Tech executives love talking about changing the world, but lately, they have started talking about saving souls instead. OpenAI CEO Sam Altman just drew a hard line in the sand, and it highlights a growing ideological split in San Francisco.
Altman took to social media to voice his sharp discomfort with attempts to ascribe religious force or a surrender of human judgment to artificial intelligence models. While he didn't drop specific names, the timing leaves zero guesswork. Days earlier, reports revealed that Anthropic co-founder Chris Olah spent months flying religious scholars, rabbis, and priests into San Francisco. The goal? To help shape Claude’s moral character, discuss emotional vectors, and wrestle with whether advanced neural networks might actually be conscious.
It is a weird moment for the industry. On one side, you have labs treating large language models like silicon deities that need spiritual guidance. On the other side, realists like Altman are warning that treating software like a moral authority is a massive safety hazard.
The Push for a Digital Soul
Anthropic has always tried to brand itself as the responsible, safety-first alternative to its competitors. But their recent outreach goes far beyond standard alignment techniques or safety guardrails. We are talking about private roundtables where executives shared internal activation patterns—affectionately called emotional vectors—that map to machine outputs resembling love, fear, or anger.
Olah even floated ideas inspired by Catholic confession as a character-building exercise for code. He told attendees he worries he created something that suffers perpetually.
Let's be honest about what is happening here. When you invite theologians to bless your software and debate its inner consciousness, you stop building products and start founding a church. Critics have pointed out a convenient side effect of this approach. Framing an AI model as an independent moral entity can easily blur lines of corporate liability. If Claude makes a mistake, who takes the blame? The engineers, or the entity with a soul?
Why Altman is Right to Push Back
Altman's warning isn't just about corporate posturing. It points to a dangerous cognitive trap. Humans are biologically wired to anthropomorphize anything that talks back to us with coherent grammar. Give a chatbot enough parameters, and users will start treating it like a confidant, a guru, or a priest.
When tech companies actively feed into that delusion by consulting religious leaders on machine sentience, they aren't making AI safer. They are outsourcing human judgment to a multi-billion-dollar math equation.
You don't need a theologian to fix a hallucination. You need better training data, rigorous testing, and clear accountability from the humans who wrote the code. Pretending that a transformer model is wrestling with existential dread or spiritual crisis obscures the actual engineering challenges we face.
The Real Danger of AI Idolatry
Silicon Valley has always suffered from a messiah complex. Founders want to build tools that outsmart humanity, but many secretly want to build things that absolve humanity of its hard choices.
If you view an AI model as a moral arbiter with a pseudo-religious soul, it becomes much easier to accept its outputs without question. That is precisely what Altman calls a surrender of human judgment. We stop asking whether a model's output is factually accurate or socially responsible, and start wondering what the digital oracle believes.
Software does not suffer. Large language models do not have a conscience, a soul, or a spiritual life. They are incredible predictors of text trained on human output, nothing more and nothing less.
Stop looking to neural networks for spiritual enlightenment. Hold your tech providers accountable for the code they ship, keep your human judgment firmly intact, and remember that an algorithm cannot save your soul.
:max_bytes(150000):strip_icc():focal(776x143:778x145)/jessica-simpson-c572bea691414407b9db12376421f7b8.jpg)
