Navigating AI Tools in Education: ChatGPT, Magic Write, & Beyond

Webinar Transcript

Dr. Jill Johnson
Lifelong Educator & Class Intercom President 

Ben Pankonin
Tech Entrepreneur, Founder & CEO of Class intercom

Editor’s Note: This transcript was originally created using an AI tool called Otter.IA, which translated a recording of the live audio/video of the webinar into text almost instantly. An editor from the Class Intercom content team then reviewed the transcript for accuracy and made corrections. Her notes are as follows: Automating the initial transcription definitely saved tons of time, providing a baseline document to edit. However, the document still needed an editorial pass. The AI sometimes missed punctuation and struggled to convert proper nouns accurately into written format. Overall, very impressed with the draft it created, but definitely would recommend a human review before publishing.

Dr. Jill: Hello everyone, Dr. Jill here and I am joined by Ben Pankonin in Lincoln, Nebraska at Class Intercom global headquarters. We’re always so excited to welcome such an awesome group and today we’re gonna be talking about AI. And I find that really exciting. For those of you that haven’t already gotten the reference, I did wear my Sheldon Cooper Green Lantern shirt today, because if Sheldon were a real person, I feel like he would be super excited about this topic. And he might, you know, call us out on a few things. But I think it’s so fun. And Big Bang Theory is one of my favorite shows. I was rewatching an episode last night, and I was showing my daughter since I told her I was wearing this shirt today. And the one where Sheldon actually is on a cart and his face is just on a screen. That was in 2012, that episode came. And I found it just really interesting, because, you know, it was using these kind of crazy far fetched tools that now we use every day, multiple times a day to talk to people really all across the globe. And so I think it’s really exciting. It is amazing. It feels like we are in the future, right? And the future is, well, it’s always changing, I guess. But yeah, it seems like it’s coming at us so very quickly. So I am excited to have you guys join us today. I know we have a ton of educators, school PR professionals, administrators, and even some students registered today. So excited. And Ben, I know I’m leaning on you heavily today. Ben is the tech guru. Ben’s our CEO and founder here at Class Intercom and so I’m gonna let you go ahead and kick it off.

Ben: Awesome. Well, I’m excited to chat about this. You know, we’ve had a lot of initiatives in the last several months where we’ve been talking about this a lot at the office, and working with, you know, a lot of our clients in different aspects of AI, as well as with our dev team. And so that’s been really fun to see that progression as we navigate how AI really works in schools. But you know, I want to address a couple things off the bat, right? We had a number of questions, even as I was talking to some of our clients about, you know, their concerns over AI and things like that. And we’ve definitely had some things that we want to address up front and have some ethical conversations in AI, because that is one of the questions we’ve had a lot lately. And, you know, those questions have really, largely revolved around things like: How will I catch students cheating in AI? And I think that is one of the areas we’re going to address. We’re going to talk about that a little bit. But you know, I hate to start in the downer side of things. But I think if we can move past that, we’re going to have a lot of things that we can talk about that sort of move the needle for all of us in our productivity, and a lot of the ways we may think about the future of technology. But also, you know, first things first, we want to address a little bit that the concern really isn’t just about cheating in AI. That’s been where a lot of schools have focused on the conversation. But it’s actually quite a bit more complicated than that. There was a story that came out just the beginning of this past month, where essentially an open AI, they were training one of their AI machines in the most advanced AI area. And they were asking it to do some tasks. And this is a real story. So what they were asking it to do involves having to fill out the Captcha, right? So, every time you’re sort of filling out that thing where you’ve got to figure out the numbers and letters that are sort of like an image in there, and then have to fill that out. Well, obviously, the machine couldn’t do that. Because it was, in fact, a robot, right? The really fascinating part of what it did in order to get past that is it actually reached out because it could reach out to TaskRabbit and hire someone. And so it actually hired a human and asked them: Would you fill out the Captcha for me? Now the funny thing is the human on the other end of it said: Well, are you a robot? And asked it back to the AI machine and it responded with: No, I’m visually impaired. And that’s why I can’t fill out the CAPTCHA information. And so the human actually filled that out, and the AI machine was able to accomplish the task it was looking to do. That’s actually fundamentally the real concern that we have with AI right now. Ethical concerns aren’t just that it might help a student cheat, but that it might essentially control a lot of thinking and it might be very harmful. So, there are guards that the responsible parties are putting around their AI machines. However, we do have to be careful about that. And I thought this was, you know, a somewhat shocking story quite honestly. 

Jill: And I think that’s one of the things that we and I know the first time when you first brought this story to my attention, I was like: Holy cow, that’s like, that’s scary. And that really is some scary stuff. You guys, you all know, I was an educator for 30 years, I’m worried about the kids cheating in school, too, right? That’s where our brains go. It’s what we think about. But those are really, I guess, kind of lower-level concerns, and they’re things that will work around. And I know we’re gonna get into some of the basics. But those are smaller concerns and things that we can do better and continue to address what you’re talking about here. Those are really major concerns that we see. Hopefully, our politicians definitely are people in the tech world who are threatened by this, their businesses, their, you know, information systems, are really like being watchdogs and trying to help take care of some of those.

Ben: Absolutely. And so when we see some of those headlines, I think all of us sort of have this caution. And we sort of put it in our own world. And we say, well, but what are the ethical concerns? This type of thing is a real ethical concern, and the way that it can spread false information and do that automatically, the way that maybe I could use it for real harm, creating malware and sending out–there are a lot of things around that that we do want to curb. And I think, you know, as is the case, and in every other aspect, you know, it takes politics a while to catch up to these things. And so these are real concerns. It’s not just that, that it would sort of like, you know, create some inflammation or something like that. So, you know, anybody that I talked to who’s in the tech space, who’s either writing AI machines, you know, some of our developers or are working with and writing code with AI machines, they’re very hyper-aware that ethics are really important to this whole scenario. So one of the questions, you know, as you think about that area of: Could they use it to cheat? I do want to throw fair warning that there are a lot of–I just saw, TurnItIn has some sort of new functionalities. And it’s tried to say that it can detect whether or not human content is available. I actually took the language on here from one of the machines that is designed to detect human content. And I thought it was actually pretty interesting because right in the, in its own information, it’s saying: Our classifier is not fully reliable. In fact, in English texts, our classifier correctly identifies 26% of AI written content. That’s a pretty low percentage, and quite honestly, in a lot of cases, not worth the time to put that in there. So it is going to change education. It’s going to change the way that we learn education. It’s going to change the way that we write and learn how to write. However, looking out at the Internet and saying: Can this detect whether or not there’s cheating? Quite honestly, my definitive answer right now is: Not with any reliability.

Jill: Same. I haven’t found any others either. And I’m getting that question every day as a former language arts teacher. Several friends that are having really interesting conversations, though, about changing. How would I do this differently if I was in the classroom today? How would I do a literary analysis? How would I ask students to do that? How would it look different than it has in the past?

Ben: Yeah, and I think what it’s going to push us to is a lot of in-class writing. It’s going to push us to evaluate things with AI. Start, maybe, with AI in a lot of cases, and then evaluate its results. And so I think there’s a lot of really valuable learning. We do have to acknowledge, though, that you aren’t going to be able to fully detect whether or not that’s written by AI, because it is relatively intelligent in the way that it’s writing and clarifying that information. So it uses all that to say, hey, we’re fine. It’s fine. There’s ethical concerns. There’s moral concerns. There’s concerns about how we use it in education. But let’s get to how it works. Because if we’re going to be using it in education, and we know that it’s a tool that students are using and we’ll be using, let’s figure out a little bit more of how it works and why it works. 

I didn’t give credit to this photo, but this photo was created fully by AI. I asked it to create a stack of papers with fire behind it and I’ll keep citing my sources for where AI is coming into play in our presentation. But when we start thinking about how we use AI, we’re seeing a little bit of where AI comes to play. So there’s a number of tools that have been popularized recently. And I’m gonna give a shout out to a number of these. One of those is Open AI. So if you’re using chat.openai.com, that would be the address for that. Open API is a project that was a number of different developers. Microsoft ended up buying a significant chunk of their company, and is rolling it into some of their technologies. So Open AI’s platform, you can also now get to with the new Bing. You do have to register for their Beta version, but there is a Bing version. There’s also a Google version, as is the case, right? Like we have in Microsoft platform and a Google platform. Google’s platform is called Bard. You can Google for that, if you’re looking for it. It is also an opt-in, so you have to register. So in both of those scenarios, you have to register, however it is free for the base version. There’s some limits to how many searches you can do in a short amount of time. And it’s a little bit limited by speed–some of that functionality, but the full paid version is $20 a month.

Jill: I was just gonna throw out, too, I know ChatGPT, when I first started trying it, I was really frustrated, because during the typical work hours, you know, 8 a.m. to 5 p.m., I was having a lot of trouble even getting in. It just kept making me wait and spinning. But I still have the free version now. And I’m able to get access about any time, and I know you have the paid version, and you said it’s very fast. I know mine’s very, very fast as well. And I’m kind of… it’s a little bit scary, scary fast. So if you haven’t tried it, if you run into any of those roadblocks, you know, give it another try. But I think you’ll be really impressed. And you will learn as you go, too, but I know Ben is going to give us some frameworks and prompts to help us be better creators of content using AI as well.

Ben: Right. There’s some other platforms that are worthwhile to mention: Jasper, that AI is largely around writing. They’ve got a lot of good functionality around that. There’s also, if you’re using Canva, Canva has several different tools now. We were just testing out one of them yesterday. We didn’t use that for this presentation. However, I had some cool presentation options, where it will generate, you get 100 characters, and it’ll generate the whole thing for you. So pretty cool. As is the case with most of the rest of the images in this slide deck, that image on the right was created by asking an open AI project called Dali, which makes images.That is what it generated for me. 

So focusing a little bit on ChatGPT, which is the Open AI project I referenced and gave you the URL on the previous link. It’s a simple text box. So, you know, when I open this up, and I go to that site, it feels a lot like using Google in the sense that I have, you know, a one line text box and have a little button at the end that sort of generates results for me. It is fundamentally different than Google. And we’re going to try to focus a little bit on the differences between those as we think about how we use AI different than how we would Google. So that’s one of the sort of initial things that I started doing was I was sort of Googling things. And then my results vary dramatically. So first things first: What is ChatGPT? It’s a generative, pre-trained transformer. Big words to essentially say generative means it’s generating a result; pre-trained meaning there is a fixed list of content that has been submitted to it to teach it. So essentially what they’ve done. These used to be that we would train an AI machine and we would say: Hey, I’m gonna give it like a book, or maybe the encyclopedia. Now ChatGPT has been given, essentially, the internet. So it’s indexed a large portion of the text on the internet and said: This is all available to you. Learn from this, and you can now transform. Funny thing is, one of my friends, his son, I was chatting with him about a month ago. He’s 16. And he’s working on programming AI himself. And he pre-trained his AI machine all on open literary examples. So all of his examples were essentially 100 years old or older. So all of his results were kind of coming in Old English, right? But you can train it on that and so he was learning how to program it. You’re an AI machine with Old English, which is, which is fun. And he didn’t give it nearly enough data to have the kind of results that you get with this type of mechanism.

Jill: Are there some limitations, though, as far as fake news, Wikipedia, whatever else that it’s going to have access to?

Ben: Good question. So it depends on which version you use. The newest and latest version of ChatGPT is version four, and it does have access to most of the internet. So there could be false information in there. One of the things I’ve found myself doing is asking it after it searches, I will ask it to cite its source, simply because I want to know was that credible? Where did you come up with this idea?

Jill: Right? Which, I think I asked that question, knowing the answer, thinking about, again, my language arts friends, and, and my research friends like thinking about how we actually verify and check, you know. And this is something we’re teaching: digital literacy. And I feel like this is just such a huge part of that now, the search has become easier and yet broader. And so you know, it’s just, it changes the lessons we’re teaching and how we teach them as well.

Ben: Absolutely. So, when we think about some of the things that we’re sort of comparing, right, the existing model that we know is: I go to Google and I Google something. So, I actually went to Google. And I went to Google, not the most recent Google. But I went to Google in 2001. So this is what happened when you went to Google in 2001. You’ll notice the funny shading on the Google logo. That’s how it used to look. For those of you younger folks, this is what Google looked like when it was launching in 2001. It was actually launched in the 90s. But this is when it started to really pick up some steam. And what was really fascinating about this is there was a button on the bottom of your Google screen that said How to Google, and it led you to this page. I found this in the Wayback Machine when we were prepping for this. And one of the most fascinating things is it said: New to Google? Learn the basics. Here’s one of the things you need to learn. So Google only searches pages that are an exact match for your search term. And so the example here it gives is, so if you didn’t find what you were looking for, with Boston Hotel, try searching for “Boston hotels.” Fascinating, right? You know, Google has advanced so much since 2001, that we forget how rudimentary that tool was in 2001. One of the things that started to happen is that Google started to interject some AI in its search results, it started to learn from the search results that had before and said, does it really matter if it’s plural or not? If you’re searching for “hotel,” and your browser happens to be in Boston, do we just want to give you Boston hotels, because now we know the GPS that’s local to you. There are a lot of things that have advanced, but we sort of take it for granted. It’s sort of like that, that frog in the boiling pot of hot water that we just didn’t realize until we go back to Google in 2001. And we say: Wow, that was a terrible search engine by today’s standards.

Jill: I remember teaching Boolean Logic. Yeah, we were going and, you know, doing queries that were like, you know, this in quotation marks, and in capital letters, this in quotation marks, right. So like, it’s amazing to me to see how far it’s come. But I think it’s important for us to go back and look and see how it’s learned and make things better for us. 

Ben: Yeah. So in the same example of the way that we’ve learned how to Google in 2001, we’re now in that phase ever since ChatGPT launched at the end of last year. We now have to learn how to AI in a different way, because the way that we Google and the way that we use AI are fundamentally different. So the way to think about it is: when I’m asking Google for something fundamentally, I’m asking typically for a list of webpages, and I’m determining the success of that webpage by its result being close to the top. So a good search engine delivers the results I’m looking for, but it links to the result I’m looking for in the top five or so. And usually they would say top four, because after four, then you start scrolling, right? But I want the best results to come in the first four results in a Google search engine. And then it’s giving me new search results. And then it’s generating, in contrast, ChatGPT and AI. If I’m asking it for something, it’s actually providing the answer to me. It’s not providing a link to the answer. It’s actually providing the answer in terms of text directly to me. So my result is fundamentally different, which means the thing I’m asking it to do is fundamentally different. And when I go on a Google search, and I didn’t find the thing I was looking for, I change my search query, right? Like, it’s sort of like, go into that search box. And I fundamentally change something in Chat GPT, the search actually goes away. And now I’m typing back to it to say, actually, you didn’t give me what I wanted. Here, alter this, or change that result, so that I get a better result.

Jill: I’m getting some questions in here. Here’s one from Rebecca, just asking how to frame questions to get the exact answer, which I think is what you’re really talking about.

Ben: Yeah, yeah. We call that a prompt. And one of the jobs of the future is prompt engineers. So people whose job is to train AI machines to get the desired results. So we have to enter in text in order to get that desired result. And we actually, in many ways, I find myself putting on my programmer hat and saying, I’m thinking more in this logical way about the outcome I want, than I am the way that I would Google things. And that’s a fundamentally different way of thinking. So when we think about learning AI, we want to ask it exactly what to do. Right? We want to ask it for the result. One of the things I found myself, by the way, doing right away is, I was really polite to it: Please give me a question format. 

So I sort of find myself saying: Hey, give me this, right before my next meeting. I have the previous information. You know, one of the first things I was testing, I was at home, my wife cooks professionally. And I was telling her about how cool it was. And like many things, my wife said, “That doesn’t sound impressive. I already have Google.” And I said, Well, you know, in order to prove this to my wife, I was like, “Well, what are you making a recipe for?” And I said, “Give me a recipe for,” and it spits out a recipe. And she responds with, “Well, I could Google for that.” And I said, “Okay, that’s cool.” I said, “Give me the shopping list.” And so I asked her to rearrange it into a shopping list. She’s like, “That’s cool. I’ve seen some apps that do that.” And then I said, “Now substitute, one of the key ingredients.” I actually didn’t know what the ingredient was. So I was like, “What’s a substitute for this ingredient?” And it gave me four examples of alternative ingredients. And then, “Describe to me how those would be different in the recipe. And she said, “Can you send it to me? Because I’m not gonna say you’re right.”

And she uses it every day. And I’ll give you some other examples of ways that she uses it also as an educator who teaches. But as we think about prompt engineering, when we talk too early, it’s natural language. So AI understands natural language–the way that we speak. It has been trained on internet language. So in the English language, it knows it has a large library of English language. So it does understand the words that you are normally used to speaking. What it doesn’t have, oftentimes, is context. So when I’m searching for a hotel, it knows hotel, and it knows where location would be. But like, I have to actually give it other things. Like I would like this kind of hotel, or this kind of thing, because I’m asking for the exact result, not a list of results. I want the results, right. So when I tell it, “Hey, I’m taking a trip. And I would like some advice for planning for that trip.” It knows that I said, I’m going to Chicago, but it doesn’t know that I like deep dish pizza and I’m allergic to some parts of seafood. And I’ve got these sort of, you know, I want it in this distance from downtown. It doesn’t know any of that information. So that becomes really important that we feed it with information.

Jill: And like I was just going to share a teaching example. Today I was asking it to create some, excuse me, some questions using Webb’s Depth of Knowledge levels one, two, three, and four, which it did, very impressively. So teachers utilize it, it’ll save you a lot of time. But then I went back and I said, “Can you do levels one and two, and put them in an objective format?” And I meant multiple choice, true/false, yes/no, right. But it didn’t understand what I meant. And so I went back and said, “Can you give me the level one and level two questions in multiple choice format? And boom, boom, kicks it out. I mean, and really excellent examples. 

Ben: Yeah, that’s fantastic. Yeah, so you know, I’m going to try to give you some examples of different types of prompts, different use cases for using AI. One of these would be things about like, you know, what if I was thinking about a press release, right? So I might have a, you know, if I did this on Google, and I asked for a press release, whatever the press release was, you know, I searched for, you know, a press release, or maybe a query that involves that, it’s going to give me a list of templates for a press release. That’s kind of how I would have done it four months ago. Now, I would never start a press release without just asking it for, “Hey, we had this incident, here’s what I need a press release for.” I’m gonna give it the context of where the school is, I’m going to give a context for. Now, this is obviously longer than I would typically search for on Google. But my result is that I want is I want it to actually write the press release as close to possible as it can. One of the other things that I find in here is when I’m writing on behalf of an organization, I also would love… you’ll see my bonus there is, if I’ve got an about text about the school, I’m going to paste that in every time. Because at the bottom of my press release, or included in that long newsletter, I want it to understand that text. And I wanted to use that both for style. I might ask it literally, can you write it in the style that I’ve written this, right? So if you’ve got an about text that you feel really confident about, that can be the information that it knows that’s the tone. That’s also some of the base information that it needs to know, in order to write a great prompt, and it will begin writing it.

Jill: Yeah. And you guys, if you’re new to this, I think it’s important to know you can say, “Write this.” I think you could say, “Write it in Old English” or something. 

Ben: Write this in pirate, as well. There you go. 

Jill: Of course, but I know you can write it in, I can say, “Summarize this in three to four sentences for a teenage audience.” I can say, “Write this in the voice of an eighth grade boy.” I can say, “Please add for grammatical errors,” because that’s one thing I’ve been impressed with is the grammar and syntax is actually quite good. From my experience with it. And so those things are also, they’re super cool. And it also, you know, is what causes us a little angst as far as teachers as well.

Ben: Yeah, yeah, for sure. But yeah, as I’m thinking about press releases, those are very programmatic in nature. They’re very formulaic, right. So when I’m writing something like a press release, even some of the newsletter types of things I might write, they’re, they’re very formulaic. And that’s a fantastic way to start thinking about, again, give it context, sometimes be creative. So you know, stretch it. Copy and paste something to give it a title and give it that style and tone. And then that relevant information that you want to make sure is included. Also things like, “Hey, I want to include a quote.” So here’s what Dr. Jill said, and include this in the press release, that would be something I might include, as I was writing a larger form, and then it’ll start to, you know, bring that in. And articulate that.

Photo creation apps, I mentioned a couple of those. Left side is Canva. Canva has a couple things that both do writing, they have one for writing, they have one for designing images, and they have one for actually creating presentations that we were playing around with this week. Those are pretty effective. You can get started for free, but then they of course, have a kind of freemium version, where Canva does charge at a certain level. Dali is free, at least as of now. There are some really cool things that dolly can do I find myself being a poor, prompt engineer sometimes. So sometimes I get images on Dali that I’m like, that is nothing like what I wanted. Also, yeah, come to think of it, I’m also not weirdly that specific about the images. So you know, when you’re thinking about something like that, hey, I, I asked it to put a cat in a parking lot with a note to take home, right? So this is a great example. If you put that same text into Dali, you will get a different image. And you’ll usually get… there’s about six images it usually generates, but I’ve had a few of those where, you know, I was texting my dad and I was like, oh, I should text him an image. So I just pulled it up and Dali made it really quick and it downloaded an image that fit what I wanted to text him so those are easy ways to just try it out. You know, there’s easy ways to just kind of like especially these more generic images. That’s kind of where it’s being successful today with Dali.

Jill: I was just gonna say I’d love to see, and I haven’t done it yet but, to have one of my art teacher friends have a go at it because you can really be much more specific in style and color and nuances that it really picks up on when you’re giving a good instruction. I my knowledge is just limited enough, and I don’t know exactly what I want to get to. 

Ben: Yes, you certainly can get a lot more artsy with it. Plus, you can also, like with Canva, you can have it actually design out some of the layouts. So you can have it include the specific green that you want, include, you know, a logo or style elements. So that’s, that’s really cool. Also worthwhile noting, there are AI opportunities for videos as well. So the way that I would certainly think about doing video, first of all, is thinking about creating scripts. Hey, what’s an idea? I need a video to promote the next snow day? What like, how would I write that, right? Or we have an opportunity to communicate with the public on these five elements at our next board meeting. Or, as a result of the board meeting, here’s the board meeting minutes; I’m going to paste those in, “Would you help me summarize these,” to then generate the video scripts that I want to read after the fact. If you can paste all of the meeting minutes in, you’re gonna have a pretty good opportunity to have a good succinct summary to that video. But read it before you actually record it.

Jill: Right? I think that’s true. Anything that’s being AI generated. One of the things we’re gonna try, I’m just gonna throw this out there, we’re going to try to take the transcript from today, throw it in AI and see if it’ll summarize and highlight and what exactly we can get it to do. We’ve done that with a few meetings, and it’s done pretty well. But seeing like, a little bit longer-form with really a lot of content, seeing what it can do. So we’ll share that out if it if it works.

Ben: Yeah, I use an application called Otter.ai. It does a great job of recording and transcribing meetings like this. It also integrates with Zoom if you’re using zoom. So you can transcribe a lot of text and it will summarize and categorize information for you. It’s a fantastic way to be thinking about that, too.

Jill: That’s a great tool. You can also then use it to, you can highlight and take notes while it’s recording. So I think especially my special education teachers out there, I think that’s a really cool talk-to-text tool that you could utilize in your classroom.

Ben: Yeah, so a couple other things you can use. There are some AI tools, most of these are paid for AI tools that will actually edit and generate a video for you. So if you think of a talking head video, that’s kind of the most successful ones I’ve seen so far, where you can essentially put a sound to them write out a script, and it will essentially have a person reading that. And it does look incredibly realistic. So but they’re typically kind of restricted to stuff like that, where it’s sort of a one person talking head because it is pretty complicated to master that in a variety of other ways. So some easy time savers that we’re kind of covering so far, rewriting event language. So thinking about, hey, we just had this event last year. So it was last May, and we had a prom. And here’s what that looks like as far as text goes, would you update that prom, we have a new user, you could paste in all the language you had from last year, paste that in. New location, king and queen. Get one and ask it to update that information. Tell it what that new information is.

I’ve used it for proofing. So I pieced, you know, several paragraphs that I just wrote. And I’m like, yes, I wrote that fast. I pasted into ChatGPT and say, “Hey, would you proof this for me for grammatical errors, spelling,” and it’ll spit it back to me.

Writing sequences of communication. So that’s an interesting way when we think about sort of longer-form, and we’ll share some examples of that.

Press releases, surveys, summarizing data, survey results, or another one, we get a bunch of survey results, taking that text, pasting it in there. Also fair warning, we don’t have great necessarily security mechanisms on these AI machines. So if you are pasting in something that would again have FERPA implications, HIPAA implications. You know, PII like there, there are things, there are definitely things that you want to be careful in that respect. So as I paste in things, I might paste results of survey data, you know, that are anonymized and then ask it to summarize that for me, and that can be a really nice useful tool if I need some natural language that can that can spit that out.

So what now? Well, one of the things we want to address is some blowback. So one of those things that we’re seeing in some instances, Vanderbilt, you know, unfortunately had a pretty significant outage in which they copied and pasted some information that they got from ChatGPT for tragic events at their school. And they left the language in that said that it was written by ChatGPT, which certainly didn’t land well with the public. That’s a really important component. So that’s one of the blowbacks is, and we have to make sure that if it’s blatantly written by ChatGPT, or it looks like it’s too formulaic in its writing style, that we are editing and reading that. That’s really critically important.

The other thing that I’ll say is that there’s a blowback for the potential that AI is going to take away my job. And I think that’s a really important thing to be thinking about. You know, is your community worried that you’re just outsourcing things to AI? Right? So if that’s a perception, we certainly want to combat that by making things not feel AI driven. But we do want to, you know, create things faster, outline things much better. I mentioned that my wife is an educator. In January this year, she was updating the syllabus for a new class she’s teaching. And she said, she was just finishing it. And I said, yeah, what’s the class you’re teaching? So I typed in the class name. And I said, how many sections are you teaching? She gives those to me? And I asked her to generate an entire syllabus, and I said, “How does this look relative to what she just spent her time doing?” She said, “Wow, that would have been really useful two hours ago, because I spent two hours writing the syllabus really contemplating that. That is 90% of what I just spent the last two hours doing.”

Jill: Yeah, so it’s a great, it’s a great first draft, right. And I was, I was going to share the example… Ben and I have been running through a lot of examples of, where might this help with school, and I was thinking about my time as an AP. And I was, we were talking about how it can really help you craft drafts of emails that you have to write home. And so even when there was an incident, like I was mentioning, if I had an incident report, which I would have already had to write, I could type that in and say, create an email for me. And what it does is takes away some bias, it takes away some of the human emotion, that we as humans a lot of times want to put into things in what we can offer. But in those situations, maybe if I’m mad at a kid, upset with the way a teacher handled it, upset with the whole way the whole day’s going, whatever might be happening in my world emotionally, ChatGPT kind of helps take that away a little bit, and almost dehumanize it and stick to the facts a little bit, which I I kind of liked. And I know we have a couple of examples of that. But then what Ben did was he goes, “Tell me about this incident report.” And then he had it write an incident report. So I just had a little paragraph in the incident report. And it actually put it in a beautiful format that made it really clear and easy to read that I thought, well, that would have been nice to share back with the teacher to share with maybe, alternatively, other teachers that that student may have who would be needing that information appropriately, or even the parents. And so yeah, I was like, that’s pretty cool. It also is going to make sure your grammar is correct, your spelling is correct. A lot of things that when we’re very, very busy, can sometimes put a bad look on us. If we’re sending something out to parents that have those kinds of errors, it can minimize our credentials. Whether or not they should be, it definitely does that.

Ben: Yeah. So you know, I threw some examples here of hey, maybe I needed to create a test or quiz. Now in this case, I didn’t actually give it the context of who this is, who I’m writing this for. I didn’t tell him like what grade, what aspect, that would obviously be a great scenario. Also, if I need to write this three different times for three different sections of the class, and I want different answers or different order of answers, different order of questions, I could always have it reorganize those. So I could be asking that after it wrote the first one, and then have it simplify that into two other steps. That’s a huge time saver.

Yeah, we talked about bias. Removing bias is really important thing that machines are pretty good at that we are not great as humans, especially when we’re highly emotional about it. So you know, I created a really ridiculous sort of incident that happened at a school. Obviously, this was reported to me as, you know, assistant principal. So I’m illustrating this event as I’m just sort of typing with, you know, not great English, and I’m just kind of like typing that freeform in the moment. I’m emotional about it clearly. So I submit all of that incident report up and I want to show you kind of what two different outcomes are having that type of report. So I’m asking it to generate… the output I’m asking for is an email. And so ChatGPT generates this for me. And I want to highlight a couple of things for you. One, so it’s generating it to Johnny’s parent’s name, right. So I need to change that. And then it gives me a pretty clear example of what’s happening. What I do want to highlight, though, is it also in the end paragraphs, it starts to actually go a little far, like, it starts to say, hey, we’re actually, you know, we request your support and discussing this matter. It also says things like, we’re addressing this behavior in accordance with our school’s code of conduct. And some of those things. Well, I didn’t really ask it to do that. I didn’t ask it to go that step. But ChatGPT is going that far, right. So that’s important when you’re thinking about the outcome, is that sometimes it actually goes a little farther than maybe you asked it to go. And it’s filling in details that maybe you may or may not want to be including.

Now, I did ask it, the follow up on this one, to Jill’s point of the incident report. And it gave me a great example that were just I asked it just give me the facts without any any bias or color. And it did a great job of doing that. Now. So ChatGPT’s case, it went a little far. Google, on the other hand, one doesn’t go nearly as far in the total content. But I want to direct your attention to the address at the top, which is Dear Mr. And Mrs. Smith. Now, I did not give it Johnny’s last name, but I did tell it that Mrs. Smith was the teacher in that scenario, right? So fatal flaw there on a couple significant levels, right. So again, great reminder: read it, after it generates that. That’s a really important miss on Google bards fault. However, the actual text that I wrote in there was was pretty solid, in most cases, so. So it’s interesting to kind of analyze those and test how those are working.

But I know we had some questions also, that were a little bit more around, now that we learned some of these tools, we understand a little bit of how they’re working, what next? Like what’s changing what’s going to happen with AI? If we thought of this as Google in 2001, Google had a pretty good run for the last 22 years. We definitely think that AI is going to continue to advance significantly. But we kind of want to understand what do we do, especially as content creators, at schools, in light of AI having a prolific impact. So a few things, one of those is the recognition that trust is a human component, right? It’s not an AI component. We don’t sort of inherently trust AI. We trust the fact that it seems more human. Like we actually… the thing that we really like and trust about AI is that it’s starting to sound more like a human. But we do trust in robots differently than we do people. So with trust being distinctly human, we’re trusting that as human, we distrust things that have selfish motives. So if you ever disagree with me, just go try to buy a car, right? Like the first person you’re talking to, you’re saying, “Hey, I don’t know that your motives and my motives are lined up, right.” But we also really value that human connection. So we kind of put these three together, and we say, we really trust that which is human. We distrust, selfish motives. And we really value human connection. What does that lead me to do? As it pertains to the communication I would have for my school or my school district, it does cause me to say, hey, what’s happening in an AI world where I can auto generate a lot of text, not really care, I can push that out at you to you. There’s certain things that are announcements that I have to get across and are great uses of AI. However, there’s, there’s a human component that I want to really bring in there. And that human component is going to stand out in the future more so than ever, because it’s the thing that’s really hard to collect and uncover.

So a few quick examples. So as we think about things that are blatantly and obviously human, obviously, the way that we connect to young people is, is one of those most profound ways to do that. So pushing that energy that students really provide, pushing that to the forefront is going to be increasingly important and so much more important as we can auto-generate a story for you, but we can’t generate that story that’s about and with students. The other thing that is happening is with texts being so easy to generate, the way that we think about that bar being raised and more content is that we have to focus more on uncovering these students stories and their specific stories. And we have to improve the way that we’re doing that visually. Like, we have to be uncovering photos and videos. So that’s where I would say, I’m really bullish on the way we think of content collection. Like putting, you know, getting students on the ground getting, spreading out where our stories come from, and getting, getting those content collectors to come to events that then become content creators, that becomes a really important component that, quite honestly, AI is incapable of doing. So you’ll see this example. I think one of the most blatant examples: Canva started doing a series of videos to promote how good their AI photo generator is. Well, of course, they can’t generate, you know, this story of the student that I met last month. But what they can do is say, “Hey, here’s a Purple Cow in front of lockers.” And if I want that, and and if that’s the message that I want, I can do that. I can ask AI to do that in a quick, easy pattern. I can have it in whatever shade of color I want. But finding that unique and valuable student, that human story, I can’t do that. And that becomes really important.

Jill: Yeah, it’s kind of those… even that idea of AI can help our teachers or school PR professionals get that some of that busy first-draft-type of work done that, that then, like, it frees up the teacher then to spend more time with the student to spend more time one on one individualizing things. Whatever it might be. As opposed to, you know, drafting emails home and all those things that really bog us down being creative. But obviously we promote student storytelling. I think their perspective is so important. And getting in there. And you’re exactly right. That’s just something that AI can’t do. And I don’t see it being able to do, either. It just can’t. It doesn’t have… it doesn’t have that type of intuitive.

Ben: Yeah, it can’t go collect the stress. But what it can do is help us to realize, hey, when it’s not about us, and maybe I need to send Drew out to gather some photos. When I talked to Drew and I say, “Hey, you’re going out to this specific event, you’re going out to visit the play,” I could ask AI to come up with a plan for it. What are the photos that you need to capture? What are the moments that you need to capture and give them a plan that’s written by AI and send them out. So I think there’s some really cool ways that we can empower young people to have that story of students to be uncovering that story for your school. And that can be distinctly human. That allow us to no longer be guiding towards, towards how do we beat the robot. But essentially, the way that we win long-term is not forcing ourselves to sort of compete with a robot, but instead to be distinctly human.

Jill: Yeah, definitely. Well, I think we’ve, you’ve definitely given us a lot of information. I learn something from Ben every time he talks about this, but we’ve covered some basics. So hopefully, if you haven’t used any type of AI tool yet, or if you’ve just dabbled in it, you can now jump into several. I know I’ve done a couple. I’ve done ChatGPT, Google, and Dali. Oh, and Canva’s Magic Write. And so I’ve used each of those a little bit… ChatGPT more extensively. So you can jump in and give some of those a try and start to learn from some of that. We’ve shown you how to frame some of those questions or, or actually asking it what to do. And being specific and learning from that. I think you’ve given us some really great efficiencies, right? So no matter what your job, but especially school PR, classroom teachers, school administrators, or whatever role you might have in the school district, I think there’s some real efficiencies there.

I know Ben has pushed us here in our office, like, sometimes I’ll be like, I think it’s just faster if I just do it. But thinking about it in a way that, is there something that could save me some time here if I allow AI to do it? For me, at least maybe that first step or a first couple of steps. So if I’m going to go do a presentation, I can ask it, “Here’s my basic idea. Can you give me an outline,” and then I can even put it in and here’s some slide ideas for you. And that really does save me some time and just organizing and then I can add, you know, I can bring my human element to that, but it saves me some definite time there. And so I really… I just really appreciate those things. So we talked about some of those things you may be afraid of, I think you put some worries at ease. To me, those high-level things that made me worry a little bit more that I wasn’t worried about because I didn’t know, you know. Ignorance can be bliss. But knowing why we click those boxes that say I am not a robot, and how systems are in place to be…

Ben: You know, I’ll share… we have a copywriter on staff here, who is one of the best writers I’ve worked with. And, you know, we had a conversation, this was about a month before ChatGPT was launched for public use. And I said, you know, we’re going to be using so much more AI. And she said to me in that moment, she said: “Doesn’t that fly in the face of everything we’re about?!” And I said, “No.” I said, “You’re too smart. You’re too capable of a writer to not be using tools like that in the future.” And, you know, we’ve had great follow-up conversation sinse. And she’s clearly using a lot of AI tools to be more efficient. And her writing is phenomenal. And so that’s, that’s really part of that story is that it does come around to: We want to focus on the truly human elements and human stories.

Jill: I have to… I kind of want to close a little bit with: My daughter was writing a research paper and she actually wrote it on why teachers should embrace tools like ChatGPT. And I said, “Include in there that you used ChatGPT to write a first draft.” And she goes, “Am I gonna get in trouble?” I’m like, “Maybe, but I don’t think so.” Because I happen to know our teacher and our teacher is fantastic. And I said, “Just put it in there, and be honest, and cite your sources. And talk about how it helped you through a certain process to get to the end.” And she ended up getting not only a good grade on her paper, but the teacher put it up on her bright board, which was also pretty cool. So that was a neat experience for me to be able to work with her on that, too.

So hang in there. I know these are ever-changing, even as we were prepping for today, like, things are constantly changing. We’re here for you. We’ll send out some follow-up resources. But that’s part of our job is to help keep you informed, and give you ideas. So we’re here. We ground everything we do in pedagogy. We want to support you as educators and school storytellers. So thanks for joining us today. And thanks so much, Ben. Thanks.

Interested in listening to the full webinar? Click the button below to watch/listen free.