OpenAI Pauses Frontier RL Training as It Tightens Defenses Against Unsafe AI Behavior
A leading artificial intelligence (AI) research organization has halted training on a high-stakes project after discovering vulnerabilities that could allow malicious actors to exploit AI systems. OpenAI, the developer behind popular language model GPT-3, has paused its Frontier RL training program due to concerns over potentially “unsafe” behavior. Frontier RL is an ambitious endeavor aimed … Read more