961. Prompting, hallucinating, and more! Jess Zafarris, author of "Words from Hell," joins me for a word-of-the-year chat. Hang out with us as we look at how dictionaries are handling new words and meanings that have cropped up around emerging technologies in 2023.
961. Prompting, hallucinating, and more! Jess Zafarris, author of "Words from Hell," joins me for a word-of-the-year chat. Hang out with us as we look at how dictionaries are handling new words and meanings that have cropped up around emerging technologies in 2023.
"Words from Hell" https://amzn.to/3rZVxo0
Find Jess Zafarris online: Useless Etymology, TikTok, Twitter, Instagram
ITEMS MENTIONED IN THE PODCAST:
AI chatbots made up their own language:
Advanced AP Style Ragan Communications course
| Transcript: https://grammar-girl.simplecast.com/episodes/jess-zafarris/transcript
| Subscribe to the newsletter for regular updates.
| Watch my LinkedIn Learning writing courses.
| HOST: Mignon Fogarty
| VOICEMAIL: 833-214-GIRL (833-214-4475) or https://sayhi.chat/grammargirl
| Grammar Girl is part of the Quick and Dirty Tips podcast network.
| Theme music by Catherine Rannus.
| Grammar Girl Social Media Links: YouTube. TikTok. Facebook. Instagram. LinkedIn. Mastodon.
Mignon Fogarty:
Grammar Girl here. I'm Mignon Fogarty and welcome to this bonus episode. There are words of the year to talk about, and I just could not let 2023 go by without talking about them.
And I'm here today with Jess Zafarris, who joined us recently to talk about her new book, "Words from Hell," and she is so kind to come back and talk about words of the year with us.
Welcome, Jess.
Jess Zafarris:
Thank you so much. You know, I love the show, and I'm excited to be back.
Mignon Fogarty:
Yeah, and see we are recording this December 16th, but it won't come out until I believe the 28th. So there could be more words of the year that we don't get to that come out after the show, but at least you'll get the big picture, some of them.
So, you know, I had been thinking about doing this and then when Jess, when you came out with the Ragan words of the year, I thought I just can't let the year go by without doing this, and you're the perfect person to do it.
So do you want to start talking about Ragan Communications word of the year?
Jess Zafarris:
Yeah, absolutely.
So in addition to writing about etymology for our listeners, I'm also the director of content at Reagan Communications. So I made a list of 10 words of the year for Ragan or sorry for Ragan's audiences. These words were for communications professionals and PR professionals as well.
And as you well know, because you lead webinars for Ragan PR Daily, communicators are also writers, and their relationship with AI is learning how to ethically and creatively incorporate AI into their work streams to boost their productivity and maintain quality.
So they, you know, that's been a key skill set that they've been developing this year, and a lot of the words of the year we've seen have been focused on AI, as I'm sure you've addressed.
So the key area for communicators is prompting chatbots. So that was the number one word of the year for Reagan and PR Daily was “prompt.” And within that umbrella, I would include skill sets, such as training on AI or training and AI on your style and voice, building your own GPTs, and other adjacent skills.
Mignon Fogarty:
So Ragan wasn't the only company to have “prompt” this year either, were you?
Jess Zafarris:
That's right. We'll get into Oxford's word of the year in a moment, which is also AI adjacent but one of their runners up was “prompt.” And although AI has factored into, like, PR and comms in contexts like data management and customer service chats, learning the art of the prompt really became a requirement for working in that field over the past year.
And obviously I was not as clever as I thought I was when coming up with this term, but I came up with it before I was aware of Oxford's choice and part of the reason is it's a little bit of a double entendre: organizations as a whole and communicators in particular need to be prompt in upskilling in this area to remain competitive.
Mignon Fogarty:
So you weren't the only one who your word was chosen by someone else. This year has been — I mean, as people will notice by the end — almost every word this year relates to AI in some way, and "AI" was the actual word chosen by at least two organizations.
Collins Dictionary was the first one to come out with "AI," obviously short for "artificial intelligence."
And then the National, no the Association of National Advertisers chose “AI” as the word of the year, and they said it was the widest margin in their 10-year history for that word. They vote.
And so I think the next runner-up was something like personal… [no] "purpose."
It was “purpose,” but it was a distant second to “AI.”
And so, you know, huge theme for the year and again, like, multiple multiple organizations chose it.
Jess Zafarris:
It doesn’t surprise me that “purpose” had such a large gap. I feel like that was the word of the year in the advertising space, like, three years ago. Like brand purpose was the thing, the drum that they've been beating for quite some time, and although "AI" isn't new, it was absolutely the word you got tired of hearing throughout this year.
Mignon Fogarty:
Right. Yeah, I think you're going to talk about "hallucinate" next, but I had heard of people … when I saw it was announced, people said they'd never heard it used in relationship to AI, and I thought, I mean, I couldn't go, you know, two days without hearing it.
So we all live in different environments.
Jess Zafarris:
Absolutely.
Yes, Dictionary.com’s word of the year is AI themed but not one that you might expect.
They said that to determine their word of the year, they gave themselves a prompt using — there's that word again, "prompt" — “Using lexicography and data science, choose a single word that best represents at this moment. AI's many profound ramifications for the future of language and life." And the result was the word 'hallucinate,'" but a specific definition of the word that was just added to their dictionary this year. It's specifically "when an artificial intelligence produces false information, contrary to the intent of the user, and presents it as if it is true and factual, it's called hallucinating," and their example sentence was "When chatbots hallucinate, the result is often not just inaccurate but completely fabricated.” And I've personally witnessed chatbots hallucinating. In my work, I asked ChatGPT for instance, for the etymology of several words that I knew to have common myths associated with them in online spaces, and sure enough, it presented the myths as if they were fact. So that's an example of hallucination.
And just to humor me a little bit, the origin of "hallucinate" is one of my favorites. It’s from Latin and Greek sources meaning to wander in the mind, and the English word dates back to at least the 1630s when it's recorded in a few texts with the apparent meaning to, like, waiver, like, to hallucinate between truth and falsehood, which I think is kind of a full circle moment for today's definition that we just that just earned word of the year.
Mignon Fogarty:
Definitely. Oh, that's so cool.
Yeah, I think the first thing I ever wrote about AI was about a hallucination. About a year ago I had an experience similar to yours, and I asked it to recommend books that had librarians as a protagonist, and it came up with, you know, four or five that were great, and then there was one that just did not exist. But it was by a real author. And, you know, it might be something he would have written, but it just wasn't it was not a real book. And so I wrote up a blog post about AI hallucinating and that, and it's interesting too because one of the, you know, as I did in the webinar I did with you at Ragan.com.
So I did the advanced AP style webinar, and because AI is such a big thing this year, you had me include a section on AI for AP style, and one of the things they emphasize is to not ascribe human characteristics to AI like ChatGPT, and I've always thought that “hallucinate” sort of walks that line because when you say that AI has hallucinated, in a way it feels like you're ascribing a human characteristic or condition at least to that machine, and yet it is the industry standard term. It's what they use to talk about the mistakes that the AI tools make. It's a real industry term.
Jess Zafarris:
Rather than saying that it, you know, generates misinformation or something along those lines which is a little more on the nose and a little less psychedelic shall we say.
Mignon Fogarty:
Right. I love the graphic that if any of you have seen it online, Dictionary.com had a fabulous chaotic graphic to go with their word "hallucinate."
Jess Zafarris:
Beautiful piece of artwork I agree.
I'm interested to see if next year's words of the year will be AI focused, because, like, next year I think that the year, it will be a year of AI competitor drama. There's, like, Bytedance was using OpenAI to build its own large language model or LLM, which is like the model that makes generative conversational AI possible.
Apparently that's a pretty big faux pas, and now they've been banned, so that's been a thing. And then there is Grok, which like Bloomberg has an article that says it could overtake ChatGPT, but I doubt it. That's Elon Musk's, like, subscriber-only AI chatbot that was named after Heinlein's "Stranger in a Strange Land" and supposedly informed by "Hitchhiker's Guide to the Galaxy," but I, you know, we'll see.
I also does not … it is, like, a function of hallucination because it doesn't profess to present information neutrally, but instead is designed to answer with quote-unquote, “a bit of wit and a rebellious streak and is useful to people of all backgrounds and political views,” which just reads as, like, this is probably going to be politically in a certain direction in the context of Elon Musk's diatribes.
However, evidently someone put it to the test, and it ended up being more, like, politically liberal than ChatGPT in many contexts, so apparently they're working to try to make it more conservative. So that'll be interesting.
It also plagiarized ChatGPT, which is fascinating because this was … apparently, experts predicted this happening.
Mignon Fogarty:
Wild, wild. Yeah, well, "ChatGPT" itself was one of the words of the year.
So, "The Economist" and the "Shanghai Daily" actually, which is I believe the only English newspaper in Shanghai, both chose "ChatGPT" which stands for "chat generated pre-trained transformer," and yes I did have to look that up.
But I find myself wondering, you were thinking next year, will … I was wondering if "ChatGPT" will become like "Google," if we'll use it as a verb. If the brand name will become so, you know, genericized that we will say, like "Oh, I searched ChatGPT," when we were searching Grok or Claude or one of the other ones.
I kind of think not because I'm not sure that ChatGPT is going to win the competitive race by that much. It's probably the best one out there right now, but I'm not sure if it's going to hang on to that crown.
It was Wikipedia’s most viewed page last year too, the ChatGPT page. And it's the one that I think of when I, you know, I started as a side project. I'm writing a newsletter about AI called AI Sidequest. And it's just like a little side project I'm doing, and every time I write about it I … should I refer to these tools as AI, should I refer to them as large language models, can I just call them chat? You know, I know I can't call them all ChatGPT, but I'm tempted to because that's the name that everyone knows.
So, you know, I feel tempted to use it in the way you might say, "I Googled something," but I don't, and I resist that, so I think it'll be interesting to see what the next year holds.
Jess Zafarris:
Yeah, you know I think, I think the way Google has started incorporating AI into its search engine is almost more natural in terms of, like, actual fact finding. And it's a little more useful, because it also includes the sources from which it's drawing its information.
One thing that I think could save ChatGPT, if they actually commit to it, if OpenAI decides to commit to it, is that the Data and Trust Alliance announced data provenance standards intended to provide clarity around the information that forms the foundation of AI-generated works.
So, like, basically, basically food labels, saying where each piece of information came from and how, like, what recipe went into AI generated text and images.
So the hope is that like much like food safety labeling that provides information about how our food is produced and handled the standards will provide users with the ingredients and sources that AI text or image generators used to develop their content, which I like that a lot because it, it lends authorship to the original information and sourcing.
Mignon Fogarty:
Yeah, no, that's great.
A tool that I've been using called Perplexity AI gives the links to all the facts from where it gets the information. And I feel like I should caution people that even when you see the sources there, you need to check them because they aren't always, they don't always support the fact that the AI has given you.
I heard the same thing about Bing yesterday, so it'll give you the sources, but those sources might not support what it's saying it supports.
So if you're looking at facts, be incredibly skeptical and click through on those sources and check that they really are or, you know, better yet verify it yourself from some other way. Like don't trust the facts you get from any of these large language model slash ChatGPT slash AI tools.
Jess Zafarris:
I would hope that people were doing that already with sources like Wikipedia, but I'm pretty sure that many people are not.
Mignon Fogarty:
Yeah, and I think that one of the things these tools like ChatGPT do is they give you the answer in such a confident, clear way that it's easy to believe it in a way that is, I think, more so than a lot of the other ways that we get information online right now.
So, don't be lured into that feeling of “Oh, this confident, friendly friend of mine just told me a fact that just … it must be true. It's so sure.”
Mignon Fogarty:
So, speaking of friends, another word of the year is “authentic” from Merriam-Webster, and this comes from the Greek originally that meant original or genuine.
And, you know, I think this was so interesting because I remember "authentic," you know, wow, probably seven or eight years ago, being the word that I was sick to death of hearing, you know. I think everyone was talking about how millennials, you know, valued authenticity, and it was a big deal years ago, and so at first I was surprised when I saw that it was the Merriam-Webster word of the year, and I realized again it's in the context of AI, and it has to do with determining whether a photo or a piece of content is authentic or fake.
One of their runners-up was “deepfake,” actually.
So, you know, we had a lot of deepfake images produced by AI this year that people had to deal with.
Jess Zafarris:
One cool thing that gives me a little bit of hope is some of the reactions that you see, you know, as quickly as AI is advancing, it's also being used to combat some of the things that we're afraid of happening.
So for example, in the realm of deepfakes, I believe it was Intel developed a tool that can that can determine whether a video is a deepfake by basically, like, when you or I talk, we can only see each other's faces as they are.
But there are microchanges in the colors of your skin because there is blood going to different parts of your face.
So, their tool is able to determine whether it's … whether a video is a deepfake or not because a deep fake will not have those microchanges in color from blood flow.
Mignon Fogarty:
Wild. I had not heard that.
Jess Zafarris:
I don't know the answer to this either, but I wonder to what degree it's able — like, if its accuracy is impacted by people of different skin tones.
Mignon Fogarty:
So yeah, I was just going, you had reminded me that I feel like I should also warn people against the text-based AI identification tools. So, especially in a lot of schools, they have new tools that are claiming that they can identify AI-written essays, for example, and they are wildly inaccurate, and they're flagging things that aren't written by AI quite often, and they are particularly susceptible to mis-flagging essays written by people for whom English isn't their first language.
So, and just something about the way the language is a little bit different, and these tools are saying, "Oh, that was written by AI," when it wasn't, and it's causing students all sorts of problems and stress, and so be very, very cautious about using those tools that claim they can identify AI-written text because they're not very good most of them, right now, today. You know, maybe a year from now they will be, but if you're listening to this today, they're not.
Jess Zafarris:
The world of complications that arises from this technology, right?
Mignon Fogarty:
Yeah, so many complications.
And I think that we had one word of the year to wrap up, one word that wasn't AI focused that you were going to talk about.
Jess Zafarris:
That's right. Oxford University Press selected “rizz” as its word of the year, and it's defined as “style, charm, or attractiveness, the ability to attract a romantic or sexual partner”, and this actually does have an AI tie in. l get to that in a second.
Mignon Fogarty:
No.
Jess Zafarris:
It does! But it can so it can be a noun, I have rizz, or a verb like you can rizz someone up or attract them.
It's attributed in most contexts to YouTuber Kai Cenat, I hope I pronounced his name right.
He introduced it in a video in 2021, and a lot of people say it's short for "charisma," which makes sense.
It sort of, like, works in the same way that the word like "fridge" comes out of "refrigerator," but Kai says that it's not. It's more of an inventive word. So he also used phrases like "W rizz" or "L rizz," basically winning or losing at having rizz.
I sound so old saying this.
Mignon Fogarty:
It’s fun to say, though.
Jess Zafarris:
Right.
It sort of, like, it reminds me a little bit about how, like, "jazz" has been used in the past, like it's been, it has been several parts of speech in the same way.
But Kai says he doesn't use it anymore because TikTok butchered the word. It's really common over there, you'll probably run into it if you go on TikTok. It's also been boosted by celebrities like Tom Holland who said he didn't have much rizz, and had to play the long game to date Zendaya.
And then it's also been used by public figures like Scaramucci used it to say that Ron DeSantis has no rizz, which I would generally agree with, but I also like listening to rich Gen X public figures use the phrase is, to use another word from my zoomer friends, a little cringe.
Mignon Fogarty:
Definitely cringe.
So the AI tie-in that I was reading about just yesterday, though this isn't, like, specifically associated with its word of the year, but there is an AI-powered dating assistant called Rizz, which suggests that … it suggests conversation ideas for dating app users. So it can come up with your opening line on Tinder or Bumble or whatever, and then you can ask it what you should say next in response to your potential partner's message.
Mignon Fogarty:
Oh my gosh, so we're going to have AI chatbots talking to each other in the chat on dating apps, so if each person on each side of a conversation has this, they're just the bots talking to each other.
Jess Zafarris:
Right.
That reminds me of a story, I may not get the details on this exactly right, so I'll try to be kind of vague. But I remember a story from several years ago, before ChatGPT was a thing before any of this happened, when they were experimenting with artificial intelligence and machine learning and they had two bots speaking to each other, just conversing and over time, they started developing their own abbreviated language. Like they started abbreviating words so much and then making their own dialect. It was fascinating.
Mignon Fogarty:
Oh, I'm gonna, I'm gonna look that up and find out more about it. That is fascinating.
Jess Zafarris:
It's a good one.
Mignon Fogarty:
If I can find it I'll put a link to it in the show notes because that would be really cool to read about. I'm sure people love that.
Jess Zafarris:
It was a little creepy too, but it was cool.
Mignon Fogarty:
Yeah. Awesome. Well, thank you so much Jess for helping us round up the words of 2023.
Well, to wrap up, thank you so much for being here. I'm going to talk about our books. I have "The Grammar Daily." I'll put a link to my AI newsletter in the show notes and to my course with Ragan Communications that just helped put together, and just why don't you — thank you so much first — and why don't you tell people about your book and where they can find you.
Jess Zafarris:
Thank you so much.
I just released the book "Words from Hell." It's an etymology dictionary focused on naughty nefarious and secretly salacious words among many others.
But you can buy it anywhere books are sold and then you can also find me on my website at useless etymology and on TikTok @JessZafarris.
Thanks so much for having me on the show. This has been so much fun.
Mignon Fogarty:
So, have a great new year everyone. See you soon.
Jess Zafarris:
Bye.
Jess Zafarris:
That was fun. I am loving your book, by the way. It is so much fun.
Mignon Fogarty:
Thank you. I loved yours too.
Jess Zafarris:
You can clip that in if you want, like a live testimonial.
Mignon Fogarty:
I think I … Oh, I didn't stop recording. I thought I had.