Formal semantics: is it relevant to artificial intelligence?

摘要

It is argued that formal semantics, in the model-theoretic style pioneered by Tarski, is appropriate for specifying the meanings of the compositional component of artificial formal languages but not of natural languages. Since computer programming languages are both formal and artificial, formal semantics has a clear application to them, but this does not mean that it is in any way relevant to the problem of meaning in AI. The distinction is drawn between what an expression in a language means, and what a person means by using it. The former is the only kind of meaning that formal semantics can ever explain, whereas for AI to succeed it is essential to elucidate, and then to recreate, the latter. No verdict is offered on whether or not this may ultimately be possible; but it is argued that formal semantics would be an inappropriate tool to use to this end.