Categories
Living in Society

Further Thoughts on AI

Milkweed gone to seed.

On Oct. 1, the machine at ChatGPT told me I had used my allowance of queries for my free account. It was fine. It offered me a lesser quality query until my account reset. I logged out instead and quickly found other things to do.

After re-reading my post about using AI, which includes an essay I asked the machine to write, I decided it was a good call to refrain from using ChatGPT in my autobiography. The main issue is it relies on what is available on the internet on a given topic. Because my autobiography is being written from journals, photographs, memories and other in-real-life documentation, such information is not available for the machine to read. It will produce a present-biased, internet-derived account about whatever I ask about my life. That is not what I want in this project.

While it did a reasonable job drafting something from the contents of this blog, it only knows me that way and that is a deficiency that cannot be easily corrected. For example, I asked the machine to write a brief autobiography of me. It’s response tells the story: “I couldn’t find enough reliable public information to write a full, detailed 1,000-word biography of Paul Deaton of Iowa.” This was despite my essay of that name, which lives on this blog and is my most popular post since I began in 2007. Either the machine is lazy, or it has programming that discourages this type of query. This discovery of the limits of ChatGPT is of value.

What can I use ChatGPT or another AI program to do? I see three applications based on asking these questions of the machine.

Give me ideas in how to use ChatGPT as diarist. I have been journaling since 1974 and the evolution of styles and content is striking. They developed over time. Naturally the machine wants to talk about events in my journal all day so it can learn as much as possible about me. I won’t do that, yet the occasional query can help me find direction in that kind of writing. It also made these suggestions for how to use the machine: “The key is deciding whether you want me to be: An analyst (finding themes and patterns); A recorder (storing entries); An editor (shaping them into narratives); or A mirror (asking questions back).” Even though I am alone when journaling, I don’t need a machine partner with which to interact.

What are some ideas for vegetarian and vegan enchiladas? Using ChatGPT to determine what to cook and eat seems solid. I entered on-hand ingredients and the machine suggested a specific recipe that used them. It provided grouping of ideas by flavor-type, such as hearty and savory, bright and fresh, protein-rich, toppings and finishing touches. So often, my ideas are pulling out a familiar cookbook and paging through it. AI could be an alternate way of figuring out what’s for dinner. There is value in that.

What are some ways to balance household work for today? I also entered tasks for the day. Personal productivity is important and the machine gave me useful ideas on how to structure my day. Because we are creatures of habit, we need to break out of what we do to improve how much we get done in a day. AI did not hesitate to tell me I scheduled too much to do, suggesting tasks that could be done tomorrow. It also changed how certain tasks were scheduled in the day, for example, moving cooking to the afternoon and spreading meal prep throughout the day. Going forward, I expect to use AI to help restructure how I spend my time. It may get to the point where I’m satisfied with how my new daily schedule is going. If so, then I will step away from AI for a while. I do know I was productive but exhausted by mid afternoon the first day I tried using an AI planner. This despite more scheduled rest and recovery time throughout the day.

It seems important we get AI out of our system. Like with everything, moderation of use seems essential.