Rich Sutton and Khurram Javed: Why AI Models Stop Learning, and How to Start It Again

Reinforcement learning pioneer Rich Sutton and co-founder Khurram Javed argue that the future of artificial intelligence lies in continual, experiential learning rather than static, human-knowledge-dependent models. While large language models represent a significant breakthrough, they remain limited by their inability to learn post-training. True intelligence requires systems that adapt to the "Big World"—an infinitely complex environment—by generating their own experiences and forming self-consistent abstractions. Sutton’s "Bitter Lesson" emphasizes that long-term AI progress depends on scaling computation and learning algorithms, not human-provided priors. Through their new venture, Oak Lab, the pair aims to overcome catastrophic forgetting and develop architectures capable of sustained, self-maintaining learning. This shift moves beyond current paradigms, prioritizing agents that evolve through interaction rather than relying on finite, human-curated datasets or static weights.