What should an AI's personality be?

How do you imbue character in an AI assistant? What does that even mean? And why would you do it in the first place? In this conversation, Stuart Ritchie (Research Communications at Anthropic) speaks to Amanda Askell (Alignment Finetuning Researcher at Anthropic) about the ins and outs of “character training” for Claude, Anthropic’s AI model. 00:00 — Introduction 01:41 — The importance of an AI’s character 03:32 — Training an AI model 05:44 — System prompts 07:24 — Impacts of system prompts 11:20 — Character vs personality 16:39 — Training good moral character 19:10 — Claude's trait: charitability 25:08 — Claude's trait: honesty 28:21 — Deciding on Claude’s personality 31:11 — Self awareness in AI 34:02 — Kindness towards AI 37:25 — Conclusion Learn more about Claude’s character: www.anthropic.com/research/claude-character Learn more about research at Anthropic: anthropic.com/research Try out Claude: claude.ai/