Generative Artificial Intelligence and the Imitation Game

Abstract

This paper discusses the recent experiments by Jones and Bergen that showed ChatGPT-4 passed the Turing test. To this end, the game proposed by Turing is placed in the proper philosophical dimension by showing how it tries to answer a well-formulated and circumscribed question. The details of the experiment are then investigated. Finally, some important consequences are drawn regarding the problem of explainability of neural networks.