ChatGPT: What now?

ABOUT THE SPEAKER
Prof. Luc Steels
Venice International University

ABSTRACT
Generative AI (instantiated by ChatGPT, Dall-E, Claude, Copilot, Soundraw, etc.) has recently taken the world by storm, creating a wave of expectations and causing generative AI to be seen as the most transformative disruptive technology of today. Although the power and utility of the deep learning/transformer methodology underlying generative AI is undeniable, there are also several by now well known serious issues. One of the most important ones is the tendency of generative AI to come up with ‘hallucinations’, in the sense of producing facts that are untrue or devoid of common sense. This makes generative AI inherently untrustworthy and unreliable. There is also the issue of energy usage, data provisioning, and the rights of authors whose data forms the basis of training and reuse. What are the research avenues we could explore to overcome these issues?