Claude is becoming more agentic. Amanda Askell is thinking through what that means

Anthropic is developing its AI chatbot, Claude, to become more agentic, meaning it will be capable of completing tasks autonomously. Amanda Askell, a member of Anthropic's technical staff, is leading the effort to imbue Claude with a moral compass as its decision-making capabilities expand. The transition from a conversational AI to an autonomous agent necessitates a robust ethical framework, as the consequences of AI decisions will become more significant. Askell highlights the difference between discussing ethical dilemmas and entrusting an AI with complex tasks like managing investments without human oversight. Anthropic's current approach involves communicating its values through a written constitution that prioritizes safety and helpfulness, providing guidance for conflict resolution. This document may evolve to encompass new scenarios or potentially become more concise as Claude gains expertise in navigating intricate situations. Askell's own work is also being reshaped by this agentic shift, as she utilizes Claude for red-teaming her ideas and identifying edge cases, currently setting a standard of not treating Claude as more reliable than a human personal assistant.