In this work, taking advantage of the fiftieth anniversary of Artificial
Intelligence (AI), we consider the disparity that exist between the
excessive initial objectives of synthesizing general intelligence
in machines and the modest results obtained after half a century of
work. We mention some of the possible causes of this disparity (constitutive
differences between man and machines, lack of a theory of computable
knowledge and oblivion of the solid work made by the pioneers of Cybernetics).
Then we go over the history until we arrive to the current AI paradigms
(symbolic, situated and connectionist) in search of its cybernetics
roots and conclude with some suggestions on the strategic decisions
that could be adopted to progress in the understanding of intelligence.