I met AI through Max Clowes, in 1969 when I was still a lecturer in philosophy at Sussex University, and soon became deeply involved, through a paper at IJCAI 1971 criticising the 1969 logicist manifesto by McCarthy and Hayes, followed by a fellowship in Edinburgh 1972--3. Since then I’ve worked on forms of representation, vision, architectures, emotions, ontology for architectures, tools for AI research and teaching, links with psychology, biology and philosophy, and most recently robotics, and have helped to build up two major AI centres for teaching and research (at Sussex and Birmingham). I invited McCarthy and Minsky to join a 'Philosophical Encounter' at IJCAI'95 [pictured below] because I believe philosophy needs AI and AI needs philosophy. Much of what philosophers write about consciousness and the mind-body problem shows their ignorance of AI and many silly debates between factions in AI (e.g. about representations, use of symbols, GOFAI) and some fashions (e.g. recent enthusiasm for `emotions’) result from doing poor philosophical analysis.
I always thought progress in AI would be slow and difficult, and that people who predicted rapid results had simply failed to understand the problems, as sketched in my 1978 book, now online.
Another reason for slow progress is the fragmentation of AI: people learn about tiny fragments of a whole system and build solutions that could not form part of an integrated human-like robot. One explanation is that we do not have full length undergraduate degrees in AI and most researchers have to do a rapid switch from another discipline, so they learn just enough for their PhD topic, and they and their students suffer thereafter from the resulting blinkered vision.
I’ve proposed some solutions to this problem in an introduction to a multi-disciplinary tutorial at IJCAI’05, including use of multiple partially ordered scenarios to drive research.
It requires a lot more people to step back and think about the hard problems of combining diverse AI techniques in fully functional human-like robots, though some room for specialists remains.