• Contribute
• What's New
• Submit News

ABSTRACT

This paper addresses the question of what the relationship may be between conscious experience and the reality which underlies it and attempts to describe what it would mean for a machine to have its own subjective (visual) experiences.

It is uniquely true of the brain (in itself a percept) that if it is damaged, then many other percepts are degraded. This implies either the brain percept causes all other percepts (including presumably itself) or that whatever it is (outside consciousness) that causes my brain percept, somehow controls the generation of, or provides access to, all my other percepts. This second hypothesis seems more viable. There is, logically, little point in expecting examination of any percept (even the one known as ‘firing patterns in the brain’) in order to understand the creation of percepts. Neural correlates can never provide a direct explanation of consciousness.

Consciousness can be thought of as providing a summary of priorities, extracted from the tidal wave of incoming sensory data which are needed to allow human-like flexibility of behaviour. For machines to have similar levels of flexibility, they will need to undertake such a prioritization based on an effective summary.

The idea that consciousness is made up of sequences of very simple instantaneous percepts may make discussion of the “qualia problem”, i.e. whether we all perceive the same thing which I call redness, somewhat easier. A complete answer to the qualia problem for machines can, it seems, only be obtained if nested worlds of experience can be generated for machine and human observer. This acid test of machine consciousness seems logically impossible to provide.

We can’t prove that another human is conscious and the same will be true of machines. It can however be envisaged that by setting up a number of qualitatively different, ordered scales, something akin to the richness and meaningfulness of human consciousness might be achieved within a machine. If we design machines to automatically apply significance-based ‘weights’ for the prioritisation of events, then they can be thought of as having conscious, even emotional, experiences: -‘quantia’ can become qualia.

1.INTRODUCTION

Nature's fundamental laws don’t govern the world as it appears in our mental picture in any very direct way but instead they control a substratum of which we cannot form a mental picture without introducing irrelevancies.

Pavlov used to fine his students if they used words like ‘voluntary’ or ‘conscious’. The field of consciousness research has become scientifically respectable only in recent years [1] -despite being the subject of longstanding philosophical debate. Many of the major theories of visual perception (e.g. ecological optics, gestalt, empiricism/constructivism) have implicitly dealt with consciousness. Those aspects of consciousness which centre on visual experience will be emphasised in this discussion. It is assumed here that the experience of visual phenomena is similar, in its underlying character, to the experience of pain or thirst or sadness.

Consciousness has thus been thought to be fundamentally different from other phenomena studied by the natural sciences because everything else that we know of has an objective existence (i.e. all other nameable things than consciousness seem to exist largely independent of anyone’s subjective experience of them). This paper addresses the question of what the relationship may be between conscious experience and the reality which underlies it and attempts to describe what it would mean for a machine to have its own subjective (visual) experiences.

It is not particularly controversial to assert that much of a sighted individual's conscious experience is concerned with visual sensations. Understanding, in an objective way, how the brain can perform visual identification, for example, is usually taken to mean being able to chart (or better, model predictively) the behaviour of neurons and correlate this with the performance in matching a given image with one from a large set of similar examples. It seems reasonable in principle to say that when making changes to the shape a letter A, we could monitor the corresponding changes in neuronal firing and thus deduce which aspects of firing ‘went with’ which aspects of the A's shape. This would provide a form of understanding of the visual recognition process: a kind of neural behaviourism. Such an objective approach is the way in which understanding the visual part of conscious experience vision has traditionally been dealt with by neurophysiology [2],[3]. Science would be significantly advanced by even an approximate understanding of which collection of neurons, pulsing in a particular way, stands for a capital letter A. This kind of understanding, and pattern-recognising machines based on it, can be achieved without addressing directly the questions of consciousness (see for example, www.foveola.com).

In fact, this powerful, objective approach has allowed investigation of the deeper problems of consciousness to be deferred. It seems worthwhile now, however, to present some thoughts on how to pose these deeper questions; at least in order to make more explicit the degree to which they may be experimentally falsifiable.

2.WHAT IS CONSCIOUSNESS?

Many basic questions exist about consciousness: what is it for?; is it the same as self awareness? [4]; why do neurons fire when percepts occur and percepts cease when neurons are damaged?; is it possible to communicate the exact contents of my consciousness to others? (It can be argued that the primary function of both verbal language and Art is to allow an approximate version of this to occur). In his theorising about the subject, Crick says anything we can describe explicitly (and presumably objectively) can be explained [5].

If machines are ultimately each to be equipped with their own degree of subjectivity, via machine consciousness, we need to find a way of describing what consciousness is in order to design the required processes (Note that as machines acquire better and better-integrated sensors interfacing to the world, they may well develop personal databases of e.g. shapes. These can’t surely in themselves be ‘experiences’ of the type which humans have.)

Craik was among the earliest scientists to make clear the requirement for a level of explanation to be specified when a scientific solution is sought [6]. We can explain the choice which someone makes, say, to eat an apple in terms of different levels -psychological, neurocellular, neurochemical etc. One day, we may achieve a form of explanation of human behaviour by using a supercomputer to predict Human actions: based on every level higher than quantum fluctuations. This reductionist process might, by the way, be inherently limited by our ability to understand very complex causal relations across levels but even a 'perfect' model of this type could not, it seems, shed (objective) light on the experience of any individual.

This is because none of our senses or experimental techniques allows us access to the underlying reality -we therefore have no way to experience it directly, in principle.

This idea usually surfaces in the form of a nagging difficulty with the central experimental observation that there is a correlation between activation of certain neurons and the perception of an object. How can my apparently rich visual experience be caused by, or even equated with, a cellular firing pattern occurring 'in my head'? i.e. if some process is causing my percept <aeroplane> what is that something [7]?

Figure 1. Conscious perception by analogy to signal transmission.

In Figure 1, what might be described as an established view of conscious visual awareness, as a form of signal transmission, is pictured. Here, light from an object in the real world, an aeroplane, is captured and processed by another real-world object i.e. neural firing patterns in the brain, in order to generate a perception of the aeroplane. Sherrington describes the passage of information in light traveling from a star to the cortex; but, he continues, “at this point the scheme puts its finger to its lips and is silent” [8].

This, perhaps the deepest mystery of neurophysiology/psychology, is addressed by some of the most eminent scientists simply by saying that the conscious experience of thought is neuronal activity and that this idea must simply be accepted (Many other counterintuitive ideas have become accepted within Science but usually only after some experimental verification e.g. Special Relativity[9]). Crick describes this astonishing hypothesis “You...are in fact no more than the behaviour of a vast assembly of nerve cells and their associated molecules. [1]. This neural correlate of consciousness continues to be sought avidly by experimentalists. For example, neurons in various regions of the visual cortex are known to project their results backwards and forwards and this has been suggested as a candidate mechanism for the neural correlate in question [10]. Consciousness is thought by a number of authors to be connected with spatially separate but synchronised neuronal firing [11]. Barinaga describes William Newsome's work with monkeys trained to react in a given way when responding to a rising stimulus. The monkeys can react in this way when the right neurons are stimulated, even though the stimulus is actually falling. These fascinating experiments are in principle unable to determine whether a monkey actually experiences upward movement or if it is just somehow caused to report it, despite actually seeing downward movement [12]. Zeki has found neurons in v4 which fire only when the subject reports seeing a red patch (irrespective of the actual wavelength emitted from the patch [13] ).

None of these neural correlates can ever provide a direct explanation of consciousness, however.

Although the idea of the Cartesian Theatre is almost universally rejected, the view that the brain breaks down the visual image and then reassembles it, is (wrongly) encouraged by misinterpretations of studies of e.g. neural channels.

The perceptions of human experimenters and the data their sensors record all have to be consistent. This contributes to the view that all such perceptions are fundamentally real, when in fact they all form a ‘pre-digested’ view of underlying reality. We can never be directly aware of this underlying world.

This difficulty seems to be exacerbated by our understandable tendency to assign priority to familiar, rich subjective experiences. In fact, both the percept of an aeroplane and the percept of the cortical firing pattern which occurs at the same time are just that: percepts...neither has a greater claim to being real or causal than the other. When anyone asks what is causing my percept of my mother's face, the answer is traditionally “a specific neuronal firing pattern within my brain” i.e. to use one particular percept to explain the production of all other percepts...which seems deeply unsatisfying.

Figure 2. Conscious perception as a transform.

An obvious alternative hypothesis, illustrated in Figure2, is that the brain (like everything else we perceive) is a creation within one's world of experience, as represented by the large oval, from an underlying reality (the lower rectangle).

It is uniquely true of the brain (percept) that if it is damaged, then many other percepts are degraded. This implies either the brain percept causes all other percepts (including presumably itself) or that whatever it is (outside consciousness) that causes my brain percept, somehow controls the generation of, or provides access to, all my other percepts. This second hypothesis seems more viable.

For the purpose of this discussion, things which are conscious events or percepts are shown bracketed thus < >, whilst the corresponding real-world entities are bracketed in this way: [ ]. It is by no means clear that a one-to-one mapping exists between < > entities and [ ] entities: indeed, for consciousness to act as a simplifying, summarising process, a several-to-one relationship may be mandatory. Here, a one-to-one organisation is assumed, in order to simplify the discussion.

Consciousness, then, as described here, is a process of transforming [underlying reality] into . One such transform is marked “b” (the one which generates the percept known as “my brain”.

All of traditional Science can be thought of as occurring quite consistently and coherently within the oval region in Figure2. This Figure expresses something not unlike Kant’s view: “Things which we see are not by themselves what we see...it remains completely unknown to us what the objects may be by themselves and apart from the receptivity of our sense. We know nothing but our manner of perceiving them”(14). Similarly, Burne quotes Ramachandran –“Could it be that we should think of our bodies as self-projected constructs, an integral part of our minds, rather than as something out there?” and in other words “...the body is a creation of the mind” [15]. Harman states the same case straightforwardly - that the perceived world is part of consciousness: it cannot be the thing itself [16].

To extend the explanation of the consciousness phenomenon to a deeper level, it may be necessary also to be able to understand the link between whatever underlies the percept <aeroplane> -symbolised by [the right-hand blob] and whatever underlies the corresponding firing pattern -symbolised by [the left-hand blob]. This type of link is shown in Figure2 by a process marked “a".

Science has established many causal links between physical entities of which we can't be directly conscious but about whose properties and behaviour indirect evidence can be gathered. We can postulate that the effects we see represented by our instruments are caused by e.g. subatomic entities and develop consistent theories, testable within our world of experience, of how these unperceived things, i.e. outside consciousness, interact. In the last century, Physics has even begun to recognise that the objective and subjective cannot be separated: the observer must be included in the observation [17].

It may not be possible to form a testable theory about process “a” i.e. the interactions between the [unperceived thing(s)] underlying <aeroplane> and those which underlie the corresponding <neural firing pattern>. This is due to the fact that normally the way to understand any transform process experimentally would be to feed in inputs, monitor outputs and try to induce their relationship. In the “a” process case, however, we have no direct contact with either input or output.

A scientific understanding of whatever mechanism (The “b" process) creates <aeroplane> from [whatever underlies aeroplane in reality] seems intractable but we do at least have access to the output in this case. Some ways to get to grips with this question may include experimental investigations at the edge of consciousness [18],[19] or via attempts to build machines which may be designed to undertake such a “b” process transform for themselves. The construction of demonstrably conscious machines would be of enormous significance [7].

3. WOULD MACHINE CONSCIOUSNESS BE USEFUL?

Aspects of behaviour which aren't detrimental to reproduction and which occur at little genetic expense to an individual, tend to be largely ignored by natural selection[20]. The ability to experience things via consciousness might even therefore occur purely as a by-product of neural processing: one which provides no advantage over an automaton or Zombie. Even such an unconscious machine would probably require many sensory subsystems in order to maintain and reproduce itself in a complex world. Any algorithm written to link its sensory inputs to its behavioural outputs would need to be fast, adaptive, and correspondingly complex. The environment might well act in such a way as to result in conflicts between whether to eat or run, mate or fight. Thus, for higher organisms which may generate many combinations of signals from a number of distinct sensory modalities, the need for intelligent data processing is much more pressing than in the case of an earthworm, for example.

A machine could, of course, be designed to avoid collisions, seek out fuel etc without experiencing any of these processes (or at least without having been deliberately designed to have experiences). Thermocouples and strain gauges could produce synthetic specific nerve energies [21]. In this way, we could, theoretically, have identical behaviour in two different systems, one conscious and the other unconscious (22) (Problem solving can of course occur even in humans without full consciousness...Kekulé is reputed to have discovered the ring structure of benzene whilst dreaming).

We know that, among many other things, consciousness provides us with information in the form of colourful visual images, emotionally-significant perceptions of music, warnings of possible bodily damage in the form of pain, and an ‘inner voice’ in which we can try out opinions before we hear them expressed. Consciousness is also often thought of as performing some kind of strategic role. Reflexes can cope with split-second reactions and stored programs can undertake skilled movements when appropriate. There is still however a need for tactical or strategic decisions to be made which will affect the chances of an organism’s surviving the next ten minutes (taking into account the goals and attitudes which may constitute an individual personality). Such a process might act as a preview or filter or brake on combinations of stored behaviour programs. Minsky suggests that separating off a part of the brain, as an internal agent of consciousness dedicated to observing and instructing the others, would have significant survival value [23]. In a similar vein, Carpenter puts forward the possibility that consciousness may be merely a spectator watching the preparation for action and fooling itself that it is in control of behaviour [24]. Some experimental support for this kind of idea has come from e.g. the work of Libet who has claimed that awareness of a wish to perform some act arises only after cerebral processes driving muscle activation have begun [25].

The question is, could humans have survived as a species of Zombies; using qualitatively uniform neural codes without the apparent richness of consciousness? Holmes, [26] reports that vision for action (i.e. visuomotor processing) such as grasping is indeed unconscious. Vision for perception, according to this interpretation, is however, conscious. It provides visual descriptions of objects which are largely independent of variations in lighting and perspective using inbuilt assumptions about the raw data. If one sees a small elephant, one assumes it's far away. Hills actually look measurably steeper if one is tired, so one sees not just the hill but also an encoding of the anticipated effort required. Furthermore, our perception of the sizes of 3-D objects and components often contain inconsistencies, which we term illusions [27]. We tend to neglect many small-scale erroneous summaries which appear in consciousness. One example is the limit on instantaneous counting of up to 4 dots in a line [28]. Even more remarkable is the phenomenon of blindsight in which individuals cannot see consciously objects in a scene but, if forced to guess, can successfully indicate their position [13].

Perhaps the answer to this question relates to the fact that such a Zombie/automaton would have to have a phenomenally fast processor in order to produce an environmental summary, of the type which consciousness seems to provide, which would be as effective in guiding survival decisions. This is especially true given the tidal wave of incoming sensory data which are needed to allow human-like flexibility of behaviour.

Humans have a very varied repertoire of possible sensations. This variety of states makes consciousness seem even more intractable ([29] see pg143 “one's visual world is shot through with hyperimmense richness"). It certainly seems that conscious states are discrete; they don't blur to form a continuum. There are some apparently intermediate states where e.g. laughing and crying are intermingled but increasing pain never normally becomes green.

The apparent continuity of the world of one's experience is particularly suggestive of many simultaneous perceptual activities being seamlessly knitted together. This poses the ‘binding problem’ for neurophysiologists who have discovered simultaneous activity in remote clusters of cortical cells. As Zeki describes it, if one set of cells is firing in response to the redness of a shape and another set firing in response to the shape itself, how can these disparate firings contribute to the overall conclusion "pillar box” [13]?

Of course, it might be suggested that the apparent richness of our perceptual world is, in itself, a form of illusion. Freedom to inspect, sequentially, all parts of the visual scene inspires (surely groundless) confidence that we have a complete view of reality at all times (even allowing for the fact that gaze control is not attention itself [30]). Similarly, the high frequency with which we can cycle between the large number of available of percepts creates the questionable belief that we can be aware of several at once. The entire experimental literature on attention and perceptual rivalry implies that this is not so [31],[32]. Several authors have written that each conscious experience is singular in time [33],[34] (i.e. one can't solve physics problems when one has a toothache). One can be unaware of even serious injury when flight is the priority.

There is evidence that the human visual system can provide efficient vision by deliberately forming simple, sequential percepts; each corresponding to a very small visual angle [35]. The suggestion is that perhaps conscious events are actually quite uncomplicated individually but that the many possible sequences contribute to the richness of conscious experience. Calvin, [36], proposes that there may be a processing core buffer just big enough to hold seven "chunks" of information, as suggested by the experiments of Miller [37]. Each of these chunks, or simple percepts, might be considered the instantaneous result of the “b”-type transform indicated above.

Consciousness can therefore be viewed as a fast, fluctuating series of comparatively simple perceptions (transforms from the real world) indicative of a moment-by-moment conclusion about what is the most important reaction to current events. This seems to be supported by the finding that severance of the corpus callosum can result in the production of separate consciousnesses, [38],[39]. Dennett comes, via different thinking, to the similar idea that the brain is a hypothesis making machine throwing up new drafts. Mental states become conscious by winning the competition to control action [40].

Figure 3. Two perceptions of the same object.

Figure 4. Confirmation of machine consciousness.

4. THE QUALIA PROBLEM FOR MACHINES

It is hard even to estimate the number of actual instantaneous experiences which are possible for a Human; but the number is almost certainly finite. A machine could be built which might register a distinct internal state, as a number, say, from 1 to 1 M, in response to things we experience as hot, dry, triangular, frightening or salty etc. These values could be described as “quantia”; but they would fail to provide the qualitative experience which the words suggest to a person. Even if we could arrange that a particular firing pattern in a brain would trigger a certain value within the machine and vice versa, this would be merely correlation, not machine consciousness, as discussed above.

It has been proposed that an instant of consciousness may be a comparatively simple event, which might be described crudely as a bitstream of firing and non-firing neural elements i.e.

<100100001001111111100011001………110101111010……………..1101001001010101>.

Making use of this representation of the event of the moment i.e. linking it to some action program might be extremely difficult to perform swiftly using such a distributed representation. To speed up the process, we could recode (the same number of bits) in terms of a much smaller number of separate registers < A , 6 ...> .

Note that each of the proposed registers still contains elements from a finite set of values (e.g. 26 or 10, here) but that now each of these registers has a qualitatively different scale (letters vs. numbers). This is similar to the apparently inefficient codes which government departments use in which references to documents are overspecified to help portray the meaning of their contents, at a glance. Each register's instruction set is now bigger, but the message in terms of symbols is smaller and therefore faster to transmit and link to other data. This is reminiscent of a complex-resident-instruction-set machine (the opposite of a RISC machine) i.e. one which can recognise and execute quickly a bigger repertoire of standard instructions in which each instruction is comparatively complicated -and therefore ‘means more’ by itself. Thus it can be envisaged that by setting up a number of qualitatively different, ordered scales, something akin to the richness and meaningfulness of consciousness might be achieved within a machine. If we design machines to automatically apply significance-based ‘weights’ for the prioritisation of events, then they can be thought of as having conscious experiences –quantia become qualia. According to the present discussion, Dennet’s competition between drafts corresponds loosely to a mechanism in which complex multidimensional sensory events are automatically rated in terms of significance. These significance weightings may be viewed as equivalent to what we experience as emotional states.

5. PROVING THAT A MACHINE IS CONSCIOUS

The question of proving whether a machine is conscious remains particularly difficult. If, as proposed above, consciousness occurs one simple percept at a time, the world of anyone's experience can contain only one thing at any time. Figure 4 illustrates that nesting is necessary to be sure that a robot is conscious. If my world of experience can contain only one item at a time, however, to envelop the consciousness of the robot would require that it would be the only thing of which I could be conscious during the demonstration. Thus, paradoxically, one would have to literally become the robot to demonstrate its consciousness. We can’t prove that another human is conscious and the same will be true of machines.

Any attempt to explain consciousness must surely provide a method of assigning locations on the same consciousness scale to earthworms, infants, brain-damaged patients and the partially-anaesthetised. The present scheme would suggest that earthworms have fewer combinations of sensory inputs, fewer possible internal states to generate by transform, a narrower range of register types and therefore enjoy much less varied consciousness than a human. Machines however could be designed with an arbitrarily large number of independent sensory modalities and tools to quickly evaluate their combinations. Such evaluations may even be taken to constitute the machine’s emotional states.

6. REFERENCES

[1] F. Crick, The astonishing hypothesis. The scientific search for the soul. London: Touchstone Books, 1995.

[2] P. Andrews, Implications for machine vision of a natural recognition algorithm, US Army Research Institute, 6881-R8-06, 25.8.921992.

[3] R. DeValois and K. DeValois, Spatial Vision. Oxford: Oxford University Press, 1990.

[4] G. Gallup, Chimpanzees: self-recognition, Science, vol. 167, pp. 86-87, 1970.

[5] F. Crick and C. Koch, Some reflections on Visual awareness, presented at Cold Spring Harbor Symposium on Quantitative Biology, New York, 1991.

[6] S. Sherwood, The nature of psychology: a selection of papers, essays and other writings by the late Kenneth JW Craik, : Cambridge university press, 1966.

[7] I. Aleksander, Impossible minds, London: Imperial College Press, 1996.

[8] P. Fenwick, The neurological boundaries of reality, in Alterations in consciousness awareness, E. Critchley, Ed. London: Farrand Press, 1993.

[9] H. Barlow, Understanding natural vision, presented at Physical and biological processing of images: proceedings of an international symposium organised by the rank prize funds, London, 1982.

[10] C. Koch and I. Segev, Methods in Neuronal modeling - From synapses to networks, , C. Koch, Ed. Cambridge, MA: MIT Press - Bradford Books, 1989.

[11] F. Crick and C. Koch, Towards a neurobiological theory of consciousness, Seminars in the neurosciences, vol. 2, pp. 263-275, 1990.

[12] M. Barinaga, Visual system provides clues to how the brain perceives, Science, vol. 275, pp. 15831585, 1997.

[13] S. Zeki, A vision of the brain, Oxford: Blackwell Science, 1993.

[14] I. Kant, Critique of pure reason, 1978 translation ed. London : Macmillan, 1781.

[15] J. Burne, Brainpower, in The Sunday Times, London, 1997.

[16] W. Harman, "The scientific exploration of consciousness: towards an adequate epistemology, Journal of consciousness studies, vol. 1, pp. 140-148, 1994.

[17] R. Penrose, The Emperor's New Mind, Oxford: Oxford University Press, 1989.

[18] P. Andrews and F. Camp bell, Images at the blind spot, Nature, vol. 353, pp. 308, 1991.

[19] R. Gregory, The (usually) intelligent eye," in What is intelligence?, J. Khalfa, Ed. Cambridge: Cambridge university press, 1994.

[20] R. Dawkins, The selfish gene, New York: Oxford university press, 1978.

[21] J. S. Albus, Brains, behaviour and robotics, Peterborough,NH: Byte Books, 1981.

[22] J. Searle, Is the brain's mind a computer program?, Scientific American, vol. 282, pp. 20-25, 1990.

(23] M. Minsky, The society of mind, New York: Simon and Schuster Inc, 1988.

[24] R. Carpenter, Neurophysiology: Edward Arnold, 1991.

[25] B. Libet, Cerebral processes that distinguish conscious experience from unconscious mental functions," in Principles of design and operation of the brain, Experimental brain research supplement, J. Eccles and M. Wiesendanger, Eds., 1989.

[26] B. Holmes, Irresistible illusions, New scientist, pp. 32-37, 1998.

[27] S. Tolansky, Optical illusions, Oxford: Pergamon Press, 1964.

[28] J. Atkinson, F. Camp bell, and M. Francis, The magic number 4+/-0: a new look at visual numerosity judgements, Perception, vol. 5, pp. 327-334, 1976.

[29] P. Churchland and T. Sejnowski, The computational brain, Cambridge,MA: MIT press, 1992.

[30] S. Tanimoto, A. Buizza, C. Marzi, M. Savini, and S. Vitulano, Allocation of attention in vision, in Human and machine Vision, V. Cantoni, Ed. New York: Plenum Press, 1994.

[31] R. J. W. Mansfield, International symposium on attention and performance, 1981.

[32] J. Atkinson and F. Campbell, The dependence of monocular rivalry on spatial frequency, Perception, Ed. Cambridge: Pion, 1973.

[33] S. Greenfield, Journey to the centers of the mind, Oxford: WH Freeman & Co, 1995.

[34] T. Cornsweet, Visual perception, London: Academic press, 1970.

[35] P. Andrews, Biometrics: the mathematics of identification, presented at Second IMA Conference on Image Processing: Mathematical Methods, Algorithms and Applications, DeMontfort University, 1998.

[36] W. Calvin, How brains think, London: Wiedenfeld & Nicholson, 1997.

[37] G. Miller , The magical number seven, plus or minus two: some limits on our capacity for processing information, Psychological reviews, vol. 63, pp. 81-97, 1956.

[38] S. Blackmore, Consciousness: science tackles the self, New Scientist, 1989.

[39] E. Schwartz, Anatomical and physiological correlates of visual computation from striate to inferotemporal cortex, IEEE trans. syst. man cybern., vol. SMC-14, pp. 257-271, 1984.

[40] D. Dennett, Consciousness explained, Little, brown, 1991.

[41] T. Nagel, What is it like to be a bat?, in Readings in Philosophy of psychology, vol. 1, N. Block, Ed. Cambridge, MA: Harvard university press, 1980.

CONCLUSIONS

There is little point in expecting examination of any percept (even the one known as firing patterns in the brain) in order to understand the creation of percepts. Neural correlates can never provide a direct explanation of consciousness.

It may be possible to obtain some further understanding of the design requirements for machine consciousness from experiments on human subjects in which the preceptual outputs are available for measurement i.e. via traditional psychophysics.

We can’t prove that another human is conscious and the same must, logically, be true of machines, because nesting consciousnesses is in principle impossible without radically changing the status of the observer.

The idea that consciousness is made up of sequences of very simple instantaneous percepts (transforms from the real world) may make discussion of the “qualia problem”, i.e. whether we all perceive the same thing which I call redness, somewhat easier.

Consciousness can be viewed as a rapidly fluctuating series of comparatively simple perceptions indicative of a moment-by-moment conclusion about what is the most important reaction to current events. This might be closely simulated by having a machine automatically assign weights to multidimensional sensory data, which could be thought of as essentially its emotional reactions.