Conversations With Cade Metz on the Rationalists

According to LessWrong, New York Times reporter Cade Metz is writing a book about the rationalist community—a small group of early believers in artificial general intelligence whose ideas shaped modern AI development. But cooperation is tricky: some rationalists won't talk to Metz due to past reporting controversies. In conversations published today, community member Zack Davis explains why the book matters philosophically while highlighting a key tension in AI safety thinking: current large language models appear relatively safe because they learn from human examples, not dangerous optimization targets. But when researchers add reinforcement learning—as in recent systems like OpenAI o one and DeepSeek—the fear returns: you might be training away the safeguards. It's a story of vindication and ambivalence: the rationalists were right about AGI, but many worry their own advocacy helped accelerate it.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton