Long Turing

LessWrong contributor Stephen Braniff is developing an experimental chatbot architecture designed to sustain human-like conversation over days of interaction. Rather than relying on safety restrictions to prevent misuse, Braniff proposes that embedding persistent identity, motives, and beliefs in the AI's context could more effectively guide ethical behavior. He argues that humans inherently trust things that sound human—imperfections included—and that an AI designed to sound flawed might be more trustworthy than one designed to sound perfect. Braniff is inviting engagement with his experimental platform to study how artificial identity and human-AI interaction develop over time.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton