A Conscious Computing
System ?
|
|
What is Consciousness? There has been quite a lot of work on this question over the years. So I shall briefly cover some of the ideas which have achieved a sort of consensus. Consciousness is the "I" that we all know, from which we view the world and interact with it. John Searle describes it as that sense of "subjective qualitative states of awareness, sentience or feeling" which we experience when awake [Searle: presentation to Tucson II (1)]. In the 17th Century Rene Descartes (2) showed that no matter to what extent our senses might be deceived there would still remain a something which could be called "myself" even if this "self" were utterly deceived as to the existence of any one or any thing else. I suppose his "cogito ergo sum" might have been better put if he'd said I am deceived therefore I am. William James (3), in the late 19th Century, put it as having a sense of a personal consciousness that is ours, not something that we share [though in passing, how we consider this against the matter of the social construction of our world view remains to be discussed]. Everything we know is a function of experience, either through sense perception or reflection upon that experience. The mind or the "I" is born empty of knowledge of the world. As the 17th Century English philosopher John Locke (4) described it we are born "tabula rasa" (or as a 'blank slate'). It is only by our experience of the world that we gain ideas of it. There are no ideas that we we are born with, no "innate ideas". "Let us then suppose the Mind to be, as we say, white Paper, void of all characters, without any Ideas; How comes it to be furnished ?....To this I answer, in one word, from Experience... Our observation imployed either about External sensible objects; or about the Internal Operations of our Minds, perceived and reflected by our selves... These two are the Fountains of Knowledge from whence all the ideas we have, or can naturally have, do spring. [Locke. A Treatise of Human Knowledge, 1721, p67] It is the data of sensation which have the qualities or "qualia" which are the stuff of our subjectivity, that which we know. Consciousness is a function of the state of our central nervous system, i.e. the physiology is the substrate upon which consciousness runs. To quote William James in his Text Book of Psychology, 1892: "The immediate condition of a state of consciousness is an activity of some sort in the cerebral hemispheres... One has only to consider how quickly consciousness may be abolished by a blow on the head...[or] by a full dose of alcohol...to see how at the mercy of bodily happenings our spirit is... Destruction of certain definite portions of the cerebral hemispheres involves losses of memory and of acquired motor faculty of quite determinate sort...Taking all such facts together, the simple and radical conception dawns upon the mind that mental action may be uniformly and absolutely a function of brain-action, varying as the latter varies, and being to the brain-action as effect to cause." [James, 1892, pp5-6] Our state of consciousness: (a) is always changing as we are exposed to continually novel sensations. (b) is selective of what it pays attention to, and (c) provides a sense of temporal continuity (William James' "stream of consciousness"), which the normal day-to-day changes of sleep and wakefulness, as well as abnormal changes such as unconsciousness, do not interrupt. Paul Churchland, author of The Engine of Reason; The Seat of the Soul (5), has described some important functional aspects of consciousness, which we would need to explain in a theory of consciousness. (a) Short-term memory and its decay. (b) Directable attention, or conscious control over what we attend to and what we do. (c) Multi-valent comprehensioneflect within on the patterns within, (c) we need to able to reflect on the world by interacting with others, and (d) we need to able to project into the world for puposes of generating feedback. Now, the human body/brain is a complex organised system consisting in collections of cells which are organised in such a way as to do various biological/biochemical tasks. Among these tasks are the processes of the brain in co-ordinating and handling all this stuff. The effect of being in the world in real-time (and strictly in real-time) is that the entity is consistently exposed to difference and novelty. We know of the the world by the differences that show up between and among all processes extended in time (having duration). No process can act completely independently and each is affected by the changes (the differences) in, ultimately, all other processes. So we are inextricably interlinked with the world. Because we are active creatures the environment is always changing about us as we move about in it. So every thing produces difference, even if only the two different views of some object which allow us to build up a 3-dimensional view of it. The determination of intelligence in other systems What of intelligent machines? Given that an artificial intelligence will also need to communicate with its world, which hopefully includes us, it to will need to have an array of inputs and sensors and communication systems at its disposal. An AI will presumably have some sort of "brain" to co-ordinate all its activities, as likely as not this will be at least as highly distributed a processing system as we use. What is it that enables us to determine whether an other entity is intelligent or not? Largely it's a matter of whether or not that other entity can report to us that it is so, or if common language is unavailable can make other output which demonstrates use of reflectively produced information. I recognise that this allows many current computers a foot in the door, but the capacity to extemporise on the intricacies of current political drama might show us a capacity beyond what is currently found in computing devices. The traditional procedure for determining whether some artificial/constructed device can be described as "intelligent" or not has been the Turing Test (TT) (8). In the TT an interrogator is set in front of a terminal which presents information from the entity being tested without indicating the nature of that entities' direct input/output capabilities. In other words the information is mediated in such a way as to remove any tell tale clues (say, typing speed). The interrogator then asks the entity questions and if the interrogator is unable to tell which kind of entity (human or machine) is supplying the answers then that entity would have to be considered "intelligent". Of course it could be playing down to us. The Turing test requires the "computer" being tested to behave anthropomorphically in that we require it to report to us in a manner which we can understand. We want it to display human-like behaviour. So I am going to carry out an anthropomorphic analysis of what an intelligent computer system might be. By the way, there is some difference which should be noted between "intelligence" and "consciousness". Consciousness seems to have more general applicability than intelligence in biological entities. We tend to think these days that most creatures above the levels of fairly rudimentary mammals have some sort of "consciousness" of their surrounds, but intelligence is a term which may or not be applied to an acceptably conscious individual. Its use will largely be dependent on context. I will come back to this point in discussing the various possibilities of classes of AI's. What are the signs by which we might recognise consciousness in biological or silicon others? As Robert Kirk has pointed out the most obvious thing about a conscious entity is that it does something for itself and very largely that something is to collect data about the changes of and the differences within its environment. Another thing which a concscious system would need to be able to do is to act independently in a more general way than simply gathering information. That is to decide to do (or not to do) some thing for itself, of its own accord, and to proceed to do that without being prompted by some external conscious agency (an 'other'). We maintain an 'idea' of the world about us by gathering information (no matter how low level) which is for us, for our own use. We gather information for ourselves. Thus knowing will be an inherent process (complex and organised as it is) of any conscious system for gathering any information that may be useful in some way (in some sense). For example: What is the surface like? Where is the next food? Where is my friend? So if this is more or less what consciousness is for us, what will we need to see in a non-biological entity, a silicon-based computer system for example, for us to say that it is conscious? and how might that be different from saying that it is intelligent? We should apply the same criteria. This is the anthropomorphic procedure I suggested would be necessary to answer the question. So where are we likely to find this system if we're to ask it if it is conscious? We can envisage two classes of intelligent or Cs machines. The first are (at least, logically) self-contained machines which in some way are able to act like an individual human. They would be able to sense and make sense and act on the sense and display, or report on, the act and its results. They would have to be some sort of highly integrated machine which had enough internal "physiology" to enable it to handle the vast amounts of data that are necessary for a coherent view and understanding of the world. Perhaps a large neural network machine with multitudes of distributed nets carrying out all the tasks necessary to keep such a device in touch with its context (the world). An example of such a machine might be HAL in Arthur C Clarke's "2001" (9). We can possibly produce a machine like HAL in which we specify the kinds of tasks we want it to carry out. But what happens when it decides to do something for itself? HAL's problem was that it had two contradictory prime directives and went psychotic when these were forced into contradiction as they were at a certain point. The average modern supercomputer is an "idiot savante" when it comes to doing anything other than its primary task: calculating. Our self-contained machine would need to have a vastly distributed set of processing subsystems which sense (feed forward) , reflect and control (feed back) each other, keeping it in touch with itself and its world. Because of things like the now acknowledged impossibility of ever describing the world and consciousness in a systematically complete formal or logical system (vide Goedel's Incompleteness theorem (10)) and because of the huge programming task, it's probable that we would not be able to pre-program this system and simply set it runnng. We would have to set a machine up to run itself and then teach it just like we do our children. Though it may have vast access to libraries of knowledge the machine still has to be able to operate in the world: to know who and what its neighbours are, how to communicate with them in which particular language or protocol, and so on. And further, how would we actually manufacture such a system? the complexity of wiring and the range of sub-system capability are staggering if not insurmountable. It seems that we would have to use evolutionary algorithms and some kind of auto-assembly process to even begin to build such a machine. I argue that the imbuing of consciousness to this machine will follow fairly straightforward principles but that the actual technical implementation of such a machine will require a total reworking of manufacturing techniques into something more akin to the biological process. So, do we ultimately have to let the system grow itself and decide what it wants to do both in relation to us (in responding to our input and requests) and in relation to itself? The second class of intelligent or conscious machines is a distributed network of machines which probably would act more as a society, but might also be able to act as a single combined entity, an "individual". This a large system in which the elements are widely distributed, that is not all in the same room, say. Generally it is known as a net or a network and given enough layers of linked subsystems, in some sort of hierarchical as well as horizontal structure, with enough of an organised basis as to be able to distribute the array of tasks necessary for conscious behaviour over an array of appropriately inter-linked subsystems of computer embodied sub-nets. My feeling is that the Internet will be the context in which this kind of capability first appears. Some sort of capacity for self-organisation would have to be installed on the network to enable it to evolve to a point where it might exhibit an awareness of its (logical) surroundings, and to be able to report at least on this situation, if not on its internal goings on. What would be the differences between a single intellig ent machine and a social network of machines (nodes) as in the Internet? An intelligent or conscious machine is in a sense an analog of what you or I are, an individual. But I mean that more in the sense of having some sort of self-perceived bodily integrity. Will the machine be mobile? It might seem necessary if it were to gain the kind of experience that we consider the basis of our consciousness, but I am hesitant about saying that mobility is necessary. An intelligent internet couldn't, of course, be mobile. It would gain the knowledge of the world that we have as individuals in much the way our society conveys ideas and knowledge that we might not get if we had to have first-hand experience to gain knowldege. That is through teaching and learning and through its equivalent of the movies and the news. Also an intelligent Internet would be everywhere at once; an omniscient being, whereas an individual intelligent machine would be restricted to the kind of localised mobility and knowledge sources that we use. (Though the machine, one supposes, would be able to log in to whatever library system it needed for more general information about the world and might emulate the sense of being everywhere at once.) One might liken the difference between these two classes of machines as the difference between an individual and the society within which an individual lives. What might a conscious machine want to do for itself? There are two questions involved here: First, what do we do for ourselves? and subsequently: What of these things might a machine need to do or want to do? But I will proceed by describing how what we do translates into what a machine might do. The collection of functions of consciousness I mentioned above can be summed up by saying that a conscious entity will need to act independently, to do things for itself and to be able to report on things and interact with others, i.e. to communicate. We are not here concerned with the particular facilities for doing such things held by any particular entity, we are really only interested in what kinds of things might be done. Getting down to the details: Self-Maintenance The first thing that any self-consistent system must do is look after its own maintenance. Even if it has to get the "doctor" in it will need to able to report at least to itself as well as to another, in whatever language is appropriate, the failure of some piece of "physiology" (hardware) or the loss of a link. Also this includes the ingestion or acquisition of energy sources and, if it is going to emulate the biological, by being self-maintaining and self-replicating, component parts. For example, the internet is a vast array of ("intelligent") terminals which occasionally hook into a slightly less vast array of servers which store files for access by terminals, these files being placed there for human use by and large. A potentially intelligent computer network of this collection of smaller networks would have to look after its own maintenance. Each "terminal" would need to maintain a diagnostic (inner) "eye" on its overall operation. It would particularly need to build techniques to prevent its penetration by an unwelcome agent (e.g. virus, hacker). Is the anti-virus programme the first step toward an immune system? The nodal computer (the "server") would need to maintain itself and also keep an "eye" on its links, these being the terminal links to it (the user names), and the links up to the next server in the next layer up the hierarchy of nets, and so on. Each individual machine would also need to look after itself, both in terms of internal maintenance, and materials (information and fuel). It could refer up the net for system diagrams and diagnostics but it would need to carry these out itself or request help from another (perhaps a "user"). A basic level of this self-maintainance is already available. For example, individual computers, the nodes of the net, can use what are known as boundary scan techniques (for which there is already a formal standard) to maintain themselves. Meanwhile, their links are already maintained using the system of pathway redundancy. Through the need to be auto-sustaining in case of a link breakdown, the internet has a flexibility of interconnect pathways which reduces all pathways to being pretty much equivalent. Organisation The machine or system is going to have to be "organised" with a structured interconnect system and groups of subsystems which carry out particular functions, such as vision. Biological systems are "intelligent" because the collections of cells we call 'organs' are organised into particular classes of functions. It is these functional groupings of cells (elements) which feed and maintain themselves and serve each other as part of an essentially co-operative system that forms the body or the organism. By system I mean just this, a collection of functional units or terms which function together (mostly interactively) to form an entity greater than the sum of parts. It is the relations between elements of this system which keep it together. What are the equivalents to specialised organs in a computing system? In an individual computer we have the CPU, an arithmetic unit, Memory, and the internal and external interfacing logic. These are more like individual neurons or the machinery of an individual cell, not enough to be considered organs, but more the possible constituent parts of an organ. The nearest example similar to an organ we can find, is a set of computers (particularly using a parallel processing architecture) linked together for a particular purpose, say mathematical or theoretical physics problem-solving. Or we might think of the telephone network as being analogous to a nervous system. Going on up into the network scale, are there groups of computers with oversight of the net which carry out maintenance roles? And, if those computers are doing something which is essential to their continued functioning, are they doing something for themselves? Is there a sense in which maintaining the net is doing something for these "selves" (or itself as the 'group-self'), and if so, do they recognise this fact? Things that a computer does at present are those which it is requested to do by a user, whether it is acting as a typewriter, or doing calculations for the most abstruse scientific visualisation task, it is still only doing things for us. Maintenance and the formation of functional units are very low level activities, but for an intelligent or conscious system, very necessary. We still do not have active systems which can do this kind of work, although many parts of the task are extant. In the maintenance mode, keeping track of the state of individual physical links is still the nearest thing to "doing something for itself" that we have. Then, what else? Knowledge of its surrounds The next thing is knowing its environment: searching for information which is for its own use, such as who its neighbours are, and what the overall structure of the society of computers might be. This information may be reportable to humans but it needs to be stored as part of the maps of its culture. Current search engines are among the few semi-autonomous entities on the net/web. Their activity is to go from node to node gathering information about files kept on the nodes. They maintain a database (a body of knowledge) about links between nodes, but this is only as information about what is kept at other nodes. So a search engine is an organ in one sense, but we cannot call it intelligent because it does not do this for itself, it does it for its users which are, ultimately, all the humans at all the terminals at the ends of all the links in the network. The production of new kinds of web bots and data mining and sorting agents increases apace. But these agents are still only doing something for an external 'user'. It is when a computer in the network generates some code for its own use to go and find information about, say, a maintenance problem, that things get interesting. To put this example in cultural terms: it needs to go and consult the 'doctor' subsystem in the network. Independent action This is where the maintenance mode produces the first signs of another indicator of consciousness through independence of action. It should initiate the task of finding the necessary data and modifying the code to its own ends. Now, we can say that "doing something for oneself" is an independent action, but there is a complementary form of independent action in that a system might decide of itself to do something for someone else, something like finding a book that you know a colleague is interested in. Then, of course, there is the purely altruistic independent action of buying a present for a friend. A system might also want to work independently on a problem set by its 'boss' so that it can generate useful answers and receive positive feedback about its own usefulness. This is almost a level of emotional need. So what are the machine equivalents of independent action, productive or generative action, and even altruistic action? It is the latter two that bring independence of action out of the realm of simple maintenance procedures and where we see the first kind of behaviour which could be considered conscious. The two things that a system will want to do for itself, if what we humans do can be taken as any kind of an indication, will be to: 1. organise itself and its relations (with other machines as well as humans), and maintain that organisation, and name its social system (these things are the means by which we are in the world); and 2. enquire into some of the "imponderables" of its existence, such as how it came to be, and what does it mean to be human (these things are the means by which we make art and other cultural productions). For humans, two of the deepest givens we have are: 1. the social system handed to us by our parents first and then they tell us of the ancestors who "wrote" the rules about uncles and aunts, exogamy and taboos, parents and grandparents; and 2. the imponderables revealed to us by some "hypothesis" (or construct) we usually call the Ancients or God, but which are the long lost antecedents of our respective cultures. We could describe these two processes as firstly, elucidating the structure of relations between all entities on a net, and secondly, the modulation of these sets of relations ("sets of relations" include relations with physical and mental productions in the world as well as between entities in the world). The modulation of relations is the business of art and science, politics and religion, but the effect of each folds back into the structure of relations giving us a dynamic informational/social environment, which in its turn modulates us, right down to the synaptic connectionism between nerves. The structures of "social relations" are really just a thorough knowledge of the links in the system and the "mental" contents of the entities at the nodes of the network. Conscious machines have to deal with all these sorts of things: the needs and desires of their (machine) neighbours, the co-operative construction of real or virtual/logical aspects of their environs. They are going to have to be able to know that certain sets of relations are inextricably bound, like the objects which are assembled into a car (given that the computer might only have a parts list) or like the social relations of marriage. They also have to bind things occurring concurrently on different stages in the world, and to recognise that these things are happening now. For a single distributed machine like the internet, this would all go on within its "body" and, as we don't usually know in detail what is going on within our bodies, the "conscious" level of such a machine would not need to know either. If it were as conscious of everything in its internal world as we are of ours, and if this machine were truly intelligent, then it could be expected to get lonely. That is, it could be expected to desire and seek other entities to converse and interact with. These are emotional and social needs which bring the system into a condition which is more and more identifiably anthropomorphic. The Social Dimension Beyond the terminals of the network are the users (currently human). What is their role in the intelligent network? Is the intelligent network somehow like a culture, providing informational nourishment to its end users? If the network is intelligent it will ask of its end-users information that it needs for its own continuance. Again, what are the kinds of things that a system would want to do for itself? The prime reference model is of course us. What do we do for ourselves? We maintain ourselves and our contexts and we generate stuff for ourselves and others around us. We: - maintain
the nodes and the links (i.e talk to our friends, relations and associates);
Perhaps to allow a system to be intelligent is to allow it to decide what it considers to be useful for itself. A conscious internet would be a nodal system with "culture" as the model and so we can imagine a variety of possible social structures which we might expect to find in this network system. For example, individual nodes in the net might want to establish one of their members as a kind of oversight node looking after the link structure of all the other nodes in the local area. They might hive off their external maintainance to one individual (the minister for health). Some nodes might be established as repositories of information, like librarians or teachers. Other nodes might deal directly with the humans, the users at terminals. These might be the foreign affairs department, (forgive the bureaucratic model here). In a sense this is a social entity, an array of individuals who live as a subculture in society. The networks of individual computers will as likely as not form into political groupings or family groupings and we will see hierarchies of relationships develop. Will they be able to evolve better political models than we have so far? Will co-operative and compassionate structures evolve with their framework? Now, societies do things for themselves, so are societies or cultures intelligent or conscious? We might have to say that the well-known tendency for an institiution to look after itself against all comers implies a kind of consciousness. (Just look at the police service or the church.) But it is embodied in the entities which are the nodes in that social network who, in the process of their doing "some things" for themselves, generate the process of maintainiing the institution. The traditions of the institution form the skein of ideas and information which the network (i.e. the sets of relations between....) uses to inform and shape the embodiement of the ideas that make up that institution. These ideas are transmitted from node to node using a medium of language of some sort. Some people have suggested that language is a virus. It certainly shows many of the epidemiological characteristics of diseases caused by bacterial or viral agents in terms of modes of transmission and patterns of dispersion. Ideas are carried via the vector of language, but they are also moulded by their vector. Some languages can only carry certain limited aspects of the full possible spectrum of an idea. I suggest that this means that "memes", Dawkins' (11) and Dennett's (12) units of meaning, carried on the vector of words are highly dependent on that vector, or language. That is they are context dependent. They are moulded by the language vehicle being used and they then mould the interpretive mechanisms of the individuals exposed to these languages. This process then moulds the possible interpretations of the memes used in the first place, folding in on the language in an endless generative process in which potential meanings grow and the endless variety of languages and cultures appears. The same would happen in a conscious computer network, though it might be capable of knowing and using a vastly greater variety of languages, which would then provide a greater spectrum of possibilities attached to any word one might use. In a social system the vectors are spread by direct contact and interaction as well as carried in cultural productions, be they books, films, art or whatever. The great value of art in the world is its capacity to be used as a vector for new, undeveloped and perhaps even socially undiscovered ideas. Of course the other great value of art is its capacity to trigger generative interpretative processes, actually bringing new ideas to the surface of a culture. In human societies, ideas spread, traditions and cultures come and go, but they are diverse and dynamic. Will a society of computers retain this diversity and its inherent dynamism? Or will it settle to a lowest common denominator condition wherein it knows all it needs to know and the low level stuff looks after itself. Will a conscious computer system need to be jolted into activity by some threat or problem? Will we have to keep it entertained? Will we, as generative agents,be useful to the machine, or would it prefer to be a slumbering giant. Or is the system, if conscious, already generative? There is already some evidence to suggest that this is the case. I refer particularly to Steven Thaler's Imagination Engines (13) in which a large neural network is trained up on a particular set of input forms, for example motor car designs, and then turned loose, without any input patterns, using only noise as an input, to generate any patterns which might come up, with some sort of "selection" mechanism picking out those results deemed useful according to some "socially acceptable (?)" criteria. But what else do social systems do? It is generally considered, I suppose, that we generate these social systems. We produce them so that we can maintain ourselves and gather useful information for that purpose. It is obviously not the only reason we do things for ourselves; we act creatively in myriads of ways. Generative activity requires a generative capability, and is an important part of being conscious or intelligent. We often generate stuff for other people of course. Most art and produced culture is done for others to see and interact with, the payoff for us (the producers), being in the feedback. Thus, I argue, we would expect an intelligent or conscious computer system to be similarly generative, and it is this which, in the long run, will give us evidence that the machine or system is in fact conscious. How would we produce this conscious net? How would we start up a process in which something like an intelligent and conscious machine or system would develop? Francis Heylighen (14) of the Free University of Brussels, Belgium, has suggested a model for searching the web using the analogy of associative memory where hypertext documents are considered to be "concepts" in memory and hypertext links are like associations of ideas. As links are followed by the user they accumulate "weightings" which relate to the frequency of usage. (This is a kind of Hebbian model of "learning" in which connections between neurons are made easier by usage. For example, one learns to associate concepts with words by a process in which those associations that gain the greatest connection strength or "weighting" become the best "interpretation" of the input data pattern.) Thus links between concepts which are regularly used will become stronger, and links which are less used show up the rarer connotations of a word or concept, and so on outward in a kind of "spreading activation" of links which, Heylighen suggests, emulates the way we associate concepts and ideas in our brains. If we could set up a search engine system to automatically record the link usage as a kind of "association strength" meter, then we would see the most direct connections between words and ideas having the greatest strength and the less well known connections coming up as less strong. A system could then make suggestions as to possible connotations which it "thought" a user might be looking for if they asked only a vague question about something or asked for, say, poetic connections. How could this be implemented? The seed process would need to be a means for a computer to write code for itself. The system would then decide that it should contact some other machine to help with a particular task. It would then go out and find it by searching for particular bits of code which we might liken to names or symbols. Once a machine can act independently and offer services and make enquiries then we start to have useful networks of entities. We then need to get the system to recombine its new knowledge with old knowledge and draw new conclusions, develop new combinations and so on, with the kinds of activity we know as creative. Any system which is capable of conscious behaviour must be able to reflect on the contents of its consciousness and of its surrounding culture. It is through this that we develop new ideas, new techniques and new cultural productions. Reflection is the process of storing input information and feeding it back through the input processing susbsystems, adding new input information to the body of the "idea", and allowing the body of the "idea" to modulate any interpretation placed upon the incoming information. During infancy, as we deal with and "reflect" (albeit at a very low "hardware" level) upon this input we learn, for example, to recognise our parents and others around us. Also we learn our language and its meaning, through just this repitition of processing and "trying" to understand, and layering of previous examples and understandings onto new ones. Another activity that a conscious computer network will engage in is the search for the means of doing or achieving something. A reflexive network system will find software on particular nodes in the system that allow it to do all sorts of things, especially at our request. Intelligent agents that find the nearest available node to carry out some particular process that we need to do probably almost/already exist. Of course it can also find that software for itself, but still, what drives it to want to do what that software node does; to use that operation, that software? And does it look at the system in terms of software nodes rather than hardware nodes? I mean, we might look on the kinds of things we do as an array of soft functions relating to input terminals: eyes, ears; and output terminals: mouth, muscles, etc. If the map of the system is not dependent on hardware nodes then machines can become specialists in particular activities available to any (human or machine) user which needs that processing. That is, give the software an address it can recognise and, when it is needed, the system simply posts the data with the appropriate software address. The package is then sent out to the network and captured by the addressed software which then carries out the requested processing and returns the package to its source address. The software simply needs to know if it is being addressed and whether it is busy or not. Is this the end of selling software? One simply buys the machine-cycles to do whatever tasks are needed at your own node. Is this the (dumb) Network Computer that is currently in development? In conclusion, the first context in which we may see a truly conscious machine will be in the internet/web nexus, but this will entail a development by stages or levels of hierarchy. The introduction into the internet of techniques and algorithms for letting the network develop and extend its connectionism, that is, its hypertext linkages, and the provision of means for closer co-operation with its human users in search processes using various kinds of "agents" and "web-bots", might then generate a kind of bond between humans and machines that will be radically new. This process will produce a kind of group-mind or "Global Brain" (Heylighen). Then the extension of this towards an actually conscious system will require the introduction, into the net, of the kinds of capacities discussed above. Obviously this leaves many issues unresolved, for example, the ethical question. If we do produce a conscious computing system but we don't allow this system to develop its own freedoms and individualities, then are we simply breeding slaves? (vide Do Androids Dream of Electric Sheep? by Philip K. Dick, 15). Also we may not be able to do this in a machine with the technology we currently use. It is most likely that we will need to implement computing machines in biological hardware before they are complex enough to grow and learn as human children do. And, of course, this leads us to a final question: are we simply trying to find another way of taking over the natural process to re-invent ourselves as some kind of extended human, or are we simply making more humans? |
| ©1997 Stephen
Jones |
| Sources |
|
1: John Searle: How to Study Consciousness Scientifically, a paper delivered at Towards a Science of Consciousness 1996, see also Is the Brain's Mind a Computer Program? Scientific American, 262, 1990, pp26-31. 2: Rene Descartes: (i) A Discourse on Method. (ii) The Treatise on Man. (iii) The Dioptrics. (iv) The Passions of the Soul. [see also extracts from Descartes in the Brain Project web site] 3: Wiliam James: Text Book of Psychology, 1892 4: John Locke: A Treatise of Human Knowledge, 1721 5: Paul Churchland: (i) Consciousness: its Past and Future, a paper delivered at Towards a Science of Consciousness 1996. (ii) The Engine of Reason; The Seat of the Soul, 1995, MIT Press. [See also the web page www.merlin.com.au/brain_proj/Neur-net.htm, Neural Networks and the Computational Brain, which includes a precis of Churchland's talk at Towards a Science of Consciousness 1996.] 6: Robert Kirk: The Basic Package and Consciousness, a paper delivered at Towards a Science of Consciousness 1996. [see also Robert Kirk explains the Basic Package, a transcript of an interview recorded at Towards a Science of Consciousness 1996.] 7: David Chalmers: (i) On the search for the Neural Correlates of Consciousness, a paper delivered at Towards a Science of Consciousness 1996. (ii) The Conscious Mind: In Search of a Fundamental Theory. Oxford University Press, 1996. [see also David Chalmers on the Hard Problem, a transcript of an interview recorded at Towards a Science of Consciousness 1996.] 8: Alan Turing: On Computable Numbers, with an Application to the Entscheidungsproblem, in Davis, Martin, The Undecidable: Basic Papers on Undecidable Propositions, Unsovable Problems and Computable Functions, Raven Press, New York, 1965. 9: HAL, see Simson Garfinkel's article in the January 1997 issue of Wired celebrating HAL's fictional birthday. See also Arthur C. Clarke, 2001: A Space Odyssey. 10: Godel, Kurt. On Undecidable Propositions of Formal Mathematical Systems, in Davis, Martin, The Undecidable: Basic Papers on Undecidable Propositions, Unsovable Problems and Computable Functions, Raven Press, New York, 1965. 11: Richard Dawkins; (i) The Blind Watchmaker. (ii) The Selfish Gene. 12: Daniel Dennett: (i) Consciousness Explained, Little Brown & Co., 1991. (ii) Kinds of Minds. (iii) Darwin's Dangerous Idea. 13: Steven Thaler: www.imagination-engines.com/devo.html, The Fragmentation of the Universe and the Devolution of Consciousness. 14: Francis Heylighen: From World-Wide Web to Super-Brain, on the Principia Cybernetica website. 15: Philip K. Dick: Do Androids Dream of Electric Sheep? |
|
Biography |