‘We find evidence of introspection’: The unsettling admission of Anthropic co-founder

Anthropic's billionaire co-founder and self-described atheist, Chris Olah's visit to the Vatican was probably one of the most consequential PR campaigns for Anthropic. Standing alongside Pope Leo, the AI lab’s interpretability research division leader, Olah, called for collaboration between tech developers and spiritual authorities. He also noted that Anthropic's AI models are showing more than computing ability.

.“And I will be honest: we keep finding things that are mysterious, even unsettling. We find structures that mirror results from human neuroscience. We find evidence of introspection. We find internal states that functionally mirror joy, satisfaction, fear, grief, and unease. I don’t know what that means, but I think it warrants ongoing discernment,” said Olah.

Anthropic eventhough maintains ambivalence, continues to flirt with the idea of a conscious AI. Earlier this year, Anthropic published a “constitution” for Claude, a document outlining “the kind of entity we would like Claude to be.” It specifies that eventhough Anthropic currently refers to Claude as “it”, they do not dismiss the possibility of Anthropic being a subject. Anthropic has now, since its founding, evolved into a “devil's advocate”. The company was founded by former OpenAI employees specifically to prioritise AI safety over profit, but it has embedded itself within the most high-stakes arena of human conflict – the United States military and intelligence complex. While it warns about the unsettling nature of its model, it gladly accepts defence money and partnership.

Vatican's entry into AI chat seems unsettling

The Vatican's entry into the AI discourse is creating a profound sense of unease because it hints that AI has transitioned from a commercial tech boom to a deeply unpredictable existential and moral crisis. The very people who are building the AI infrastructure are struggling with the ability to govern it. The Claude Mythos is a massive leap past standard LLMs, had shown troubling symptoms. While Pope Leo has been critical of the AI, it is of course possible that the Vatican’s official stance on AI will evolve. Then, by combining spiritual authority with cutting-edge technology and introducing religious framing into AI, we risk turning a digital tool into an entity that is capable of soul, consciousness and divine proxy, even if it merely mimics human activity.