OpenAI's Astra model sparks AI safety concerns with opaque recurrence

openai's new astra model utilizes a reasoning technique called opaque recurrence, also known as recurrent depth, which enables it to process queries in a less linear approach. this technique has sparked concerns among ai safety experts, including redwood ceo buck shlegeris and longtime ai safety advocate zvi mowshowitz, as it may make the model's chain of thought harder to monitor. while astra's use of the technique appears to be limited, experts worry about the potential risks and the need for laws to prevent a 'race to the bottom' among ai labs. openai has emphasized its commitment to legible chains of thought and announced plans for extensive chain-of-thought monitoring systems.