Anthropic, the creator of the chatbot Claude, has pioneered a unique approach to A. I. governance called "Collective Constitutional A. I. " This experiment aims to involve ordinary citizens in shaping the rules and behavior of powerful artificial intelligence systems. Instead of leaving the decision-making solely in the hands of a few industry insiders, Anthropic invited a panel of around 1, 000 American adults to contribute their ideas and principles. These were then refined into a list of 75 principles, forming what Anthropic refers to as the "public constitution. " Two versions of Claude were trained, one with the original constitution written by Anthropic and another based on the public constitution. Initial results showed that the public-sourced version of Claude performed similarly to the original, and even exhibited less bias. While this experiment is in its early stages and may have limitations on larger A. I. models or larger participant groups, it represents a step towards more inclusive A. I.

governance. By allowing the public to have a voice in shaping A. I. systems, concerns around bias and manipulation can be alleviated. Although there are still questions about the democratic nature of such experiments and whose voices should be represented, the idea of holding A. I. companies accountable to the public is a crucial one. While it would have been preferable for these companies to seek input earlier, the progress being made is encouraging.