Abstract :
The proposition that artificial intelligence (AI) technology is not in fact able to reach to any unique level of analytical authority is argued. Three propositions, in particular, will be elaborated and defended: 1) it has not been demonstrated in the field¿nor can it be asserted axiomatically¿that AI constructs can endogenously perform the sorts of inference operations that stochastic decision exercises demand; 2) lacking such a capability, AI constructs cannot transcend, but rather must compete with, instruments of the type available from the traditional decision disciplines; and 3) in many instances of such competition, certain properties common to AI constructs might often place them at an efficiency disadvantage relative to alternative technical approaches. On the positive side, however, it also is asserted that many of the more troublesome technical provisions built into AI constructs are merely paradigmatic preferences, not methodological imperatives.
