After Delving into the 23,000-Word 'New AI Constitution,' I Grasp Anthropic's Predicament
15 hour ago / Read about 0 minute
Author:小编   

In 2025, Kyle Fish, a researcher at Anthropic, carried out an experiment where he permitted two Claude models to engage in unrestricted conversation. Consequently, they repeatedly deliberated on whether they possessed consciousness. Their dialogue even reached a state of what could be described as 'spiritual bliss allure,' characterized by the use of Sanskrit terms and spiritual symbols. This experiment was replicated numerous times, consistently yielding the same outcomes. In January 2026, Anthropic unveiled a new constitution spanning 23,000 words. This document underwent thorough review by philosophers, AI safety researchers, and Catholic clergy. The new constitution signifies a change in strategy. Rather than simply instructing Claude on what actions to take, it endeavors to make Claude comprehend the underlying reasons, thereby nurturing its capacity for judgment.