|
Back to list of 2000/01 seminars
In 1962 Rosenblatt described connectionist approaches to Artificial Intelligence, AI, primarily in terms of formal pattern matching tasks [1]. Alternatively, in a seminal paper fourteen years later, Newell and Simon defined symbolic AI in terms of the ratiocination of symbol manipulation and search [2]. Common to both, seemingly disparate, methodologies is unspoken use of a computational 'metaphor' underlying intelligent action. But cognition is not computation... In this talk I will give a brief critical account of knowledge representation and processing in both connectionist and symbolic AI, before outlining a new metaphor for intelligent action grounded upon stochastic communication in a population of agents [3]. I will illustrate this metaphor in action by demonstrating how a simple network of stochastically communicating agents can both dynamically represent hi-arity knowledge and self-organise to perform best-fit constraint satisfaction search on distributed sub-symbolic information. I conclude the presentation with a practical discussion of how to apply this methodology to the problem of how to locate the best Indian Restaurant when visiting a new town...
|
http://www.cs.ukc.ac.uk/dept_info/seminars/2000_01/abs.2001.03.06.html
Last modified Friday June 22 16:01:37 BST 2001
Problems with this page?
Contact the CS Webmaster