If someone is able to put words in an order that sets my soul to singing, I don't care how they arrived at it -- unless they got it from someone else and are acting like it is their own.

If llms can be prompted to produce truly insightful writing that changes how I understand the world and the colors I see, then the prompters are welcome to the credit of such writing (because it certainly isn't a one-shot right now and requires thought to make it produce insight).

Given the trajectory of llms, it is likely they will in short order be producing writing that is very difficult to distinguish from genuinely thoughtful writing. It may still be slop, but it will be well-hidden slop -- which is to say it will not be so different from most of the writing currently produced by humans.

If, however, llms get to the point where very little prompting is required to open new fields of discovery in my mind, then we are confronted with a new problem: are such models deserving of some sort of status beyond a dumb tool?

If it is a problem for a person to post llm writing without attribution, it means llms are more than a dumb tool. There has to be a person to whom the attribution needs to be made.

This, for me, is the core of the debate: are llms a tool or are they an entity? If tool, no attribution necessary -- if an entity, attribution should be made - but also !!!!!!

I think any pre and post production use of AI doesn't need to be disclosed (which looks to be everything but (1) and (4)). Although, if I were to ask a friend for production help, I'd cite them out of courtesy and you could argue we owe that to the LLM researchers.

My line is when you begin copy and pasting things from LLM output as if it's your own work unless it's something formatting related like make this list a markdown list. I just see it as dishonest to share thoughts/content where it's assumed to be from your own brain when it isn't from your own brain and not disclose it. I read most things knowing that I'm taking an attention-risk and that calculation is affected by how much attention-cost the writer paid to produce a work.

As you say, this may just be where norms are currently. Eventually, maybe no one will think their own thoughts and produce their own content beyond the prompt and there won't be this mismatch of expectations. Instead, real writers will disclose the tools they didn't use.

I don't think I would do this. From what I've heard about current models is that they are often wrong but also very confident.

I ask ai to write me a xxx word post about xxx and then I edit the slop to suit my needs, it ALWAYS needs attention, editing and change.

I find that there is a lot of sentence construction I like and it would take me xxx time to create that organically.

So if I edit ai slop, is it still slop or is it original, or is it a hybrid?!

If it can't be enforced then I'm not really sure what the point of the rule is