Caitlin Dewey’s early interest in artificial intelligence probably came from science fiction movies. (Shout out to anyone old enough to remember the ’80s show “Small Wonder” about a not-at-all creepy robot girl.)
“I do remember writing a story for the Washington Post food section in 2015 about an AI cooking application that IBM developed,” said Dewey, an enterprise reporter for The Buffalo News who's about to go freelance full time. “The concept was so cool — here's this technology that can supercharge human creativity in all these unexpected ways! — but the actual output was not terribly appetizing. I fed at least one of the dishes I recipe-tested to my dog. Even then, though, you could see the potential of the technology, once it got farther along.”
We have seen recent results of more not terribly appetizing AI output.
But Dewey has been testing it. (She’s also the author of the newsletter Links I Would Gchat You If We Were Friends, and an adviser with the Poynter-Koch Media and Journalism fellowship, where she recently taught a session on using Chat GPT.)
“My feelings about ChatGPT (and generative AI writ large) are actually pretty conflicted. I'm not naive to the very significant risks/flaws/drawbacks of these tools. I'm also very concerned about how they've been deployed in some media organizations,” she said. “At the same time, I have found ChatGPT to be enormously useful and exciting in my own work. I think those things are sometimes in tension … but not mutually exclusive."
Dewey wrote about how journalists can use it in their work for Poynter. I asked her to share how it’s worked for a specific story. Here are six things she asked the generative AI program “while drafting a recent freelance piece about the barriers that some patients face when trying to access medication to manage a miscarriage,” Dewey said.
-
"A comprehensive list of any and all themes a national newspaper story on the topic might cover. (This is consistently a very solid initial brainstorm.)"
-
"Relevant peer-reviewed research, via a plug-in called ScholarAI, which is available to paid subscribers."
-
"Question ideas for a series of interviews. I write my own questions first, but ChatGPT usually comes up with at least a couple things I missed."
-
"The meaning of the post-nominal initials 'KM' — it's 'Knight of Malta,' for what it’s worth — and how to politely address someone using those initials."
-
"How to look up the current status of a federal court case. (This example is a little embarrassing, but I'm including it to illustrate ChatGPT's usefulness for jogging your brain when it blanks.)"
-
"Other ways to say 'pharmacy chain' and 'health care provider'; also, a list of adjectives meaning 'not publicized' or 'not publicly available' with more neutral connotations than 'hidden' or 'secretive.'"
“Once I had a draft," Dewey said, "I also asked ChatGPT for style and grammar feedback using what I call the ‘helpful editor prompt’ — it's a multistep recipe that instructs ChatGPT to proofread for specific issues and errors,” Dewey said. “I also asked it to identify gaps in the draft and predict what stakeholders with different ideological viewpoints would think about it.”
Have you tried using AI as a tool in your work? Are you enraged we’re even talking about this? Let me know what you think, and check out more ways Dewey suggests experimenting with generative AI here.
And here are Dewey’s tips:
|
