“Prehistory” of Neocybernetics

by Heikki Hyötyniemi

Some old AI publications

Study of neocybernetics, being a (more or less) scientific enterprise, is also a cybernetic system of theorizing: innovations are random, but discrepancies as the derived models are compared to natural systems determine the “tensions” of evolution. The following set of papers reveals how things within neocybernetics have matured, and, in retrospect, it may be interesting to glance them through. I am grateful to Finnish Artificial Intelligence Society (FAIS) for offering the opportunity of publishing the clearly “raw” material, ideas still searching for their form.

Starting (somewhat arbitrarily) from the Finnish Artificial Intelligence Conference 1998 (STeP’98) in Jyväskylä, there are the following three papers in the proceedings book “Human and Artificial Information Processing”, edited by P. Koikkalainen and S. Puuronen:

• “Structure from data: AI approaches to systems modeling”, pp. 31–40.
• “On mental images and ‘computational semantics’”, pp. 199–208.
• “Semantic feature extraction: reader-specific text document classification”, pp. 55–60.

The theme of the year 2000 Finnish Artificial Intelligence Conference in Espoo was “Millennium of Artificial Intelligence”, and the proceedings was edited by H. Hyötyniemi:

• “Expertise: an optimized data model?”, Vol. 2: pp. 95–102.
• “Independent components of expertise”, Vol. 3: pp. 37–44.
• “System theory vs. theory of mind: Towards a synthesis?”, Vol. 3: pp. 25–34.
• “AI in practice: Case study on a flotation plant”, Vol. 2: pp. 159–166.
• “Explorations on ‘naturalistic formalisms’”, Vol. 3: pp. 123–131.

The year 2002 Finnish Artificial Intelligence Conference (STeP’02) was held in Oulu. The editors of the proceedings “Intelligence, The Art of Natural and Artificial” were P. Ala-Siuru and S. Kaski:

• “Intelligent Information Management – The Challenges of a SOM-Based Mail Organizer”, pp. 10–18.
• “Towards perception hierarchies”, pp. 111–123.
• “Life-like control”, pp. 124–139.
• “HUTCH model in information structuring”, pp. 241–255.
• “Studies on emergence and cognition – Part 1: Low-level functions”, pp. 286–299.
• “Studies on emergence and cognition – Part 2: High-level functionalities”, pp. 300–312.

As a glimpse to the future ... my inaugural lecture, March 12, 2002: “Complex Systems – Searching for Gold”.

The new science of neocybernetics was officially introduced at the Finnish Artificial Intelligence Conference, Sep. 1–3, 2004 (STeP’04), in Helsinki. The basic material is presented in the following STeP papers:

• The basic intuitions are presented in “Cybernetics: Towards a Unified Theory?” (Proc. STeP’04, Vol. 1). It is shown how concrete analyses of a Hebbian neuron grid can offer intuitions on how distributed systems can coordinate themselves in general.
• More general views of how a cybernetic system is connected to its environment and what kind of universal properties it has is explained in “Information and Entropy in Cybernetic Systems” (Proc. STeP’04, electronic version). The concepts of entropy and information offer a good framework for understanding the properties of such systems.
• In “Higher-Order Balances: Unifying Views of Data?” (Proc. STeP’04, Vol. 1) it is explained what kind of consequences there are on observation data when a system being observed is cybernetic; it is also shown how a neuron can constitute the atom of semantics in mental representations.
• The cybernetic networks are studied in “Cybernetics in Analysis and Synthesis of Networked Systems” (Proc. STeP’04, Vol. 3). Often, the dependencies are multiplicative – such probabilistic networks can be also studied in the same framework, resulting in non-strictly-power-law distributions.
• As a concrete application example, in “Emergent Coordination in Distributed Sensor Networks” (Proc. STeP’04, Vol. 2) it is shown how the cybernetic intuitions can be applied in a distributed agent system – so that distribution brings some real added value that cannot be predicted applying centralized approaches.
• The above studies concentrated on populations – but the balances can also be based on constraints, so that the emergent models are based on the remaining degrees of freedom. In “Cybernetic Intuitions in Control Engineering” (Proc. STeP’04, electronic version) these approaches are applied to complex process optimization.
• The homeostatic balance within a cell is an extreme example of constraints-governed systems. The processes of life (metabolic and genetic) are discussed in “Processes of Life: Towards a Unified Model?” (Proc. STeP’04, Vol. 1). It may turn out that even the questions concerning the origin of life can be studied from a fresh point of view.

... One of my earliest publications (Proc. STeP’92, Vol. 1) was, incidentally, also about Artificial Life.

Back to recent publications

About neocybernetics