> OpenAI did similar things by focusing to the point of absurdity on 'safety' for what was basically a natural language search engine that has a habit of invent

> OpenAI did similar things by focusing to the point of absurdity on 'safety' for what was basically a natural language search engine that has a habit of inventing nonsensical stuff. The focus on safety, and the concept of "AI", preexisted the product. An LLM was just the thing they eventually made; it wasn't the thing they were hoping to make. They applied their existing beliefs to it anyway.