#Issue 12, served up hot.

On today’s quest:

— My current favorite use for ChatGPT (text formatting)
— AI-related words of the year (so far)
— AI in the workplace news

Tip: Using ChatGPT to change formatting

The way you format words used as words varies by style guide, so I often find myself changing italicized words to roman type surrounded by quotation marks (and vice versa), which ranges from tedious to soul-crushing.*

Imagine my joy when I discovered ChatGPT-4 can do it in seconds.

My starting test document had multiple words in italics:

Every italicized word was properly put in quotation marks. I went from Chicago word formatting to AP word formatting in seconds.

You may notice the errant quotation marks in the third line above. It turned out I had an italicized space! As with all AI edits, you’ll have to watch out for errors, even those created by problems with your starting document.

An interesting note is that Claude did not put the italicized space in quotation marks. It seemed to do a better job of limiting its actions to just “words or phrases” as directed in the prompt.

Finally, to begin this process with italics, you have to use ChatGPT-4 (paid) or Claude (free) because they let you upload documents with formatting. If you paste italicized text into the GPT-3.5 chat box, the formatting is lost (and it will put random words in quotation marks if you use this prompt).

However, if you start with text in quotation marks, ChatGPT-3.5 (free) can output italicized text.

Addendum: I tested this again right before I was going to send out the newsletter, and it worked differently. Now, it’s putting two quotation marks on each side of each word, to which I can only say WTF? It’s easy enough to fix with a quick find-and-replace, and it’s still easier than doing it by hand, but it’s weird — like so many other things with AI. It’s important to remember that prompts can give different results on different days even when they seem straightforward.

I also had another interesting experience when ChatGPT failed “due to a high volume of requests.” I went through a series of emotions:

1. I was annoyed because, hey, I pay for this thing.

2. I was amused because it cheerfully replied, “Let's try a different approach. I will provide you with instructions on how to make these changes yourself using Microsoft Word.” And I thought, ”Gee, thanks. I think I know how to do basic formatting changes, you condescending glitch goblin.”

3. But then, whoa … it told me I could do an advanced find-and-replace in Word to globally make the changes I want! I don’t use Word, so I can’t do this unless I pony up and switch programs, but still, I had no idea this was possible. Now I see why hard-core professional editors use Word.

So, as is often the case these days with AI, I learned something interesting and potentially quite useful.

News

AI invades word-of-the-year lists at Oxford, Cambridge and Merriam-Webster

It’s word-of-the-year season, and it’s no surprise that many of the choices so far are related to AI:

• Cambridge: Hallucinate (as in what AI is known to do)

• Merriam-Webster: Authentic (as in the opposite of AI). Runner-up: Deepfake

• Oxford: Prompt (as in the instructions you give AI). “Prompt” was actually a runner-up. The winner was “rizz,” short for “charisma.” — TechCrunch

AI skills in demand by employers

Consider the source, but a new survey by Amazon Web Service showed that 42% of employers are actively looking for people with AI qualifications, but 75% say they are having trouble finding qualified candidates. But it’s not all bad news: “Employers estimate that workers who acquire AI expertise could see their paychecks jump by 35% or more, depending on their departments with IT (47%) and sales and marketing (48%) seeing the highest bumps.” Overall, “88% of workers expect to use AI in their daily work by 2028, and of those, one in four (25%) expect to be using it “extensively” or in over 60% of their job tasks.” — Amazon Web Services (PDF)

AI Content Is Preferred Over Human Content

This one hurts: when people didn’t know whether text was written by GPT-4 or a human, they preferred the AI-written text. Researchers at the MIT Sloan School of Business showed 1,203 people ad copy and persuasive copy and asked about satisfaction, willingness to pay, interest for the ad content, and persuasiveness in the persuasion content. GPT-4 content did better than human content in every category. When people were told which content was written by humans, they rated it a little bit higher, but it still didn’t beat the AI-written content. — Almost Timely Newsletter (and original study PDF)

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.

^{* A few months ago I had hacked together a process using markdown that saved some time, but it was temperamental and less than perfect.}

^{Written by a human.}