San Francisco-based Anthropic is one of the companies leading the artificial intelligence revolution. Its primary large-language model product is Claude, and Claude has offshoots that write computer code and can essentially reflect upon and rewrite themselves.
Anthropic’s founders split from OpenAI, another San Francisco AI company, primarily because they were concerned that reckless AI development could lead to unfortunate unintended consequences. These range from the elimination of humans by an artificial intelligence that no longer needs us to the development of AI-controlled autonomous weapons engineered for war or terrorism. Less apocalyptically, it means imagining profound effects on job markets, education, inequality and privacy.
Anthropic was founded explicitly to develop AI more responsibly. With forethought and commitment, it has developed a “constitution” that spells out what it, and AI, should and should not do. The company has published its constitution under a license that makes it freely available to other AI developers and encourages the industry to adopt its principles.
Development of the Claude Constitution has been a company-wide priority. Anthropic hired not an engineer to lead the effort but a philosopher, Amanda Askell. Many engineers at Anthropic also participate. And Claude itself reviews the constitution and suggests revisions. It is creating, essentially, a pact between the company and its artificially intelligent product to behave ethically.
As business writer Melissa Daimler observes, it’s also an inspiration for non-AI firms. Anthropic’s first constitution established a detailed list of safety and ethics rules for Claude and for software engineers. “But as Claude became more capable and the stakes got higher, they realized a list wasn’t enough. The model needed to understand why it should behave in certain ways, not just what to do. They rewrote it with a focus on judgment and context, not just rules.” Daimler suggests that there is much to learn from Anthropic’s collective effort that can apply to other firms’ efforts to build a culture of ethics, collaboration and creativity.
This is a more complicated subject than one post can cover. We encourage you to start exploring with the links above. This is the beginning of something with unimaginable outcomes. But it’s starting, it’s happening, and Anthropic’s engineers are trying to bend it toward responsible development.