Full text of the speech by Google's Chief Scientist Shanahan
Google DeepMind's Chief Scientist Shanahan explored the mental attributes of Large Language Models (LLM) at the International Conference on AI and Philosophy at the University of London. He analyzed the applicability of LLMs in areas such as understanding, belief, agency, self, and consciousness, emphasizing the impact of multimodality and embodiment on conceptual evolution. Shanahan believes that the identity of LLMs is peculiar, with role-playing and superposition making their self transient and discontinuous. Regarding consciousness, he advocates focusing on the engineered interaction with LLMs rather than whether they possess consciousness.
GoogleShanahanlarge language modelsmindunderstandingbeliefs