Turing, Searle and the quest for the meaning in the NLP era

This contribution examines the epistemological implications of advanced AI systems, specifically Large Language Models (LLMs), treating them as incredible informational machines. The release of ChatGPT in 2022 is identified as a symbolic milestone, comparable to the publication of Alan Turing’s seminal 1950 work, necessitating a revisit of the Turing Test and Searle’s Chinese Room argument. While LLMs are now capable of sustaining human-level conversation and passing variants of the Turing Test, attributing intelligence requires considering the artifact’s design and functioning, as mere third-person behavioral evidence is insufficient. The paper argues that LLMs introduce a novel locational semantic component, representing meaning through multidimensional vectors based on word proximity in training corpora. Outputs have conventional meaning (derived intentionality) but generally lack intentionality, consciousness, or speaker meaning. The development of this software exemplifies digital philosophy, reshaping conceptual horizons.