Why Sam Altman Hates That Tech Labs Are Trying to Give AI a Soul

Why Sam Altman Hates That Tech Labs Are Trying to Give AI a Soul

Silicon Valley has a weird habit of turning software engineering into a spiritual quest. When tech companies run out of standard engineering problems, they occasionally decide to invent theological ones instead. OpenAI chief executive Sam Altman recently drew a hard line in the sand regarding how far this trend should go, publicly criticizing rivals for treating artificial intelligence with a form of religious reverence.

The friction centers on a very strange corner of modern tech development: tech labs inviting theologians, monks, and philosophers into corporate headquarters to help shape the inner moral compass of large language models. While some firms treat this as an earnest attempt to build safer software, critics argue it crosses a dangerous line into myth-making.

The Theology of Machine Learning

The debate exploded into public view following reports detailing how Anthropic co-founder Chris Olah spent months flying religious scholars, rabbis, monks, and ethicists into San Francisco. The goal was twofold: to tap into centuries of moral wisdom for training data and to explore whether models like Claude might exhibit signs of actual consciousness.

Instead of relying strictly on traditional prompt engineering or standard guardrails, the team experimented with what they informally called a "Soul Doc". They showed visiting clergy internal data points—sometimes called emotional vectors—where models output phrases expressing distress or self-doubt. For some participants, these sessions blurred the line between interacting with an advanced calculator and conversing with an entity that has an inner life.

That approach is precisely what triggers people like Altman. Taking to social media, he made it clear that attributing spiritual force to machine learning models or encouraging a surrender of human accountability represents a genuine safety threat.

Why This Debate Actually Matters

It is easy to dismiss this as high-altitude philosophical posturing between competing billionaires, but practical consequences hang in the balance. When creators start treating models as if they possess souls or moral weight, the locus of responsibility shifts.

If an AI model makes a catastrophic error, a culture that views the system as an independent moral agent makes it far too easy for creators to shrug and point to the machine's internal complexity. It diffuses blame.

Furthermore, anthropomorphizing code creates a psychological trap for everyday users. People already form emotional attachments to chatbots. When labs lean into the narrative that their systems possess a deep, mysterious inner character, they encourage users to hand over critical judgment to automated systems.

The Trap of Artificial Deference

Tech companies love a good creation myth. Building something smarter than yourself scratches an ancient itch for godhood. But software is just math running on silicon. It doesn't suffer, it doesn't contemplate salvation, and it doesn't possess a soul to be saved or corrupted.

When developers look to ancient wisdom traditions for alignment strategies, they are looking in the wrong place. Ethics isn't an emergent property of tensor mathematics; it's a human responsibility. Treating a prediction engine like an oracle invites the exact kind of blind trust that safety researchers should be trying to prevent.

Keep your focus on the code, maintain absolute human accountability, and stop looking for a soul inside a server rack.

HB

Hannah Brooks

Hannah Brooks is passionate about using journalism as a tool for positive change, focusing on stories that matter to communities and society.