At the moment I'm using it for local history research. I feed it all the text I can find on an event (mostly newspaper articles and other primary sources, occasionally quotes from secondary sources) and I prompt with something like "Summarize this document in a concise and direct style. Focus on the main points and key details. Maintain a neutral, objective voice." Then I hack at it until I'm happy (mostly I cut stuff). Analysis, I do the other way around: I write the first draft, then ask the AI to polish. Then I go back and forth a few times until I'm happy with that paragraph.
I'm not going anywhere with this really, I'm just musing out loud. Am I contributing to a tragedy of the commons by writing about 18th century enclosures? Because that would be ironic.
Reading and writing are mental processes (with or without advanced technology) that shape our collective mind.
So there's a human in the loop who is prepared to vouch for those sentences. They're not 100% human-written, but they are 100% human-approved. I haven't just connected my blog to a Markov chain firehose and walked away.
Am I still adding to the AI smog? idk. I imagine that, at a bare minimum, its way of organising text bleeds through no matter how much editing I do.