Three lessons I learned from 10,000 generative AI prompts

For the month of October, I used generative A.I. writing tools to help me write. I used them to help me write sections in an academic book, emails, an academic article, and brainstorming. I wanted to make sure I got to about 10,000 prompts (but really, it’s hard to define what a prompt is, so it’s probably lower than that if you define a prompt as more than clicking the run button. Sometimes, I clicked “run” repeatedly to see if I got different results. (The short answer is no, but the longer answer is yes.) I used a variety of programs, some free, others were paid. I learned a lot and got really frustrated. As many have written, generative AI really isn’t going to replace writing. It certainly could, maybe in 5-8 years, but that’s not really the point.

Neither is plagiarism. I am bored with the plagiarism conversations for a lot reasons, including the moving target of plagiarism (people thought spell checkers constituted cheating in the 1980s) and because if you can get a machine to write something of value for you, that’s valuable. I’ll mention this again at the end of this post. Here are three lessons I learned.

First lesson learned: Generative A.I. inculcated five habits when I used it to write. These habits were iteration, waiting, tinkering, editing, and randomness. I’m pretty sure I do all of those when I write anyways.

1)     Iteration: You have to keep trying, tweaking, and experimenting. The more you iterate, the better the results. This is true for all of writing, of course, but it’s especially true for generative A.I. writing.

2)     Waiting: Sometimes, you have to let the machine do its thing, and trust it will come up with something good.

3)     Tinkering: Tinkering is important. You have to fiddle with settings, inputs, and outputs. The more you tinker, the better the results. Imagine using Word or Google docs but never touching any of the formatting.

4)     Editing: Editing is important. Generative AI consistently adds all sorts of weird extra words in random places. But I do this on my first draft.

5)     Randomness: Sometimes, the machines get repetitive that I found myself adding in random thoughts other stuff just to see what it wrote. But….this is true for most writing.

I found that these habits are all useful for writing even if I’m not using a machine.

Second Lesson learned: The best way to use generative A.I. is to use it as a tool, not a dictator.

I found it was best used when I tired and had already written a lot. In fact, this is when I learned the most because I was able to generate a bunch of random ideas. I typically write my best in the morning. In the afternoons, I run out of gas. The afternoons became tinkering time.

Third lesson learned: What is the purpose of writing?

Using generative AI made me think about what I was trying to do with my writing. I began to feel the machine was my audience and that I was trying to convince the machine to say something novel or unexpected. The machine didn’t respond.

I began, then, to reflect about what writing is and what it means. This is where I return to the idea of doing something with machines is valuable for its own sake. I have never lamented writing an essay by hand (but maybe I should). I was taking a tool-based approach to generative AI but realized writing is also a tool and mediates much of my world.

I was valorizing writing while missing writing is about cognitive expansion. Writing mediates my thoughts, and it (re)shapes my thoughts, and it’s about using tools to do so engage in that nonlinear, recursive process of revision. It’s frustrating and it will always be. Or, simply put, it’s writing.