Sam Altman Says AI Does Not Need a Congregation

Sam Altman has a warning for anyone getting a little too spiritual about the chatbot.

On Saturday, he said treating AI models as religious authorities, or surrendering human judgment to them, creates a real safety problem. He did not name a rival or identify a particular incident.

The timing is interesting. OpenAI has just introduced dots, assistants designed to work around the clock and learn their users' preferences. Its own launch notes tell people to review consequential work.

More capable assistants will need more trust. Altman's point is that trust should still have a ceiling. The machine does not get a halo.