ABSTRACT During the 1960s and 1970s, Keith Donnellan ([Donnellan, K. S. 1966. “Reference and Definite Descriptions.” The Philosophical Review 75 (3): 281–304. https://doi.org/10.2307/2183143]; Donnellan, K. S. 1970. “Proper Names and Identifying Descriptions.” Synthese 21 (3–4): 335–358. https://doi.org/10.1007/BF00484804]; Donnellan, K. S. 1974. “Speaking of Nothing.” Philosophical Review 83 (1): 3–31. https://doi.org/10.2307/2183871] and Saul Kripke [Kripke, S. 1972. “Naming and Necessity.” In Semantics of Natural Language, edited by Donald Davidson and Gilbert Harman, 253–355. D. Reidel] promoted a ‘historical picture’ of reference as an alternative to previously prevailing description theories. The historical picture has since gained broad acceptance. Its most recent frontier is the apparent reference of names (and other words) produced by large language models (LLM). A number of authors have pointed out that the historical picture might support the view that this reference is not merely apparent, but is constitutively determined in the same way as human reference. A sticking point, however, is the role that linguistic intentions have been thought to play in the historical picture. This paper urges that in applying the historical picture of reference to LLMs we need not and should not focus on the plausibility of ascribing linguistic intentions to LLMs. Instead, the critical issue is whether LLMs can achieve a kind of basic reference that is prior to and independent of propositional attitudes such as intentions.
Read full abstract