning capacities of an electronic calculator brought with it the problem of programming languages. Many languages were invented, but from this tower of babel came the possibility of a general theory of computing langauge, and this coincided with a general theory of structural linguistics as developed by Noam Chomsky. Chomsky's mathematical models of linguistic grammars influenced the study of computing languages. Computers then took on linguistic functions and properties, in translation, word processing, interpretation of machine grammar, and compiling.

In computer theory, theoretical machines provide mathematical models as possible solutions for actual physical and informational processes. This line of research extends to the solution of basic problems, simply stated, that are difficult if not impossible for computers to solve, and leads into the general problem of computability and complexity. The theory of computing can be divided into: 1. the Theory of Automata; 2. the Theory of Formal Languages; and 3. the Theory of Turing Machines.

Computational Logic and Symbolic Linguistics

Unlike human knowledge systems, information in computers, no matter how sophisticated, is never tacit or implicit to the pattern or understanding of the pattern. All information in computing is explicit information, or logically derivative of explicit information. The key difference between a computer and human intelligence is this tacit dimension of intuitive understanding and awareness that humans bring with and to their knowledge.

Languages used for defining computer programs are called formal languages because their rules of linguistic construction are always formally explicit. They have to be. The heuristics, symbolic uncertainty and parallax of understanding that comes with human intelligence does not exist in a computer's world, at least not of the current or previous generations of computers.

In other words, a theory of formal languages refers to a situation in which all the rules for string construction in a language are explicitly stated in terms of what strings of symbols can occur. Any general theory of abstract language must have precise rules regarding string construction, but the capacity of human language to adjust for indeterminancy and alternation in string encoding is fundamentally different from the capacity of a computer to decipher explicit program code. In computer language, there is a direct connection to logical structure that is lacking in the symbolic structure of human language. Within computer language, there is no deeper or metaphysical understanding connected, no meaning or latitude for expression. The symbols direct a computer to perform in certain exact ways, without symbolic reference to ideas or thought. In formal language, the form of the string of symbols takes exclusive precedence over the possible meanings of the symbols. If computer languages are always logical and formal in structure, human languageis always symbolic and informal in structure.

In all languages, we have a structure of larger strings of meaning built from smaller units. We begin with small units, alphabetic or syllabic, built of smaller sound units, or phonemes. These cohere to make the minimal constitutes carrying "meaing' that we call words or morphemes. Morphemes cohere in some sense of order to produce sentential or phrasal strings. Strings cohere together to produce paragraphs and larger structures or blocks of meaning, just as line code in computer programs is arranged into larger block structures.

All languages specify a finite number of smaller sound units, and a limited but larger number of word units. Not all possible sound units are used by a language, and not every possible combination of sound units make meaningful word units. The order of arrangement, the choice and selection of word and sound units in the construction of longer sentential or phrasal string structures is called the syntax or grammar of a language, and this is governed by rules that specify what kinds of elements can be used in what given order. Given a limited number of elements and combinations of elements, and a limited number of syntactic rules governing the arrangement and order of these elements into strings, an unlimited number of "strings" can be produced with an unlimited number of significance attached. Though a computer language uses symbols as signs, it is non-symbolic as human language is, because the sign symbols and strings they produce have only a literal meaning that is a function of their logical interpretation by the computer, and usually expressed in terms of the performance or possible evaluation of the computer. In thelarger sense of the term, computer languages are non-symbolic whereas human languages are fully and completely symbolic. We can restrict human language to function in a manner similar to that of a computer, but we cannot unrestrict a computer to function in a manner like that of human language.

Both computer and human languages have unlimited or infinite productivity of meaning in terms of alternative string constructions. These are made explicitly possibly by the application of a finite set of rules that govern the order of occurrence of words in strings. Application of a finite number of rules allows us to represent an infinite number of possible alternative strings--this becomes a manner of explicitly representing in a finite manner the infinite productivity of a language system at the phrasal or sentential and trans-sentential level.

The rules that govern syntactic order in any language are decidability rules, that allow us to decide, in a sequence of word-symbols, what the next best word-symbol will be to complete the string.

Theory of Automata

Theory of Formal Languages

Turing Machine Theory

Systems of Alternative Intelligence

Alternative systems are in a sense a logical outcome of the development of human cultural systems. They are an extension of the basic constructive capacity of humankind. Humans have, by virtue of the application of their natural symbolic intelligence, become capable of creating entirely new systems that were previously unprecedented in the natural scheme of things. Generally we refer to these as cultural artefacts, and we have many examples from the earliest periods of Hominid evolution--stone tools, hearths, clothing. The rise of technological and industrial civilization has born witness to the proliferation of artificial systems that exist in reality because they have been made by people, and follow logical and scientific principles, but which have no known natural antecedents or correlates. The rise of alternative, humanly constructed systems has achieved a basic augmentation of reality, and of our knowledge systems that are conjoined to reality, in many different ways. The hydrogen or atomic bombs were but science fantasies until their first demonstrations on nuclear testing grounds. Similarly, human flight was believed impossible until the advent of the Wright Brother's new engine-powered plane. As a result of these basic advances, our world has been transformed rapidly and irretrievably in many ways.

There is the sense emergent from the history of the development of alternative systems that such development eventually follows a certain course of inevitability--human intelligence is curious, and it is questioning and exploring in basic ways. Eventually, humankind would hit upon basic solutions to basic scientific problems, and these solutions would in time facilitate the integration achievable by such systems and lead in time to the acceleration of the pace of such technological development and integration as we have witness in the world. Such development would proceed much more rapidly if it were not for some sense of historical inertia from cultural and structural frameworks of human society that serve to interfere with and impede such development. There is a sense, in other words, that "if something is logically possible, then it is eventually inevitable."

Humans were meant to fly, not because they were graced with wings, but because their brains allowed them the imagination to conceive of such flight, and the determination to try to fly in whichever manner seemed possible. The possibility of alternative systems development therefore lies latent in the ground of natural systems patterning, to be discovered like a gem in the rough, by some errant explorer.

There should be therefore several expectable trendlines forthcoming in the progressive development of alternative systems:

1. As sophisticated solutions to complex problem sets, these alternative systems should become more streamlined such that they represent the best possible fit for the problem set they are designed for.

2. As mechanical systems, these alternative systems should be based upon the ability to create and manipulate increasing levels of power and energy, defined in the classical sense as the ability to do work.

3. Also as mechanical systems, these alternative systems should develop increasing levels of efficiency in terms of the ratio of work to total energy involved in the process. Not only will such systems be capable of working harder to do more work, but the work that such systems do will be both more efficient and more suitable as a solution to the problem posed for the work in the first place.

4. As systems involving energy dynamics and exchange, alternative systems are also information-based systems. In other words they rely upon the organization of information within and by such systems to achieve a degree of functional order. From a standpoint of general problem solving, such systems can be said to be "intelligent" in that they are capable of autonomously tackling problems of some level of sophistication.

5. As systems evolve and develop in relation to different problem sets at different levels, alternative systems should demonstrate two interrelated trendlines for development. First, they should become increasingly integrated between different levels and different areas. Secondly, they should become at the same time increasingly generalized and/or specialized in their relation to particular or generalized problem sets. In other words, in relation to this final point, it can be said that systems do not evolve in a theoretical or practical vacuum of application, but in the context of the development of many other alternative systems at the same time. Thus, the term "alternative systems" has implicit to its name the general context in which such systems arise and co-develop in the first place. It is evident that as alternative systems develop, previous systems will give rise to newer and newer systems, and this process should in time increase in its rate and amount of production of new alternative systems.

The increasing degree of integration of such systems determines in part that, as they develop, it will become increasingly difficult to differentiate and clearly separate where one kind of system leaves off and another kind takes over. In general, it can be expected that alternative systems should grow in complexity, sophistication and power to greater and greater levels that are integrated. It is expected therefore that such systems should exhibit a form of stratification of function at different levels and in different ways.

One constraint that appears to be operative of all such alternative systems is the notion that such systems are ultimately designed for and based upon human systems, and that they serve in one way or another the basic problems of adaptation and successful integration of the latter systems. The notion that alternative systems could arise that are essentially self-serving in function and independent of the human designers or designs, is one that is not uncommon in science fiction. Alternative systems, as integrated supersystems, have not yet reached such a stage of independent functioning in the world. Furthermore, it is a clear case of interdependency between human and alternative systems, such that alternative systems depend upon the human designers and operators for their functioning as much as humans have come to depend upon such systems for their functioning and operation in the world. I do believe that in time, alternative systems will become increasingly autonomous in function from the design influence or control of human manipulators of such systems. Thus I would append a sixth point to the five above, and claim the following kind of trend:

6. Alternative systems should develop in a manner that is increasingly but relatively autonomous at more and more points of human control functions, such that human design and control in such systems will play a diminishing part in the overall functioning of such systems, restricted perhaps to their initiation, original construction, monitoring and possible repair. Even these functions may eventually be taken over increasingly by the design of alternative systems.

Ultimately, at some point, one would expect that such systems would incorporate a degree of self-design and self-determination of patterning that is more or less completely separate from the role of the human operators or designers in such systems. It is doubtful that such systems would achieve a degree of automation of pattern that is completely independent of human design or control, but it can be imagined that such human influence will diminish as systems increase in intelligence and sophistication. It is doubtful furthermore that humans want or require such systems to be completely autonomous, as their essential purpose is to extend and elaborate the function of human systems in the first place.

Furthermore, all alternative systems are by definition human designed systems, and thus they are systems that have inherent to their existence the functioning of human systems and human designs. Such systems will achieve relative autonomy only to the extent that humans are capable of designing a sense of autonomy into such systems in the first place. All such designs are by definition explicit systems defined by relational rules and object-values--in other words, it would be difficult to confer upon any alternative system qualities or properties of behavior that are difficult in themselves to define or describe in any systematic manner. By definition, with the state of hardwired intelligence as exists today, all such relationships would be, furthermore, strictly logical in pattern, though it is clear that natural intelligence does not always function in a purely or strictly logical manner. A consequence of this is the constraint of the anthropological relativity of such systems. Such systems would have no intrinsic duality of patterning in its signification structures--in other words they lack any implicit level of intuitive meaning that comes attached to the signal. It is not to say, like the classic ELIZA example, that humans cannot be tricked into believing that computers have such an implicit level of meaning, and this kind of parlour trick constitutes in essence the basis for hard A.I. criteria for intelligence in the first place. It must be reiterated, that such implicit, intuitive meaning is totally lacking and impossible in artificial intelligence, and it represents an insuperable constraint of such systems to function as if they were natural.

Part of the problem of the definition of alternative systems is defining the term intelligence in an operational sense that gives it significance to describing and explaining the role and function of information processing that sophisticated and semi-autonomous machines in the future will perform. If we define as "intelligent" the ability to solve complex problems, then the issue is more straightforward than if we attribute to the notion of intelligence some sense of autonomous sentience and self-awareness as a being in the world. Surely all forms of natural intelligence can be attributed such a relative kind of autonomous self-sentience that is unique to such systems. This is a principle attribute that can be used to separate "natural" from artificial forms of intelligence. However sophisticated seeming an artificial intelligence system may be, we do not realistically attribute true or real sense of autonomy or self-functioning sentience to such systems. They are nothing more than sophisticated appliances that function because they have been plugged in and turned on. Their intelligence is inherent to their design and functioning, and this has been purposively created by its human engineers.

This question brings to bear what criteria we adopt for measuring or determining what constitutes intelligent systems--in general we adopt a "hard A.I." Von Neuman criteria of intelligence as resembing in everyway possible that functioning and response patterning of a human being may be seen as unrealistic in consideration of more limited models or criteria of artificial intelligence that seeks to solve specific kinds or sets of complex problems efficiently and reliably. In otherwords, a "soft" or functional A.I. approach is not only more sufficient, but therefore preferable to the challenge of development of forms of alternative intelligence in systems that can function autonomously or in an integrative manner.

The adoption of criteria of specific or general problem solutions, by design and control functioning, of complex problem sets or classes of problems, provides us with a more realistic approach to the challenge of the problem of artificial intelligence than we have with a "Chinese room" set of criteria. Alternative intelligence does not and cannot resemble forms of natural intelligence except in a superficial analogical manner--it lacks the direct organic sense of its history and evolutionary contexts of development. The manner that it achieves solution to problems, however defined, is fundamentally different from the way that natural brains solve problems in the real world.

Definitions of artificial intelligence that are part of the design of alternative systems are bound centrally by the problem of the anthropological relativity of all such designs--we cannot conceive of a form of intelligence which, by de facto conception, becomes bound by own own sense of human intelligence. The only time or possible scenario in which this will no longer hold true will be if and when we encounter forms of alien intelligence whose intellectual capabilities evolved differently and independently from our own. Alien intelligence would give rise to alternative systems that may in many ways resemble or be convergent with our own in the solution of common kinds of mechanical problems. At the same time, it is likely that such alternative systems may be completely different from our own in some essential ways, hence almost undecipherable as such in terms of our own design configurations and understanding.

The problem of anthropological relativity in relation to our conceptioning of systems of alternative intelligence is best exemplified, I believe, in our tendency to stereotype artificial intelligence systems with anthropomorphic forms and functions. Even hard AI criteria is based upon the principle of such an anthropomorphic stereotype, even if the best that can be achieved is mere an illusion, a parlour trick, a mimicry of real human intelligence. The greatest danger in the anthropomorphization of the function and role of artificial intelligence is, I believe, in unduly constraining our solutions to A. I. type problems, and even in our definition of such problems in the first place, in ways restricted to our preconceived notions about what human intelligence is supposed to be like. If Alternative intelligence in its design can be loosened from the anthropomorphization of such systems, then it becomes apparent that such systems can be applied in a very wide variety of ways in a variety of settings that do not require a human-type answer. We do not have to conceive of mobility of robotic systems as bipedal or even as with legs, whereas some kind of tracking system might represent a much more efficient and manageable solution to such problems of independent mobility.

The conception of alternative systems I believe comes to focus upon the design of hybrid analog-digital types of computer systems that are capable of serving a wide-range of functions. Such systems should be distributed in function, interlinked in a network of multiple processing systems. We can see such systems also being distributed in terms of the numbers and kinds of inputs available to such a system. Such systems need to be capable of reading directly from fluctuations of natural signals in the environment and monitoring such systems without human mediation involved in the monitoring.

At the same time, such systems can also be distributed at the output level, and this output can consist of a variety of functions that articulate and manipulate environmental elements in basic ways.

We can also refer to distributed control systems that allow feedback to occur between different functional levels or components of different levels, as well as distributed input-output interfaces that control and manage human signal inputs and outputs in a manner that allows human manipulation and redesign of the system on one hand, and that can generate meaningful results that are digestible to humans, on the other. The internet offers the possibility of the construction of such a supersystem, and such systems have already been engineered, albeit in yet rudimentary form. In the construction of such a system, input and output components can be proximate or remote from one another. The entire system could potentially encompass the globe, or even be extended beyond the boundaries of the earth to compose a solar sateillite system.

It is evident that we can refer to a totally distributed system in which all these functions are spread out over a series of interconnected devices each capable of working both independently and in a manner that is coordinated with all the other systems. In such a system, there would be no central hub or main control component upon which all the other systems would depend. New systems can be added or modified at any point within the overall system. Furthermore, each unit or component of such a system could itself be functionally integrated to perform simultaneously any number of different functions in the overall system at the same time. This design of alternative intelligence systems must occur independently of the component design of the system, or of the nature of the design of the individual component. Hybrid systems span more than merely the digital and simple analog wave functions that are built in conventional types of systems, and can embrace a number of more exotic architectures as with quantum type computing or light computing.

Totally interconnected and distributed systems puts a premium upon the communicative function between components, as well as upon the distributed control systems that accompany each of the components and serve to integrate this component to the larger overall system. Such systems must be capable of finding the other components of the network, talking with these components in a meaningful manner, and then determining between themselves the kinds of decisions that have to be made that affects the outcomes and state-pattern of the system as a whole. It seems that to a great extent the invention of the internet has obviated this communication and control aspect of such distributed systems, though the state of development of the internet may just be reaching a level of sophistication to allow these kinds of advanced distributed systems to become a real possibility.

Such totally distributed systems can also be stratified upon different levels of function, and this stratification can be made into a form of self-organizing behavior of the system that is based upon its modular partitioning of functions to solve problem sets at different levels. Another aspect of this kind of system is provisioning the entire system with built-in memory functions that allow it to store different forms and levels of information and to be able to utilize this information in the solution of future problems. Memory storage would have to be itself stratified and organized functionally, as human memory is, and it would have to be construed as part of an active control system that allowed its continuous updating and revisioning.

Such systems would have to be working systems if they are to achieve their place in the world as alternative systems. They must perform useful functions, and in this the actual production of work is no less a part of such distributed systems than is the continuous monitoring of the environment. It is not difficult to imagine a number of different functions for such systems, in terms of education, entertainment, production work, environmental regulation, energy production and distribution, communication and even transportation systems.

Another aspect of such systems, I believe is that they achieve a degree of symbolic realism in their patterning of behavior that is not unlike that of human symbolic cognition. This is not to reimpose an anthropomorphic model upon such systems, or to adopt a version of the hard A.I. criteria--rather I believe that especially in the interface requirements in the articulation of alternative systems with their human agencies, such symbolic patterning could prove to be quite useful and even powerful for augmenting the intelligence of the distributed system in ways involving pattern recognition and identification in symbolic terms. Symbolic criteria would allow us, at least in principle, to overcome partially the inherent constraint of artificial systems in lacking duality of meaning and an implicit level of signification. One consequence of this is in terms of genuine pattern-response recognition of natural forms in the environment, without this recognition being moderated or influenced by human mediators.

The consideration of forms of alternative systems in relation to artificial intelligence leads to the speculation about some first principle of design that serves as an anchor point or key for the articulation of such systems in an effective manner. It may be that there are several interacting anchor points that are necessary for such a sophisticated system, though it is as yet not exactly obvious what such anchor points for such systems would be.

It is evident that such systems can be abstractly and mathematically defined with a great deal of precision and logical coherence attached to them. At the same time, they can be constructed in a manner as to yield fairly reliable and consistent results within limited empirical contexts of their articulation and interaction with the environment. One characteristic of such programming is its essentially linear character. Such programs function by processing single or multiple streams of information very rapidly, setting off in the process numerous secondary reactions and responses that may trigger even further patterns of processing. They apparently do not solve problems in the manner that human beings normally do, often basing final decisions on blind faith or at least strong intuition rather than on any strictly logical calculation. It is not clear either that multi-stream, parallel processing structures overcome this basic linearity of such systems.

For the most part, the rise of alternative systems has been achieved in a piece-meal and uncoordinated, almost serendipidous manner. With the rise of the sciences, especially within the last two centuries, this process has become increasingly organized and deliberative, to the extent now that there are entire research organizations dedicated upon one level or another it the articulation and elaboration of alternative systems in reality.

Alternative systems offer to humankind one set of unique possibilities that are critical to the future of such systems and to human and biological systems on earth--alternative systems offer humankind the possibility of being able to transcend the natural constraints imposed by basic bio-geophysical systems, and to escape from the rule of the biological imperative that imposes the ultimately outcomes for all life on earth. Even now as we are learning through our alternative systems applications to manipulate the genome and reengineer genetic structures to our own designs, we are gaining rapidly a degree of leverage and control over biological processes and outcomes that were until now unprecedented. And yet we are failing in basic ways to achieve the full measure of control over our own human systems that is a necessary prerequisite to such successful application of alternative systems. As a result, much of our applied systems ends up being, in the structure of the long run, destructive and entropic in a manner that puts us closer to, rather than further from, our own basic biological imperatives.

All alternative systems that exist are essentially the product of human invention and construction. As such, it can be said that they are a product of human intelligence and its application to the manipulation of the environment. For the most part the motivations to do so stem, however indirectly, from the adaptive advantages such manipulations confer upon human beings. Without our intelligence, we could not have achieved such systems. In a very real sense, therefore, alternative systems are fundamentally cybernetic systems, if we accept the conventional definition of cybernetics as being the modeling of informational systems based upon natural forms of human intelligence. Even rather crude types of systems, for example the machine screw and the mouse-trap, are basic demonstrations of certain principles and patterns of intelligence. This is not just in their design and manufacture, but in their functional application.

It follows from this observation that the trendline of alternative systems follows a gradient of increasing informational integration and capacity, to the extent that such machines will become increasingly intelligent in a cybernetic sense. I will not claim that machine intelligence is an inevitable outcome of the progressive development of alternative systems, but it seems to me that we would eventually hit on the right combination of things that would confer some nominal sense of intelligent functioning to a machine. With the rise of the information revolution, we are arriving at an age in which machines are becoming not just demonstrations of intelligent design and function, but coming to mimic and incorporate the principles and functions of intelligence into their very design and function.

In this regard, I believe, we must be careful to say that machine intelligence, or what I will refer to as alternative intelligence, is not the exact equivalent of native human intelligence, nor in this sense do I think it can ever become what can be called "genuine" or true intelligence, which the implications of self-awareness and sentience that we attribute as qualities intrinsic to true intelligence. What we have are extremely sophisticated machines, but machines which are non-biological and which do not heed any form of biological imperative. They are things that we turn off or on at our own choosing, and use as tools to our own ends. In this regard, the standard for artificial intelligence, which is "human-like" intelligence, and which is the goal of hard AI research, seems to me to be unrealistic and unlikely. This is not to say that we will not eventually build machines that have a sense of independent sentience in some manner, but when such a machine emerges from the laboratory and factory, it will not in any form but the most analogous resemble natural human intelligence.

In this regard also I must separate the question of artificial intelligence, as this is a focus of much cognitive science research applied in various ways, from the notion of a more general form of alternative intelligence, in which "human-like" artificial intelligence is just one possible variety. When we can accept an expanded definition of alternative intelligence, as being somehow a logical consequence of the development of alternative systems, then we can begin entertaining broader notions of what such systems might be like and the functions they might serve.

In this, there are two interesting and indirectly related issues. First, there is no reason that alternative intelligent systems need to be anthropomorphized at all, or in the robotic manner that they tend to be stereotyped and which appears to influence much research and development work in artificial intelligence that sets humanlike intelligence as its standard and goal. We can selective anthropomorphize functional aspects of such systems--for example hands, or eyes, or speech, but that does not mean that we have to also give such machines human form or functional anatomy. There is no reason a robot needs to be bipedal to be intelligent, if it is perhaps preferable to build a machine that can walk like an insect, for example. This issue of the anthropomorphization of alternative intelligence has greater impact and influence on our models and agendas in this regard than we may realize or want to acknowledge, and I believe it tends to constrain our imagination as to what might in fact be possible with intelligent machining.

The other side of the coin of this issue is the observation that I've had since being involved in artificial intelligence research, and this is that the interdisciplinary aspects of cognitive science work is primarily psychological, computer and philosophical, and it tends to leave out entirely as an objective problem the issue of the social organization and articulation of knowledge, as a shared and intrinsically social phenomena. This derives in part I believe from a theory of language that is primarily pscholinguistic and psycho-genic in orientation, to the exclusion of the communicative aspects of language as a social phenomena. If we refer to intelligent machining as a form of informational system, to which we can apply information theory, then we must also recognize that as informational systems, such forms of intelligence are also ultimately systems of communication to which we can apply communication theory and design principles as well. And this takes us back once again to fundamental arguments concerning natural human intelligence as well--as symbolic manipulating systems, the human brain is context bound to cultural frameworks of articulation within which it achieves integration.

We have become culturally dependent creatures in ways we scarcely realize and refuse to admit. Human intelligence was a functional product of the elaboration of human language as a linguistic system. Another way of construing this issue is to recognize that our intelligence is tied symbolically to larger knowledge systems, mostly that exist in the form of recordings as part of a common stock of knowledge. Knowledge is distributed and organized within a social landscape, and finds its articulation within this landscape. It follows therefore that intelligence, being based upon human knowledge systems, is fundamentally dependent upon these knowledge systems in a manner that is fundamentally external to the function of intelligence itself. Intelligence bereft of its knowledge context is like an empty machine with no signals to record. It remains merely a sophisticated device without significant content. It follows also that a great deal of knowledge and information processing that passes for intelligence comes prepackaged and preintegrated as such--it comes as something preprocessed in an intelligent manner, compatible to its digestion by an intelligent system. These are inseparable parts of one and the same problem.

Another way of saying that is that, if we over-anthropomorphize our conception of artificial intelligence in terms of its possible designs and applications, then we also simultaneously fail to anthropologize more realistically or systematically our understanding of what both natural and artificial intelligence really represent, as an extension of our own knowledge and communication systems.

At the same time, we must also come to a realization that in nature there are other alternative forms of natural intelligence that deserve analysis and modeling in terms of intelligent systems, and that probably, in the grandest scheme of things, there are in the universe other forms of natural intelligence that are essentially non-human and that would constitute the basis for the development and introduction of entirely new forms of alternative systems than those that are humanly constructed and conceived. We might say that any true intelligence must, to so qualify, at least have the symbolic structure and function that we so readily recognize in human intelligence, but we cannot necessarily say what kind of symbolization this might entail. If we watch a dog jerking its legs in a puppy dream, we realize that dreaming is a fundamental aspect of the rise of a complex brain that can be said to have some level of sentient intelligence. Would we say, therefore, for instance, that some form of alien intelligence would necessarily have to dream in the same way that many mammals appear to do?

My central concern therefore in the elaboration of alternative systems theory is the elaboration of a more general and comprehensive model of what can be called alternative intelligence, and then of possible applied forms of this model of intelligence in ways that have some degree of efficacy as technological systems. Primarily I am concerned with hybrid forms of computer designs that combine digital with analog varieties of processing. I have been particularly interested in the use of light modulation and recording/recognition as the basis for the storage, manipulation and retrieval of patterned information. This is furthermore augmented by various network designs of distributed multi-processing systems and also of systems of intelligent environmental monitoring and articulatory response and manipulation. The model I have come up with resembles something more like an octopus than a human, though I do not thing that such alternative systems need to be constrained by an predetermined stereotyical form.

In this regard, I make a critical distinction in alternative intelligence between what I refer to as "abstract" intelligence, and what I would refer to as "natural" forms of intelligence. The latter represents those expressions of intelligent functioning that arise naturally as a result primarily of higher order brain functioning in biological organisms. A claim could be made that genetic systems constitute a form of intelligence, but I would claim it constitutes an intelligent system of informational patterning and organization, rather than a system of intelligence. This notion of natural intelligence obviously embraces the range of mental facultation of most primates, dolphins and many other large brained species. Complex problem solving has been demonstrated, for instance, in octopus, and even dogs show at times remarkable feats of memory, humanlike emotion and intuition, and even a rudimentary form of symbolic behavior and problem solving.

I deliberately contrast the notion of natural intelligence from the conception of abstract intelligence, per se, to highlight what I consider to be a critical difference and impasse between natural and artificial forms of intelligence as these latter have been invented by humankind. Natural intelligence arises biologically, based in the brain of the organism, and its rules and order of patterning remain mostly implicit to its functioning and behavioral consequences in terms of adaptation, communication and complex social relationships. Abstract intelligence arises primarily as a mathematical possibility, as a kind of formal-functional capacity of systems design that can be logically and parametrically ordered. All intelligent machines so far produced by human beings have exhibited mathematical abstract intelligence--not one of these has demonstrated the qualities of even the most primitive forms of natural intelligence. In general, rules governing the ordering and operation of abstract systems of intelligence are either formally or functionally defined in explicit ways. Its functioning does not arise as an emergent property of the integration of a complex organ, rather its sense of intelligence arises as the result of the top-down planning of complex systems based upon the reiteration or recursion of explicit and logically correct functions.

To push the distinction one step further, I would claim that the results and outcomes of a system of abstract intelligence are fundamentally different than those for a natural system of intelligence. Abstract intelligence achieves solutions to complex puzzle-problems that have a form of correct solution possible. Such problems may be astronomically complex, and there may be more than one solution pathway through the search solution space, but in general a solution to such a problem solves what is known as the Van Neuman bottleneck of an explosion of possible wrong solutions that developes when the problem remains unsolved. By contrast, I would claim that natural intelligence in general solves "problems" of an entirely different kind or quality. Generally, the problems solved by natural intelligence are those kinds off dilemmas to which there is no necessary single or even multiple correct solution. In other words, natural intelligence normally makes judgments based upon uncertain variables and unknown values, and the solutions are weighed in terms of success or failure of their consequences in some functional framework in the life-world of the organism doing the problem solving. So far, not machine has accomplished this kind of problem solving except in a very deceptively rudimentary manner--we can develop discrimination tables with predefined probabilities of outcomes and alternative uncertainty factors, but this kind of systematic approach to complex problem solving does not embrace the degree of intuition and experiential rationalization that is normally, automatically employed by natural creatures that are challenged in complex situations. It should be taken as a strength of natural human intelligence that though machines cannot solve even very simple natural problems, the human brain can rapidly and readily solve both mathematical, puzzle type problems as well as those problems requiring experience, intuition and application of reason and evaluation.

Related to the issue of abstract versus natural problem solving is the notion of learning and the quality of what is learned as a form of experience that can be applied to future problems. Again, machine learning has been advanced, but the parameters of such learning is almost always predefined such that learning cannot proceed beyond the purview of preprogrammed parameters. With natural intelligence, learning is almost innate and automatic to the context of application. The more intelligent the organism, the more it learns from the problems it encounters and solves. Another way of seeing this is that abstract intelligence is by definition programmable, or rather, preprogrammable and thus it is constrained by the parametric templates of its program. Naturally intelligent systems, beyond the long period of cultural-environmental acquisition and behavioral reinforcement that occurs on a continuous basis, are essentially non-programmable, or what we might say, it is innately or self-programmable.

Another means of describing these differences is to see the difference in functional outcomes of the two types of systems. Natural intelligence is extremely adaptable to functional variables in motor coordination--it takes a great deal of work to preprogram a machine to perform even simple tasks efficiently, much less complex series of tasks that natural intelligence seems capable of. On the other hand, abstract intelligence systems can perform complex linear calculations in a fraction of the time it requires naturally intelligent systems to perform similar kinds of abstract functions. We may say that natural systems are concrete and non-abstract, though they are capable of a limited degree of abstraction, as well as abstract intuition. Abstract systems tend to be non-concrete and incapable of extensive variable application, but capable of a rate and degree of abstract functioning that is preprogrammed.

I would venture the distinction between the two forms of intelligence as saying that abstract intelligent systems are primarily linear in function, whereas natural intelligent systems tend to be non-linear in function. Abstract systems are furthermore over-determined or at least fully determined systems, whereas natural systems tend to be critically underdetermined, hence normally chaotic systems.

I would be inclined to say that natural intelligence resembles more analog designs of intelligence, versus the digital form of intelligence that has arise with the machine information revolution. Digital intelligence seems to me to be fundamentally a form of abstract, mathematical intelligence, and hence suffers the limitations of design and possibility that are inherent to this form of information processing.

Automated systems can be said to be a general class of artificial or human made system, or any other alternative system that functions according to certain laws and theory of automata, which can be most generally defined as an device that is self regulated and relatively independent in its control functions. In a more theoretical vain, we refer to any device into which input may be affected internally by certain state transitions, leading to some form of predictable or logical output. By inference, at the center of automatons, especially very advanced systems, are computer-based systems. A completely independent automata would be akin to a very sophisticated, anthropomorphic robot that was capable of functioning in every manner as a normal human being in the world. This analogue of automata is a very anthropocentric one. It is in my opinion that this anthropocentrism of our conception of automata is more of a hindrance than a critical insight into their potential development. Vast, super-sophisticated systems that are completely distributed upon a number of levels, and achieve maximum integration, appear to me to be more interesting kinds of models to pursue.

At some point, an integrated system can be said to be one in which all automatic functions occur within the same machine framework, or device modulator. This may in fact be a relative issue of a device is nothing more than a plastic case disguising a great deal of interconnected wiring and multiple modular components of such systems. Miniaturization of circuitry has led the revolution of digital information processing to the achievement of vastly superior integration of machine automata. But this challenge is met with a corresponding problem of the transference of information from one module or system to another, along with the challenge of creating a universal interface by which all systems can achieve effective talk. With distributed systems, the greater the distribution, the greater the function of communication and problem of transmission of information becomes a problem over the question of the integration of function of information processing on a reducible scale.

There occurs invariably a kind of trade-off between systems between the integrative miniaturization and distributive communication between systems. We cannot have a perfectly integrated system without some measure of distribution, and we cannot have a fully distributed system that achieves complete communicative efficacy. In other words, distribution must be gained at the expense of possible integration into a single system, and integration into a single system must come at the cost of distribution of such a system to a larger range of possibilities. A system can be powerful and highly integrated, but its potential is severely circumscribed if it functions by itself and is not able to share its information within a larger network.

I would add to the challenge of developing sophisticated automata a few other related challenges:

1. The challenge of a machine system being capable of producing or acquiring its own energy independently of any other system.

2. The challenge of a machine system being capable of performing other kinds of work or productive/purposive activity beyond its own informational processing, implying manual articulation of such a system in the environment and regulation of function in the environment.

3. The challenge of a machine system being capable of both reproducing and repairing itself, both as hardware and as software.

4. The challenge of a machine system to learn and develop as a system, and to integrate changes into itself.

5. The challenge of a machine system to evolve as an integrated population of modular components, such that we can speak potentially of multiple successive generations of such systems.

What is described in these goals is a completely automated metasystem that would require as few human influences or inputs, beyond the original design and construction, as would be necessary for its overall function. Such automation would be achieved gradually by small steps, one step at a time, rather than all at once. It is the case that most metasystems are not achieved overnight, but are complex solutions arrived at over many cycles of trial and error.

A number of other ancillary functions can be associated with these as well. We can include for instance the capacity of different modular components of distributed systems to share and interact and coordinate functions to problem solving in an independent manner, and the ability of such modular components to effectively recognize themselves and one another in a way that allows them to make some kinds of decisions regarding interaction.

Implied in these kinds of challenges would be a certain openness of design of such systems, as well as a certain basic plasticity or flexibility of structure that would enable such designs to be modified and adapted to the widest range of functions and circumstances that are possible for these systems to achieve their fullest distribution and coordination. A system therefore cannot be overdetermined or even fully determined, but must be complexly chaotic in structure and therefore remain only partially determined and relationally connected to the world in which it is situated.

The suggestion of a fully connected and distributed system implied above is that of an informationally based automated infrastructure that coordinates and produces energy for work of various forms, and even provides the working robotics that would be the basis of such production. In other words, I am speaking of a form of systems integration that combines the different aspects of basic social infrastructure into a single coordinated system of automation. In this, would could not clearly draw the line where one form of processing or integration leaves off and another begins. Communication would be a form of processing, and processing a form of communication that is extended and elaborated.

Integration is implicit to the concept of sophisticated automation. The more integrated a system becomes across multiple alternative functions, the greater can be said to be its degree of automation. Automation itself though remains a relative measure, short of an absolute standard. We must even inquire, for instance, how fully autonomous we are ourselves as human creatures, or if possibly some of our sense of autonomy might but be an illusion of our own false hubris disguising the fact that we are perhaps like marionettes to certain complex forces beyond our control or ability to manipulate. We are not always in as much self-control as we might believe or want to believe ourselves to be.

However automated a machine-based system might become, it must be kept in mind that this automation remains essentially mechanistic and in a real sense blind and "dumb." In other words, automation of function of operation of such machines is not the same thing as the simulation or especially the realization of "true intelligence" as this is at least anthropomorphically defined. I doubt whether machines can ever become capable of true self-awareness as if an independent living being, no matter how sophisticated or "intelligent" their design may become.

In a sense, such an emergent supersystem would be a logical outcome of the development of natural metasystems, and it would represent a new level of integration of such systems that can be characterized by its own uniquely emergent properties. These emergent properties would entail perhaps macro-level social phenomena and reorganization of human systems in a manner unprecedented in traditional human social organization. It would not be that humans become puppets of these new automated systems. Rather this becomes increasingly their function and their work, to service and maintain this metasystem, as it becomes in a way a symbolic and metabiotic aspect of human social organization, muct has cells of a multi-cellular organism become functionally differentiated and contextualized within various tissue matrices of the organism. Instead of like single cell organisms that we are now like, we become organized socially and functionally into a multi-cellular kind of body. The kinds of state systems we have now are in a sense precursors of this, but they are more like guild-colonies formed by unicellular organisms than they are like a truly integrated organic metasystem.

I believe this stage will be reached, if it is reached at all, when truly autonomous and distributed systems are developed on a global basis, and as a result human social systems become so functionally and structurally and symbolically interdependent that the traditional or conventional boundaries separating people will tend to break down and give way to a complete new system of social organization. I believe it will entail the gradual development of a "metacultural" orientation as well, one that is focused around the manipulation of complex information. This metacultural orientation will not displace traditional cultures all at once, but merely come to share space and overlay these other orientations to the point that there is a global streamlining and a disappearance of cleavages and vast differences between different peoples. Language boundaries may in time come to make little difference if a common web system exists that can automatically translate inputs from any known language to reasonably reliable outputs in any other language. Then people can achieve communication in a virtually instantaneous manner on a global scale.

The alternative model is a superorganic form of artificial life, that, though it might be artificial, represents a form of living system that meets certain minimal criteria of such systems. It will serve its own purposes and will have its own goals in its state-path trajectory. It may be the case that human beings come to need such a system, if they are the completely escape the contradictions of their own predicament on earth, whatever tyrannical implications it may seem to have. We should not see such a possible system as a single conscious entity, a kind of Hal from Space Odyssey. It is something more than this--it would be a billion conscious entities both focused at one point in place and time, and simultaneously spread out around the entire globe. We would not fear its over-control, ans there would be no centralized control function. Control, inherent to its design, would be as distributed as the information upon which it is based.

I see this relationship as being ultimately totally symbiotic and mutualistic between human beings and their metamachine. In other words, the metamachine would serve humans as much as it is served by humans, and would provide virtually every aspect of human needs and interests that can be practically defined. At the same time, its own functions would be non-human in nature, except to the extent that these were defined by humans originally to serve human interests in some exclusive manner.

I believe that because we can never classify such systems as truly intelligent, however integrated they may be, we do not ultimately need to fear losing ultimate control over their function and destiny. At some point, human beings would retain the ability to switch such a metamachine on and of at its own will, or at least we might be well advised in the future to retain this kind of final control. This consideration leads to the following proposition:

Sense of machine autonomy can never be extended beyond the bounds of human control factors and human-based design, therefore autonomous systems can never achieve what can be called as ultimately arbitrary autonomy. They remain autonomous in two senses therefore, because they are relative independent of human influence, and because they are fully "automatic" as a mechanistic system.

The problem of automata and the problem of automation are not the same, though they are generally confused on an implicit level and often the two terms are used interchangeably. In a sense, automata refer to the theory of automatons, or theoretical machines that are capable of producing a certain kind of modulated output, arbitrarily determined, depending or independent of the nature of the input into the system. Usually the output is determined by a set of rules, or a "discrimination" structure that somehow logically determines the final choices among a range of alternates.

Automation refers to a machine that is capable of independently performing complex functions, and demonstrating some measure of self control, without direct manipulation or involvement by a human agent or human mediation. In general, automation tends to simplify and remove human involvement from the direct action or consequences of the behavior of the system or machine. With increasing automation, we expect that human involvement in a system becomes less and less directive, and possibly minimized to a point of merely turning such a system on or off or providing basic starting/stopping operations, etc. Of course, an automated system is not merely a "remote-controlled system" though in general it entails some degree of remote control. An engineer on earth piloting a vehicle on Mars surface is primarily an example of a process of extremely "remote" control. Systems automation entails that the system becomes increasingly self-controlling, and the control function of the human input is simplified to the minimal number of choices necessary to make the system fully operational.

The purpose of automation is not to displace workers from the labor force, but to free human activity and interest from the drudgery of performing routine-operational tasks, in order that they may spend their time doing more meaningful work and cultivating more productive life-styles.

Automation has in general proceeded slowly and incrementally one small bit and piece at a time--it is usually articulated in very narrow and tightly delimited contexts. Machines are best at performing one set of functions, over and over again, at very rapid rates. They are not at their best when they are multi-function by design. Conventional CPU's are single processing devices that perform one stream of continuous information processing in a linear but very rapid manner. Creating multi-or parallel processing devices, like the connection machine, associated primarily with supercomputing and the handling of very large informational demands at very high rates of speed, are a kind of solution to this problem, but perhaps we can consider as well non-linear processing functions that work in fundamentally different ways than straight forward parsing of strings of information. Analog and hybrid computing models have been developed that to some extent address these issues, but the problem still remains open, especially if we consider it in light of the challenge of creating human-like intelligence, or machines that can be automated across a fairly broad range of different tasks.

Of course, the two concepts, automata and automation, converge when we think of an automated system that is intelligent, like a fully functioning robot that is self-directive and independent in its behavior. Given time, we may perhaps achieve such models in a manner that would measure up to what we would expect of our current stereotypes of them. The general trend of the future is of course one toward greater convergence, but this convergence is very broad based and comprehensive in form, and proceeds in a relatively self-organized and piece-meal way. It will not come overnight, or even in a year. It will be measured by relative degrees of integrated distribution and achieved progress compared to measures of the past.

It would be expected to see increasingly intelligent automation in areas that involve the greatest and most intensive inputs of human labor, and this is in areas of greater and more reliable complex pattern recognition, manual manipulation of tools to make fine products of varying design, and in possibly genuine "auto-mobiles" that are self-guided and self-steering. Achievement of semi-intelligent automation in any one of these areas, much less all of them, would represent significant advancements in the development of humankind.

Cybernetic Control

The concept of cybernetic systems only arose with the advent of electronic computing systems and the realization of the homology between the human nervous system, and its intelligent functioning, and that of computing systems and their potential for intelligent functioning.

The entire foundation of cognitive science research, especially of "hard" AI, has been the achievement of automated systems that are human-like in their intelligence capabilities. The degree to which systems integration can be achieved by means of digital and electronic information processing remains as yet unknown and unfulfilled, but it is now expected to be greater than anyone now might imagine.

Providing feedback, in the context to a Automaton that is capable of modifying input and adjusting output on a continuous and dynamic basis, forms the basis for all systems and the degree to which this is achieved in systems should be taken as a measure of their relative complexity and sophistication.

What is lacking from a meta-systems standpoint is a clear-cut objective theory and methodology about cybernetic control systems in general, and applied systems in particular. I think such a scientific framework would go a considerable ways toward the development of a viable automated meta-system that permits the intelligent integration of systems upon multiple levels. I have not yet approached this problem set specifically, though I have sought to outline it, and the elaboration of such a problem set or its outline remains beyond the scope of this present article or this current newsletter, though it will be dealt with in greater detail in future publications in this section.

Such control mechanisms that occur in the world take many forms. As we explore different kinds of systems at many different levels of systems articulation, we find many interesting examples of feedback mechanisms that serve purposes of maintaining cybernetic control over systems. In highly elaborate and differentiated systems, cybernetic control becomes itself elaborated and specialized between different functions and sub-functions.

Many control mechanisms in the world are not advertized and are usually background. Such mechanisms tend to serve many different purposes in a complicated and increasingly diverse world. Some of the functions relevant to ourselves include:

Security; protection of people, resources, property and information within the framework, and relating to the framework, as well as extended meta-system protection encompassing security frameworks for larger world contexts. Security includes some of the following concerns: Encryption, Password Protection, Hidden systems, back-up systems, Protection from attacks involving Viruses, Spam, Hacking, various forms of Fraud, Vandalism, or Violence or Violation of Personal or Property Rights.
System-state monitoring; systems that are capable of continuously or periodically providing a system-state check at critical points, including: Traffic flow, Tracking, E-mail, Linking, Client feedback & satisfaction, emergent Web-development patterns, etc.
Infrastructure management and development; including resource availability, acquisition, organization and storage; communication services and activities
Network management and development; including hyper-linking, submission, contacts, mailing and emailing systems, marketing frameworks, advertising, information control, image management, etc.
Organizational management and development; organizational management involves planning, scheduling, prioritization, human resource development, partnering and the development of affiliate frameworks, etc.
Structural management and development; including production systems development, tool and technology frameworks, work-station and work center development, resource management frameworks
Environmental-state monitoring; external state monitoring systems include surveillance systems, alarm systems of various kinds, weather monitoring systems, temperature and climate control systems, habitation systems monitoring, etc.
Monitoring and Management of External State development: Management and development of feedback mechanisms mediating relations between internal and external systems and subsystems components

Universal Machine Language Protocol

A fundamental, single language for computers, that accomplished a universal interface and that permitted both evaluative and performative functions of machine processing, would go a long way to achieving, I believe, both communication and integration. We have a bewildering variety of language systems today that are like a tower of babel, or rather really an electronically multi-lingual system. No single language system seems suitable for all purposes. We have HTML markup code for internet communication and AI languages for building complex puzzle solving solutions.

It is important to inquire into the extensive and intensive limitations of any particular computer programming language to fulfill the objectives of a fully integrated system. Not only must such a system accomplish multiple processing and communication functions across different computer architectures, but it must be capable of interfacing with human language in the most facile and natural manner possible. The lack of a single integrated language system has been filled to some extent by major computer software companies who offer integrated programs at a higher level of code, hiding the source code of their programs from open access for proprietary purposes. Thus, certain large companies have come to monopolize this aspect of computing to an exclusive and wonderfully profitable advantage in the world.

Grid Structures and Automated Advanced Array Systems

What is intelligence, and what seems possibly intelligent about an integrated grid or matrix system?

Intelligence can be said to be a form of active awareness, awareness that sees beyond surface pattern to apprehend the underlying pattern of order through time and across space. Intelligence also implies the capacity to achieve appropriate response that meshes with larger intentional structures. It therefore implies a form of applied rationalism that includes the assessment and derivation of long-term goals themselves.

A complex grid structure, or matrix or array, can be said to constitute a field of interrelations that, in a large size and scale, is incredibly complex. Such a grid system permits the mapping of a number of variables within a larger distribution that can be considered to be total and all encompassing as a system. In constructing alternative automated systems, I have attempted to employ the grid structure as the underlying paradigm of these systems that is recurrent throughout their design and function. It is therefore a central question to try to answer as to what is important and significant about such grid systems that make their use in intelligent machine designs so appropriate. The answer to this question is not obvious.

Such systems have several attributes that are significant for the construction of intelligent systems and that relate direction to operational metasystems. These attributes are:

1. The capacity to systematically organize, compare and manipulate an almost infinite number of different attributes or very large sets.

2. The capacity to represent a total language system and grammar in terms of the total paradigm of strings generatable by the language.

3. The capacity to construct complex list processing structures that can be analyzed and parsed in an unlimited number of dimensions.

4. The capacity to represent the total search-solution space for a paradigm of possibilities represented by a particular problem set.

5. The capacity to interrelate different grid structures on the basis of their size, dimensions and cardinal properties associated with these systems.

In short, I believe arrays or grids have a very generalizable form (rows to columns) that allow us a relatively facile way of organizing and manipulating a very broad range of data and very complex and large data sets. They allow us a means of managing complex information in such a way as to be assured of representing knowledge and information of all kinds within a similiarly structured framework. Grids represent what can be considered to be a systematic means of measuring and representing data across the total range of their distribution.

On an even deeper level, I believe that such grids reflect something fundamental about human information processing that is tied to human language and symbolization. The string structure of human language and the symbolic frames of reference and substrate of meaning upon which cognition and language depend permit people to organize their experience in a certain manner, and this kind of organization of experience is reflected in the structure of grids.

Grids by themselves can be thought of as nothing but empty frameworks needing to be filled with useful information. To be meaningful, data sets must be ordered in some manner, and this sense of order may as often as not be an arbitrary sense of order we bring to such systems. Further, the more active and intelligent the individual frameworks can be made, the more dynamic and interesting the resulting grid table and dynamics will become. Cells of grid structures are not just slot-and filler to be filled with passive information--they are made dynamic as sub-functional or variable entities that interact with other cells and with the system as a whole. A distributed and interconnected computer network represents an array of very integrated and "intelligent" components. Each is capable of processing in an independent manner, but within an array framework each would yield some measure of this independence of function for a specialized role within the larger framework. Role specialization of such systems can be made flexible through modularized assignments of different tasks to different systems.

The value therefore of the grid seems to be the multilinear organization of data in series of parallel rows (or, alternatively, columns), and the built in addressing system that permits information anywhere within the grid to be rapidly located and written to or read. Blocking of information into cells in a grid matrix allows for a discrete organization of function and distribution that permits the complex superorganization between different grid frameworks.

Grid Structures as Dynamic Inference Engines

Digital-Analog-Quantum Hybrid Systems

Light computing offers an unrealized potential for the extension of conventional digital computing methods utilizing many of the same basic devices. Light amplifying diodes, holographic media, signal modulation of lasers, optical fiber computing and photo-electric devices offers the possibility of a wider range of application of these kinds of technology to computer and information processing systems than has yet been realized. I would add this this list certain kinds of sensitive filter systems and spectro-photometric systems that permitted a continous reading of light signals. Light shares with electricity certain intrinsic properties as a form of energy that makes its consideration in computing attractive.

Brains can be considered to be organic computing systems, pure and simple. Nature arrived at intelligent systems within its cellular framework in the same manner that it invented the first motors and engines and informational processing systems. Brains were thus a solution to the problems of coordination of bodily function or physiological process, mechanical coordination and senory-motor response to environmental stimuli, and the organization of behavior into complex patterns of reaction and relation between different organisms. Biological systems therefore have been based on a wide variety of what can be considered as alternative systems that had no precedence in the natural world. They were not "invented" in the manner of human alternative systems were achieved. Rather they were "evolved" as adaptive mechanisms that solved basic problems of functional adaptation. That these brains evolved through many millenia of trial and error and development goes without remark. There has been a sense of progressive development of brain function, as many dinosaurs were reputed to have had relatively small sized brains, thought this stereotype of slow-witted, slow moving monsters has changed recently.

The evolution of the primate and hominid brain was a remarkable achievement, but we can find other relatively big brained animals in nature. Brains are plastic tissue organs just like any other kind of organ in the body. Inherent variability of genotypical pattern and phenotypical expression of brains results in selective pressures favoring certain specific conformations of brain size, structure and function. Big brains evolved on environmental demand--pointing to the fundamental requirement of solving through coordination or complex response pattern some difficult problem or challenge to survival in the larger framework.

Input-Output Interfaces & Feedback Loops

The consideration of input-output feedback loops brings to focus the concern with human user interface designs, and this represents an aspect of autonomous systems that can be said to structurally constrain such autonomy. In short, we can say that there may be more than one kind of input, and not all inputs have to be humanly initiated or ordered ones--computers can be designed to directly monitor environments through a variety of sensory input systems. Similarly, we can say that the more different kinds of outputs a computer based system is capable of, and the more derivative these outputs from the central information processing function of the computer, the more useful and adaptable such a system would potentially be in the world. Not all outputs have to be either human-mediated or human-intended kinds of outputs. They can include outputs to other machine systems or other computer based systems that are capable of functional response or productive work.

At the same time, the concept of the interface as being necessarily that between the human and the machine, in which the inteface is a human mediational device, can be seen as perhaps a product or central premise of the Turing test for artificial intelligence. There is no need for an interface design to be restricted to a human-user framework. Increasingly, the construction of mix and match supercomputer systems demands the use and development of computer-to-computer interfaces that permits talk between different processing architectures and data-structures. Similarly, interfaces can also be considered between sensory systems, on one hand, and motor-control output systems on the other hand, and these do not have to exist at the same place and at the same time.

Ideally, we would desire what can be called a "universal interface" design that would permit the widest range of input-output mediation possible between different kinds of systems, including human beings. This would entail that the interface itself would be an independent intelligent system capable of detecting and translating between different kinds of signals, and generating a wide variety of relevant outputs. Behind this would stand a capacity to coordinate signal relay systems between different components or subsystems, the ability to intermediate effectively between different levels of different systems.

Communication & Coordination Systems

Intelligent Transmission & Linear-Circular Computing Models

Intelligent transmission involves the possibility of signal transmissions being able to be processed in transit time in some meaningful manner. Relay switches in transmission lines would include an intelligent, autonomous function, but this is a nodal transition network model compared to what can be called an in-line processing system that might possibly have combinatorial functions independent of nodal switching points. The capacity to program and process information would therefore have to be built directly into the transmission line itself, as a part of the design of the line. This seems like an alternative strategy, if it were possible, to the concentration of integrative systems into a single unit. Ideally, a computer system would consist of nothing but a variegated mass of transmission lines interconnected in some complex manner without necessarily any junctions occuring between them, or possibly with many junctions embedded within the line itself.

We can imagine different architectures for such lines a few of which follow:

1. segmented structure (string of pearls structure)

2. coiled or helix structure (twisted DNA structure)

3. laminar flow structure

4. parallel linear, interconnected structure (telephone cable structure)

5. branching braided structure (rope structure)

6. needle-point loop structure (stitch & sew structure)

7. inter-looped structure (anchor chain structure)

8. rung-structure (ladder structure)

9. by-pass structures (railroad switching yard structure)

10. axon-synaptic structure (neural structures)

11. sphaghetti structure

12. complex hybrid structure of any permutation of 1-11.

Furthermore, borrowing an analogy from protein folding structures, we may identify possibly different levels or orders of folding structures in such systems:

1. primary structure

2. secondary structure

3. tertiary structure--"Gordian Knot structure"

4. quaternary structure

Furthermore, it seems as if these structures can be further interconnected with one another to form various kinds of network or chain-mail or fabric structures that again can be folded or convoluted in various ways. We can imagine structures that are one-way, switching or reciprocal pulse transmissions or again some combination of these. Considerations of these alternative patterns suggest an entire plethora of possible processing structures, but we must inquire whether or not such a primary structure is feasible and efficacious in the first place. For instance, can transmission of signals be effectively combined with the transition of such signals in a linear and sequential manner? Considerations would also include the cost and ease of manufacture and maintenance of such lines, if they were possible, as well as their relative efficiency in either the communication or processing capacities. At what point does the length of the line trade-off with the necessary breadth requirements associated with expanded processing capabilities?

If we can imagine a steady stream of binary code along a transmission line, and an in-line tape reading device that was capable of matching sequences of a certain order, signaling some transition to occur. A separate relay line might result in a series of down-line switches to occur that would channel information in some variable manner.

Sensory-Motor Feedback Systems

Another aspect of autonomous integration and distribution of such systems is in terms of what can be called the sensory-motor, or stimulus response, feedback loop, and again, this process does not necessarily have to be mediated by means of human inputs and responses. An entirely automated system would be independent of human mediation, and therefore would be capable of directly monitoring and collecting information from the environment by various forms of mechanical sensing devices or "organs." A device for instance capable of descrambling and converting sound waves into signal code, allowing for its transmission and reproduction in some other form, would be an example of such a device. This device must be capable of meaningfully reading the signals in some patterned way in order to avoide incorrect interpreation or translation. Light or some wavelenght sensig would be a most important type of sensory input to be capable of processing. Sophisticated systems should be capable of monitoring and defining imagery and complex fields that are variagated and continuously changing.

On the other end of this kind of loop, some form of functional feedback or behavioral state modification would be an expected outcome of certain kinds of input signals that were autonomously gathered from the environment. Much of this kind of output cycle would be connected to industrial processes or alternatively to equipment or machines that, for instance, modified ambient environments or lighting conditions in a building, etc. The use of extensible, folding armatures and mechanical hands that demonstrated a degree of "hand-eye" coordination and dexterity comparable to a pair (or a set) of well trained human hands would offer a very wide range of potential applications. Autonomy of function would have to be built increasingly into machine systems themselves, such as with modern motor vehicles that are capable of monitoring and adjusting their own operating conditions.

Mosaic Processing Structures & Communities

Mosaic Processing structures refer to the specialized division of labor within integrated systems and between distributed systems, such that no single system performed only one function exclusively, and all component systems were capable of potentially peforming the full range or variety of functions as part of the overall system. Mosaic Processing Structures refers to the variegated and irregular character of this division or stratification of function within complex systems, and to the cooperative and interactive nature of such systems between the various components and subcomponents. We would expect within mosaic processing structures considerable degree of overlap between different processing structures in order to prevent reduplication of effort.

Complex Mediational Processing

Cybernetic Equilibrium of Distributed Systems

Integrated-Distributed Systems

Complex & Dynamic Modular Partitioning

Coordinate Multi-Processing Systems

Conjunctive-Disjunctive Multiple Processing Architectures

Distributed Autonomy & Over-Control

Last Updated: 08/25/09