25 November 2024 – 16:00 – 18:00 (UTC+1) (Zoom)
Pr. Bernie HOGAN (Oxford Internet Institute, University of Oxford) : “Hallucinating the Future: Theoretically disentangling Sincerity and Authenticity in the usage of generative AI”
Abstract:
Generative AI systems use inference in order to provide responses to users. Whether in the text, audio, or image space these technologies, the answers provided are synthesized from the existing model and provide a ‘best guess’ to the users. Yet, public framing of AI has shown considerable dissatisfaction with AI systems that confuse dates, facts, and make untoward assumptions. Yet what is an appropriate assumption? If I ask an image generator for an American President, how likely am I to get an image of a woman or a person of colour? What would be the right proportion?
In this talk, I disentangle the notions of sincerity and authenticity with respect to AI systems, highlighting how authenticity is used as a benchmark of accuracy with the past and sincerity as a benchmark of accuracy with future intent. I provide examples that challenge the notion that an AI system can come up with the ‘right answer’ where there might not be a right answer. I cover some recent examples of trust and safety behaviours that show the challenges in this context and provide a framework for how to reflect on these challenges within AI systems. I conclude by reinforcing the importance of establishing both user intent and factual context for dealing with issues where there is no necessarily right answer but many potentially wrong ones.
Registration link : https://univ-amu-fr.zoom.us/meeting/register/tZYrceuhpj0sGtOLstaLYTmtjc2NwtXm7Olr