Haikonen’s Doctoral Thesis – Part III

Pentti Haikonen is one of the most salient researchers on Machine Consciousness. His PhD Thesis entitled:

“An Artificial Cognitive Neural System Based on a Novel Neuron Structure and a Reentrant Modular Architecture with Implications to Machine Consciousness”

is one of the first doctoral dissertations in the field of Machine Consciousness. In this thesis, Haikonen introduces the Haikonen Associative Neurons and his Cognitive Architecture.

Part III of Haikonen’s thesis is available here:

Haikonen, Pentti O. A., An Artificial Cognitive Neural System Based on a Novel Neuron Structure and a Reentrant Modular Architecture with Implications to Machine Consciousness. Helsinki University of Technology, Applied Electronics Laboratory, Series B: Research Reports, Espoo 1999, 156 pp. ISBN 951-22-4730-5, ISSN 1456-1174.

Abstract

The author proposes that artificial cognition, as opposed to rule-based artificial intelligence, should emulate human cognition. A human-like artificial cognitive system should be able to duplicate cognitive processes like inner imagery, inner speech, sensations, etc., perhaps even consciousness. However, this kind of machine would not necessarily be a model for the human brain, instead it would be a creation in its own right and could be tailored to suit actual practical applications as needed.

As a starting point cognitive functions and processes are reviewed and discussed. The important issues relate to information acquisition, the representation of information and the grounding of the meanings for these representations, learning and memory and finally, the actual information processing and response generation. The following elements are identified: Distributed signal representation with meaning and significance, perception process, attention, associative learning and recall, match/mismatch/novelty detection, pain/pleasure as rewarding and motivating functions, processing by inner representations, the faculty of introspection.

A novel non-numeric associative neuron, suitable for distributed signal representation and associative learning is proposed as the basic processing unit. This neuron preserves the point-of-origin meaning of the signals and thus allows consistent internal representations and the build-up of interconnected modular cognitive neural network architectures. A cognitive neural system that processes information by inner representations; inner imagery and inner words is presented. These inner representations have meanings that are grounded to sensory information and to syntactical order. This system is based on the above associative neurons and consists of a number of parallel reentrant perception/response loop modules that are associatively crossconnected. The system supports the following cognitive processes: perception, attention, learning and memory, control by match/mismatch/novelty detection and pleasure/displeasure functions. The system has several features that are commonly attributed to consciousness: It perceives; it has inner imagery and inner speech; it is introspective, the inner representations are perceived by the system via reentry to perception process; there is distinction between the externally evoked and internally evoked inner representations; there is cross-module reportability, the activity of one module can be reported by one or all the other modules in their own terms; there is attention and short-term memory. The operation of the neuron and the cognitive system is verified by computer simulations incorporating visual and linguistic modalities.

Leave a Reply

Your email address will not be published. Required fields are marked *