OpenAI Deploys AI "Control Room" to Block Model Output Within 35 Seconds
18 hour ago / Read about 0 minute
Author:小编   

OpenAI recently announced the deployment of a centralized control system called "Control Room" within its internal research infrastructure. When a model task poses security or cybersecurity risks, the system can block model output within 35 seconds of receiving a stop command.