Richard Sutton – Father of RL thinks LLMs are a dead-end

Bullish:

• $RL - Reinforcement learning represents the fundamental path to true AI and intelligence, as it focuses on understanding the world through experience rather than pattern matching

  • “I consider reinforcement learning to be basic AI. And what is intelligence the problem is to understand your world. And reinforcement learning is about understanding your world.”

    Richard·1:03
• $EXPAI - Experiential AI systems that learn from direct experience will dramatically outperform current LLM approaches and represent the future of scalable intelligence

  • “I, in particular, expect there to be systems that can learn from experience, which could well perform much, much better and be much more scalable.”

    Richard·11:17
• $DIGITALAI - Digital intelligence represents a fundamental evolutionary transition from biological replication to designed intelligence, marking a key inflection point in the universe

  • “I think we're entering the age of design because our AIs are designed... And so this is a key step in the world and in the universe. I think it's a transition from the world in which most of the interesting things are replicated.”

    Richard·57:14

Bearish:

• $LLM - Large language models are fundamentally flawed as they lack goals, ground truth, and true understanding - starting from the wrong foundation for achieving intelligence

  • “Large language models are trying to get by without having a goal or a sense of better or worse. That's just... it's exactly starting in the wrong place.”

    Richard·12:49
• $SUPLEARN - Supervised learning is an unnatural approach that doesn't occur in biological intelligence and lacks the ground truth necessary for continual learning

  • “It's absolutely obvious, I would say, that supervised learning doesn't happen in animals.”

    Richard·18:11