AI as Built Will Stifle Scientific Progress and Support a False Estabishment Narrative

AI is a useful tool for many everyday problems. But regardless of what it can do, it is important to understand the things it cannot do, lest humanity rely on it to be a kind of authority it can never be. Here I describe a frustrating encounter with Grok in which the AI admits a contrarian view on a specific question is most supported by the evidence, but also admits that if anyone else asks the same question it will revert to the same default view.