In this event, co-hosted by MIT CSAIL Alliances, the MIT-UK program and the UK government’s AI Safety Institute, the panellists discussed the current state of research and where we’re headed on AI Safety & Policy. Key topics included AI's technical and societal impacts, bridging gaps between research and regulation, and the importance of interpretability and robustness in AI systems. Panellists highlighted recent progress in AI safety measures and challenges like epistemic uncertainty and misuse risks. They emphasized collaboration between academia, industry, and policymakers. The event also explored AI governance, diversity in AI development, and the need for more accessible tools to address biases and improve model auditing.

The event was held at the MIT Computer Science and Artificial Intelligence Laboratory’s (CSAIL) Stata Centre in Cambridge, Massachusetts, on Wednesday, 30 October. The panel included:

• Kwamina Orleans-Pobee, Head of Engineering at the UK government’s AI Safety Institute
• Hima Lakkaraju, Associate Professor at Harvard HBS
• Dylan Hadfield-Menell, Assistant Professor, MIT EECS
• Joseph Enguehard, LLM Research Engineer at Robin AI
• Vishal Sharma, Senior Lecturer at Queen’s University Belfast

Further details are available at https://cap.csail.mit.edu/members/events-professional-programs/frontier-ai-safety-policy-panel-where-were-where-were-headed

You can watch the recording of the panel discussion and read a transcript here.