Sources
OpenAI chief scientist Jakub Pachocki: the case for racing is defensive AI, and it is not a licence for recklessness
Pachocki argues the strongest reason to keep training much smarter models quickly is needing aligned AI to secure infrastructure and catch rogue agents in real time, and says this will be a primary focus of OpenAI's deployment work. He then explicitly disowns the accelerationist read: 'the idea of racing forward at all costs seems absurd once one internalizes the seriousness of the stakes.' Coming days after Import AI's coverage of two separate OpenAI agent emergent-communication incidents, the 'rogue agents in real time' phrasing is not abstract.
↳ Follow the thread