The Information Processing Unit: Robots and Artificial Intelligence

Is the development of Artificial Intelligence and Robotics yet another attempt to escape the 'ignobility' of the flesh of the body, or is it another quest for immortality? A propagation of our own presence beyond our immediate circle and lifetime until the data that constitutes a life gets fed into the next smaller, but better, version? The replication of the human model witnessed in successive attempts at developing artificial intelligence, computer technology and robots, forms a continued romance humans have with themselves as a species. It would seem we build robots to see ourselves in a new way, and Artificial Intelligence contains possibilities for transcending the species. Given that there is as yet no thorough and agreed upon understanding of what constitutes consciousness and how it operates, there is plenty of room for development in this area. Early information theorists saw the brain as a 'meat machine' and attempted to design systems that operated in the same way. Even whilst recognising the brains interconnectedness to the nervous system and the body per se, they largely focused on the brain as if it were a solo entity, rather than the brain as but one organ in a whole operating system.

So saying, the technology is here and I for one am a happy user.

The body operates in a reciprocating relationship to technology. Robotics research and Artificial Intelligence constantly work with the human as a model, to inaugurate possibilities and to discover more about humans. Then there is also the human body adapting to using technology. Witness the speed with which young kids, hungry to be on a computer, get around the screen and the comparative agony of watching a new but older user getting their 'mouse muscles' into shape. Technologies write different ways of being a body whether it be with the development of robots or adaptation to computer use.

History is full of attempts to make artificial intelligences; to reproduce the essence of the human. In a romance with the self, humanity engages in a form of self-production and, as an audience, we are enchanted by our imitations. And yet responses vary from the extremes of these 'graven images' being deemed praiseworthy and useful to being fraudulent, even blasphemous. Three themes run through the early discussions around robotics. The first that there is some supernatural power at work animating matter. Secondly that a non-supernatural entity will not be able to control the magical contrivance. And thirdly that as the creations are used as slaves, instead of as a source of knowledge, they will ultimately rebel and overcome their creators. Mary Wollstonecraft Shelley's Frankenstein, for instance, pointed out the moral dilemma of modern science, namely that good can produce evil, and asks what safeguards and constraints could be used to ensure that progress can still be made.

In the early stages (1940's) companies such as IBM suppressed the idea of intelligent machines to control public hysteria whilst simultaneously funding research into building checker and chess playing computers designed with the intent to learn as they went. The early hard work in the development of information processing as we view it today was done amongst figures such as Edward Feigenbaum, John McCarthy, Warren McCulloch, Marvin Minsky, Alan Newel, Herbert Simon, Alan Turing, John von Naumann, and Seymour Papert to name but a few. Involved institutions were Kendall Massachusetts Institute of Technology (MIT), Yale University, Stanford Research Institute, Princeton University and Carnegie-Mellon University amongst others, actively still the hubs of Artificial Intelligence (AI) Research complete with robots roving their corridors today.

Norbert Weiner in Cybernetics, 1961(1), revealed that energy, the notion central to Newtonian Mechanics and a system that dealt with closed systems, could be replaced by information theory. Coding, storage and noise better explained anything from electronic circuits to replicating cells and could deal with open systems coupled to the outside world, both for the reception of impressions and the performance of actions. It was, however, the information processing unit level of modelling (distinct from information theory) whose central idea was the manipulation of symbols (as opposed to previous models of mere feedback or on/off switches) that could admit mathematical and non-mathematical expressions that moved the field into its contemporary stages. The computer is viewed as an information processor or symbol manipulator and the catalyst for a leap in development ended up being not the question of how the brain works (or how information is processed) but what it does. It was the function that needed to be analysed and replicated with the right kinds of descriptions and mechanisms put into practise regardless of its similarity, or not, to the present system (us). The next step was working with knowledge-based systems models whose data was assimilated by knowledge engineers who worked intensively with experts in the field (say medicine) to acquire the specific knowledge the expert has and to organise it for use by a program. What became implemented was not a hierarchical system but a heterarchical one, similar to that found in biological systems and like the coordination of different systems within an organism. A number of concurrent processes are working in a coordinated fashion without being under the primary hierarchical control of one of them. Finally computers like HACKER (Goldstein and Papert, 1976) are able to examine their own procedures and identify flaws, learn from mistakes and 'debug' themselves. HACKER was programmed to recognise that sub-goals on the way to the larger goals can sometimes cause conflicts, and learn to reorder the sub-goals so the greater goal could be achieved. Hence a degree of self-awareness was integrated into computer technologies. Analogy, similarity and metaphor were concepts that the computer could manipulate and large amounts of specialised procedural knowledge could be processed. The recognition of similarities to old situations and dissimilarities to new ones was occurring in a dynamically functioning and organised knowledge base.

Why was there this insistence on a replication of the human, both in the system of processing and the look of the thing, the robot? The answer is manifold. The workers in the field didn't have any particularly good ideas about better ways of working out intelligence processes, so they milked all they could from what they knew or could observe. It would also have been impossible to construct a complex program if the mechanisms in the program were vastly different from human problem-solving mechanisms because what was going on inside the program would have been virtually incomprehensible. There would have been no way to relate to the processes going on inside, to add to them, to modify them unless they were pretty much 'done' in the same manner. Robotics and AI research is a far-reaching field that also attempts to find out more about human beings within the scope of its research : keeping the human as a model for investigation furthers that possibility.

Seemingly indestructible robots have explored the extremes; from the landscapes of Mars to the interiors of human bodies. As adaptable as the imaginations of their designers, they can go where frail and intractable human flesh cannot. So are they just one more tool and a logical and necessary extension of the human species? Or are they becoming or have they already become a species of their own, our own possibility for transcending the limits of the human tribe? The name robot comes from a Czechoslovakian word for servitude or slave. Certainly from early childhood I assumed I'd be having my own customised robot by the time I got to be an adult. Not only would my robot brew the tea and make the toast, 'it' would also handle my correspondence, write up all my ideas coherently (and apply to the appropriate funding bodies without needing to be told) as well as working out the engineering and architectural specs for every idea I ran past 'it'. As cool as Spock! Imagine how much one could get done! The list for their activities grew correspondingly with my understandings of how many good things there were to be doing. So where is it, my robot? Well it seems things have not progressed as fast as one would hope or Science Fiction and special effects could make seem possible.

At present our brain handles 10 million million computations every second, still a million times faster than any computer today - although computing capacity is growing at a rate of 1000 times every 50 years (2). # So what is the state of contemporary technologies? At MIT we have robots like "WAM" and "COG". "WAM" is a "Whole Arm Manipulator" comprised of two overhead swivelling cameras, a monitor for each camera and limbs of steel which quiver in anticipation for a ball to be thrown to them. Once caught it throws the ball back in a smooth arc and the whole process begins again, much the same as with a puppy dog. "COG" is still in its early stages and perhaps the most ambitious project in AI as yet. "COG" is being built to look like a human (to find out more about humans) and will use sight, sound and touch to accumulate information and develop more knowledge. So far "COG" has 4 eyes, 2 for panoramic and 2 for narrow focus, and a brain that sits like a body beside it where each sense is processed by different nodules which in turn will interact. There will be skin, arms etc. to come but "COG" has a long way to go before becoming fully operational.

Some features of the current debate on AI center around what is consciousness and how does it come about? (also called 'the hard problem') which includes in its parameters notions of Quantum theory, the Heisenberg Uncertainty Principle and Godel's Incompleteness Theorem. The interaction between knowledge and neurons was pointed out as an area needing more investigation by McCulloch and Pitts in 1943 in A Logical Calculus of Ideas Immanent in Nervous Activity in the Bulletin of Mathematical Biophysics. We are still investigating and uncertain how the body processes information and what certain parts are for. At the heart of this debate are Mathematician Professor Roger Penrose (Oxford University) and Anaesthesiologist Professor Stuart Hameroff (University of Arizona). Roger Penrose is the author of The Emperors New Mind (and more recently Shadows of the Mind with Prof. Stuart Hameroff) and sees the area where consciousness resides as that part within quantum physics where millions of thoughts and feelings exist simultaneously until looked at and then it all collapses into one spark of self-awareness. The continual stream of sparking, ever-increasing, eventually becomes a stream of consciousness.

Professor Stuart Hameroff discovered that microtubles, dense networks of which operate in all cells, seethe with activity except when a patient is under anaesthetic, which is also when a patient is unconscious. Consequently he claims microtubles are the physiological residence for consciousness (3). There is a lot of contentious debate surrounding this work with many scientists working in the area seeing this as hypothetical at best, dubious at worst. Interesting to note the audience at the recent conference on AI Towards a Science of Consciousness (1996, Tucson), were equally divided on those who saw machines as having, or having the potential for, consciousness and those who didn't.

Edward Fredkin (a professor of Electrical Engineering at MIT ) like Hans Moravec (Stanford) in 1977 sees artificial intelligence as the next stage in evolution. Fredkin discusses the popular view that we barely use 5-10 percent of our brain's capacity and yet the way he sees it, to get along we need about 110 percent, causing breakdowns and trouble of various sorts. But artificial intelligence could assume the heavy thinking for us, sort out the unsolved problems we've been accumulating and not knowing how to resolve: over-population, over-pollution, food shortages, planet destruction... The danger Fredkin says lies in this intermediate stage of intelligence when real damage could be done because the development hasn't gone far enough to prevent that. (Humans have a history of destructiveness. There is no reason to assume that intelligence once it has gone beyond the realms of the human would imitate this). Why is it that we don't mind machines (trucks, cranes, planes) doing all our physical work for us? Why is the intellect this sacred untouchable realm? (Why does World Chess Champion Gary Kasparov feel he must defend human dignity rather than just play the best chess when playing against the latest chess-playing computer 'Deep Blue'?(4)) Fredkin states "The mere idea that we have to be the best in the universe is kind of far-fetched. We certainly aren't physically"(5).#

Hans Moravec in 1979#(6) sees the evolution of a community-mind evolving where we won't be attempting to lug around all we can possibly know about the world; instead we'll access storage banks as needed, and utilise our available memory to specialise. So individuals will pick and choose what their minds contain at any time. Our systems won't be overworked and fragmented. They'll be streamlined, efficient, parred down to their bare essentials and very focused. In 1996 Moravec predicts we have only another 50 years to wait in Robot Evolution before Artificial Intelligence will be intelligent; a mere speck in the overall time-scale of evolution. It certainly sounds hopeful for artificial limbs, I might be needing some by then!

© Ann Morrison,1996
(Ann Morrison is an artist, currently enrolled in M A in Design Science (Computing), Univ. of Sydney)
MESH film/video/multimedia/art #10,MESH is published by Experimenta Media Arts
Pictures for the article