Abstract: The normative challenge of AI alignment centres upon what goals or values to encode in AI systems to govern their behaviour. A number of answers have been proposed, including the notion that AI must be aligned with human intentions or that it should aim to be helpful, honest and harmless. Nonetheless, both accounts suffer …
Abstract: AI agents based on Large Language Models (LLMs) demonstrate human-level performance at some theory of mind (ToM) tasks (Kosinski 2024; Street et al. 2024). Here ToM is roughly the ability to predict and explain behaviour by attributing mental states to oneself and others. ToM capabilities matter for AI safety because, at least in humans, …
Evaluating AI Agents for Dangerous (Cognitive) Capabilities Read More »
Date: February 7, 2025 (Friday) Time: 13:00 – 15:00 Venue: Rm 10.13, Run Run Shaw Tower, Centennial Campus, The University of Hong Kong Registration: here Speaker: Ms Ninell Oldenburg, University of Copenhagen Chair: Dr Frank Hong, The University of Hong Kong Abstract: While most technical and philosophical research on LLMs tries to find a translation from human functions to computational …
Date: February 21, 2025 (Friday) Time: 13:00 – 15:00 Venue: 3/F, MPZ Room 2, HKU Main Library Registration: here Speaker: Dr Leonard Dung, Ruhr-University Bochum Chair: Dr Frank Hong, The University of Hong Kong Abstract: Does consciousness require biology or can systems made out of other materials be conscious? I develop an argument for the view that it is (nomologically) …
Date: February 28, 2025 (Friday) Time: 13:00 – 15:00 Venue: 3/F, MPZ Room 2, HKU Main Library Registration: here Speaker: Dr Jeffrey Howard, University College London Chair: Dr Frank Hong, The University of Hong Kong Abstract: Social media platforms’ AI systems learn from user data to predict what content will keep users engaged. This content is subsequently amplified, increasing …
Date: February 14, 2025 (Friday) Time: 13:00 – 18:00 Venue: 11/F, Cheng Yu Tung Tower, The University of Hong Kong Registration: here Organizers: AIH Lab, Programme on Artificial Intelligence and the Law and Hong Kong Ethics Lab A Matter of Principle? AI alignment as the Fair Treatment of Claims (co-authored with Geoffrey Keeling) Iason Gabriel, Staff Research Scientist, Google Deepmind Evaluating AI Agents …
Abstract: Maybe AIs have alien contents and we need a fundamentally different kind ofmetasemantics for AI systems than we do for humans. The aim is to explore a cluster of closely related topics: (i) What would it be for a system to have alien (non-human) contents and howcould we recognize these as contents? (ii) What …
Abstract: The speaker will distinguish some different conceptions of alignment, exploring how each conception relates to safety and existential risk.
Abstract: Central to the inquiry of cognitive science is the idea of agents for whom there is the possibility of both success and failure. If we think that LLMs are the sort of thing that fall within the domain of cognitive science, then we should also ask whether they are agents of this kind – …