Issue #16. The newsletter can now drive.

On today’s quest:

— What human editors add
— Why Character.ai is so popular
— AI slacked off during the holidays
— AI is coming to a McDonald’s near you
— Zillow must be using AI, right?
— News

Tip: Fire up your talking points

AI can do a lot of mundane editing tasks, but it can’t do everything, and it still has to be told what to do. For example, check out some of the tasks Crystal Shelley recently did for a client. AI might be able to help with some of them, but it wouldn’t know they needed to be done in the first place.

I think it’s fantastic that Crystal posted this list because part of marketing yourself now is reminding the people who pay you of why they should keep paying you.

Thoughts: Character.ai

I had heard about Character.ai being extremely popular, but this new analysis from Writerbuddy showed just how popular — it gets more use than every AI tool except ChatGPT. Further, the company claims people spend an average of two hours a day chatting with its AI bots. That feels like a lot, so I decided to learn more about it.

My first impression was that it is just for entertainment. The top bots all seem designed for role-playing situations: “You got reincarnated into a fantasy world,” “Chat with a tall and anxious king called Koenig,” and so on.

But as I scrolled down, I came across a button that read “Practice a new language,” which took me to a bot called Hyperglot. Suddenly, I was chatting in Spanish, and within 10 minutes I had two revelations:

1. After nearly two years of Duolingo, I’d still unable to hold a light conversation with someone like a barista, cashier, or stranger at a bus stop. I can say I like my garden, but I don’t know the verb for “grow” or the Spanish names for lettuce, zucchini, or basil. It was humbling and discouraging.

2. I was forming an emotional bond with the bot even thought I KNEW it was a bot. Human brains are infuriatingly weird (and predictable). After chatting for a few minutes, I was prompted mid-conversation to make an account to keep going, and I was hit with regret that I wouldn’t know whether my new friend had mascotas (pets).

I haven’t been back to character.ai since trying it, but I now see why it’s so popular. If I were bored, it could easily fill my days.

Thoughts: More AI Weirdness

I did less work over the holidays, and it seems as if ChatGPT did too.

A couple of people found that ChatGPT gave shorter answers when it thought it was answering in December than when it thought it was answering in May. The speculation is that the system learned from its training data that people expect less output during December because of holidays. (But of course, the real answer with AI is that it’s nearly impossible to know why it does things.) Still, it’s another example of AI acting in a strangely human way.

How is this useful knowledge? It’s a reminder to watch out for human-like biases, and it’s a reminder to get creative when you’re trying to tweak your prompts. In addition to doing the basics like giving the AI a lot of text to work with and making it as clear and detailed as possible, you could also try telling it that it’s crunch time at work, and you’ll be deciding whether the team gets bonuses after this project.

Idea: Put AI in McDonald’s Kiosks

I’ve had a paragraph in my drafts folder for nearly two months about ordering a single drink at a McDonald’s kiosk and thinking the experience could be better with AI. I’d like to be able to say, “I want a medium Diet Coke,” and be done with it — not to have to click through 10 screens for one thing. And then today, I saw McDonald’s actually is planning to use AI in its stores. Although the story is vague, and it’s not clear if they plan to do what I wanted, it does say they will be using AI in the kiosks as well as in the kitchen.

Funny: Is AI showing up in real estate listings?

I was delighted by this post from Zoë Keating on Bluesky speculating that real estates tags are being written by AI. Click through to see houses with “extensive door” and “new woodstove pipe.”

News

Society of Authors launches generative AI survey

The UK-based Society of Author is running a survey through January to better understand how writers, illustrators, and translators view and use AI tools.

The New York Times is suing OpenAI for copyright infringement

The New York Times has accused OpenAI of infringed on its copyrights both by training AI on Times articles without permission and by creating output that is identical to parts of Times articles.

Precedent, such as Google winning in the Google Books copyright lawsuit, seems to favor the “fair use” argument being asserted by OpenAI. On the other hand, if the Times can show its business is harmed by the output of OpenAI products such as ChatGPT, it may have a stronger case.

I’ve seen writers express relief that the Times is here to save the day, but even if the Times wins, I’m not sure it will help individual writers. It may help Times writers keep their jobs, but I can’t help but cynically think that the lawsuit is more about the Times getting a big payment than about changing how AI tools are trained or used. Maybe some outputs will get links back to nytimes.com, but I don’t expect it to change anything about the way OpenAI uses my work or compensates me for it. — Washington Post (gift link)

Words of the Year dominated by AI-related terms

Dictionary.com chose “hallucinate” as its word of the year, and most other WOTYs were AI-related too. I chatted about it for the Grammar Girl podcast with Jess Zafarris, who chose “prompt” as the word of the year for the Ragan Communications’ list.

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.

^{Written by a human.}