OpenAI CEO Sam Altman announced a new program to grant 'employee level' access to outside researchers for evaluating AI safety. However, the pilot program appears much more restrictive than promised, raising concerns about its effectiveness. The program involves staff from four AI safety and research organizations becoming 'embedded assessors' to provide a clearer view of how labs are evaluating and managing risks. Despite Altman's vow in September that independent evaluators would have 'employee-like access' to internal company systems, the actual program allows only two assessors to talk to employees, and none have access to internal Slack. Researchers say the limited access could prevent them from finding blind spots and understanding what the company is missing. OpenAI stated that there are currently no standards for embedded evaluations and that they expect to deepen these relationships over time.
También lo cubren: NPR Topics: News.
🔗 Leer la noticia completa en npr.org - NPR Topics: Home Page Top Stories
Ver más en NewsToday