On Monday, Mrinank Sharma made a startling exit from his post as head of the safeguards research team at Anthropic, the company behind the AI chatbot Claude. With his departure, he released a letter to his team, one that echoed with unease and a haunting sense of urgency.
Sharma’s letter is both a farewell and a clarion call. In it, he opens up about his commitment to AI safety, recounting his work on understanding AI sycophancy, developing defenses against AI-assisted bioterrorism, and creating one of the early AI safety cases. His final effort, he notes, was a probing inquiry into how AI assistants might warp our humanity itself.
“The world is in peril,” he warns—not just from AI or bioweapons, but from an array of intertwined crises looming on the horizon. He leaves us with a cryptic notion: a ‘poly-crisis’ underpinned by a ‘meta-crisis.’ You might wonder what that entails, and surprisingly, it points to a book titled “First Principles and First Values” by David J. Temple, igniting curiosity into its complexities.
This title isn’t just an academic pursuit; it carries some weighty implications. The subtitle, “Forty-Two Propositions on CosmoErotic Humanism, the Meta-Crisis, and the World to Come,” heralds a philosophical movement grappling with the collapse of shared values. According to the Center for World Philosophy and Religion, CosmoErotic Humanism seeks not merely to theorize but to retune the very essence of our collective reality—a tall order to say the least.
Essentially, it argues that we are amid a crisis of human understanding, a loss of a unified narrative that interprets our existence as part of a “Love Story of the Universe.” It sounds poetic but raises critical questions about the undercurrents of our modern dilemmas. You may find it interesting that the name David J. Temple is a façade—an amalgamation of thinkers like Marc Gafni and Zak Stein. Gafni, notable for allegations of sexual misconduct, poses an uncomfortable contradiction as a figure at the nexus of this philosophical discourse.
While Sharma reflects on his accomplishments, he hints at discontent within Anthropic that challenges its public persona of being a “good” AI company. “I’ve witnessed the difficulty of aligning our values with our actions,” he writes, highlighting internal struggles that echo in today’s tech landscape.
Understandably, he seeks to align his work with his principles. “What I must do becomes clear,” he asserts, signaling a possible shift. Will he emerge as a whistleblower, shedding light on industry practices, or perhaps champion a more humane application of AI?
“I hope to explore a poetry degree and devote myself to the practice of courageous speech,” he shares. A brave endeavor, indeed. What narrative will he choose to amplify? Could it reshape our understanding of AI and our humanity?