An "Anthropic Principle" for Formulations of AI Alignment

In a new essay on LessWrong, computer scientist Adam Chlipala proposes a way to define AI alignment without needing to formalize human values. He suggests focusing on agents with enough computational power to think about alignment, drawing an analogy to the anthropic principle. The idea is that there's a narrow window of intelligence capable of addressing alignment before an intelligence explosion changes everything. Chlipala also explores how layered computational systems, like cells forming organisms, might inform this approach.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton