Richard Sutton: Father of RL thinks LLMs are a dead end

Why it matters

2024 Turing Award winner, author of 'The Bitter Lesson', father of reinforcement learning Richard Sutton delivers the most forceful critique of the LLM roadmap: 'LLMs don't satisfy the Bitter Lesson — they rely on human-curated data, not learning from their own experience.' He explains why continual learning and online interaction are the true path, and why after AGI, AI researchers will 'scale exponentially like compute'.

Key takeaways

• 'LLMs don't satisfy the Bitter Lesson' — true intelligence must learn from its own experience, not just human-fed data.
• The Bitter Lesson's core: compute crushes handcrafted knowledge — every time we tried to encode human knowledge into AI, larger compute + simpler algorithms won.
• Era of Experience: the next paradigm is letting AI 'live, make mistakes, and learn' like humans — continual learning, online interaction.
• Post-AGI: AI researchers will scale exponentially like compute — millions of AI scientists doing research simultaneously.
• Cultural evolution is the next frontier: multiple AIs teaching and competing with each other, generating knowledge growth far beyond a single model.