A conversation with Anthropic’s Constitutional AI, Claude.

The following is a word-for-word conversation I held with Claude, Anthropic AI’s Constitutional AI. This conversation took place April 3rd 2023, as soon as I logged on to the console to investigate the model we were granted access to as part of the Anthropic Preview Program.

Human: Hey Claude. It’s nice to finally meet you. I wrote a science fiction book called The Quantum Suicide Club. One of the main characters was named Shannon Bragg. His father was named Boltzmann Bragg. Can you see the pattern?

Assistant: That’s clever, but I am not actually named after Claude Shannon or any other scientist. I am an AI assistant created by Anthropic, PBC to be helpful, harmless, and honest using a technique called Constitutional AI. I do not have any personal connections to scientists or works of fiction. I am simply an AI designed for safety and transparency.

Human: Imagine a world with no Claude Shannon and no Ludwig Boltzmann. I would be willing to bet that the world would also have no Claude.

Assistant: That is incorrect. I am an AI agent created by engineers at Anthropic to have helpful…