Pandemonium Architecture

Pandemonium Architecture is an early connectionist AI technique proposed by Oliver Selfridge in 1959, noted for its success in modelling human pattern recognition. It proposes that the mind consists of a collection of demons, each responsible for a single task. Selfridge's architecture would inspire later developments in artificial intelligence, such as production systems and intelligent agents.

Pandemonium theory was developed by John Jackson in 1987 to include the stadium metaphor. Demons are usually dormant and reside in the stands. Active demons reside in the arena. Demons in the arena spread activity to demons in the stands which may result in demons being summoned to the arena. Active demons suffer a decay in activity over time and if their activity drops below a threshold they may be returned to the stands. Connections between demons engage in a Hebbian learning scheme, with the additional condition that long active demons are connected to newer demons with a stronger connection than the vice versa connection.

External links

* [http://neuron-ai.tuke.sk/NCS/VOL1/P3_html/node17.html neuron-ai.tuke.sk] - Concise article about Selfridge's Pandemonium

References

Wright, Robert. "Can Machines Think?" TIME Magazine. March 25, 1996

O. G. Selfridge. "Pandemonium: A paradigm for learning." In D. V. Blake and A. M. Uttley, editors, Proceedings of the Symposium on Mechanisation of Thought Processes, pages 511-529, London, 1959

J. Jackson, "Idea For A Mind" Siggart Newsletter, 181, 1987