A colloquium organised by BCS-SGAI
At the BCS London Office
Ground Floor, 25 Copthall Avenue, London, EC2R 7BP (near Moorgate underground)
BCS-SGAI, the British Computer Society Specialist Group on Artificial Intelligence was founded in June 1980. Its mission is: `To foster achievement, capability and awareness in both business and research in Artificial Intelligence, and to promote the interests of the related community`. It is one of Europe`s longest established groups working to support the community of artificial intelligence developers and users and is the organiser of one of the longest running annual series of AI conferences in Europe: the AI-20xx series.
Generative AI has progressed at an extraordinary pace, becoming an increasingly pervasive engine in communication platforms, creative industries, and organisational decision‑making. As deployment accelerates across high‑impact sectors, the risks associated with misleading content, systemic failures, and misuse underscore the need for stronger technical, ethical, and regulatory safeguards. Despite impressive breakthroughs in capabilities across text, vision, audio, and multimodal reasoning, current generative models remain prone to unpredictable failures and exhibit unresolved challenges in alignment, grounding, and robustness techniques, including hallucinations, instability under distribution shift, hidden internal representations, and limited theoretical understanding of how large‑scale models encode and propagate knowledge.
This colloquium aims to create a forum for discussing cutting edge advances and emerging challenges in the trustworthiness and safety of generative AI. It will bring together researchers and practitioners to examine the scientific, engineering, and societal foundations of trustworthy and safe Generative AI.
The colloquium will feature a combination of invited talks and contributed talks enabling interdisciplinary discussions to identify key research gaps, strengthen methodological approaches understanding, and seek actionable directions toward greater reliability, transparency, and resilience.
|
