What I don't do

• I will not use LLMs to produce any text that makes it into my final product. That means no asking it to "find a better way to word this point", and no "can you write this section for me". The writing on the page is fully my own. If I didn't want to write, I wouldn't be a writer.
• I do not formulate any thesis with an AI tool. The points I make should be my own.
• Additionally, I do not produce outlines with LLMs, either.

What I might do

• I don't trust outputs of chatbots or other LLM tools as being useful knowledge. However, I may occasionally use an LLM as an advanced search engine. The state of even the best search engines is pretty pitiful these days. LLM tools are able to ingest a large number of web pages in short order, then spit back up a link to the one that most closely aligned with the question I asked it.
  • When doing this, I absolutely, under no circumstances trust the output of the LLM. Instead, I click the link, see if the source seems trustworthy, and then listen to the source.
  • Doing this too much can warp your brain, so I usually only do it when I'm looking for something obscure that I can't find, or looking for a source that I forgot to bookmark and need to resurface based on my memory.
• I might occasionally feed my transcripts into an LLM to see if it finds any obvious syntactical errors. Things like repeated words, incorrect words that aren't misspelled ('there' instead of 'their'), things like that. When doing this, I do not accept corrections regarding my thesis, outline, or delivery. I don't usually do this at all, but it's an option on the table, thus including it in my policy.

Images and quotations

• I will not knowingly include AI generated images in my works, unless done to demonstrate some point about the AI system that generated it, or the people who use it, and it will be clearly labelled as such if so.
• I include quotations of other writers in my works sometimes. It's impossible for me to know if these authors used LLMs to generate their works, though anything clearly LLM generated is probably not worth my time to reference in my own writing.
• I might knowingly include LLM-generated writing as a quotation in my works, in order to demonstrate something about LLM systems, or perhaps about the people who use those tools in their writings. These quotations will be clearly labelled as being partially or fully LLM-generated, if so.