[Help] [Aide] [Up]

Science Tribune - Article - May 1997

http://www.tribunes.com/tribune/art97/peru1.htm

Neural networks, quantum systems, and consciousness



Mitja Perus

National Institute of Chemistry, POB 3430, 1001 Ljubljana, Slovenia and Society for Cognitive Sciences, Slovenia.
Website : http://kihp6.ki.si/~mitja/index1.html
E-mail : mitja.perus@uni-lj.si.


Summary

Could hybrid "neuro-quantum" models help explain consciousness ? This question is inspired by mathematical analogies between quantum theory and models of associative neural networks. We indicate the gaps in studies on consciousness that quantum physics might fill.

Those readers unfamiliar with this topic might like to read our introductory comments.
__________________________________________________________________________________________


Over the last few years, both scientists and laymen have become increasingly interested in studies on consciousness in the belief that this state of being could be explained by modern computation and modelling techniques.


Brain processes as explained by neural networks

With the development of neural network theory, the mathematical basis needed to model processes as complex as those of the human brain and to simulate these processes in computers has become available. Neural network models successfully account for (1) :
- recognition, storage and recall of patterns-qua-attractors (a),
- associative connections and contextual relationships among ideas,
- extraction of novel stimuli from amidst known stimuli,
- optimal condensed storage.

Neural nets can also partly explain the background processes involved in attention, meaning and higher thought processes (intuition, reasoning, and also language). In addition, because the neurons of a neural net interact, a neural net can self-interact and thus account for our mind's interaction with itself, in other words, for information on our own thoughts.


A role for quantum processes ?

However, neural nets alone probably cannot explain consciousness. Unlike neural nets, we can also feel the qualities (qualia) of the objects and ideas we perceive, i.e., colours and smells, or pain and pleasure. These feelings form a relatively stable and unified experience. Even though a neural net may recognize objects automatically, it cannot answer for a unified experience (2); the holistic aspect of consciousness (3) evokes a global organizational process. For example, although we can locate the memory patterns encoding objects we have perceived within certain large, ill-defined areas of the brain, we cannot locate the conscious experience of these objects. Our experience is, of course, related to the patterns encoding these objects but it is also 'something more' and as yet unknown that embraces individual and social mental life in its natural environment.

Quantum systems might account for this 'something more' by creating what could be described as a unified 'sea' of processes in which concrete objects are sunk, just like our sensations are embedded in consciousness. They would form a low-level backdrop that is common to all material processes and might also incorporate subtle unidentified information processes.

However, consciousness is not explained by quantum systems alone just as it might not be explained by neural networks alone. What may be required is a delicate interplay of several different types of processes (neuronal, subcellular and quantum) each with its own hierarchical organisation of information (4).


Mathematical analogies between neural networks and quantum systems

The whole system, and not just its component parts, is needed to process information and, in this respect, our knowledge of neural-network-processing at a macro-level may prove handy in the search for mind-like information processing systems at more subtle quantum levels.

The mathematical models and computer simulations of neural networks have enabled pioneering research into the complex-systems-dynamics that could underlay consciousness. For instance :
- These models can reproduce the information processing that is characteristic of the microstructure of cognition (5).
- Hierarchical neural networks are suitable for modelling early perception.
- Associative neural networks (networks of bi-directionally densely connected formal neurons, which might also represent modules of neurons within the brain cortex) are appropriate models for associative processes. They manipulate neuronal patterns (a) that have been preprocessed by hierarchical neural networks and that act as attractors (a) within the dynamic system.

A comparison of the mathematical formalism of associative (or attractor) neural networks, on the one hand, and of quantum theory, on the other, reveals significant similarities in the structures of the basic equations (6). In mathematical terms, the complex-systems-dynamics of the quantum world is very similar to the collective dynamic processes of neurons connected by synapses. Thus, the interactions within neural and quantum networks obey similar general laws. The dissimilarities lie in the differences in the internal structure of neurons and of quantum particles.


The unity of quantum systems

Quantum systems have a more holistic nature (7) than neural networks; they act as an indivisible whole (for experimental evidence, see (8)). Whereas they retain most of the information processing capabilities of neural networks, they can also merge patterns together. Quantum unity is more profound than the unity of the virtual attractor (a) emerging in all complex systems and is deeper than the coherence of neuronal 40 Hz oscillations (b). It is not only a new category but a melting-together of particles (Bose-Einstein condensation). Could it be related to the unity of conscious experience? (2)

Human beings do not experience network components nor their exchange of signals. They experience the objects which are represented by patterns-qua-attractors (a) within the multilevel networks of the brain as "real" objects that are part and parcel of a conscious unified image. An essential feature of the quantum world is that particles are not only connected but "smelted". Thus, understanding the interplay of quantum particles with an all-embracing whole may help us understand how perceptions are bound together.

If we wish, we can take this even a step further. Claims of mystical, meditational and parapsychological experiences suggest that consciousness might not be confined to the neural brain. The conscious minds of individuals might be weakly connected into a "collective consciousness" - of an intuitive (not cognitive) nature - within the quantum world (2) that gives rise to all physical processes including the material objects of classical physics (9).


A question of observation scale

The question would thus no longer be whether quantum systems are relevant to research on consciousness but to what extent do they contribute to consciousness. Could anything prevent the quantum world from affecting its offspring, the classical world? The division into classical and quantum complex systems is artificial and primarily a question of observation scale and not of function. The collective processes of both these systems are described by coupled equations of dynamics, coherent oscillatory phenomena, uncertainty principles, pattern-extraction processes (the recall of a neuronal pattern or the 'collapse' of quantum wave-function (10), respectively) etc.


Why have a fractal-like multilevel brain structure to process information ?

For consciousness, quantum systems might need to be triggered and organized by neural networks that regulate the coding/decoding of information from/to the macroscopic environment.

Because the 'sociology' of an entire neural network, and not the 'anatomy' of its parts, underpins information processing, the principles governing neural networks could be applied to the fractal-like (c) biophysical networks of the brain. (We speak of a fractal-like brain structure because of similar collective dynamics that might occur at different levels). Several biophysical networks are potential candidates for processing information in the way that neural networks do (e.g. dendritic networks, the molecular networks of microtubules, photonic and other quantum networks ..... (11) (12)). Crucial questions that arise are : What is the division of labour among these networks existing on different scales ? (13); why does the brain need so many levels? Probable answers are : Cognition is an encoding-decoding process that is most effectively performed by a multi-layer complex system. Since this system evolved gradually over many tens of thousands of years, it may be partially redundant with structural repetitions and duplications.


The backdrop to consciousness : a multilevel cooperation between neural and quantum processes ?

A fairly simple illustration of cooperation between neural and quantum processes is given by conscious (not cognitive) visual perception which can be best described by the quantum holographic (d) model (14). The mathematical principles of this model are very similar to interference in quantum or neural networks. A dynamic optical (quantum electromagnetic) network between the eyes and the objects incorporates information flows. The brain's hierarchical neural network behind the eyes "mechanically" perceives the object-images offered by the photonic network (light) but cannot produce the experience of objects located in an external environment. Only when cooperating with the quantum-optical (i.e., photonic) network in the visual field, does the brain's network project the perceived virtual image back to the object's original location.

Thus, if neural networks are a macroscopic replica of quantum processual structures, they could be an interface between the macro-world of man's environment (and man's motorics) and the micro-world of his non-local consciousness, a consciousness based on quantum processess. This is why a hybrid neural-quantum model is needed to explain consciousness (6) (15).


To summarize, neural networks can already more or less explain micro-cognition. From here, we can proceed to the higher-order symbolic or conceptual and semantic networks (5) which are no longer physiological ('hardware') but virtual ('software'). With quantum networks (12), we might be able to go further and trace fine information processing and the binding together of perceptions, and also explain some rare non-local effects of consciousness such as external virtual projections in vision or even intersubjective extrasensory phenomena. Moreover, since self-perception arises from self-interaction of network structures, i.e., interaction between their basic elements, neural and quantum networks might account for the background processes of self-awareness and the I. However, qualia (qualitative experience, feelings) cannot be considered merely as the result of complex-system-dynamics (2). We do not know why we experience the colour green exactly as we actually feel it - green. Nor can we explain why somebody feels himself as he does. It is still not possible to test or verify scientifically what it is like to be somebody !


Notes

(a) Neural attractors are special virtual structures that emerge from collective states of neuronal activities, so-called neuronal patterns. Usually, a specific neuronal pattern is formed as a network's response to a specific external stimulus. Because this pattern is correlated with the environmental state, it is more stable than other possible neuronal configurations and thus acts as an attractor, i.e., the other possible configurations gradually transform into a pattern-qua-attractor.

(b) Coherent oscillations are oscillations with the same phase (i.e., that are "in rhythm"). If the oscillations of two neural or quantum assemblies encoding two perceived objects are coherent, this indicates that these objects have some features in common. Coherent oscillations thus bind visual, auditive and other perceptions from different brain regions into higher-order categories.

(c) Fractal structures are structures that are replicated on different scales.

(d) Holography is an optical procedure for storing and reconstructing real and virtual images of objects in a medium (hologram).


References

1. Arbib M. (Ed.) The handbook of brain theory and neural networks. MIT Press, Cambridge (MA), 1995.

2. Hameroff SR, Kaszniak AW, Scott AC (Eds.) Toward a science of consciousness I. MIT Press, Cambridge (MA), 1996.

3. Kafatos M, Nadeau R. The conscious universe. Springer, New York, 1990.

4. Jibu M, Yasue K. Quantum brain dynamics and consciousness. John Benjamins, Amsterdam / Philadelphia, 1995.

5. McClelland JL, Rumelhart DE, PDP Group. Parallel distributed processing. MIT Press, Cambridge (MA), 1986.

6. Perus M. Neuro-quantum parallelism in brain-mind and computers. Informatica 20, 173, 1996.

7. Bohm D, Hiley BJ. The undivided universe. Routledge, London, 1993.

8. Davies PCW, Brown JR (Eds). The ghost in the atom. Cambridge Univ. Press, 1986.

9. Bohm D. Wholeness and implicate order. Routledge & Paul Kegan, London, 1980.

10. Penrose R. The shadows of the mind. Oxford Univ. Press, 1994.

11. Perus M. Multi-level synergetic computation in brain. Advances in Synergetics 9 (1997) (in press).

12. Stern A. The quantum brain. North Holland / Elsevier, Amsterdam, 1994.

13. Pylkkanen P, Pylkko P (Eds). New directions in cognitive science. Fin. AI Soc., Lapland, 1995.

14. Schempp W. Quantum holography. Nanobiology 2, 109, 1993.

15. Kak SC. On quantum neural computing. Information Sci 83, 143, 1995.


[Up]