**[OpenAI Chief Scientist Responds to Astra Controversy: New Architecture Hasn't Suddenly Turned Models into Black Boxes]** Beating AI News Flash: OpenAI Chief Scientist Jakub Pachocki has refuted some external interpretations of the Astra recurrent architecture. He stated that the computational graph depth of OpenAI's cutting-edge models, including Astra, differs from GPT-4 by at most a factor of 2, without any sudden increase by several orders of magnitude. This was primarily in response to a report by *The Information*. The report revealed that Astra employs recurrent depth, allowing the same set of Transformer layers to compute repeatedly. As a result, some reasoning can occur internally within the model, without needing to fully articulate it as a chain of thought (CoT), raising concerns about whether models are becoming increasingly difficult to monitor. Pachocki acknowledged that CoT monitoring is indeed fragile and is trending toward becoming harder to oversee. However, he explicitly stated that this is not due to changes in model architecture. OpenAI has been retaining and utilizing CoT monitoring since the first generation of reasoning models and continues to prioritize enhancing this capability as a core research goal. [Original Link]