I'm working on a little writing project and I'm a terrible editor, especially of my own work. I'll likely make a mistake in this toot. I let ChatGPT proofread it. The revision was good.
I tried it two different text prompts: 1. "Proofread this: xxx," 2. "Proofread this in the style of Bryan Hansel: xxx."
Surprisingly, the second is closer to how I write. I still had to change things a bit, but it did catch errors that my head just voiced right over.
#AI is so wild.
@bryanhansel it’s going to be interesting to see how my students papers sound this spring
@daphsci I recently read about a tool that detects AI. If I remember right, it was designed to sniff out AI in college papers.
@bryanhansel I will look into that. I think I will take the approach a colleague suggested— talk to the students about it from the very start. And structure some exercises where they have to evaluate the validity and accuracy of the AI generated answers.
@daphsci as an experiment, I had ChatGPT generate an article on the parts of a compass. I wrote a navigation column for a magazine for a couple of years and used to teach navigation, so I know the topic well. It got a couple of things wrong. I revised what it got wrong and posted it on my canoe and kayaking website as an experiment. Parts of a compass are so simple that I was surprised it got things wrong.