Thomas L. Clarke

Institute for Simulation and Training, University of Central Florida

3280 Progress Drive, Orlando, FL 32826

The ultimate challenge for simulation is consciousness. While great progress has been made in simulating many aspects of intelligence for applications such as ModSAF, the full simulation of consciousness remains elusive. From the time of Descartes to the early 1990's, science had regarded consciousness as the domain of philosophers and mystics, but this is rapidly changing, and recently scientists have begun to investigate the nature of consciousness and how it might be simulated.

It is now becoming clear that the many separate aspects of intelligence modeled by AI: planning, target recognition, sensor fusion, etc, can only be integrated through a full scientific understanding of consciousness. In addition to the reasoning aspects of intelligence normally associated with AI, consciousness research suggests consideration be given to affective aspects such as emotion. This line of research may lead to more effective weapons such as smart bombs that "want" to hit their target and will do whatever it takes to reach that goal.

Many professional-level conferences have been started to support research in consciousness. The biannual Toward a Science of Consciousness conference held at Tucson by the University of Arizona is the preeminent meeting in the field. The Sante Fe Institute is another center where non-linear dynamics and complexity theory are brought to bear on problems related to consciousness.

The talk will discuss visits to Tucson and will make connections with CGF research in military simulation. Suggestions will be about how research in consciousness studies will provide the basis for a new generation of CGF's for the next century.

The ultimate challenge for simulation is consciousness. Great progress has been made in simulating many aspects of intelligence in applications such as ModSAF and similar computer generated force (CGF) systems, but these systems fall far short of being able to model the full range of human capabilities made possible by the conscious mind.

Consciousness until recently has been a taboo subject in science. Talk of consciousness would bring to mind images of gypsy women peering into crystal balls. But a combination of advances in neuroscience such as MRI (Magnetic Resonance Imaging) and PET (Positron Emission Tomography) scans with new approaches to computation such as neural networks have made scientific investigation of consciousness a reasonable topic of research. In the past decade great detail about the functional nature of brain structures has been determined via MRI and PET (Figure 1). Where much of the brain was once considered silent terra incognita, the broad functional outlines are now clear.

The entry of such eminent scientists as Francis Crick (of DNA fame, 1993) into the field of consciousness studies has led to the establishment of major conferences devoted to consciousness studies. The spirit and content of one of these conferences is discussed below.

One of the most exciting developments in the field is the synergism between consciousness studies and the developing field of quantum computation. Like consciousness, quantum computation was a field that brought to mind spooky new age gurus, rather than hard science, until recently. But exciting new developments have led to actual quantum computations in surprising media. Recent developments in the field of quantum computation are discussed below.

This section discusses insights gained from attendance at two "Toward a Science of Consciousness" conferences. The third biannual conference was held in Tucson May 1998. An alternating year companion conference is being inaugurated in Tokyo in 1999. The Tucson conference papers are collected in the volumes edited by Hameroff et al (1996, 1998). In addition other professional-level conferences have sprung up to support research in consciousness.

Journals are also being started to support research in consciousness studies. Chief among these is the Journal of Consciousness Studies (http://www.zynet.co.uk/imprint/jcs.html). Research in this field is also reported in the full range of academic journals ranging from Physical Review Letters to Brain and Behavioral Sciences.

The current renaissance in consciousness got its impetus when several Nobel laureates lent their prestige to the field. Sir John Eccles (Nobel Medicine 1963) was the earliest to test the waters with his 1977 book co-authored with the philosopher of science Karl Popper: The Self And Its Brain. With Edelman’s (Nobel Medicine 1972) Bright Air, Brilliant Fire (1992) and Crick’s (Nobel Medicine 1962) The Astonishing Hypothesis (1993) the field was truly well launched. Francis Crick is still active in the field of consciousness studies collaborating with Cristof Koch of the Salk Institute.

Currently the most strongly debated question in consciousness studies is how hard the consciousness problem really is. The split here is between the members of the "classic" AI school and advocates of the "hard problem." The classic school argues that achieving consciousness is just a matter of running the correct program on an appropriate machine. See the classic book by Newell and Simon (1972). Lately the idea of an appropriate machine has been expanded from the von Neuman stored program computer to include neural networks (Churchland and Sejinowski, 1992). Hollywood has adopted these putative architectures so that the quintessential conscious weapon, The Terminator T800 Model 101 as played by Arnold Schwartzenegger (Figure 2), has neural chips for its brain.

If the classic, or functionalist, approach is correct then eventually computers as they are currently designed and constructed will become fast enough so that with clever programming they might achieve consciousness. The simulated entities in ModSAF or descendent CGF system will be cognitively indistinguishable from the human combatants being simulated.

Opposed to the classic approach, the hard problem school argues that there is something irreducibly hard about the problem of consciousness. The term "hard problem" was coined by philosopher David Chalmers (1996), but other philosophers such as Dreyfus (1986) have raised similar objections. Perhaps the most famous argument against computer consciousness is Searle’s Chinese room (1992), a thought experiment. In the Chinese room an English-only person carries out the steps of an algorithm for translating Chinese to French but has no understanding of what the steps accomplish. After accomplishing a translation the person has a French translation of the Chinese but has no idea of what the text may mean. Searle argues that this shows that algorithmic computers can only accomplish syntactic manipulation but never semantic understanding so that a computer can never be conscious. Searle’s arguments and other philosophical arguments have stirred what many computer scientists consider a tempest in a teapot. The computer scientists tend to ignore the philosophers and continue with classic research programs despite the presence of surprising figures such as Jaron Lanier (the originator of virtual reality) in the hard camp.

But allied with the philosophers, some physicists have come to think that there is more to consciousness than algorithmic computation on a von Neuman or other Turing-machine equivalent computer. These physicists typically draw attention to quantum physics, which has historically had a love-hate relationship with the conscious observer. For example, Eugene Paul Wigner (1963 Nobel Prize for Physics) thought that consciousness was absolutely necessary to make sense of quantum physics and posed many thought experiments illustrating this necessity.

More recently, Roger Penrose (1994) has written extensively on the connection between consciousness and quantum physics. These physicists, unlike the philosophers, do not just say computers cannot be conscious, but instead are proposing that a conscious computer must be based on quantum principles. This is a positive suggestion about what direction research into conscious machines should take. In the next section some current research into quantum-based computation will be briefly discussed.

The field of consciousness studies is rapidly developing and the "hard problem" and "classic AI" camps are not nearly as polarized as the above might imply. In particular the philosopher Dennett (1991) describes himself as a traffic cop, somewhere in the middle of the dispute. Other researchers think the neural net approach is not equivalent to a von Neuman computer when chaotic effects are properly taken into account. Other investigators approaching consciousness through chaos theory include Karl Pribram (1991) and Robertson and Combs (1995). Since both quantum and chaos-theory approaches involve the effects of small perturbations, it is likely that these two approaches will converge. The sensitivity of non-linear chaotic processes to small perturbations may provide the amplification needed to bring quantum phenomena to the macroscopic level.

The military applications of truly intelligent or conscious machines are fairly evident. Intelligent science-fiction-like weapons such as the Terminator leap to mind. However, even before consciousness is completely understood to the point of engineering truly intelligent machines, the insights gained in researching consciousness will pay big benefits.

For example it can be argued that one of the main roles of consciousness is in fusing the multitude of sense data that the organism receives. Dr. David Nagel of the Naval Research Lab who gave a post paper at Tucson III [Nagel, 1998] described how he thought of himself as an advance scout looking for emerging technologies for NRL. Sensor fusion is a difficult problem for military systems. Combining data from FLIR, visual, radar and other sensors is proving surprisingly hard. It is, however, a task that the lowliest animal performs with ease so research into consciousness can be expected to suggest new approaches and solutions.

Within the simulation community, the same problems are faced as with operational systems so that consciousness studies will be significant for simulation. As simulation becomes a more integral part of the acquisition process, the ability to fuse simulated data from a variety of sources will be crucial to achieving maximum fidelity.

In addition, simulation faces unique challenges not faced by the operational community. Training simulations must simulate human beings, the ultimate conscious animal, in the form of CGFs. The Army’s motto is "We train as we fight"; to meet this goal the simulated enemies must be just as savvy and wily as real human opponents. Anything short of full human consciousness will be found wanting for training purposes. While Deep Blue did defeat Gary Kasparov in the narrow domain of chess, Kasparov’s next move should of course to challenge Deep Blue to a game of poker, sending IBM back to a very expensive drawing board. This illustrates that humans are always constantly changing context and creating new challenges for their opponents. Nothing short of simulation of consciousness will serve to train as rigorously as a real fight.

In the previous section it was noted how many consciousness researchers think quantum computation is the key. Simulation should pay close attention to developments in quantum computation. Where simulation has ridden the exponential growth in computer power described by Moore’s Law (Clarke, 1995), quantum computation researchers believe a truly quantum computer will provide an unprecedented leap forward in ability to compute (Bennet, 1995). This "quantum leap" in computational ability may be necessary for implementation of truly intelligent CGFs, but in addition it will provide greater capability to simulate processes of all varieties (Lloyd, 1995). There are also interesting connections between quantum computation and other logic and computational paradigms (Clarke, 1998).

The Stanford-Berkeley-MIT-IBM Quantum Computation Research Project

In brief, quantum computation operates by utilizing qbits (quantum bit) instead of bits. A qbit is the state of a quantum entity like a nuclear spin that under the correct circumstances can be regarded as neither true or false, but as a combination, a superposition, of both. According to quantum mechanics, which has not been proven wrong since 1925, the qbit is really in both states. Properly utilizing this strange but true fact of nature enables the quantum computer to simultaneously execute both branches of a conditional. This ability to follow all branches of a computation simultaneously expands enormously the power of computers that use quantum effects.

As strange as quantum computers seem, even further out are the transpersonal and parapsychological investigators who also attend the Tucson conferences. While the first tendency may be to dismiss these as mystic touchy-feely nonsense, there may be some value in these investigations. For example one session at the latest Tucson meeting focused one lucid dreaming which is an "altered state" wherein you are asleep and dreaming but aware of the fact. Lucid dreaming can be learned and is sometimes assisted by certain (legal) drugs. Experiments have shown that subjects do indeed have volition during lucid dreaming. They can change their breathing pattern or give some similar signal to indicate that they are aware while still in REM sleep.

In addition to purely academic interest, phenomena such as lucid dreaming may have practical training implications. Other research on dreaming suggests that dreaming plays an assimilative role in learning. The brain replays the days activities during dreaming, editing and organizing data for permanent storage or for discard. The insights being gained by lucid dream researchers may thus provide techniques for enhancing the retention of training by modifying how the sleeping brain processes the training data.

To date the nature of consciousness has remained elusive. Only recently have scientists begun to discuss the nature of consciousness. From the time of Descartes to the early 1990’s, science had regarded consciousness as the domain of philosophers and mystics, but this is rapidly changing and will lead to developments that cannot be ignored by the simulation community.

It is now becoming clear that the many separate aspects of intelligence modeled by AI, planning, target recognition, sensor fusion, etc, can only be integrated with a scientific understanding of consciousness. In addition to aspects of intelligence normally associated with AI, consciousness research suggests consideration of affective aspects such as emotion. This line of research may lead to more effective smart bombs that "want" to hit their target and will do whatever it takes to reach that goal.

Within the simulation world consciousness research should result in methods for building more robust and capable CGFs for the 21^st century. Sensor fusion techniques should also enhance design simulation making simulation more widely applicable in the design process. Quantum computation, in part motivated by consciousness research, is poised to revolutionize the art of computing and thereby of simulation.

Thus there should be a synergy of simulation with consciousness research that will yield very valuable fruit during the 21^st century.

Bennet, C. H. (1995): "Quantum Information and Computation," Physics Today, Vol. 48, No. 10, pages 24-30; October 1995.

Chalmers, David J. (1996): The Conscious Mind: In Search Of A Fundamental Theory, New York: Oxford University Press, 414 pp.

Churchland, Patricia Smith, and Sejnowski, Terrence J. (1992): The Computational Brain, Cambridge, Mass: MIT Press, 544 pp.

Clarke, Thomas L (Ed) (1995); Distributed Interactive Simulation Systems for the Simulation and Training in the Aerospace Environment; SPIE, Bellingham, WA. 327pp.

Clarke, Thomas L. (1998): "A Comparison of Linear Logic with Wave Logic"; Fifth International Symposium on Artificial Intelligence and Applied Mathematics.

Crick, Francis (1993): The Astonishing Hypothesis: The Scientific Search For The Soul, New York: Scribner, 317 pp.

Dennett, Daniel C (1991): Consciousness Explained: Boston, MA: Little, Brown and Co, 511pp.

Dreyfus, Hubert L., Dreyfus, Stuart E. and Athanasiou, Tom (1986): Mind over machine: the power of human intuition and expertise in the era of the computer: New York: Free Press, 231pp.

Edelman, Gerald M. (1992): Bright Air, Brilliant Fire: On The Matter Of The Mind, New York, N.Y.: BasicBooks, 280 pp.

Gershenfeld, Neil and Chuang, Isaac L. (1998): "Quantum Computing with Molecules," Scientific American, Vol. 276, No 6.

Hameroff, Stuart R., Kaszniak, Alfred W, and Scott, Alwyn C. (1996): Toward a Science of Consciousness, The First Tucson Discussions and Debates: Cambridge, Mass: MIT Press, 790pp.

Hameroff, Stuart R., Kaszniak, Alfred W, and Scott, Alwyn C. (1998): Toward a Science of Consciousness II, The Second Tucson Discussions and Debates: Cambridge, Mass: MIT Press, 764pp.

Lloyd, Seth (1995): "Quantum-Mechanical Computers," Scientific American, Vol. 273, No. 4, pages 140-145.

Nagel, David J, (1998): "Quantification of consciousness," Tucson III, Toward a Science of Consciousness, poster paper.

Newell, Allen and Herbert A. Simon, Herbert A. (1972): Human problem solving, Englewood Cliffs, N. J., Prentice-Hall, 920 pp.

Penrose, Roger (1994): Shadows Of The Mind: A Search For The Missing Science Of Consciousness, Oxford: New York: Oxford University Press, 457pp.

Popper, Karl R. and John C. Eccles, John C. (1977): The Self And Its Brain, New York: Springer International, 597 pp.

Pribram, Karl H. (1991): Brain And Perception: Holonomy And Structure In Figural Processing, Hillsdale, NJ: Lawrence Erlbaum Associates, 388 pp.

Robertson, Robin, and Allan Combs (Eds.) (1995); Chaos Theory in Psychology and the Life Sciences; Lawrence Erlbaum Associates, New Jersey; 399 pp.

Searle. John R. (1992): The Rediscovery Of The Mind, Cambridge, Mass: MIT Press, 270pp.