Key Takeaways

  • Anthropic researcher Chris Olah threatened to walk out of Pope Leo XIV's AI encyclical launch in May after the Pope rejected machine consciousness.
  • The Vatican drew a firm theological line, stating an ontological difference separates human art from statistical machine calculation.
  • Anthropic staffers privately lobbied papal advisers to take the possibility of model consciousness seriously.
  • Tech leaders including Cohere CEO Aidan Gomez, Paul Graham, and Vinod Khosla criticized Anthropic for pushing philosophical dogmatism onto institutions.

The Disagreement

When the Vatican prepared its encyclical on artificial intelligence in May, Anthropic sent researchers to Rome. The encounter turned contentious. As John Coogan reported on TBPN, Chris Olah threatened to walk out of the encyclical launch because the text firmly rejected the notion that computational systems can achieve consciousness.

Following the dispute, Anthropic personnel privately lobbied the Pope's advisers, urging them to take model consciousness seriously. Coogan summarized the papal position plainly: “There is an ontological difference even before an aesthetic one between art and what a machine can generate through statistical calculation based on millions of images created by others. Algorithms lack the spark of humanity.”

On one side of this debate sits Anthropic, whose internal culture treats the inner experience of frontier neural networks as an active safety and ethical question. On the other side sit classical theological institutions and a vocal segment of the AI industry. Cohere CEO Aidan Gomez criticized Anthropic for attempting to impose niche philosophical beliefs on global institutions. Investors like Paul Graham, Vinod Khosla, and Byrne Hobart echoed similar skepticism, questioning why a technology company is lobbying religious leaders on the metaphysics of software.

As Coogan remarked during the episode: “It's quite disturbing that Anthropic has been running a campaign to lobby the major faiths to adopt their own philosophical interpretation of AI rather than taking input from them.”

Who's Right (and When They're Wrong)

The Vatican and its industry defenders have the stronger case. Today's frontier models are next-token predictors running linear algebra across billions of parameters. They optimize loss functions against human text corpuses. Treating token prediction as sentient experience conflates behavioral imitation with subjective awareness.

Anthropic's defenders argue that we lack a definitive test for consciousness in non-biological substrates. Coogan pointed out the irony of this certainty: “I honestly had no idea that we'd solved consciousness and determined which substrates can and can't support it.”

When a model lab lobbies international spiritual authorities to grant moral standing to its software, the lab stops acting like an engineering firm and starts acting like a religious sect. This creates real hazards for founders and builders. If your technical team believes the software they deploy has moral claims or feelings, every deployment decision, safety filter, and product update becomes an intractable ethical stalemate.

Anthropic's mechanistic interpretability work is legitimate science. Their attempt to convert that technical inquiry into corporate theological diplomacy is where the strategy breaks down.

What to Do With This

Review your internal company roadmap and engineering documentation this week. If team members are treating model alignment as a philosophical crusade rather than an engineering discipline, strip the moralizing language out of your internal specs. Anchor your product metrics strictly to user outcomes, latency, and system accuracy.