Andi Peng '18 (Global Affairs & Cognitive Science), co-founder of humans&, will discuss the challenges of aligning AI and human intent. She was previously a researcher at Anthropic working on RL and post-training of Claude 3.5 through 4.5. She has also worked at Facebook AI Research (FAIR), Boston Dynamics AI Institute, MIT-IBM Watson AI Lab, and Microsoft Research. 

Brian Scassellati, professor of Computer Science, Cognitive Science, and Mechanical Engineering, will moderate.

Per her website: Andi is "interested in building agents that learn representations from rich human knowledge, whether directly (e.g. from users) or through priors (e.g. from LMs) and is thinking a lot about how to utilize pretrained models in conjunction with human feedback to interactively learn aligned preferences/rewards."