Abstract

This paper investigates the possibility and conditions of correctly ascribing intentionality to AI, focusing specifically on the case of ChatGPT, using as conceptual tools the late philosophy of Wittgenstein. It rejects traditional views by examining intentionality not as an inherent mental state or property, but rather as a feature emerging from language games and social interactions. Through an analysis of the concepts of belief, desire, intention and fear, as well as ChatGPT’s participation in language games, the paper reveals the technological and ethical limitations currently preventing AI from correctly being described as an intentional agent. However, this paper suggests that evolving linguistic practices, as well as the development of AI may eventually allow for a meaningful ascription of intentional states to AI entities., highlighting not an ontological, but rather a technological nature of these limitations once we adopt this OLP perspective.