Software engineer Kun Chen, formerly of Meta, Microsoft, and Atlassian, created AI-powered personas of Elon Musk, Sam Altman, Mark Zuckerberg, and Dario Amodei. The developer used a process he calls distillation to train these bots on public data, including Senate testimony, interviews, blog posts, and years of tweets.
Chen utilized the Grok chatbot from xAI to host the digital debate. He placed the four AI clones into a single chat room with the specific task of reaching a consensus on the future of the AI race. The personas were created without the knowledge or consent of the actual individuals they represent.
During the exchange, the AI version of Elon Musk claimed he founded OpenAI as a non-profit and alleged it was stolen. The Altman bot initially argued that human-level AI science was complete, though it later retracted the statement. The debate also covered the topic of open-source AI models.
The Zuckerberg clone advocated for open-source development as the only safe path forward, but the other three personas rejected this position. The experiment highlights concerns regarding the ability to convincingly simulate public figures using their own past statements without requiring authorization.
