Issue #33. A meticulous and notable newsletter

On today’s quest:

— AI has some favorite words
— More bad AI-generated product reviews
— Beware AI-generated cookbooks
— You may be able to opt out of training data (next year)
— But you can’t delete your best posts from Stack Overflow
— More survey data about job vibes

You’re not imagining it. AI uses different words

If you think you can “just tell” when text was written by AI, you won’t be surprised to learn that researchers have found that AI really does use certain words more often than humans — words such as “intricate,” “commendable,” and “pivotal.” And now, a new report in Scientific American shows that these words are showing up more often in scientific papers, suggesting some researchers are using chatbots to help write their papers.

Major publications were posting AI-generated product reviews

Futurism, the outlet that broke the story about “Sports Illustrated” running AI-generated content a few months ago, is back with a much bigger list of publications using the same shoddy AI-content company to create poorly written “best of” lists filled with affiliate links and recommendations for products nobody has ever touched.

Many of these companies have since stopped working with the vendor, but not all, and the list includes “USA Today,” “US Weekly,” “PC Magazine,” “Real Simple,” “Modesto Bee,” “Los Angeles Times,” “Miami Herald,”* “The Sacramento Bee,” and more. The list goes on and on.

Even worse, the owner has another company that sells placement on these “best of” lists so customers can make claims like “named best ab roller by ‘Sports Illustrated.’” Suffice it to say I’ll never trust another online review again.

Beware AI-generated cookbooks

This is from a couple of months ago, but in case you missed it, Alex posted images on Bluesky of what appears to be an AI-generated cookbook with gems as “the veggies” under recommended proteins and popcorn cooked al dente under snacks. From the post, it sounds like a hospital purchased this book for use by patients and returned it after seeing the inside.

The bio touts Rhonda C. Anderson MS RDN as a well-regarded author, speaker, and contributor to major outlets, but when I put the author photo into TinEye (a great reverse image search tool — bookmark it for future use), the results show it’s stock art. Other people have done different searches and have also come up empty for any real Rhonda Anderson with the credentials in the bio.

I’ve seen a lot of people talking about crappy AI-generated books flooding Amazon and about bad AI-generated recipes online, but this is the first I’d seen of a bad medically-oriented cookbook. It’s not surprising, but it does seem especially harmful.

OpenAI will let content creators opt out of training

Open AI is developing a tool that will let content creators and publishers flag their work so it isn’t used as training data for the company’s AI models. OpenAI plans to launch the tool, called Media Manager, by 2025. TechCrunch reports that the move is meant to fend off future copyright infringement lawsuits.

The company already has a tool that allows visual artists to designate images that shouldn’t be used for training, but some artists say the tool is onerous because it makes them upload each individual image with a description. So although Media Manager seems like a good step from the company, we’ll have to wait and see how useful it really is.

On the other hand…

A Mastodon post by ben ui says that Stack Overflow is refusing to let people delete their posts on the site after the company announced that it made a deal with OpenAI to let the site data be used for training.

According to the writer, Stack Overflow said it ”does not let you delete questions that have accepted answers and many upvotes because it would remove knowledge from the community.”

The person then tried to change the content of their existing post to a protest message, but Stack Overflow supposedly changed the text back and suspended the account for seven days. Although it seems like a rotten way to behave, it sounds as if Stack Overflow is within its rights to preserve the posts. Tom’s Hardware has more details.

More survey data about job vibes

According to Stanford University Human Centered Artificial Intelligence, “Globally, most people expect AI to change their jobs, and more than a third expect AI to replace them. Younger generations — Gen Z and millennials — anticipate more substantial effects from AI compared with older generations like Gen X and baby boomers.”

“Specifically, 66% of Gen Z compared with 46% of boomer respondents believe AI will significantly affect their current jobs. Meanwhile, individuals with higher incomes, more education, and decision-making roles foresee AI having a great impact on their employment.”

What is AI sidequest?

Using AI isn’t my main job, and it probably isn’t yours either. I’m Mignon Fogarty, and Grammar Girl is my main gig, but I haven’t seen a technology this transformative since the development of the internet, and I want to learn about it. I bet you do too.

So here we are! Sidequesting together.

If you like the newsletter, please share it with a friend.

^{* I am annoyed that none of these newspapers has the word “the” in their formal names.}

^{Written by a human.}