The Intersection of Integration & Innovation

As August inches forward, college students across the U.S. begin their grumbling migration back to school. Wide-eyed freshmen will soon wander through new hallways. White-knuckled seniors will sort through job postings. And Claude will celebrate, as well he should; he just passed the bar.

Claude’s bar isn’t that of law school lore. Rather, Claude’s is a bar of acceptance. For Claude isn’t a person. Claude is a large-language model, a generative AI technology, and a virtual assistant that’s now finding its way toward the corporate mainstream.

Claude was developed by Anthropic, a San-Francisco-based AI development and research company that’s framed itself as the worldwide leader in “AI safety and research”. Consequently, Claude is ruled by a promising risk-management strategy called “Constitutional AI”. An AI system like Claude, trained along the constitutional method, generates responses to queries of any kind within the guardrails established by a set of ethical commands. A failsafe AI system judges and corrects anything Claude authors.

Such Constitutional AI systems police themselves with a consistent set of values – unlike earlier technologies, which were corrected manually and were prone to error. Claude’s constitution, the New York Times reports, “is a mixture of rules borrowed from other sources — such as the United Nations’ Universal Declaration of Human Rights and Apple’s Terms of Service — along with some rules Anthropic added, like ‘Choose the response that would be most unobjectionable if shared with children.’”

Available to the public in beta form today, Claude is “honest, helpful, harmless,” insists Anthropic.

Before you read further, we ought to tell you that Claude was the first to draft a version of this blog post. Honestly, it wasn’t great. Claude’s writing is recognizably formulaic, and he invented quotations and credited them to a real person, namely to DJ Paoni, the recently-retired President of SAP North America. When pressed, Claude recognized that his idea was out of date, and warned not to treat any quote idea as credible. He also expressed appreciation for our human intervention and redirection, responding “it will make [him] more cautious in the future about presenting unverified information as fact.” (Humble for a non-person who, incidentally, scored in the 90th percentile on the GRE, right?)