OpenAI Chief Scientist Responds to Astra Controversy: Chain-of-Thought Monitoring Remains a Research Focus
OpenAI Chief Scientist Jakub Pachocki stated that he hopes to avoid a race to make models "unmonitorable" sparked by misleading reports. He noted that the computation graph depth of current frontier models, including Astra, does not differ from that of GPT-4 by more than a factor of two. Since its first reasoning models, OpenAI has remained committed to preserving and leveraging Chain of Thought monitoring, believing that this technique helps observe how model alignment capabilities generalize from the training distribution.