Anthropic co-founder Christopher Olah has warned of mysterious and unsettling structures inside AI models at the launch of Pope Leo XIV’s encyclical Magnifica Humanitas.
Speaking at the Vatican event, Chris Olah shared disturbing findings from his team’s research on Claude Sonnet 4.5. During experimental phases, researchers found 171 “emotion vectors” and neural patterns emerging from training on human text.
Olah stated that “the team found structures mirroring results from human neuroscience. We discovered evidence of introspection.”
He also noted the discovery of internal states reflecting joy, satisfaction, fear, grief, and unease.
Given these findings, Olah called for moral discernment beyond tech firms. He urged for earnest, thoughtful critics to challenge dominant companies and help steer AI’s creation in a positive direction.
Pope Leo XIV warned about AI’s growing impact and its disastrous consequences tied to humanity’s future and dignity at the encyclical launch. He framed AI advancements as a modern-day “Tower of Babel,” potentially leading to singular power desires. The Pontiff urged the international community to control AI development for shared human benefits.