Anthropic AI, a San Francisco-based AI startup backed by Amazon, has introduced a crowd-sourced AI constitution to govern its chatbot. The constitution, which emphasizes ethics and accessibility, was created using public input and consists of 75 guiding rules and principles.
The new constitution urges the chatbot, named Claude, to prioritize balanced and objective answers. It also emphasizes the importance of remaining harmless, ethical, and avoiding toxic, racist, sexist, or unethical behavior in its responses.
Anthropic AI partnered with the research firm Collective Intelligence Project to survey 1,000 Americans from diverse backgrounds, including age, gender, income, and location. Participants were given the opportunity to vote for or against existing rules or create their own, with new principles reflecting widely shared sentiments being added to the constitution.
The survey revealed that users want the AI chatbot to prioritize responses that are honest about admitting flaws, promote good mental health, and demonstrate the least jealousy towards humans. Existing principles that discouraged racism, sexism, and promoted reliability and honesty in responses were also popular among voters.
The new constitution takes into account over 1,000 statements and more than 38,000 votes. Anthropic’s spokesperson stated that since AI can significantly impact people’s lives, the values and norms that govern these systems are crucial.
The release of the AI constitution comes at a time of increasing concern about AI safety. Technology leaders such as Elon Musk and Satya Nadella have highlighted the potential risks associated with AI. Government leaders have also taken notice, with Anthropic’s co-founder Dario Amodei joining other industry executives in discussions with White House officials about plans to address AI safety concerns.
By adopting a democratic approach and incorporating public input, Anthropic AI aims to ensure that its chatbot aligns with the values and expectations of users. The company’s commitment to ethics and accessibility is reflected in its newly unveiled constitution, which underscores the importance of responsible AI development and deployment.
