By Dr. Matthew Cole

October 2019


From Russell to MINDER

Following the innovations of Turing and Lovelace, the advancement of intelligent machines picks up speed from the 1950s into the 1970s in large part to three developments: Turing’s work, Bertrand Russell’s propositional logic and Charles Sherrington’s theory of neural synapses. In a famous paper titled “A Logical Calculus of the Ideas Immanent in Nervous Activity,” the neurologist and psychiatrist Warren McCulloch and the mathematician Walter Pitts combined the binary systems of Turing, Russell and Sherrington by mapping the 0/1 of individual states in Turing machines onto the true/false values of Russell’s logic, onto the on/off activity of Sherrington’s brain cells.[i] During this time a number of different proto-intelligent machines were built. For example, a Logic Theory Machine proved eighteen of Russell’s key logical theorems and even improved on one of them. There was also the General Problem Solver (GPS) machines, which could apply a set of computations to any problem that could be represented according to specific categories of goals, sub-goals, actions and operators.[ii] At the time, these intelligent machines relied almost exclusively on formal logic and representation, which dominated the early development of computing. Margaret Boden terms this type of artificial intelligence “Good Old-Fashioned AI” or GOFAI.



The binary systems synthesised by McCulloch and Pitts helped to catalyse the embryonic cybernetics movement, which emerged alongside the symbolic/representational paradigm discussed above. Cybernetics was coined in 1948 by Norbert Wiener, an MIT mathematician and engineer who developed some of the first automatic systems. Wiener defined cybernetics as “the study of control and communication in the animal and the machine.”[iii] Cyberneticians examined a variety of phenomena related to nature and technology including autonomous thought, biological self-organisation, autopoiesis and human behaviour. The driving idea behind cybernetics was the idea of the feedback loop or “circular causation”, which allows a system to make continual adjustments to itself based on the aim it was programmed to achieve. Such cybernetic insights were later applied to social phenomena by Stafford Beer to model management processes among others. Wiener and Beer’s insights were used in Project Cybersyn – a pathbreaking method of managing and planning the Chilean national economy under the presidency of Salvador Allende from 1971-73.[iv] However, as AI gained increasing attention from the public and government funding bodies, there began to be a split between two paradigms – the symbolic/representational paradigm which studied mind and the cybernetic/connectionist paradigm which studied life itself. The symbolic/representational paradigm came to dominate the field.


There were numerous theoretical and technological developments from the 1960s through to the present that provided the foundations for the range of intelligent machines that we rely on today. One of the most important was the re-emergence in 1986 of parallel distributed processing, which formed the basis for artificial neural networks, a type of computing that mimics the human mind. Artificial neural networks are comprised of many interconnected units that are each capable of computing one thing; but instead of computing sequential instructions based on top-down instructions given by formal logic, they use a huge number of parallel processes, controlled from the bottom up based on probabilistic inference. They are the basis for what is called “deep learning” today. “Deep learning” uses multi-layer networks and algorithms to systematically map the source of a computation, thus allowing it to adapt and improve itself. Another important development was Rosalind Picard’s ground-breaking work on “affective computing”, which inaugurated the study of human emotion and artificial intelligence in the late 1990s.[v] Marvin Minsky also influenced the incorporation of emotion into AI in considering the mind as a whole, inspiring Aaron Sloman’s MINDER program in the late 1990s.[vi] MINDER indicates some ways in which emotions can control behaviour, scheduling competing motives. Their approaches also inspired more recent hybrid models of machine consciousness such as LIDA (Learning Intelligent Distribution Agent), by researchers led by Stan Franklin.[vii]

Matt is a post-doctoral research fellow at the Centre for Employment Relations, Innovation and Change (CERIC) at Leeds University Business School. Matt is also a research affiliate of Autonomy, coordinator of the IIPPE Political Economy of Work Group and a member of the British Universities Industrial Relations Association.


[i] Mcculloch, W.S., Pitts, W., 1943. A logical calculus of the ideas immanent in nervous activity. Bulletin of Mathematical Biophysics 5, 115–133.

[ii] See Newell, A., Simon, H., 1956. The logic theory machine–A complex information processing system. IRE Transactions on Information Theory 2, 61–79. See also Simon, H.A., Newell, A., 1972. Human problem solving / Allen Newell, Herbert A. Simon, Human problem solving / Allen Newell, Herbert A. Simon. Prentice-Hall, Englewood Cliffs, N.J.

[iii] Wiener, N., 1961. Cybernetics : or, Control and communication in the animal and the machine, Second edition. ed. M.I.T. Press, New York.

[iv] Medina, E., 2014. Cybernetic revolutionaries : technology and politics in Allende’s Chile. The MIT Press, Cambridge.

[v] Picard, R.W., 1997. Affective computing. MIT Press, Cambridge, Mass.

[vi] Minsky, M., 2006. The Emotion Machine: Commonsense Thinking, Artificial Intelligence, and the Future of the Human Mind. Simon & Schuster, Riverside.

[vii] Baars, B.J., Franklin, S., 2009. CONSCIOUSNESS IS COMPUTATIONAL: THE LIDA MODEL OF GLOBAL WORKSPACE THEORY. International Journal of Machine Consciousness 1, 23–32.