F2025 - Technical Paper Reading Group Week 2 - Scheming
UBC AI Safety Technical Paper Reading Group
UBC AI Safety has launched a biweekly technical paper reading group focused on cutting-edge AI safety research.
Sessions will engage with recent papers across topics including mechanistic interpretability, AI control, scalable oversight, capability evaluation, and failure mode identification. The group emphasizes critical analysis and discussion.
Session 2: Scheming
Our second meeting will feature a deep dive on scheming: when AIs deliberately conceal misalignment in order to achieve their goals. We'll examine a few papers that ask whether and how LLMs exhibit this behavior. Dinner will be provided!
Prereading:
This blog post (~30 minute read) summarizes Joe Carlsmith's 2023 report on scheming. It includes all of the needed technical definitions and outlines the main considerations for and against expecting scheming to emerge. We highly recommend you read this before coming to the session. There is an audio option as well!
Location: IKB 263
Who Should Attend:
Meetings are open to anyone interested in technical AI safety research. While no prior experience is required, participants with working knowledge of AI Safety and machine learning concepts will get the most out of discussions. If you're unsure whether you have sufficient background, check out this preparation document which gives resources on topics you should be familiar with for maximum engagement with the material.