k.Work
WorkOpenAI Blog
KeyRank 45Faulty reward functions in the wild
A foundational deep-dive into a critical AI safety/ML failure mode that leaders need to understand as RL systems scale into real-world deployment. This is the kind of technical governance issue that belongs in board-level AI risk discussions.
Why it ranks · · OpenAI research post on reinforcement learning failure modes · Dec 19 – 25, 2016
Read full story