AI Papers | Roundtable (Deception) [members only]

​Let's get together to read + discuss AI papers we're interested in! This can include LLM training, architecture, fine-tuning, safety, memory, search, STT, TTS, AI + X domain, or any area of study related to AI. Drop me a comment if you want to cover a particular paper for the next session.

​This week's papers are on the topic of Deception, or when an LLM knowingly tries to manipulate the user it's talking to.

• ​Alignment Faking in Large Language Models

• ​How to Catch a Liar

​[6:30-7:30]: Reading Time (Optional).

For anyone who wants a dedicated hour to get through their paper for the session. Find a nice, quiet spot anywhere in the Commons.

​​[7:30-8:30]: Discussion Time.

​​Let's meet back at the Astrology Lounge for the discussion.

​​Share an explanatory brief on the paper you've read. Go into as much detail as you'd like. Focus on what you found interesting, what didn't make sense, what might be limitations.

​​The intention: Build a space where we learn from each other, build more resilient mental models, get insights on research & development interests we have.