AI Disclaimer

My use of generative AI in creating content for Substack and other media.

Of late I’ve seen some articles published with an “AI Disclaimer” included at the bottom of the work.

I decided against doing that because I’m afraid that I will probably forget to regularly include a disclaimer. I worry that lack of consistency will make the whole body of work look disingenuous. Instead, I will just lay it all out here and update the use as is warranted.

Types of Uses as of September 2023:

1. Images

   1. Almost all of the images included are AI generated. The visual art I create and use is not meant to be the focus of the work and I rarely spend more than a few minutes on it. When a work of art or design that is not of my own creation is used, that work and its creators are credited.

   2. (Edit 9/14/2023) I have often said that I have not found generative artificial intelligence to be biased. In fact, I’ve noted that it seems that it is often more conscious of bias and potentially problematic areas than I am or find others to be. However, a recent report by Bloomberg on internal research demonstrated that the generative AI models used for images are incredibly biased, to the point of entrenching misogyny and racism. (For example, while 38% of American physicians are female, only 7% of AI generated images of physician were female.) While I don’t believe that a computer program has the capacity to feel bigotry, it’s foolish to assert that it doesn’t have the power to further it. My experience is extensive with LLMs, but not so with the image generators. From here on out I will acknowledge the difference when I write about it. The report can be found here.

2. Evaluation of Work

   1. Organization

      1. I am not, beyond perhaps the molecular level, terribly well-organized. And that character flaw is often a weakness in my writing. I have occasionally used generative AI to better organize the flow of my writing so that someone need not be, for example, a caffeine-addled hummingbird in order to find the work more coherent.

   2. Segues

      1. I again refer you to the hummingbird analogy. I am prone to making logical (or not-so-logical) jumps of meaning or analogy. Asking generative AI to evaluate my work has been helpful in that it often suggests clearer (or simply the creation of) segues between ideas.

   3. Grammar

      1. I think grammar is a lot more flexible that Mr. Strunk and Mr. White might accept. Sometimes it’s because I think punctuation should serve me and not some higher, judgmental power. Or because I begin a sentence with a contraction. But I’d never let an incomplete sentence go uncorrected! Perish the thought.

      2. I admit LLMs have been fairly tolerant of my pliable relationship with the English language, which I appreciate.

3. Idea Generation

   1. Not really my style to plagiarize. I’ve toyed with the possibilities of idea generation (I’ve definitely asked a chatbot for ideas) but I can’t recall any that really excited me enough to lead to an article. I won’t swear to it, but skimming the things I’ve written I’m fairly confident I can point to specific inspiration from the real world that led to each piece.

4. Demonstrations and Use Cases

   1. At the risk of stating the obvious: Sometimes I demonstrate use of LLMs in the work. I endeavor to be completely straightforward and transparent about the times when it’s used for such demonstrations.