Democratizing AI: A Constitutional Approach by Anthropic and Collective Intelligence Project

Posted by

Empowering Americans in AI Development

In a quest to democratize AI tool development, Anthropic, the creator of Claude AI, partnered with the non-profit Collective Intelligence Project. Their mission? To chart a novel path for AI development by engaging 1,000 Americans in drafting an AI constitution.

A Transformational Impact of AI

Divya Siddarth, co-founder of the Collective Intelligence Project, emphasized AI’s potential to reshape the world. Yet, very few have the privilege of shaping this future. Their endeavor is to change that status quo, introducing processes to give everyone a say.

Collective Intelligence Project: Fostering Progress

Launched in January, the Collective Intelligence Project, led by Divya Siddarth and Saffron Huang, seeks to promote collective intelligence for collective progress. Their collaboration with Anthropic, the Collective Constitutional AI initiative, aims to design an AI model with external inputs and examine the influence of democratic processes on AI development.

Involving the Public in AI Decision-Making

Anthropic’s initiative involved 1,000 members of the American public in choosing rules for its AI chatbot. The diverse sample aimed to represent U.S. adults across age, gender, income, and location. Participants could select rules from a list or propose their own.

Collective Decision-Making Process

According to Anthropic, participants contributed 1,117 statements and cast over 38,000 votes on critical questions. These questions included whether AI should discriminate based on race or sexual preference. Remarkably, despite differing opinions, a high degree of consensus emerged on most statements.

Beyond the U.S. Borders

While the survey aimed for diversity, Siddarth acknowledged the need to expand experiments beyond tech-savvy and affluent Americans. Inclusivity is a key consideration for future projects.

AI Developer Collaborations

The Collective Intelligence Project has partnered with OpenAI on other democratic AI projects, but the AI Constitution project exclusively utilized Anthropics’ Claude AI models.

Unexpected Findings and Future Prospects

A noteworthy revelation in the experiment was the desire to prevent AI discrimination based on disability, a perspective not present in the Anthropic constitution. Siddarth believes this wasn’t intentional but may have been overlooked in the drafting process.

AI Developers in the Spotlight

Anthropic, a developer in the AI race, has garnered substantial investments from industry giants like Google and Amazon in 2023. The collaboration between Anthropic, OpenAI, and others reflects a commitment to responsible AI development.

A Bright Future for Democratic AI

Despite the experiment’s limitations, Siddarth remains hopeful about the Collective Intelligence Project’s role in shaping AI with public input. Their aim is to continue refining these processes through collaboration with various stakeholders, including civil society groups and policymakers.

In summary, the collaboration between Anthropic and the Collective Intelligence Project marks a significant step toward a more inclusive and democratic approach to AI development, with an eye toward learning from past experiences to create a better future.

Leave a Reply

Your email address will not be published. Required fields are marked *