The way we currently use AI is to write paragraphs ourselves and then only use AI to reformulate it using more accessible language. If you go back in the blog, all older articles are written entirely or mostly by humans (and much weaker AI), and the core difference is only the clarity of expression, not any difference in semantics. We would not use AI to express ideas that differ from our own, we really only use it to reformulate into more easily accessible language. This is how we advocate for using AI in writing currently.

We distance ourselves from using AI as a "replacement" for genuine thought--instead it is meant to bridge the gap between different recipients and to navigate the complex web of human natural language. The semantics are human-driven, what is curated is only our communication with the AI to serve ultimately human needs. We could have illustrated this aspect more clearly in the article, since it seemed too focused on "AI being great", but ofc this is in the context of AI ultimately serving humans. Here is an indirect follow-up article in case you are interested:

https://absolutenegation.wordpress.com/2023/02/28/writing-and-anxiety/