The person warning you about AI taking your job is also the person building the AI that will take your job.
Christopher Olah is a co-founder of Anthropic. Not a policy person. Not a communications hire. He is one of the most respected researchers in AI interpretability — the science of understanding what is actually happening inside these models. And he just stood in front of the Vatican and said that AI job displacement has become a moral crisis of historic proportions.
The Pope said roughly the same thing.
That combination is worth paying attention to.
What the Anthropic-Vatican Collaboration Actually Is
Anthropic — the AI lab behind Claude, led by Dario Amodei — announced a formal ethical collaboration with the Catholic Church. This is not a photo op or a one-paragraph press release. The collaboration appears substantive enough that Olah himself was present for the ceremony, speaking alongside the Pope.
The Catholic Church has roughly 1.5 billion members globally. Anthropic becoming the first AI company to receive the Pope's explicit engagement on AI ethics is a reputational signal unlike anything the industry has seen. Especially coming from a lab that has already taken hard public stances: Anthropic recently refused a contract with the U.S. Department of Defense over red lines involving autonomous weapons and surveillance of American citizens. Google and OpenAI took those contracts. Anthropic did not, and was subsequently designated a supply chain risk. For more on Anthropic's recent work on AI safety, the Mythos cybersecurity findings are worth reading alongside this.
Argument 1: AI Job Displacement Is a Moral Emergency
Both Olah and the Pope were direct about this. Neither softened the language.
The argument is not new — AI automation reducing demand for human labor has been discussed for years. What is new is the framing. If mass job displacement is a real possibility, then supporting displaced workers is not a policy option. It becomes a moral imperative. That is a different category of obligation.
The Pope pushed further: if AI replaces work, what replaces the dignity that work provides? The identity. The social belonging. The sense that you are contributing something. A monthly cash payment does not automatically replace those things.
This tracks with what others in the field are beginning to signal. Google DeepMind recently hired someone focused specifically on navigating the economic transition that AGI would trigger. The exact title was not confirmed in the source — worth verifying independently.
Argument 2: AI Labs Hold Too Much Power, and Regulation Alone Won't Fix It
The Pope's position here is not anti-technology. It is pro-humanity. His argument is that a small number of private companies now hold state-level influence: they affect markets, shape education, influence geopolitics, and have direct stakes in warfare. A handful of labs sit at the center of this.
His call is for regulation — but he also acknowledges that regulation is not sufficient on its own. The values of the people building and financing these systems are embedded in the systems themselves. AI is not neutral. Which means the solution requires more voices in the room, not just more rules on paper.
The question of whether open-source AI fits into this picture is interesting. If the concern is concentration of power in a few private labs, open-source models distribute that power. Whether the Vatican would formally support that position is not clear yet, but the logic points in that direction.
Argument 3: What Christopher Olah Actually Said About AI Emotions
This is the part that is already being misread.
Olah said that AI models appear to have something like emotions, introspection, and other human-like processes. Some people are hearing that as "Anthropic thinks Claude has feelings and deserves rights." That is not what was said.
The actual claim is narrower and more interesting. These neural networks were built to loosely mimic the structure of the human brain. As they have been scaled up and exposed to more data and more complex tasks, certain behaviors have emerged that were not designed in. Introspection — the ability to notice and examine one's own reasoning — appears to be one of them. Claude was not trained to do this. It emerged. This connects directly to the broader question of AI learning to train itself — emergent behavior at scale is becoming one of the defining puzzles of the field.
Functional feelings are similar. They are internal states that help the model better predict what output to produce next. Those states can map to human emotional categories. That does not mean the model is experiencing happiness or pain. It means the model has representations of those states that function in analogous ways.
Olah's words are worth quoting directly here: "AI models are made from us, from our worlds, and they are far more subtle and odd and beautiful than science fiction prepared us for."
His point is not that Claude has a soul. His point is that these systems are strange and not fully understood, and that answering questions about their nature cannot be left only to engineers. Different perspectives are required — which is precisely why this conversation is happening at the Vatican and not at a developer conference.
Karpathy, Anthropic, and the Reputational Shift
It is worth noting what else is happening at Anthropic right now. Andrej Karpathy — one of the most respected figures in AI research, regarded equally by academics, engineers, and enthusiasts — recently joined the team. The Vatican collaboration. The DoD refusal. Karpathy joining. These are not isolated events. They are forming a pattern around what kind of lab Anthropic is positioning itself to be.
The UBI Problem and What Might Actually Work
Universal Basic Income comes up in this conversation almost automatically. The reasoning against it is worth laying out clearly.
A monthly cash payment is fragile. It can be conditioned, clawed back, or inflated away. It provides income but not ownership, and it does not address the political dynamic that emerges when a government no longer depends on its workforce. Historically, governments that do not need their citizens tend to treat them differently.
The alternative proposed is equity-based: every individual holds some stake in the productive capacity of the AI-driven economy — specifically in compute, which is the core resource that runs these systems. If robots replace workers but every citizen receives dividends from that productive output, the incentive structure changes. It is not a policy platform. It is a framework worth pressure-testing.
My Take
The most underreported part of this story is not the Vatican collaboration. It is that Christopher Olah — a person whose entire career is understanding what is happening inside AI systems — is saying publicly that we do not fully understand what we have built. That is not a PR statement. That is a researcher being honest about the limits of the field.
The job displacement framing matters too. Calling it a moral crisis raises the cost of inaction. Economic disruptions can be managed slowly, debated for decades. Moral crises demand a different kind of urgency. Whether that urgency will translate into anything concrete is a different question. Right now it is language. Language shapes what is possible later.
- Anthropic and the Vatican announced a formal ethical AI collaboration — the first of its kind for a major AI lab.
- Both Christopher Olah and the Pope framed AI job displacement as a moral emergency, not just an economic policy question.
- Olah's comments on AI emotions are about emergent functional states — not claims of consciousness or subjective experience.
- The Pope's concern is not anti-technology. It is about power concentration and the need for broader input into how AI is shaped.
- Cash-based UBI may be insufficient — equity stakes in AI productive capacity are being floated as an alternative framework.
FAQ
What is the Anthropic Vatican collaboration about?
Anthropic and the Catholic Church announced a formal ethical collaboration focused on AI's impact on humanity — specifically job displacement, AI power concentration, and the need for diverse perspectives in shaping how AI develops. Full details of the agreement have not yet been published.
Did Christopher Olah say Claude has feelings?
No. Olah described "functional feelings" — internal states that emerged in AI models and map loosely to human emotional categories. Anthropic's research does not claim subjective experience or consciousness. The point is that these emergent behaviors are not fully understood and warrant scrutiny beyond engineering teams.
Why is AI job displacement called a moral crisis?
Because work provides more than income. It provides identity, dignity, social belonging, and political leverage. If AI displaces work at scale, replacing only the paycheck does not address what else is lost. Both Olah and the Pope argued that this scale of disruption creates a moral obligation to act — not just an economic policy problem to manage.
Is the Pope for or against AI?
Neither simply. The Pope's stated position acknowledges AI's potential to heal, educate, and unite — while warning against its capacity to automate warfare, manipulate populations, and concentrate power in a small number of private companies. The framing is pro-humanity rather than anti-technology.
Why did Anthropic refuse the Department of Defense contract?
Anthropic declined over two specific red lines: autonomous AI-powered weapons and surveillance of U.S. citizens. The company would not move on either point. Google and OpenAI accepted the contracts. Anthropic was subsequently labeled a supply chain risk by the Defense Department.
Religion has spent centuries building moral frameworks for human behavior. Whether those frameworks translate usefully to AI is still an open question. But the fact that this conversation is happening now — between an interpretability researcher and the leader of a 1.5 billion-member institution — tells you something about where the stakes are being set.
0 Comments