Estimated reading time for this article:
35
minutes.
Join Liza Long of the Idaho State Board of Education for an engaging conversation on artificial intelligence and interactive content in open publishing. Hosted by Amanda Larson of The Ohio State University, our most recent session of Pub101 for Authors helped participants explore opportunities to incorporate GenAI, consider pedagogical objectives when including AI and interactives, gain key insights on how to cite, integrate, and track AI content, and much more.
Audio Transcript
Speaker:
Amanda: Hey everybody. Welcome to Open Education Network's Pub101 for Authors. Thank you for joining our session today. I'm Amanda Larson. You all probably know me by now, but I am the AERI program coordinator at the Ohio State University and I am the Pub101 chair and I'm going to be the host and facilitator today. Soon I'm going to hand it off to Liza and she's going to talk to us about AI, which is super exciting, AI and OER. And in addition, there might be many of you who have experience in this topic. In addition to our guests, and we invite you to share your experiences and resources in the chat as well.
A few housekeeping things. As always, today's session will complement the Pub101 for Authors curriculum and related resources, specifically unit four on AI. And we have an orientation document that I should have the link to here for you. I'll paste that in the chat. Slides, session recordings and curriculum connections will be included in that document. And we are recording today's session. We already pressed that button, so we're good to go. We are committed to providing a friendly and welcoming environment for everyone that's aligned with our community norms. So please join us in helping create that constructive space.
To kick off today's conversation, we have a brief reflective question that I'm going to ask you. What excites and/or worries you about using AI to create OER? I'm going to also put that question in the chat. That AI help create Skynet and/or end my profession. Yes. Hilarious. But true and a concern. My peers' thoughts about using AI and their evaluation for my work because I used AI. I think that is definitely a super great concern. Got another one. I'm worried about crediting the use of AI in my work. I use it to edit, not create, but I am worried about people not taking my work as seriously. Yes.
We're going to talk about all kinds of things, but one of the things that we talk about in the curriculum is about being transparent. And so you could say, "Hey, I used AI to do this editing, but all of the text is my own." Appearing, seeming, or even being hypercritical after telling my students they should never use it to create their own work. The environmental costs of widespread generative AI use. All of these are great things to be concerned about and interested in as we think about how we might use AI, and that's further complicated by copyright. I'm going to hand it over now to Liza to talk to us about those things. Take it away.
Liza: Oh, welcome everyone. And I want to thank the Open Education Network and Amanda and Karen for inviting me to present on a topic that I am pretty passionate about. And I want to thank all of you for being here. This was a great way for us to kick things off to see where we are, and I'm not surprised to see that we're all over the map. So I'm going to go ahead and share my screen. We're going to walk through some of the basics that you learned about maybe through the curriculum today. I've got an overview of that, and then I want to demo a few things, and I'm hoping also, I'm assuming you may have some questions. So as those questions arise, you could put them in the chat. I'm hoping maybe we'll have some time at the end. Are you all seeing the Open Education Network? Okay. Yeah.
Amanda: Yes.
Liza: Before we even get started, I just want to introduce myself really briefly. I'm Liza Long. I am the director of digital learning and AI at the Idaho State Board of Education, and I work with our eight institutions of higher education on issues related to artificial intelligence and other emerging technologies. So this is an area of interest to me. I am also a PhD student in Idaho State University's English and Teaching of English program. And I am an instructor at the College of Western Idaho, an open access community college where I teach asynchronous online courses. So I have a, I guess, somewhat unique perspective on this particular topic.
We're going to be looking at how artificial intelligence can be used to create content and also interactive content, because for me, that's some of the most exciting possibilities that we have. But as we kick things off, before we get started, I want to clarify that when I use the term artificial intelligence, what I'm actually talking about are large language models. So we hear AI thrown out a lot, got Skynet, got all those good things you all are worried about. I mean, it's out there, but what I'm specifically going to be talking about is large language models. I want to clarify that and how they can be used with open education resources.
So what are large language models? Well, most of us know ChatGPT by now. Claude is my own personal favorite. Gemini from Anthropic. There are also other versions such as Qwen, DeepSeek. Meta has open LLMs that you can download and run yourself. And there's certainly many others. I saw, if you haven't seen it yet, Talkie is a large language model that its training data is cut off at 1930 because it's trained only on things that are in the public domain. And as you might expect, it has some interesting hot takes about life. So when we talk about large language models, we're talking about that specific subset of AI.
Also, before we get too deep into it, I like to start my presentations on AI with an acknowledgement statement that I originally found in January of 2023 from Donna Lanclos and Lawrie Phipps. They're amazing open education resource folks to follow. I follow them on BlueSky and they gave this offering to the open education community very early on in the introduction of these large language models to our lives. So I do want to acknowledge that this presentation was prepared using Google Gemini's agentic browser and ChatGPT's image generator. And I acknowledge that AI chatbots do not respect the individual rights of authors and artists and ignore concerns over copyright and intellectual property in the training of the system.
Additionally, I acknowledge that the system was trained in part through the exploitation of precarious workers in the global south. In this work, I specifically used AI tools to generate a draft of the slide deck from previous presentations and Open Education Network content and to demonstrate OER creation techniques. So when I'm teaching with AI, I always like to start with an acknowledgement statement and sometimes a conversation, often a conversation with my students so that we can do exactly what you all did in your reflection in the chat. We can get some of those concerns out there and we can acknowledge, because it's important to acknowledge that these concerns are real.
So now someone once asked me, "Well, but you still use it." Yeah. Yeah, I do. And I fly on airplanes and I'm really trying to wean myself from hamburgers, but I occasionally have a hamburger or a steak. So we all have to make choices, but I think it's important, especially with this technology to keep these results and these things in mind. Okay. So now that we have that out of the way, I want to look at my learning objectives for today. I'm hoping that we can explore some ways and opportunities that we can use AI to play with experiential learning.
We can identify ways to incorporate GenAI into the textbook creation process. I'll demonstrate some of my earlier attempts and then where I'm moving now. We'll consider pedagogical objectives when including AI and interactive materials. So as always, your pedagogical goals, your learning outcomes should drive your choice of whether to use this technology or not. And then finally, and this came up in the chat, we will learn how to cite, integrate, and track AI content. So those are my goals for today. Hopefully we'll get all of that done.
So when we think about AI, when I talk to professors, the concern that I hear most is that students will use AI to cheat, and I continue to hear that most in 2026. I highly recommend the book, The Opposite of Cheating by Bertram Gallant and Rettinger, if you haven't read it yet. They point out that students who use AI to cheat are just like students who have always cheated. They're either not invested in the assignments that we're giving them or they are afraid of failure. Those are the top two reasons that students cheat, and that has not changed in the age of AI.
So as we think about that, what I'm hoping we can do today is maybe shift our mindset to some potential benefits rather than focusing on, "Well, my students are going to use this to cheat." Maybe instead think, "How can I use this to help my students become a more effective learner even through some personalized learning?" So we can use AI in OER creation in a few different ways. And the first way is efficiency. We can rapidly generate text, images, and it's great with coding assistance. And I'm curious, even maybe you can put in the chat if you've vibe coded something or if you have a link to something you've vibe coded, it would be fun to see. Now, if you haven't, that's fine.
But I was at a conference at Cleveland State a few weeks ago, and there was a wonderful presentation on someone sharing using Google Colab to vibe code. And at my table, everyone seemed to shut down when that code popped up on the screen. And I was like, "Oh no, guys, you can do this. You can do this." I pulled up my own Colab notebook and they were like, "Wow, you're an English professor." I'm like, "Yes, the future is now." So efficiency, if you haven't tried coding something, go try it. Even Claude makes it really easy to create a fun artifact.
I initially came to AI through ... I was curious, it was a research question. So I was teaching an English 211 class and I needed to create an open education resource for that course because my very expensive textbook was going out of print and also any replacements, we're talking 80, $90 range. So I'm very committed to open education, wanted to create that textbook, had a semester to do it. And I thought, there's just no way I can actually do this. So this is January of 2023. I'd been playing around with ChatGPT since it was released in 2022, and I said, "All right, let's take this puppy for a spin. Let's see what it can do."
And I want to show you really quickly, because this was my first attempt with AI to do that first step, that efficiency step. So this is the book, and I would say you can look at the book or not, but AI was used throughout. And I have a philosophy statement. This is baby me just starting to work on this textbook. And I noted that ChatGPT came out. I was already working on the textbook and I decided to use it. And you can see that little acknowledgement statement. I was really glad that I found that early on. But my favorite part, what I would recommend you look at is the links to ChatGPT chats used in writing this book, along with my prompts.
And what I love about this experience, if you're new to AI and you want a training course, don't take a training course. Pick a really large open education project like this and just dive in. I was so bad at prompting. It's just really fun for me to look back. You'll notice prompt number 10, provide a list of 10 peer-reviewed articles, prompt number 11, these are not real articles, prompt number 12, these are also not real articles. So really fun way to get introduced to what AI could do and what it couldn't do. And when we talk about transparency and citing, as I've been working with my students, I always point to this and I say, "Look, you don't have to be good at this at first. You can learn by doing."
But then the other way that I used it for my students is by giving them, showing them, "Hey, guess what? I know that AI can write your essays for you. So here, I had AI do it and here's what AI wrote. This sounds pretty good." And then I critiqued it and we kind of flipped this particular classroom where they would do the same thing. They would try to write the very best essay they could, and then they would critique it. Now I'm going to tell you, this particular assignment is getting harder and harder to do because AI is getting better and better. Those of you who've used it to write an essay, but I still don't think it's impossible.
But as I'm teaching now, I've shifted more to like, "Hey, go find your voice. Find places where this voice feels flat or where you're doing a rhetorical analysis, a close reading of AI text." Why does it always say it's not this but that or those types of things. So that's just to show you how it could be used in creating an open education resource. So I created this whole book and what I realized through that process as I worked on it is, oh, these tools are absolutely most useful for experts, right? So we want to be careful as we think about where to introduce AI and where not to introduce AI.
The next place that we'll look at for benefits is AI as a creative partner. And I was just in a conversation yesterday where someone said, "I'm not sure that it is a great idea to use chatbots for brainstorming." I kind of agree. It feels like early on we thought, oh, using chatbots for brainstorming is a great idea. What I prefer to do is have students do some preliminary brainstorming first, or for me, if I'm working with a chatbot, I want to have a very clear vision of what I want before I go into any kind of session with a large language model. So absolutely effective as a thought partner though.
Rather than brainstorming, I like to think of them as thought partners, and I'm the one who starts with my own thoughts, right? But then they can test them and we can have conversations about them and test those out. Also, you can do rapid prototyping. Let's say you want to develop a new course or a new assignment. It's really easy to prototype a new assignment using these large language models.
One area that I'm very fascinated by, and I've seen others do some really great work with this, I just worked with a social work team through Rebus who was putting together a case study book, and they were able to translate and localize content within that case study book very quickly by using AI tools. So if you haven't looked into that, that's another great use. And then finally, personalization, if you have not heard of NotebookLM yet, that is a really fun way to get students to take ownership of the course and to bring that personalization in. So lots of fun things here, right? These are all ways that we can really improve learning.
Keep in mind, one thing that is very top of mind for me, especially my current role, the paid models are all much better, and in some cases, the only ways to do some of these things. So when I think of personalization, for example, I'm putting on my student hat, I'm just wrapping up an old English course and I created a custom GPT BeowulfBot. I'll show it to you in a minute. The BeowulfBot helps to tutor me in old English forms and test me. So it's a personalized learning bot. It also tracks my particular errors, keeps an error log for me so that I can look for trends. I know that I'm really struggling with the difference between weak and strong adjective forms because BeowulfBot tracks that, and then I can do more practice in the areas where I'm struggling.
I think there's a metacognitive benefit to having students create these types of personalized bots themselves rather than you creating them. However, that requires a paid account. And the paid accounts are all $20 a month right now for those premium models. And I'm here to tell you they're heavily subsidized. That's not really covering the cost of those tools. So for some students, out of reach. We've always been concerned in the open education community about digital equity. Let's keep our eye on the ball here. This is a huge concern for me that we'll see an increasing digital divide between students who have access to these premium models and students who do not and have to use the free models.
So reality check, that leads us right into it. I already demonstrated ... Oh, I'll share. Melissa, it is amazing. I have ChaucerBot too. I started with ChaucerBot for middle English, but I started already demonstrating some of this through my critical worlds book because I learned this by doing. And if you can find a way for your students to learn by doing, it can be really effective for them. So maybe having them pick a topic that they're really expert in and asking the AI questions, ask it to write a paper about that topic, some obscure lore or anime or something really obscure that they're just really into. That helps them to see AI hallucinations. I mentioned the rates of those are down, but they're never going to be zero with this particular type of technology.
And then we also have to watch for misinformation. I didn't list algorithmic bias as one of my huge concerns, but it is actually my biggest concern with AI tools right now, large language models. That's where it perpetuates some of the biases that are built in the training dataset. I mentioned the one that cuts off at 1930, Talkie. Boy, if you want to see some examples of bias baked into the training dataset, be aware that if you interact with Talkie, you are going to get some views that might seem like really, really out of place in 2026. I'll just kind of leave it at that. But that's a good way, again, maybe to even demonstrate to students that idea of misinformation, algorithmic bias.
Integrity, academic integrity I talked about upfront. I think it's important for us to have those conversations. And I also think it's important for us to know where we are positioned in those conversations. So I will position myself, just so you all are aware. I really subscribe to the ideas of Sarah Elaine Eaton, and I'm kind of post-plagiarist in how I view text. That is a result of my work in the open education community. That is not an inevitable result of working with OER. One of my mentors and favorite people in the world, Jonathan Poritz, is very diametrically opposed to my position on this with respect to AI.
So did AI train on my content? It did. I have a published book that it scraped without my consent. Yeah, that's fair, Amanda. He's very staunch. We've had a lot of good, productive conversations about it. But so my book was scraped. I'm a member of the class that is for getting a settlement from Anthropic. I don't really care. I mean, it would've been nice if they'd asked, but I'm like, yeah, words belong to everyone. Have at it. I realize I'm coming to that position from a place of privilege though. So these are legitimate conversations, people who have concerns over these, I think it's important to acknowledge them.
I absolutely do not think these tools are going to take out our creative class. I think creative writers are still going to shine, maybe even more so. I think when I look at AI art, the only AI art I can personally stand is created by artists who are actually artists and using AI as a medium. The rest of it just look like slop to me. So I know that these are big concerns because I hear them from my own writer and artist's friends. And some of you hinted at this in your reaction to the first question where you talked about how you feel like you might be judged if you used AI, right? I've certainly felt that in my own work.
So I'm a published author. I have a large corpus of work prior to AI. Most of the writing I do anymore is hybrid. That's just become my new method. I'm very confident as a writer. I can definitely write by myself by hand if necessary, but I like to use AI. It's fun. It's more fun for me to write. It's like talking to someone and writing at the same time. Probably talking to myself, if we're being honest. Oh, good. I'm glad Tiara agrees. It's fun. Yeah, it's fun.
The concern that also popped up in some of your responses, and it's really real, and I want to give you some tips actually today for how to address that if you do use AI tools, because it's a concern of mine. The environmental impact. I do not personally ... I think we will see job displacement for sure, but I actually think we're going to see new cool jobs. Ezra Klein has an op-ed about that this week in the New York Times. Gary, were you the one who shared that in our meeting? But yeah, I thought you did. In our Monday meeting, Gary shared this op-ed from Ezra Klein and said, "Liza, this is how you feel about it." And I do. I feel like it's not inevitable that we'll lose jobs. I think the smart companies instead are figuring out how to upscale their existing workers.
We could see an explosion of productivity, right? So in the past, we've seen technologies impact work, but overall it's been a benefit generally to workers. I think we should really be advocating though for mental health. So I published a blog post, I blog at Artisanal Intelligence, and I'm a heavy user of AI tools. I'm absolutely noticing brain changes. I'm also a mental health advocate. I'm on the International Bipolar Foundation Board, and this is a huge concern to me. It's an untested, untried tool, just like social media was, but even more immersive, I think. And people are talking about this new kind of burnout. I've personally experienced it, this kind of AI burnout. So for me, it's sustainability for my own work life. Not necessarily will I lose my job, but am I going to be expected to do an unsustainable amount of work in my job?
For the environmental impact, what we'll talk about today, because it's real, is token use. So you're going to want to start thinking about strategies to optimize and reduce how many tokens you're using. The tokens are what is using up that compute. The compute is what is what we need those data centers for and all that water. So as we reduce our impact in how we interact with these tools, we can personally make a difference. Now, if you're thinking that's like putting your plastic water bottles in the recycling bin, you may not be wrong, but I still feel better myself with trying to do what I can to manage my environmental footprint with AI.
And there's some pretty easy ways you can do it. It can also save you some money. For those of you who've hit cloud rate limits, me five times a few weeks ago on one day, learning to optimize your token use can be good for the environment and it can be good for your bank account. So we'll look at that.
Finally, equity. I have, again, already talked about some of these things. Gary put in the chat, Darren Olson, Darren's been doing this since day one. I think Darren's wrong. So Darren is not being creative in thinking about these tools, but certainly we as educators need to be thinking about job impact. We need to take it very seriously. In fact, I will try to find it. I didn't have it up. I've got a bunch of resources up I wanted to share with you all, but we had an industry panel at our Innovate Idaho conference a few weeks ago that was fascinating to me. They kind of agreed with us recline. It's not that we're going to see job loss, but they are expecting students to come into the workforce AI ready.
Entry level jobs are in fact disappearing. They're expecting students to be able to upskill themselves to level two or three right from the start. So that changes how we teach. But that final piece ties right into that, to equity. Students who've had access to working with premium tools are going to be in a different place than students who have not had that access. And simultaneously, different global regions have different levels of access. I am fascinated though to see the innovation, especially that's coming out of the Chinese models. Some of you may follow the geopolitics of this whole situation.
And from a geopolitical standpoint, the United States had tried very hard to keep our best chips and technology out of China's hands. China worked around that and found a different way of improving their large language models so that we see Qwen and DeepSeek actually really high performing models with research capabilities. So I think that's something to keep in mind. I'm loving this chat, by the way. We need to help people create meaning and purpose around AI. And I'm also, Tiara, I'm really glad that you're saying that you have mixed opinions. I am the same. I call myself a skeptical enthusiast or an enthusiastic skeptic, really kind of depending on the day, right? And yes, Maha has talked about borrowing access. Yes. So that's a great example of that equity.
I also want to talk about the legal landscape, which is unsettled, but we are starting to see some case law being made around this. And back to that open education resource I shared. So Creative Commons is working on this as well, as you're probably aware. And because Creative Commons functions within existing copyright law, I have to put the huge caveat here that I am not an attorney and any legal question should go to a copyright attorney, but I have taken the Creative Commons course. Jonathan was my certificate course mentor, and it's something I follow with interest and certainly around AI. So in the United States, the cases that we're seeing require human authorship. So if you completely generate a text with artificial intelligence and try to copyright it, that's a no-go right now.
What I'm more interested in is hybrid writing, because again, I'm a hybrid writer, so how much is AI? How much is me? We should probably all be paying attention to that, back to how we audit our use of AI, right? And there's some strategies you can use such as using projects to help see where you are in the mix. Ethan Mollick's Co-Intelligence book, I think from 2024, continues to hold up very well around this idea of always being the human in the loop. And I think it's important for us as authors and also as educators to make sure that we're paying really close attention to how to model that for our students.
In Canada, likely in the public domain, but again, this is a rapidly emerging legal area. And hybrid works right now, you can register those for copyright, and that's even if it has AI-generated images, human parts are protected, AI parts are not. Again, how do you tell? For some of us who have really developed a workflow like me, I have a bot that's trained to write like me. How much of that is me? How much of that is the bot? Also, I was a generation one mommy blogger, so I could argue that the large language model's trained on my content, so it's basically me anyway, right? I'm kind of kidding, but you get where I'm going with that. So we want to pay attention to that with any work we're creating with AI.
Then I want to talk ... I'm going to show you specifically this first one, accessibility. I'm presenting on this at OpenCon next week. This has been an area I'm very, very interested in, and I've specifically been exploring agentic browsers with this. So I like to build things while just in public, very publicly. This is very new. It's not finished. It's a work in progress, but what I've shared here so far works and it works great. I'm continuing to work on refining the amount of tokens it takes. This is one way I can sometimes really hit my token limits, but I think this can be a good introduction.
Sometimes when somebody ... I've realized, sometimes when someone like me talks, some of you are like, "Yeah, agentic browsers, I use that. " Some of you are like, "What are you talking about?" So an agentic browser in this case is just me turning on Claude. Claude is an agent now that can act on my behalf on these websites, right? So you'll notice my Claude agent says high risk right now. If I'm working with Pressbooks and I've signed in and given Claude permission to mess with Pressbooks, I don't want to be stopped at every step of the way. So you can make this lower risk when you click ask before acting. And if you do that, Claude will ask you every step of the way if it's allowed to do something.
For the workflow that I have developed here, I use act without asking, and that's where that high risk thing comes in. This book I think is kind of helpful as well though. So I did go ahead and put these core AI concepts. And something super fun that I just added, but I haven't added it to the workflow. Those of you who work in Pressbooks, I created a agentic browser workflow to provide glossary definitions. So the glossary tool in Pressbooks, it was able to just ... I didn't do this manually. This was me telling Claude what to do and that agentic browser over here went in and did this for me. Isn't that fun? So just kind of a good introduction.
Maha, again, came up in the chat. Maha pointed out to me that she likes to include a command when she's working with stuff just to tell it to make sure it's AREA, that it's following AREA guidelines. Yeah, Amanda, I know it was so fast, right? So I liked that. So there's a guide to working with chatbots in here as well if you're new to it. Some stuff that you can do, what the process looks like. I need to actually update. I just mentioned that. I'm going to go in. This is what I love about prospects. I'm going to go in and add Maha's advice to add an AREA tag when you work with Claude so that it's following those principles.
So the book just works through the accessibility stuff. These are the different prompts. And I start with a huge warning that I want to make sure you're all aware of if you decide to use these agentic browsers. Another one, by the way, if you have Google One, this is also an agentic browser. You guys, this is how I created your slide deck today. I had the Open Education Network's that you're Publishing 101 textbook open. I had several of my former content presentations open, and I told Gemini just to look at all that stuff and start creating the slide deck. Super helpful.
All right. I've still edited it and reviewed it, but just getting that draft in was so helpful. So that's what an agentic browser looks like. I'm going to delete that group for now so that we don't have it open. But let's see, it's back to our book and back to that warning. Whenever you're working with ... Oh, shoot. Now I lost my accessibility when I deleted that. Sorry guys. Pay no attention to the man behind the curtain while I just go back to where we were.
Okay. So the warning I want to share, just to make sure if you are new to agentic browsers, leave that little tag on that I showed you where you have to check with each one and be aware that you need to limit access. So don't be giving it full permission to do everything in your life, right? Stay the human in the loop. I do highly recommend most of the agenic browsers work that I do, I actually do on a separate browser tab. That's why I didn't start working with agentic browsers in front of you. I'll do it on a completely separate browser tab that's walled off from other things, or I'll do it on a laptop that I have that's running Linux so that it's walled off. I don't like to use them on my work computer, which is what I'm on right now.
So just again, because of the risk, you'll want to pay attention. And then be aware that the alt-image task particularly is token intensive. So one of the things I'm experimenting with now is trying to create more of a local language model that will do that specific task, but that is the one piece that doesn't really work well for a local language model. It really needs that Claude's reasoning in order to create robust alt text. So yeah, that's the workflow. And then also in this book I'm sharing ... Yes, yes, you can, Tiara, I know. I love using the Gemini one. But Gemini, I can't get Gemini to go in and edit my Pressbook for me though.
But another thing to be aware of too, this was so funny to me the first time I did it. I always recommend that you download an XML copy of your book before you ... So it's as is before you let any agentic browser start working on it. The very first time I tried doing this, it took me about two weeks to figure these workflows out. The first time, it went in and started rewriting my book. So I did not want it to do that. And I realized after that, oh, I need to break this up into discrete tasks. I need to make it work in the HTML code view, not in the visual editor. After I did that and put in a lot of prohibitions about don't change my content, focus very specifically on these HTML tags, then I was able to get it to work quite well.
And like those of you who use the Ally Checker, you can see I do. This Ally Checker, I went from having 19, 27, whatever, tons of alerts to zero. Yes, Amanda, you could. I haven't written the table remediation one yet, but you absolutely could. Yeah. So any of these workflows, it can do. I started with heading levels because that was pretty easy to figure out how to do. Then I did the image attribution. I've written one for TASL attribution. And then here's the glossary terms that I just showed you. So really any type of work that you want to do, those agentic browsers can do within Pressbooks, and it's really fun.
Yes, you could do that in the code. You absolutely could. So the one other thing, this is what I mentioned Colab. Oh, no, so the VPAT, the agentic browser will do a really nice VPAT analysis of a website. This was really helpful for us when we were trying to get ready for the Title II deadline that was a few weeks ago that then they bumped. But this one, it did a really good job of identifying the issues in our state board website. So we were able to go in pretty quickly and get those fixed.
And then the Colab one, this is something that I'm running not through an Agentic browser. So this is a little bit different. And it was my first Colab project. It scared me to death. It was great. Not only that, Colab was just really, really fun for me because I like to learn by doing. So this is, again, a way to think about how you might incorporate these tools as interactive tools. Having students go in and vibe code something and then using a tool like Colab, Gemini will help you. But also what I liked was I could say, "Hey, Gemini, why did you do that? Explain what that piece of code is doing." So I really enjoyed that process. That's how I like to learn, and it was kind of fun.
So this one ... Oh, there's no accessible text. I'll have to get that fixed. But yeah, so I think this book might be useful for you if you are thinking about accessibility issues, maybe some fun things to think about. And then some of these tools were actually shared in your textbook, but I don't know as much about them. So I'm not going to demo this one, but I wanted to include it because it is included in your textbook. And then the structural outline, I use this for critical worlds. So it was really, really helpful with the structural outline. I strongly recommend using it for that type of work.
And also really kind of fun. Again, back to Tiara had mentioned, she thinks it's fun to work with them. I like talking back and forth when I'm doing structural outlining, and it just feels like I have a really smart colleague that I can work with on that. So chapter structures, subtopics, really fantastic use of time. Or if you have an existing one, having it review it, look for flow, look for scaffolding, make sure you give it the role of an instructional designer or a curriculum developer, and it will do a great job of helping you with that. Oh, so this is the script because I don't have it. This is the Magic ToDo one that's mentioned in your resource, but I don't personally use this one, but I wanted to make sure you knew what it was.
Text generation. So how many of you have created ... Okay, great. Amanda put, she's the Goblin tools expert. I thought it was super cool, Amanda. I just haven't had a chance to test it out myself. And in full disclosure, I mostly work with the frontier models, but there's lots of cool tools out there, different tools that you can use. So don't feel like you're limited to those frontier models, and now we know who to contact if we have questions. So for content creation and multimedia, if you have not written a custom GPT that writes like you, do it. It's just such a time saver.
I use it a lot for drafting correspondence. So yes, Gary, you have a lot of cool content experience. Actually, Gary just presented on this at Innovate Idaho. Gary took my English 199 writing with AI experimental course and created a 30-page graphic novel, which ... Yeah. Did you know, Gary, now that Gemini has a storybook feature? Have you seen that one? Oh, you'll laugh because I wonder, I have Gemini up. Let me show you really quick. I think this would be a really fun thing too. Gary will laugh about my topic. So where is this? This is the storybook one.
So it will create a picture book for you. I've only done one. I was in Bali when this tool came out and I was AI free, but I was like, "Hey, this might be fun. I'm going to test it out." It's so wrong, but that's supposed to be me. So I gave it a picture. That's me and Bali. There's my guide. But I learned really quickly. So terima kasih. It's true that Bali is an Indonesian country, but they do not say terima kasih. That's how you say thank you in Indonesian. They say suksma. So I kept saying terima kasih and people were like, "What? Suksma." I was like "Mewali." So yeah, so there's things that are wrong, but I think that would be a fun thing.
I thought of you, Gary, when this tool came out. It might be fun for you all to play with for content creation and even creating some kind of assignment around that. I haven't done that yet, but I think it would be fun. Data and code. So I have used it for analyzing patterns in large datasets, and it's really great for that. Has anyone else used it for that? If you haven't, just be aware, know your data, and be careful always with personally identifiable information, right? So the first time I tried this with Claude, Claude couldn't read the data that I gave it, but it didn't want to admit that it couldn't read it. And so it made up a complete data analysis. I wish I were kidding.
And I was looking at it and I was like, "Those gender numbers look really off to me." You've had Copilot fail too. Yeah, exactly. I was like, "That looks really off to me." And then Claude's like, "You're right. I didn't actually use your data. I couldn't read it, so I just made this up." So build in verification, make sure it's not doing that for you. It looks like we've had ... Tiara's doing this. Yeah. I actually ended up moving toward code in that data analysis process. I didn't trust Claude enough and I needed verification. So I had it write code. I was doing some coding of some qualitative data, and I needed a verification process for that.
So we actually ended up moving away from just doing the pattern recognition to me saying, "Okay, Copilot, let's create an Excel spreadsheet where I can really see what you're doing and you can show me your work." I felt much more confident in that, but it's getting better all the time. So if it didn't work today, try it again. And one tool that some friends have shown me for data analysis, but I've not tried it myself is Julius.ai, but I've seen some really cool demos on Julius from some friends.
Multimedia. So we can absolutely generate all these things, right? I just showed you storybook. That was all generated by Gemini. The image of me looking really overwhelmed in our presentation was created with the caricature challenge for ChatGPT. But can I tell you something about that challenge? When I first did it, do you all remember this one? It was a few months ago and you were like, "Hey, based on all of our interactions, create a caricature of me." And so I did that and it created a white guy with a beard. I was a tech bro. I was like, okay. And ChatGPT and I have been through that before, so I typed in response OMFG and it was like, "I'm sorry, I know you're a woman. I'll try again."
I should find it. If you look at my Substack, I didn't have that one up, but if you follow my Substack, I wrote about it. I was like, "Really? Again? Again, I'm a man?" Okay, whatever. And also I mentioned early on that I really stand by this. I think the best AI art is created by actual artists. I see so much AI slop in the art space and it's exhausting. Anyone else? I'm exhausted by it. I'm exhausted by the writing. I'm exhausted by the art that's just so busy and weird. Yeah, yeah. And I will say the new ChatGPT, they updated it. That's why they were doing the caricatures, I think. It's better. It's definitely better, the image generator, but still not perfect.
So be careful. Remember they're not copyrightable, all that stuff. This is one area where I do worry a little bit about jobs, Gary. I think you can create some pretty photorealistic things, but keep in mind, garbage in, garbage out. And then we're getting close to time. I'm watching the time. I'm like, you all have so many cool things to share. This chat is amazing. So this is an example of AI-generated art. And the thing is, now, if I were to ask it right now, let's try it. I have chat up.
I would like a picture of a centaur and a cyborg shaking hands. Use a 1950s illustration style. Make it colorful and have a rainbow and mountains in the background. This is for a book cover. I bet I'm going to get ... This is what I mean. It's always fun. So I did that image back in 2024 for context. Oh, I love that, Amanda. Let's see how it does. I'm testing this in real time. And again, they've just had an update. Oh, and you may notice, by the way, real quick note on transparency, it can be really hard to know when things are created. Those of you who've interacted with ChatGPT, this is a Chrome extension I have that I've linked to. Someone was like, "Hey, I want timestamps for my ChatGPT." Whoa. Man. Huh. And it is a man. Of course. I didn't gender that centaur, but there we go.
Yeah, that's an upgrade over my book, I would say. I don't know. What do you guys think? Upgrade? Yeah, so this is 2024, and this is today. It's an upgrade. It definitely got the '50s book art. I agree, Amanda. I just think this is the funniest thing ever. Yeah, but while I was doing this, this was a really fun experiment, and I think you can still use images for this to really show how AI can get things wrong. We're not seeing six fingers on people's hands. We're not seeing a lot of the garbled words anymore, but we still see subtle differences where things will be just not quite right in an image.
But this one, this was so funny because I could not get it to do a centaur. Every centaur it did was a unicorn. And I kept saying, "Take the horn off." And it would be like, "I took the horn off." And I'd be like, "You didn't though. It's there." So these types of iterations I think can be really helpful with students. And when students do something like what I just did, it can also help them ... I know for real. It can also help them with prompt engineering. Well, I hate that term prompt engineering. With creating, it's really writing, right? So they need to know audience purpose and exigence when they're asking the AI for something. Who is my audience? What is my purpose? And what is the reason this content needs to exist?
That students have to get really clear on that in order to create a good prompt. So those of us who are writing instructors, I think we have an edge up, but no matter how ... So I just demonstrated it. Now the images are much better, right? So a couple of other ways we can be the human in the loop and we need to be the human in the loop. I mentioned AI slop. This is such a concern for me, you all. I really feel inundated by it. I feel like it's even kind of rude. Is anyone else feeling that way? Everywhere you look, you're seeing this stuff, it's everywhere.
So when you work with it, keep in mind that you are the ultimate arbiter of whether that content is good or not, right? You can refresh out of-date textbooks. I just did this so that we have Adam's iterative prompting process. I just did this for ... Where is this one? Sorry, you guys, I have so many things up. Let's see. Yes, for this book. So this is a book that we put out. Jason Blomquist and I were state fellows for AI, and we created this book, but obviously so many things had changed. So again, I use the Gemini agentic browser on this one to review each page and suggest updated content and check all the links for me.
So that's a really great, fast way to do things. But because it was Gemini, I was chatting back and forth with it. Yes, update that. No, don't update that. And that way I'm really staying in there. So as you're working with this, and especially ... Oops, sorry, not that. As you're thinking about how to work with AI, remember this bottom line, this is the most important thing. It's a tool, these large language models, but it's not an author. I still find them to be most useful for experts. And my biggest concern still is how do we train the next generation of experts, right?
I feel very confident in evaluating an AI output if it's art or if it's writing, just because of the years and years of training I've had in those two areas. I feel much less confident with coding. And that's another reason I'm glad I tried coding, by the way, because I can experience why a coder might think a piece of writing that's crap and AI slop is okay, right? That's not their subject matter expertise. So remember that subject matter experts are still important.
And then finally, thinking about using AI, it is important for you to check what your institutional policies are. So for example, Affordable Learning Georgia and Cleveland State University have provided specific guidelines, but Open Oregon has a hard no. And I think Marco is on here. He gave a keynote for Open Oregon, I think that really addressed the environmental impacts of AI, which I highly recommend to everyone. I think Marco, it's linked in that ... It's probably linked in this AI resource guide. I think I put it there, and if I didn't, I'll get it in there.
So I wanted for our live demo, do you have questions or things that you want to see? I've already showed you a few things, but is there something like maybe ChaucerBot is one that I had thought we could look at. Let me show you a couple things really quick, and then if you're thinking of questions, put it in the chat. I mentioned writing a personal writing assistant. I created this and trained it on my own writing. This is not a publicly available bot. Boy, wouldn't that be fun to have my students get at it? So it's just private for me, but I use this to "Create a reply to this email..." a lot, and I use "Write a blog post ..." a lot.
I'll share my Substack out. I meant to have that so you guys would have it. It's really fun for me. We had questions about where AI is used and how we acknowledge it. On this blog, I use AI a lot. I use that bot to write a lot. And this was a fun experiment in the humanities. I've been working with poetry since 2023 with ChatGPT. I've run some kind of poetry experiment for National Poetry Month every single year. And this year I gave it a handwritten sonnet. I wrote this by hand, and so I've linked to the previous years if you're ever curious, but I had to analyze the sonnet and tell me whether it was any good. And I was like, "Yeah, it's okay. Competent, sincere poem." Oh, that's so sweet. Thanks. That's Claude.
And Gemini is like, "Charming, technically perfect." Now note, ChatGPT said that the meter was wrong. And I was like, "You're wrong, ChatGPT. My meter is excellent." I'm the queen of iambic pentameter, so Gemini and Claude know this. Then I was like, "All right, you do it. You write on the same theme." And these are the poems they wrote. They're getting really good at poems, you all. I think I like Gemini the best. And then I was like, "Hey, did I write it?" I originally was tricking them. I'm like, "Hey." And Gemini knew I did because I've let Gemini know everything about me. And I was like, "Oh, you wrote that for Andrew. That's so sweet." I was like, "Yep, sure did."
And then what does it mean for the future of human writing? This was an interesting conversation. So these are always fun for me. You can look through this, but what I wanted to show you ... And there's some takeaways for instructors. AI acknowledgement: "In this poet, I wrote everything myself, including the original sonnet that is not quoted from an LLM," so I have directly quoted and made that clear where I'm quoting from LLMs, "but I also included a lot of quotes because I wanted to show you how LMs are writing and thinking about poetry in 2026." So that is an AI acknowledgement statement. It's on every single one of my posts.
Sometimes people get angry at me. I had a colleague who came in and was like, "You tricked me. I thought that was you the whole time and it was a hundred percent lies along persona bot written." And I was like, "Good, that means my bot is working as planned." It's not producing AI slop. It feels human. It feels like I wrote it. That's right. BeowulfBot is here. This one's kind of fun, right? So look, it's at strong, weak verbs. I can share a passage and it will not translate for me, but it will check my translations. I have in the instructions don't translate for me. And that's another important thing I think to teach your students, right?
Oh, and there's those horrible adjectives. Yes, let's do that. And then if I want to, I can put in ... I have it trained on my particular textbook so that it's working with that. ChaucerBot, super fun. This one, again, it tutors in the same way. TLDR, you saw that demoed. So I use TLDR in my Substack just at the top.
Amanda: Hey, Liza.
Liza:Yeah.
Amanda: We are very close to time.
Liza: Oh, you're right. We are.
Amanda: One minute.
Liza: And that's it. Thank you. See, I'm so bad. Well, I thought we're just down to questions. So yeah. And you guys, if you have more ... So yeah, that's what I wanted to show you. Lots of fun stuff. And then if you have more questions, obviously I could talk about this all day long, but I think they're fine. Oh, and that's the other thing I wanted to show you real quick. You can build those same bots over as Gems in Gemini. So those little custom GPTs, this is the Latin practice one I created for myself, so.
Amanda: And for everybody else, I have thrown our survey in the chat. We would love if you'd give us feedback on this session. And thank you so much to Liza.
Liza: Thank you so much for having me. Thank you all for being here.
END OF VIDEO
Join Liza Long of the Idaho State Board of Education for an engaging conversation on artificial intelligence and interactive content in open publishing. Hosted by Amanda Larson of The Ohio State University, our most recent session of Pub101 for Authors helped participants explore opportunities to incorporate GenAI, consider pedagogical objectives when including AI and interactives, gain key insights on how to cite, integrate, and track AI content, and much more.
Watch the video recording of this May 6, 2026, session or keep reading for a full transcript.
Audio Transcript
Speaker:
- Amanda Larson (Affordable Educational Resources Initiative Program Coordinator, The Ohio State University)
- Liza Long (Director of Digital Learning and AI, Idaho State Board of Education)
Amanda: Hey everybody. Welcome to Open Education Network's Pub101 for Authors. Thank you for joining our session today. I'm Amanda Larson. You all probably know me by now, but I am the AERI program coordinator at the Ohio State University and I am the Pub101 chair and I'm going to be the host and facilitator today. Soon I'm going to hand it off to Liza and she's going to talk to us about AI, which is super exciting, AI and OER. And in addition, there might be many of you who have experience in this topic. In addition to our guests, and we invite you to share your experiences and resources in the chat as well.
A few housekeeping things. As always, today's session will complement the Pub101 for Authors curriculum and related resources, specifically unit four on AI. And we have an orientation document that I should have the link to here for you. I'll paste that in the chat. Slides, session recordings and curriculum connections will be included in that document. And we are recording today's session. We already pressed that button, so we're good to go. We are committed to providing a friendly and welcoming environment for everyone that's aligned with our community norms. So please join us in helping create that constructive space.
To kick off today's conversation, we have a brief reflective question that I'm going to ask you. What excites and/or worries you about using AI to create OER? I'm going to also put that question in the chat. That AI help create Skynet and/or end my profession. Yes. Hilarious. But true and a concern. My peers' thoughts about using AI and their evaluation for my work because I used AI. I think that is definitely a super great concern. Got another one. I'm worried about crediting the use of AI in my work. I use it to edit, not create, but I am worried about people not taking my work as seriously. Yes.
We're going to talk about all kinds of things, but one of the things that we talk about in the curriculum is about being transparent. And so you could say, "Hey, I used AI to do this editing, but all of the text is my own." Appearing, seeming, or even being hypercritical after telling my students they should never use it to create their own work. The environmental costs of widespread generative AI use. All of these are great things to be concerned about and interested in as we think about how we might use AI, and that's further complicated by copyright. I'm going to hand it over now to Liza to talk to us about those things. Take it away.
Liza: Oh, welcome everyone. And I want to thank the Open Education Network and Amanda and Karen for inviting me to present on a topic that I am pretty passionate about. And I want to thank all of you for being here. This was a great way for us to kick things off to see where we are, and I'm not surprised to see that we're all over the map. So I'm going to go ahead and share my screen. We're going to walk through some of the basics that you learned about maybe through the curriculum today. I've got an overview of that, and then I want to demo a few things, and I'm hoping also, I'm assuming you may have some questions. So as those questions arise, you could put them in the chat. I'm hoping maybe we'll have some time at the end. Are you all seeing the Open Education Network? Okay. Yeah.
Amanda: Yes.
Liza: Before we even get started, I just want to introduce myself really briefly. I'm Liza Long. I am the director of digital learning and AI at the Idaho State Board of Education, and I work with our eight institutions of higher education on issues related to artificial intelligence and other emerging technologies. So this is an area of interest to me. I am also a PhD student in Idaho State University's English and Teaching of English program. And I am an instructor at the College of Western Idaho, an open access community college where I teach asynchronous online courses. So I have a, I guess, somewhat unique perspective on this particular topic.
We're going to be looking at how artificial intelligence can be used to create content and also interactive content, because for me, that's some of the most exciting possibilities that we have. But as we kick things off, before we get started, I want to clarify that when I use the term artificial intelligence, what I'm actually talking about are large language models. So we hear AI thrown out a lot, got Skynet, got all those good things you all are worried about. I mean, it's out there, but what I'm specifically going to be talking about is large language models. I want to clarify that and how they can be used with open education resources.
So what are large language models? Well, most of us know ChatGPT by now. Claude is my own personal favorite. Gemini from Anthropic. There are also other versions such as Qwen, DeepSeek. Meta has open LLMs that you can download and run yourself. And there's certainly many others. I saw, if you haven't seen it yet, Talkie is a large language model that its training data is cut off at 1930 because it's trained only on things that are in the public domain. And as you might expect, it has some interesting hot takes about life. So when we talk about large language models, we're talking about that specific subset of AI.
Also, before we get too deep into it, I like to start my presentations on AI with an acknowledgement statement that I originally found in January of 2023 from Donna Lanclos and Lawrie Phipps. They're amazing open education resource folks to follow. I follow them on BlueSky and they gave this offering to the open education community very early on in the introduction of these large language models to our lives. So I do want to acknowledge that this presentation was prepared using Google Gemini's agentic browser and ChatGPT's image generator. And I acknowledge that AI chatbots do not respect the individual rights of authors and artists and ignore concerns over copyright and intellectual property in the training of the system.
Additionally, I acknowledge that the system was trained in part through the exploitation of precarious workers in the global south. In this work, I specifically used AI tools to generate a draft of the slide deck from previous presentations and Open Education Network content and to demonstrate OER creation techniques. So when I'm teaching with AI, I always like to start with an acknowledgement statement and sometimes a conversation, often a conversation with my students so that we can do exactly what you all did in your reflection in the chat. We can get some of those concerns out there and we can acknowledge, because it's important to acknowledge that these concerns are real.
So now someone once asked me, "Well, but you still use it." Yeah. Yeah, I do. And I fly on airplanes and I'm really trying to wean myself from hamburgers, but I occasionally have a hamburger or a steak. So we all have to make choices, but I think it's important, especially with this technology to keep these results and these things in mind. Okay. So now that we have that out of the way, I want to look at my learning objectives for today. I'm hoping that we can explore some ways and opportunities that we can use AI to play with experiential learning.
We can identify ways to incorporate GenAI into the textbook creation process. I'll demonstrate some of my earlier attempts and then where I'm moving now. We'll consider pedagogical objectives when including AI and interactive materials. So as always, your pedagogical goals, your learning outcomes should drive your choice of whether to use this technology or not. And then finally, and this came up in the chat, we will learn how to cite, integrate, and track AI content. So those are my goals for today. Hopefully we'll get all of that done.
So when we think about AI, when I talk to professors, the concern that I hear most is that students will use AI to cheat, and I continue to hear that most in 2026. I highly recommend the book, The Opposite of Cheating by Bertram Gallant and Rettinger, if you haven't read it yet. They point out that students who use AI to cheat are just like students who have always cheated. They're either not invested in the assignments that we're giving them or they are afraid of failure. Those are the top two reasons that students cheat, and that has not changed in the age of AI.
So as we think about that, what I'm hoping we can do today is maybe shift our mindset to some potential benefits rather than focusing on, "Well, my students are going to use this to cheat." Maybe instead think, "How can I use this to help my students become a more effective learner even through some personalized learning?" So we can use AI in OER creation in a few different ways. And the first way is efficiency. We can rapidly generate text, images, and it's great with coding assistance. And I'm curious, even maybe you can put in the chat if you've vibe coded something or if you have a link to something you've vibe coded, it would be fun to see. Now, if you haven't, that's fine.
But I was at a conference at Cleveland State a few weeks ago, and there was a wonderful presentation on someone sharing using Google Colab to vibe code. And at my table, everyone seemed to shut down when that code popped up on the screen. And I was like, "Oh no, guys, you can do this. You can do this." I pulled up my own Colab notebook and they were like, "Wow, you're an English professor." I'm like, "Yes, the future is now." So efficiency, if you haven't tried coding something, go try it. Even Claude makes it really easy to create a fun artifact.
I initially came to AI through ... I was curious, it was a research question. So I was teaching an English 211 class and I needed to create an open education resource for that course because my very expensive textbook was going out of print and also any replacements, we're talking 80, $90 range. So I'm very committed to open education, wanted to create that textbook, had a semester to do it. And I thought, there's just no way I can actually do this. So this is January of 2023. I'd been playing around with ChatGPT since it was released in 2022, and I said, "All right, let's take this puppy for a spin. Let's see what it can do."
And I want to show you really quickly, because this was my first attempt with AI to do that first step, that efficiency step. So this is the book, and I would say you can look at the book or not, but AI was used throughout. And I have a philosophy statement. This is baby me just starting to work on this textbook. And I noted that ChatGPT came out. I was already working on the textbook and I decided to use it. And you can see that little acknowledgement statement. I was really glad that I found that early on. But my favorite part, what I would recommend you look at is the links to ChatGPT chats used in writing this book, along with my prompts.
And what I love about this experience, if you're new to AI and you want a training course, don't take a training course. Pick a really large open education project like this and just dive in. I was so bad at prompting. It's just really fun for me to look back. You'll notice prompt number 10, provide a list of 10 peer-reviewed articles, prompt number 11, these are not real articles, prompt number 12, these are also not real articles. So really fun way to get introduced to what AI could do and what it couldn't do. And when we talk about transparency and citing, as I've been working with my students, I always point to this and I say, "Look, you don't have to be good at this at first. You can learn by doing."
But then the other way that I used it for my students is by giving them, showing them, "Hey, guess what? I know that AI can write your essays for you. So here, I had AI do it and here's what AI wrote. This sounds pretty good." And then I critiqued it and we kind of flipped this particular classroom where they would do the same thing. They would try to write the very best essay they could, and then they would critique it. Now I'm going to tell you, this particular assignment is getting harder and harder to do because AI is getting better and better. Those of you who've used it to write an essay, but I still don't think it's impossible.
But as I'm teaching now, I've shifted more to like, "Hey, go find your voice. Find places where this voice feels flat or where you're doing a rhetorical analysis, a close reading of AI text." Why does it always say it's not this but that or those types of things. So that's just to show you how it could be used in creating an open education resource. So I created this whole book and what I realized through that process as I worked on it is, oh, these tools are absolutely most useful for experts, right? So we want to be careful as we think about where to introduce AI and where not to introduce AI.
The next place that we'll look at for benefits is AI as a creative partner. And I was just in a conversation yesterday where someone said, "I'm not sure that it is a great idea to use chatbots for brainstorming." I kind of agree. It feels like early on we thought, oh, using chatbots for brainstorming is a great idea. What I prefer to do is have students do some preliminary brainstorming first, or for me, if I'm working with a chatbot, I want to have a very clear vision of what I want before I go into any kind of session with a large language model. So absolutely effective as a thought partner though.
Rather than brainstorming, I like to think of them as thought partners, and I'm the one who starts with my own thoughts, right? But then they can test them and we can have conversations about them and test those out. Also, you can do rapid prototyping. Let's say you want to develop a new course or a new assignment. It's really easy to prototype a new assignment using these large language models.
One area that I'm very fascinated by, and I've seen others do some really great work with this, I just worked with a social work team through Rebus who was putting together a case study book, and they were able to translate and localize content within that case study book very quickly by using AI tools. So if you haven't looked into that, that's another great use. And then finally, personalization, if you have not heard of NotebookLM yet, that is a really fun way to get students to take ownership of the course and to bring that personalization in. So lots of fun things here, right? These are all ways that we can really improve learning.
Keep in mind, one thing that is very top of mind for me, especially my current role, the paid models are all much better, and in some cases, the only ways to do some of these things. So when I think of personalization, for example, I'm putting on my student hat, I'm just wrapping up an old English course and I created a custom GPT BeowulfBot. I'll show it to you in a minute. The BeowulfBot helps to tutor me in old English forms and test me. So it's a personalized learning bot. It also tracks my particular errors, keeps an error log for me so that I can look for trends. I know that I'm really struggling with the difference between weak and strong adjective forms because BeowulfBot tracks that, and then I can do more practice in the areas where I'm struggling.
I think there's a metacognitive benefit to having students create these types of personalized bots themselves rather than you creating them. However, that requires a paid account. And the paid accounts are all $20 a month right now for those premium models. And I'm here to tell you they're heavily subsidized. That's not really covering the cost of those tools. So for some students, out of reach. We've always been concerned in the open education community about digital equity. Let's keep our eye on the ball here. This is a huge concern for me that we'll see an increasing digital divide between students who have access to these premium models and students who do not and have to use the free models.
So reality check, that leads us right into it. I already demonstrated ... Oh, I'll share. Melissa, it is amazing. I have ChaucerBot too. I started with ChaucerBot for middle English, but I started already demonstrating some of this through my critical worlds book because I learned this by doing. And if you can find a way for your students to learn by doing, it can be really effective for them. So maybe having them pick a topic that they're really expert in and asking the AI questions, ask it to write a paper about that topic, some obscure lore or anime or something really obscure that they're just really into. That helps them to see AI hallucinations. I mentioned the rates of those are down, but they're never going to be zero with this particular type of technology.
And then we also have to watch for misinformation. I didn't list algorithmic bias as one of my huge concerns, but it is actually my biggest concern with AI tools right now, large language models. That's where it perpetuates some of the biases that are built in the training dataset. I mentioned the one that cuts off at 1930, Talkie. Boy, if you want to see some examples of bias baked into the training dataset, be aware that if you interact with Talkie, you are going to get some views that might seem like really, really out of place in 2026. I'll just kind of leave it at that. But that's a good way, again, maybe to even demonstrate to students that idea of misinformation, algorithmic bias.
Integrity, academic integrity I talked about upfront. I think it's important for us to have those conversations. And I also think it's important for us to know where we are positioned in those conversations. So I will position myself, just so you all are aware. I really subscribe to the ideas of Sarah Elaine Eaton, and I'm kind of post-plagiarist in how I view text. That is a result of my work in the open education community. That is not an inevitable result of working with OER. One of my mentors and favorite people in the world, Jonathan Poritz, is very diametrically opposed to my position on this with respect to AI.
So did AI train on my content? It did. I have a published book that it scraped without my consent. Yeah, that's fair, Amanda. He's very staunch. We've had a lot of good, productive conversations about it. But so my book was scraped. I'm a member of the class that is for getting a settlement from Anthropic. I don't really care. I mean, it would've been nice if they'd asked, but I'm like, yeah, words belong to everyone. Have at it. I realize I'm coming to that position from a place of privilege though. So these are legitimate conversations, people who have concerns over these, I think it's important to acknowledge them.
I absolutely do not think these tools are going to take out our creative class. I think creative writers are still going to shine, maybe even more so. I think when I look at AI art, the only AI art I can personally stand is created by artists who are actually artists and using AI as a medium. The rest of it just look like slop to me. So I know that these are big concerns because I hear them from my own writer and artist's friends. And some of you hinted at this in your reaction to the first question where you talked about how you feel like you might be judged if you used AI, right? I've certainly felt that in my own work.
So I'm a published author. I have a large corpus of work prior to AI. Most of the writing I do anymore is hybrid. That's just become my new method. I'm very confident as a writer. I can definitely write by myself by hand if necessary, but I like to use AI. It's fun. It's more fun for me to write. It's like talking to someone and writing at the same time. Probably talking to myself, if we're being honest. Oh, good. I'm glad Tiara agrees. It's fun. Yeah, it's fun.
The concern that also popped up in some of your responses, and it's really real, and I want to give you some tips actually today for how to address that if you do use AI tools, because it's a concern of mine. The environmental impact. I do not personally ... I think we will see job displacement for sure, but I actually think we're going to see new cool jobs. Ezra Klein has an op-ed about that this week in the New York Times. Gary, were you the one who shared that in our meeting? But yeah, I thought you did. In our Monday meeting, Gary shared this op-ed from Ezra Klein and said, "Liza, this is how you feel about it." And I do. I feel like it's not inevitable that we'll lose jobs. I think the smart companies instead are figuring out how to upscale their existing workers.
We could see an explosion of productivity, right? So in the past, we've seen technologies impact work, but overall it's been a benefit generally to workers. I think we should really be advocating though for mental health. So I published a blog post, I blog at Artisanal Intelligence, and I'm a heavy user of AI tools. I'm absolutely noticing brain changes. I'm also a mental health advocate. I'm on the International Bipolar Foundation Board, and this is a huge concern to me. It's an untested, untried tool, just like social media was, but even more immersive, I think. And people are talking about this new kind of burnout. I've personally experienced it, this kind of AI burnout. So for me, it's sustainability for my own work life. Not necessarily will I lose my job, but am I going to be expected to do an unsustainable amount of work in my job?
For the environmental impact, what we'll talk about today, because it's real, is token use. So you're going to want to start thinking about strategies to optimize and reduce how many tokens you're using. The tokens are what is using up that compute. The compute is what is what we need those data centers for and all that water. So as we reduce our impact in how we interact with these tools, we can personally make a difference. Now, if you're thinking that's like putting your plastic water bottles in the recycling bin, you may not be wrong, but I still feel better myself with trying to do what I can to manage my environmental footprint with AI.
And there's some pretty easy ways you can do it. It can also save you some money. For those of you who've hit cloud rate limits, me five times a few weeks ago on one day, learning to optimize your token use can be good for the environment and it can be good for your bank account. So we'll look at that.
Finally, equity. I have, again, already talked about some of these things. Gary put in the chat, Darren Olson, Darren's been doing this since day one. I think Darren's wrong. So Darren is not being creative in thinking about these tools, but certainly we as educators need to be thinking about job impact. We need to take it very seriously. In fact, I will try to find it. I didn't have it up. I've got a bunch of resources up I wanted to share with you all, but we had an industry panel at our Innovate Idaho conference a few weeks ago that was fascinating to me. They kind of agreed with us recline. It's not that we're going to see job loss, but they are expecting students to come into the workforce AI ready.
Entry level jobs are in fact disappearing. They're expecting students to be able to upskill themselves to level two or three right from the start. So that changes how we teach. But that final piece ties right into that, to equity. Students who've had access to working with premium tools are going to be in a different place than students who have not had that access. And simultaneously, different global regions have different levels of access. I am fascinated though to see the innovation, especially that's coming out of the Chinese models. Some of you may follow the geopolitics of this whole situation.
And from a geopolitical standpoint, the United States had tried very hard to keep our best chips and technology out of China's hands. China worked around that and found a different way of improving their large language models so that we see Qwen and DeepSeek actually really high performing models with research capabilities. So I think that's something to keep in mind. I'm loving this chat, by the way. We need to help people create meaning and purpose around AI. And I'm also, Tiara, I'm really glad that you're saying that you have mixed opinions. I am the same. I call myself a skeptical enthusiast or an enthusiastic skeptic, really kind of depending on the day, right? And yes, Maha has talked about borrowing access. Yes. So that's a great example of that equity.
I also want to talk about the legal landscape, which is unsettled, but we are starting to see some case law being made around this. And back to that open education resource I shared. So Creative Commons is working on this as well, as you're probably aware. And because Creative Commons functions within existing copyright law, I have to put the huge caveat here that I am not an attorney and any legal question should go to a copyright attorney, but I have taken the Creative Commons course. Jonathan was my certificate course mentor, and it's something I follow with interest and certainly around AI. So in the United States, the cases that we're seeing require human authorship. So if you completely generate a text with artificial intelligence and try to copyright it, that's a no-go right now.
What I'm more interested in is hybrid writing, because again, I'm a hybrid writer, so how much is AI? How much is me? We should probably all be paying attention to that, back to how we audit our use of AI, right? And there's some strategies you can use such as using projects to help see where you are in the mix. Ethan Mollick's Co-Intelligence book, I think from 2024, continues to hold up very well around this idea of always being the human in the loop. And I think it's important for us as authors and also as educators to make sure that we're paying really close attention to how to model that for our students.
In Canada, likely in the public domain, but again, this is a rapidly emerging legal area. And hybrid works right now, you can register those for copyright, and that's even if it has AI-generated images, human parts are protected, AI parts are not. Again, how do you tell? For some of us who have really developed a workflow like me, I have a bot that's trained to write like me. How much of that is me? How much of that is the bot? Also, I was a generation one mommy blogger, so I could argue that the large language model's trained on my content, so it's basically me anyway, right? I'm kind of kidding, but you get where I'm going with that. So we want to pay attention to that with any work we're creating with AI.
Then I want to talk ... I'm going to show you specifically this first one, accessibility. I'm presenting on this at OpenCon next week. This has been an area I'm very, very interested in, and I've specifically been exploring agentic browsers with this. So I like to build things while just in public, very publicly. This is very new. It's not finished. It's a work in progress, but what I've shared here so far works and it works great. I'm continuing to work on refining the amount of tokens it takes. This is one way I can sometimes really hit my token limits, but I think this can be a good introduction.
Sometimes when somebody ... I've realized, sometimes when someone like me talks, some of you are like, "Yeah, agentic browsers, I use that. " Some of you are like, "What are you talking about?" So an agentic browser in this case is just me turning on Claude. Claude is an agent now that can act on my behalf on these websites, right? So you'll notice my Claude agent says high risk right now. If I'm working with Pressbooks and I've signed in and given Claude permission to mess with Pressbooks, I don't want to be stopped at every step of the way. So you can make this lower risk when you click ask before acting. And if you do that, Claude will ask you every step of the way if it's allowed to do something.
For the workflow that I have developed here, I use act without asking, and that's where that high risk thing comes in. This book I think is kind of helpful as well though. So I did go ahead and put these core AI concepts. And something super fun that I just added, but I haven't added it to the workflow. Those of you who work in Pressbooks, I created a agentic browser workflow to provide glossary definitions. So the glossary tool in Pressbooks, it was able to just ... I didn't do this manually. This was me telling Claude what to do and that agentic browser over here went in and did this for me. Isn't that fun? So just kind of a good introduction.
Maha, again, came up in the chat. Maha pointed out to me that she likes to include a command when she's working with stuff just to tell it to make sure it's AREA, that it's following AREA guidelines. Yeah, Amanda, I know it was so fast, right? So I liked that. So there's a guide to working with chatbots in here as well if you're new to it. Some stuff that you can do, what the process looks like. I need to actually update. I just mentioned that. I'm going to go in. This is what I love about prospects. I'm going to go in and add Maha's advice to add an AREA tag when you work with Claude so that it's following those principles.
So the book just works through the accessibility stuff. These are the different prompts. And I start with a huge warning that I want to make sure you're all aware of if you decide to use these agentic browsers. Another one, by the way, if you have Google One, this is also an agentic browser. You guys, this is how I created your slide deck today. I had the Open Education Network's that you're Publishing 101 textbook open. I had several of my former content presentations open, and I told Gemini just to look at all that stuff and start creating the slide deck. Super helpful.
All right. I've still edited it and reviewed it, but just getting that draft in was so helpful. So that's what an agentic browser looks like. I'm going to delete that group for now so that we don't have it open. But let's see, it's back to our book and back to that warning. Whenever you're working with ... Oh, shoot. Now I lost my accessibility when I deleted that. Sorry guys. Pay no attention to the man behind the curtain while I just go back to where we were.
Okay. So the warning I want to share, just to make sure if you are new to agentic browsers, leave that little tag on that I showed you where you have to check with each one and be aware that you need to limit access. So don't be giving it full permission to do everything in your life, right? Stay the human in the loop. I do highly recommend most of the agenic browsers work that I do, I actually do on a separate browser tab. That's why I didn't start working with agentic browsers in front of you. I'll do it on a completely separate browser tab that's walled off from other things, or I'll do it on a laptop that I have that's running Linux so that it's walled off. I don't like to use them on my work computer, which is what I'm on right now.
So just again, because of the risk, you'll want to pay attention. And then be aware that the alt-image task particularly is token intensive. So one of the things I'm experimenting with now is trying to create more of a local language model that will do that specific task, but that is the one piece that doesn't really work well for a local language model. It really needs that Claude's reasoning in order to create robust alt text. So yeah, that's the workflow. And then also in this book I'm sharing ... Yes, yes, you can, Tiara, I know. I love using the Gemini one. But Gemini, I can't get Gemini to go in and edit my Pressbook for me though.
But another thing to be aware of too, this was so funny to me the first time I did it. I always recommend that you download an XML copy of your book before you ... So it's as is before you let any agentic browser start working on it. The very first time I tried doing this, it took me about two weeks to figure these workflows out. The first time, it went in and started rewriting my book. So I did not want it to do that. And I realized after that, oh, I need to break this up into discrete tasks. I need to make it work in the HTML code view, not in the visual editor. After I did that and put in a lot of prohibitions about don't change my content, focus very specifically on these HTML tags, then I was able to get it to work quite well.
And like those of you who use the Ally Checker, you can see I do. This Ally Checker, I went from having 19, 27, whatever, tons of alerts to zero. Yes, Amanda, you could. I haven't written the table remediation one yet, but you absolutely could. Yeah. So any of these workflows, it can do. I started with heading levels because that was pretty easy to figure out how to do. Then I did the image attribution. I've written one for TASL attribution. And then here's the glossary terms that I just showed you. So really any type of work that you want to do, those agentic browsers can do within Pressbooks, and it's really fun.
Yes, you could do that in the code. You absolutely could. So the one other thing, this is what I mentioned Colab. Oh, no, so the VPAT, the agentic browser will do a really nice VPAT analysis of a website. This was really helpful for us when we were trying to get ready for the Title II deadline that was a few weeks ago that then they bumped. But this one, it did a really good job of identifying the issues in our state board website. So we were able to go in pretty quickly and get those fixed.
And then the Colab one, this is something that I'm running not through an Agentic browser. So this is a little bit different. And it was my first Colab project. It scared me to death. It was great. Not only that, Colab was just really, really fun for me because I like to learn by doing. So this is, again, a way to think about how you might incorporate these tools as interactive tools. Having students go in and vibe code something and then using a tool like Colab, Gemini will help you. But also what I liked was I could say, "Hey, Gemini, why did you do that? Explain what that piece of code is doing." So I really enjoyed that process. That's how I like to learn, and it was kind of fun.
So this one ... Oh, there's no accessible text. I'll have to get that fixed. But yeah, so I think this book might be useful for you if you are thinking about accessibility issues, maybe some fun things to think about. And then some of these tools were actually shared in your textbook, but I don't know as much about them. So I'm not going to demo this one, but I wanted to include it because it is included in your textbook. And then the structural outline, I use this for critical worlds. So it was really, really helpful with the structural outline. I strongly recommend using it for that type of work.
And also really kind of fun. Again, back to Tiara had mentioned, she thinks it's fun to work with them. I like talking back and forth when I'm doing structural outlining, and it just feels like I have a really smart colleague that I can work with on that. So chapter structures, subtopics, really fantastic use of time. Or if you have an existing one, having it review it, look for flow, look for scaffolding, make sure you give it the role of an instructional designer or a curriculum developer, and it will do a great job of helping you with that. Oh, so this is the script because I don't have it. This is the Magic ToDo one that's mentioned in your resource, but I don't personally use this one, but I wanted to make sure you knew what it was.
Text generation. So how many of you have created ... Okay, great. Amanda put, she's the Goblin tools expert. I thought it was super cool, Amanda. I just haven't had a chance to test it out myself. And in full disclosure, I mostly work with the frontier models, but there's lots of cool tools out there, different tools that you can use. So don't feel like you're limited to those frontier models, and now we know who to contact if we have questions. So for content creation and multimedia, if you have not written a custom GPT that writes like you, do it. It's just such a time saver.
I use it a lot for drafting correspondence. So yes, Gary, you have a lot of cool content experience. Actually, Gary just presented on this at Innovate Idaho. Gary took my English 199 writing with AI experimental course and created a 30-page graphic novel, which ... Yeah. Did you know, Gary, now that Gemini has a storybook feature? Have you seen that one? Oh, you'll laugh because I wonder, I have Gemini up. Let me show you really quick. I think this would be a really fun thing too. Gary will laugh about my topic. So where is this? This is the storybook one.
So it will create a picture book for you. I've only done one. I was in Bali when this tool came out and I was AI free, but I was like, "Hey, this might be fun. I'm going to test it out." It's so wrong, but that's supposed to be me. So I gave it a picture. That's me and Bali. There's my guide. But I learned really quickly. So terima kasih. It's true that Bali is an Indonesian country, but they do not say terima kasih. That's how you say thank you in Indonesian. They say suksma. So I kept saying terima kasih and people were like, "What? Suksma." I was like "Mewali." So yeah, so there's things that are wrong, but I think that would be a fun thing.
I thought of you, Gary, when this tool came out. It might be fun for you all to play with for content creation and even creating some kind of assignment around that. I haven't done that yet, but I think it would be fun. Data and code. So I have used it for analyzing patterns in large datasets, and it's really great for that. Has anyone else used it for that? If you haven't, just be aware, know your data, and be careful always with personally identifiable information, right? So the first time I tried this with Claude, Claude couldn't read the data that I gave it, but it didn't want to admit that it couldn't read it. And so it made up a complete data analysis. I wish I were kidding.
And I was looking at it and I was like, "Those gender numbers look really off to me." You've had Copilot fail too. Yeah, exactly. I was like, "That looks really off to me." And then Claude's like, "You're right. I didn't actually use your data. I couldn't read it, so I just made this up." So build in verification, make sure it's not doing that for you. It looks like we've had ... Tiara's doing this. Yeah. I actually ended up moving toward code in that data analysis process. I didn't trust Claude enough and I needed verification. So I had it write code. I was doing some coding of some qualitative data, and I needed a verification process for that.
So we actually ended up moving away from just doing the pattern recognition to me saying, "Okay, Copilot, let's create an Excel spreadsheet where I can really see what you're doing and you can show me your work." I felt much more confident in that, but it's getting better all the time. So if it didn't work today, try it again. And one tool that some friends have shown me for data analysis, but I've not tried it myself is Julius.ai, but I've seen some really cool demos on Julius from some friends.
Multimedia. So we can absolutely generate all these things, right? I just showed you storybook. That was all generated by Gemini. The image of me looking really overwhelmed in our presentation was created with the caricature challenge for ChatGPT. But can I tell you something about that challenge? When I first did it, do you all remember this one? It was a few months ago and you were like, "Hey, based on all of our interactions, create a caricature of me." And so I did that and it created a white guy with a beard. I was a tech bro. I was like, okay. And ChatGPT and I have been through that before, so I typed in response OMFG and it was like, "I'm sorry, I know you're a woman. I'll try again."
I should find it. If you look at my Substack, I didn't have that one up, but if you follow my Substack, I wrote about it. I was like, "Really? Again? Again, I'm a man?" Okay, whatever. And also I mentioned early on that I really stand by this. I think the best AI art is created by actual artists. I see so much AI slop in the art space and it's exhausting. Anyone else? I'm exhausted by it. I'm exhausted by the writing. I'm exhausted by the art that's just so busy and weird. Yeah, yeah. And I will say the new ChatGPT, they updated it. That's why they were doing the caricatures, I think. It's better. It's definitely better, the image generator, but still not perfect.
So be careful. Remember they're not copyrightable, all that stuff. This is one area where I do worry a little bit about jobs, Gary. I think you can create some pretty photorealistic things, but keep in mind, garbage in, garbage out. And then we're getting close to time. I'm watching the time. I'm like, you all have so many cool things to share. This chat is amazing. So this is an example of AI-generated art. And the thing is, now, if I were to ask it right now, let's try it. I have chat up.
I would like a picture of a centaur and a cyborg shaking hands. Use a 1950s illustration style. Make it colorful and have a rainbow and mountains in the background. This is for a book cover. I bet I'm going to get ... This is what I mean. It's always fun. So I did that image back in 2024 for context. Oh, I love that, Amanda. Let's see how it does. I'm testing this in real time. And again, they've just had an update. Oh, and you may notice, by the way, real quick note on transparency, it can be really hard to know when things are created. Those of you who've interacted with ChatGPT, this is a Chrome extension I have that I've linked to. Someone was like, "Hey, I want timestamps for my ChatGPT." Whoa. Man. Huh. And it is a man. Of course. I didn't gender that centaur, but there we go.
Yeah, that's an upgrade over my book, I would say. I don't know. What do you guys think? Upgrade? Yeah, so this is 2024, and this is today. It's an upgrade. It definitely got the '50s book art. I agree, Amanda. I just think this is the funniest thing ever. Yeah, but while I was doing this, this was a really fun experiment, and I think you can still use images for this to really show how AI can get things wrong. We're not seeing six fingers on people's hands. We're not seeing a lot of the garbled words anymore, but we still see subtle differences where things will be just not quite right in an image.
But this one, this was so funny because I could not get it to do a centaur. Every centaur it did was a unicorn. And I kept saying, "Take the horn off." And it would be like, "I took the horn off." And I'd be like, "You didn't though. It's there." So these types of iterations I think can be really helpful with students. And when students do something like what I just did, it can also help them ... I know for real. It can also help them with prompt engineering. Well, I hate that term prompt engineering. With creating, it's really writing, right? So they need to know audience purpose and exigence when they're asking the AI for something. Who is my audience? What is my purpose? And what is the reason this content needs to exist?
That students have to get really clear on that in order to create a good prompt. So those of us who are writing instructors, I think we have an edge up, but no matter how ... So I just demonstrated it. Now the images are much better, right? So a couple of other ways we can be the human in the loop and we need to be the human in the loop. I mentioned AI slop. This is such a concern for me, you all. I really feel inundated by it. I feel like it's even kind of rude. Is anyone else feeling that way? Everywhere you look, you're seeing this stuff, it's everywhere.
So when you work with it, keep in mind that you are the ultimate arbiter of whether that content is good or not, right? You can refresh out of-date textbooks. I just did this so that we have Adam's iterative prompting process. I just did this for ... Where is this one? Sorry, you guys, I have so many things up. Let's see. Yes, for this book. So this is a book that we put out. Jason Blomquist and I were state fellows for AI, and we created this book, but obviously so many things had changed. So again, I use the Gemini agentic browser on this one to review each page and suggest updated content and check all the links for me.
So that's a really great, fast way to do things. But because it was Gemini, I was chatting back and forth with it. Yes, update that. No, don't update that. And that way I'm really staying in there. So as you're working with this, and especially ... Oops, sorry, not that. As you're thinking about how to work with AI, remember this bottom line, this is the most important thing. It's a tool, these large language models, but it's not an author. I still find them to be most useful for experts. And my biggest concern still is how do we train the next generation of experts, right?
I feel very confident in evaluating an AI output if it's art or if it's writing, just because of the years and years of training I've had in those two areas. I feel much less confident with coding. And that's another reason I'm glad I tried coding, by the way, because I can experience why a coder might think a piece of writing that's crap and AI slop is okay, right? That's not their subject matter expertise. So remember that subject matter experts are still important.
And then finally, thinking about using AI, it is important for you to check what your institutional policies are. So for example, Affordable Learning Georgia and Cleveland State University have provided specific guidelines, but Open Oregon has a hard no. And I think Marco is on here. He gave a keynote for Open Oregon, I think that really addressed the environmental impacts of AI, which I highly recommend to everyone. I think Marco, it's linked in that ... It's probably linked in this AI resource guide. I think I put it there, and if I didn't, I'll get it in there.
So I wanted for our live demo, do you have questions or things that you want to see? I've already showed you a few things, but is there something like maybe ChaucerBot is one that I had thought we could look at. Let me show you a couple things really quick, and then if you're thinking of questions, put it in the chat. I mentioned writing a personal writing assistant. I created this and trained it on my own writing. This is not a publicly available bot. Boy, wouldn't that be fun to have my students get at it? So it's just private for me, but I use this to "Create a reply to this email..." a lot, and I use "Write a blog post ..." a lot.
I'll share my Substack out. I meant to have that so you guys would have it. It's really fun for me. We had questions about where AI is used and how we acknowledge it. On this blog, I use AI a lot. I use that bot to write a lot. And this was a fun experiment in the humanities. I've been working with poetry since 2023 with ChatGPT. I've run some kind of poetry experiment for National Poetry Month every single year. And this year I gave it a handwritten sonnet. I wrote this by hand, and so I've linked to the previous years if you're ever curious, but I had to analyze the sonnet and tell me whether it was any good. And I was like, "Yeah, it's okay. Competent, sincere poem." Oh, that's so sweet. Thanks. That's Claude.
And Gemini is like, "Charming, technically perfect." Now note, ChatGPT said that the meter was wrong. And I was like, "You're wrong, ChatGPT. My meter is excellent." I'm the queen of iambic pentameter, so Gemini and Claude know this. Then I was like, "All right, you do it. You write on the same theme." And these are the poems they wrote. They're getting really good at poems, you all. I think I like Gemini the best. And then I was like, "Hey, did I write it?" I originally was tricking them. I'm like, "Hey." And Gemini knew I did because I've let Gemini know everything about me. And I was like, "Oh, you wrote that for Andrew. That's so sweet." I was like, "Yep, sure did."
And then what does it mean for the future of human writing? This was an interesting conversation. So these are always fun for me. You can look through this, but what I wanted to show you ... And there's some takeaways for instructors. AI acknowledgement: "In this poet, I wrote everything myself, including the original sonnet that is not quoted from an LLM," so I have directly quoted and made that clear where I'm quoting from LLMs, "but I also included a lot of quotes because I wanted to show you how LMs are writing and thinking about poetry in 2026." So that is an AI acknowledgement statement. It's on every single one of my posts.
Sometimes people get angry at me. I had a colleague who came in and was like, "You tricked me. I thought that was you the whole time and it was a hundred percent lies along persona bot written." And I was like, "Good, that means my bot is working as planned." It's not producing AI slop. It feels human. It feels like I wrote it. That's right. BeowulfBot is here. This one's kind of fun, right? So look, it's at strong, weak verbs. I can share a passage and it will not translate for me, but it will check my translations. I have in the instructions don't translate for me. And that's another important thing I think to teach your students, right?
Oh, and there's those horrible adjectives. Yes, let's do that. And then if I want to, I can put in ... I have it trained on my particular textbook so that it's working with that. ChaucerBot, super fun. This one, again, it tutors in the same way. TLDR, you saw that demoed. So I use TLDR in my Substack just at the top.
Amanda: Hey, Liza.
Liza:Yeah.
Amanda: We are very close to time.
Liza: Oh, you're right. We are.
Amanda: One minute.
Liza: And that's it. Thank you. See, I'm so bad. Well, I thought we're just down to questions. So yeah. And you guys, if you have more ... So yeah, that's what I wanted to show you. Lots of fun stuff. And then if you have more questions, obviously I could talk about this all day long, but I think they're fine. Oh, and that's the other thing I wanted to show you real quick. You can build those same bots over as Gems in Gemini. So those little custom GPTs, this is the Latin practice one I created for myself, so.
Amanda: And for everybody else, I have thrown our survey in the chat. We would love if you'd give us feedback on this session. And thank you so much to Liza.
Liza: Thank you so much for having me. Thank you all for being here.
END OF VIDEO