The Chatbots Are Now Talking to Each Other

WIRED’s Will Knight reports on companies using chatbots with detailed human personas to research audiences and develop product ideas, with Fantasy’s synthetic humans as the lead example. Fantasy gives each agent dozens of characteristics drawn from ethnographic research on real people, runs them on commercial large language models such as OpenAI’s GPT and Anthropic’s Claude, and convenes focus groups that mix synthetic and real participants while clients watch.

• Generating product ideas. In a session for Major League Soccer, two synthetic personas proposed an app with augmented-reality views of games and points earned while watching. Cole Sletten, MLS’s VP of digital experience, called the technology truly incredible.
• Exploring ideas at scale. BP asked 50 synthetic humans to discuss smart city projects. BP’s global head of design noted that a synthetic human keeps going long after a person would tire.
• Making people more creative. Peter Smart says synthetic humans have produced novel ideas for clients and prompted the real people in the conversation to be more creative.
• Knowing the limits. Stanford’s Michael Bernstein cautions that LLM characters can be more stereotypical and less varied than real people, and making them more faithful is still an open research question.