Awareness as a proxy for AI consciousness, with Xiaojian and Ailin, Fangcun AI

Chapters: 00:00 How to make the most of AI safety research 00:40 About Directory Conversations 0:10 Introducing Xiaojian (Luke) Li, Ailin Wang, and Fangcun AI 04:07 The value of research by Fangcun AI regarding AI alignment 10:19 What is AwarenessBench 12:49 Why model awareness is a practical proxy for AI consciousness 15:36 Awareness-focused evaluations reveal gaps that general-purpose benchmarks don’t surface 19:47 Real-world AI safety outcomes of researching AI awareness 23:25 Could AwarenessBench flag potential rogue behavior of AI models 26:10 Recommended directions for future research on language model awareness 28:17 More research and opportunities to collaborate with Fangcun AI 30:41 Join the AI safety community For more information on the research, see: AwarenessBench: Assessing Cognitive Capabilities of Language Models (https://aclanthology.org/2026.acl-lon...) Fangcun AI (https://www.fangcunleap.com/) The AI Safety Research Directory is a community for AI safety researchers to share information about their research, find collaborators, and inspire others to build safe AI systems. To view the directory, visit: https://aisafety.anthroforge.io/