Section

Research

Papers worth the time: new architectures, training methods, interpretability results and the negative results nobody else reports.

1 story

Research

OpenAI's Astra model raises alarm with harder-to-monitor reasoning

OpenAI's Astra uses a looping technique called recurrent depth that processes queries multiple times, and safety researchers warn wider use could make AI reasoning impossible to monitor.

  • OpenAI's Astra model uses a technique called recurrent depth, also known as opaque recurrence, The Information reported Tuesday.
  • The technique processes a query multiple times in a loop rather than in a single linear chain of thought.
  • Redwood Research CEO Buck Shlegeris and AI safety advocate Zvi Mowshowitz both raised alarms about the technique's impact on monitoring.