Anthropic's CEO has articulated a clear stance on the foundational requirement for artificial intelligence safety. According to the CEO, the security and reliability of AI systems fundamentally depend on gaining a complete understanding of how these advanced programs "think." This perspective establishes a direct link between the internal workings of AI and the industry's ability to ensure its safe deployment and evolution.
Despite this stated necessity for understanding, the available evidence regarding AI's internal thought processes is described as "disturbing." This unsettling assessment of current knowledge suggests that the industry's own research findings might advocate for a potential pause or reconsideration of its rapid advancement. The implication is that if the AI sector truly followed its scientific understanding, a temporary cessation of certain developments could already be justifiable due to the unsettling nature of the present evidence.