 That's right. I'm a senior education consultant for the CTLT. And as my director likes to say, I think we're, some of us are three months ahead of other people with generative AI. So I'm not an expert, but I've been taking courses. I've been doing research and quite a few workshops and resource development. So yeah, I'm excited to meet everyone and talk to y'all today. Judy, do you wanna introduce yourself? I also work with Lucas in the Center for Teaching and Learning Technology. I'm an education consultant. I used to say that I'm a month ahead, but now I'm thinking I'm along, working along with the majority right now, but I do spend time to play with gen AI. Over to you, John. Hi everyone, I'm a learning designer at the Center for Teaching and Learning Technology. I work alongside Lucas and Judy. And I will also say that maybe I'm a few months behind Lucas, but a few weeks ahead of everyone else, maybe. On where I'm at with gen AI, but I'm really excited to be facilitating this with all of you today. And Bosong. And Bosong. Hello everyone, my name is Bosong Kim. I'm also with the CTLT. I'm a learning designer. I try to say I'm two months ahead of people. I may say that. And I'm quite excited to talk about and listen to how gen AI can support and enhance teaching and learning. I'm going to share just one way of using gen AI in education in this workshop. Great, wonderful. Thanks for the intros. And I wanna give a land acknowledgement. So we are privileged to have this workshop on Musqueam territory that we live, work and play and is unseeded. I also wanna pull up this quote from Michael Runningwolf. And Michael is a data scientist and a university PhD student who's been looking at generative AI. And it's a good reminder to me. I'll just give you a chance to read it. So we know that generative AI is scraped the internet and captured a lot of people's IP. But I think it's also appropriated a lot of culture and it's part of the underlying exploitation. So an example of this is if you go on to it and ask it about indigenous learning at UBC, you'll get answers from all of the different web resources here without acknowledgement of where it came from. So I think it's a challenging space right now to work on and just thinking about what it's scraped and how we can acknowledge that and how this might be part of a colonial system. A couple of resources for the session that's John's gonna share into the chat right now. So we have a guide that we created for the session. And what this guide has is a lot of the prompts we're talking about today, as well as a lot of the examples and resources we mentioned. So we encourage you to follow along using the guide and it helped to kind of give things like more concrete lens for you. Secondly, we do want you to do some hands-on work today if possible and maybe I could get a thumbs up if you actively use chat GPT now using your virtual thumbs or physical thumbs, that's okay. Great, thanks Emily. I'll just give everyone a minute or so to do that. Perfect, so a few people do, perhaps a few people didn't put their thumbs up. If you didn't, please feel free also you can use Bing if you use that, Bing AI. There's also a site here called talkai.info.chat and that allows you to use chat GPT without signing in if you're uncomfortable doing that. Beware it's a little bit spammy so it's easy to click on an ad link but it does give you access to the tool. So you don't have to go hands-on but I think it'll be more valuable for you. So our agenda today, 30 minutes interactive presentation followed by 30 minute discussion and really encouraging you during our discussion to use this as a space of sharing. And our learning objectives, three learning objectives. Again, I'll give you a moment to read them. The first learning objective is kind of around understanding how it can be used by students as part of their learning. The second objective is to think about different types of prompts that can help make tools like chat GPT a better coach, a better tutor, a better simulator. And thirdly is engaging an interactive discussion at the end of the presentation. So let's start with the interactive presentation and I'd like you to open your comment boxes now and please share in the chat one way that you've used chat GPT or you would like to use chat GPT in your own learning. So what I'm gonna do is rather than click enter I'm gonna get you to type in the way that you've used it but not click enter until I count down and tell you all and then we'll kind of get a waterfall of answers. So again, one way that you've used chat GPT in your learning or you would like to use chat GPT in your own learning don't click enter yet, just type it into the comment box. And when you've done that give me either a physical thumbs up or a virtual thumbs up so I know you're ready. All right, so I'll give everyone a minute I've seen a couple of thumbs there. Thanks Annie, thanks Brandon, thanks Francis. Thanks Stella, great. I'm seeing a few more thumbs here I'll just wait another 30 seconds or so and in a moment I'm going to count down. So I'll just count down one, two, three, go. So one, two, three, go. Wonderful, thank you for sharing. So to explore different career paths Katie says coding. Another coding as well. Early free chat GPT for draft announcements to improve writing and to do writing rephrasing to get an overview of a topic especially when I'm stuck interesting to generate practice questions in the style of my existing questions. I'm working with robots. I usually tell chat GP to give me some real life tasks to teach robots to do them. Also it's useful to improve in my writing. So using robots to teach robots. I think that's the first time I've heard that to gather ideas about a topic. So lots of different ways that we're using it in our own learning. And I wanna switch gears now and think about what it might mean to student learning. And starting with a couple ways that I think Gen AI can help students success. And there's three ways. And we're gonna go into these in a little bit more detail. First of all, we know one of my other hats as I work on a project called the Chapman Learning Commons. And from that work, it's quite clear as is in the research that students come to university with a variety of different skills for studying and for metacognition and for academic achievement. So in what it's interesting to think of ways that Gen AI can help students become better in these areas. Secondly, is students have diverse abilities and different barriers to learning? So how can these tools help them overcome those barriers or mitigate these barriers at different times? And thirdly is from research like Bloom Sigma problem, we know that students learn significantly better through one-on-one tutoring than in classroom environments in many cases. So how can we leverage these tools to help students use them as tutors? And we're gonna go into these in more detail now. As we're going through this, please feel free to throw your questions and comments in the chat. But I wanna start by talking about the constraints and the limitations of these tools. First of all, there's the issue that I'm sure you've heard about is student privacy and faculty privacy, staff privacy, institutional privacy and organizational privacy. We know that these tools save our data. We know that they have ambiguous terms of reference and they haven't had privacy impact assessments done on them. We also know they've had data leaks and it seems that they're using our data for training themselves. So how do we protect student, faculty and staff privacy? I think an easy way to begin with is not putting PII or personal information into it, but it's worth thinking about where's the boundaries of what we can put into these tools. I don't put people's names in this tool, for example, if I'm writing an email to Emily, I will say name one and then write the rest of the email and fill that in later. But how are we thinking about privacy? If we're asking students to use it as a tutorial tool, how are we helping them protect their own privacy? Secondly is choice and equity. I think there's an opportunity here in terms of equity, students, more students having access to a tutor, but right now we're already seen in equity in these tools. Chat GPT-4 is much more powerful. It gets 90% on the standard bar exam in law compared to 20th percentile by Chat GPT-3.5. Chat GPT-4 costs $30 Canadian a month. How do we ensure equity between our students? And related, how do we give them choice? Not all students are going to want to touch these tools. I'm already working with students who don't wanna sign into them. So if we're thinking about using it, if we're thinking of asking students to use it as a tutor, how do we give them an equitable choice to not use it? And thirdly is cognitive offloading. I think this is quite a challenging situation we're in now. What sort of work is it okay for students to offload into these tools? If they're using it to develop their metacognition, does that take away from their metacognitive development? If they're using it to write resumes, does that take away from their resume writing skills? So how do we teach them? How do we work with them? And how do we think ourselves about what we wanna offload and what we wanna make sure we don't offload so we don't de-skill ourselves and our students? And accuracy and hallucinations. I took this picture of mid-journey and you'll see that the professor there, which just an aside, a mid-journey is quite biased. Professor is always a white male with glasses, it seems. The professor is teaching the student but what he's writing is nonsense. So how do we get our students to use this as a tutor if sometimes it lies to them and tells them things that are inaccurate? We know these tools hallucinate. We know that they're very confident. What does a tutor look like if it's not always telling the truth? So what we're gonna do now is look at three different ways that it can be used in student learning and Judy's gonna launch us off talking about a framework we can think about. Go ahead, Judy. Thank you very much. We draft most of the workshop based on the work done by Molek and Molek. Sly events? Yes, okay, I can do it. The idea behind it is we really want to encourage and challenge our students to stay in the loop when they're using generative AI. We know it's really powerful. It has a lot of ability like what we shared earlier that many of us are actually using it for our own learning. So we really want students to be able to harness the ability and also want students to recognize the error, the inaccuracy. What is the tutor? ChatGPT is not giving the accurate information. We really want to encourage students to be able to check the resources at the personal voices and the insight as we ask our students to work with gen AI tools. So in the framework, Molek and Molek introduced seven different ways to use AI to support students learning. So again, this is to support students learning. This is not to replace a human. So Bosa, Lucas and John will be talking about how we can use them as a learning coach, a tutor and a stimulator. And but Molek and Molek also proposed that we can use gen AI as a teammate so that students can learn how to collaborate with different type of personality. Students can also do self-reflection and thinking about and help them understand their own learning. Gen AI can also be used as a mentor and to give feedbacks to students and also as a tool to help them get some work done, get some of the work done. So seven different ways to use gen AI and over to my colleagues to talk about how we can use gen AI as a coach first. Thanks, Judy. So like Judy said, one way we can introduce AI in student learning is to use it as a coach. We can build prompts in AI that provide opportunities for students to develop study skills for self-regulation and for metacognition or to self-reflect about their own thinking and their own learning. So why does it matter in education? Students need support developing study skills so that they can plan for future assessments and learning activities. And metacognition plays a pivotal role in learning and enabling students to reflect, digest, retain and apply newfound knowledge. And so we can use AI to build interactions that kind of help facilitate this kind of coaching in many different disciplines and contexts. And so in this one example of using AI as a coach, we can ask the AI to respond to a prompt as a specific character and or respond as a specific, or and or respond as a specific character. And this is known as a persona pattern. And by prompting this way, it helps the AI produce specific datasets related to the persona. And in this case, I asked the AI to be the persona of an introductory French instructor and capture responses from the persona of a student, the audience. And so here I'm asking the AI to produce three study approaches for students which speak to their outcomes, these outcomes and preparing for a vocabulary and grammar quiz which is the context for this prompt. And I'm going to share with you the AI response and output in the chat. So if you link to that, it'll open up the chat GPT window which I inputted the prompt and the output that it produced. In this next example of ways of using AI as a coach to improve study approaches, I've asked the AI to be the persona of a mobility coach training an audience of middle-aged students to have better mobility as an outcome. However, unlike the last example, I've asked the AI to ask questions to the audience one at a time, to the learner one at a time. So instead of prompting the AI for a response initially, I've asked the AI to ask the learner questions which will help deepen its understanding of the answers and create a more comprehensive response. This kind of prompting where questions are being asked of the learner and is inputting responses is called a flipped interaction approach. And then based on this flipped interaction, you can see how in this example, the AI has built a more contextualized and maybe even a more customized training approach. And I'm going to share with you the prompts and the output that it produced and all the follow up responses that I inputted as a result of the questions that the AI was asking me. And so in this next piece, let's just try this out. And if you have chat open, chat TPT open and our registered user use the included prompt templates or example prompts and have chat TPT act as a learning coach supporting learners and learning strategies in your discipline or area of expertise. And I'm just going to insert this prompt template in your, in the chat which you can use and replace with your contact. I'm just going to give you guys a few minutes to try this out. And if you would like to share the output in the chat, please feel free. How did you manage to get a URL that shared the conversation with us? Thanks, Andy. So at the top of the conversation, yeah, Lucas can show this as well. At the top of the conversation, there's an ability to share the link for the specific conversation that you're generating inputs and outputs for. Jens, you can also, so we shared a guide with you at the start of the workshop. Okay. If you go to that guide, you'll see there's multiple prompt templates under coaching. So John, are you able to share the link to the guide again? I don't have access to it. Yeah, so if you go into that document, you'll find a prompt template as well as a couple of example prompts under coaching and you can use one of those if you would like. And just for time, we'll probably keep going and you can do that in the background. Thanks, Lucas. Great. So let me jump in and again, feel free to use those prompting approaches in the background as we go. I think Zoom has a setting around file uploads and disabling copy-paste. So please feel free to copy and paste from the Google doc as well. And hopefully these templates, you can keep using them. I love the idea of templates and thinking about sharing it between folks so that we start building up libraries of prompts, libraries of approaches to using these tools. So let's talk briefly about AI as a tutor now. And I think this is a really powerful way that I know I've been using AI. I have a son who's in going into eighth grade and I'm going to be forcing him to do about half an hour a day of AI tutoring. I can force him. I don't have to worry about his privacy as much. But it's a really interesting way to do back and forth. And I've mentioned this a little bit already, why tutoring matters in education, why this tool might matters. There's robust evidence that tutoring is effective in learning. So all the way back to 85, Bloom Sigma II problem, multiple articles about the value of tutoring, the one-on-one relationship. Secondly, we know there's unequal access to private tutors. And I mentioned equity earlier, but there's also already an equity built into the tutoring system where some people can pay for private tutors. Some people can't pay for private tutors. So can this be a way of ensuring equity? And overall, it's an opportunity for customized tutoring. So here's an example prompt. And I'm actually going to demo this prompt with chat GPT in a second. So you're an experienced tutor. So I used, again, this specific persona to allow it to get to more specific data. And you can layer the specific persona. You can say you're an enthusiastic, experienced university level physics instructor with 20 years of experience who specializes in working with students one-on-one. So you can kind of layer in specificity there. Explain Newton's three laws using examples. The second thing that's helpful for students is to connect abstract concepts with examples. And GPT and generative AI is really good at generating examples. And then thirdly, and this is that flipped interaction piece that John talked about, getting the tool itself to gauge your understanding. So once you've done this, ask me questions to figure out if I understand it. Let me just demo this for you now. Gonna stop sharing for a moment here. One moment. And I'll do a quick demo of that using chat GPT, just a second of patience here, while I get my screen up. So I'm on chat GPT and I'm using GPT for, if you haven't used for, it does cost, but it's a lot more effective. So I'm gonna paste in this prompt here. You're an experienced tutor in a university level physics course, and let's see what we get. Great. So certainly be happy to, it's interesting because it's always different. The answer I'm never exactly sure how this is gonna come out. So it's going over the first law of inertia, the law of acceleration, and it's giving me these concrete examples each time. Now here's where the challenge comes in. I'm not good at physics. I didn't even do high school physics. So I don't know if it's true. And this is where the tutoring comes in. I hope there's some physics folks in the room. So it's giving me some questions now based on this. And I can answer the questions. If a car is moving at constant speed in a straight line, what can you say about the net external force acting on it? Can someone share the answer in the chat? I have no idea. I'm just gonna open the chat there. Nothing yet. Okay, I'm gonna say air resistance. I hope Judy's right. So I'm just sharing one answer and we'll see what we get. Great. So it's gonna give me an answer and it's going to start kind of doing this back and forth as a tutoring tool. So that's just a demo of what GPT or these tools might look like as a tutor. And I'm gonna turn it over to Bosung now who can talk about the last category that we're looking at is AI as a simulator. Thank you, Lucas. So I'm going to discuss and then share some examples on the use of AI as a simulator for student learning. Simulation or simulated learning that is really widely used in the fields of medical education, social work, low education, counseling disciplines because it's a great way to engage a student with real world scenarios where a student can or encouraged to interact with the complexity of the problems and thereby encourage them to critically analyze and like react to those complexity and act to provide the responses. The idea of using AI as a simulator is to have AI simulate real world interactions by playing a role that we specify. And we can also ask AI to provide feedback and suggestion for improvement on our action and responses and decisions. So it can generate various scenarios across various discipline and topics. So there are not wide mirrors in education. Here are a couple of reasons. Using gen AI as a simulator, it allows students to practice those skills and apply knowledge in a safe but challenging spaces. Students will have less anxiety because it's okay to make mistakes and they can also learn from their mistakes. Gen AI can also be prompted to offer feedback, as I said before, which is really important for continuous learning and improvement. Lastly, it is quite cost effective, like well-designed assimilation. They can be costly, but using gen AI three from five, as we know it's free. So it might be more equitable platform for practice. And as far as I understand, it has no and or less administrative task involved in from instructor end, but it is still necessary to provide a clear guideline and instruct the student on its usage for their learning. So the first example here I provide is job interview simulation. As seen in the previous example by Lucas and John, we need to tell AI who it is persona and along with some context information, we can layer some specifics. We can make it more specific to make gen AI produce more well-designed questions. We also need to inform what tasks you need to perform and then how it should behave. I'm here also asked feedback for improvement that is a must for simulated learning. So here I told AI to act as a job interviewer in a tech company industry, hiring a software development position. AI was also prompted to ask a series of questions one at a time. Otherwise AI may list out all the questions in their first reply. So when we design prompt, we need to include important requirement or sometimes constraint for simulated instruction interactions. Here I said that after four interaction, AI need to provide constructive feedback on my responses and then inform me where I can improve. Here you see below that is the voice control. I installed, you can install this voice control that is Chrome extension. I installed it to engage in more conversation based simulation. The mic on this voice controller will allow me to use my voice to have chat GPT convert my response to text. And also the speaker there when it is on it will also read the question by AI allowed to me. Here the important risk to mention is that student need to think about the privacy. So not to share their two personal information with Gen AI. The next example is the patient assessment simulation. So similar to the previous example, I told AI to perform a specific role here as a patient and inform that I'm a healthcare provider. This scenario was developed to help the student in healthcare field to develop their patient assessment skill not in all area but specifically in mental health. It was prompted to provide basic patient information and wait until I read the patient and provide feedback on my assessment and treatment plan for improvement. So before we ask the student to use Gen AI for simulation in this way, it will be very important for the instructor to important instructor to check its accuracy and relevancy to their field. As Lucas mentioned before, it's very good at policy hallucination and then it may provide inaccurate information. Also, we can ask a student to analyze its feedback output and which feedback they decide to take and or not taking and their reasons. So now we want to hear from you. We are from different disciplines. So do you see Gen AI has some potential in your discipline to support your student learning as a simulator? So if you think about how you might use this approach in your course, in your discipline, so please use the chat to share. I like the Lucas approach that let's wait until everyone have time to think about and then please thumbs up when you put some salt into the chat. So for example, in business field, you may think about some role play simulation in HR meeting. I think it has some possibility potential for low education. Okay, so if you have some idea, let's hear, I will just say one, two, three and then can you hit enter? One, two, three. Okay, great, assimilate a group of learners when designing new teaching activities that create idea and then ask another student to explain a topic and then assimilate in case of study about academic integrity. That's really interesting idea and acting as tutor. Yes, that's one example of the simulation. Emily, you're a librarian and then you think that they can use Gen AI for creating effective research questions and getting started with the research plan. Michigan, University of Michigan has some great idea about how people might be using Gen AI to help with their research. I'm happy to share the link and then how to use a specific career headset. Is it a question? And I mean question to us and generating problem for material design and providing a template for answer, that's great idea. I think we shared great idea. I hope that later we can also save this chat and share with others. Okay, back to Kass. Wonderful, thanks, Bosung. So the last thing I wanted to mention really briefly before we move into discussion and activity is just where this is going. And I think right now we have these open tools where we're using chat GPT or entering it, it's these huge language models but what I'm already seen is the customization of these tools. So I've put a link on the side to a tool called Poe on the left where you can create custom bots that use chat GPT. In this case, I made a learning support bot that uses complex prompts on the backend and helps students with learning support, but that's all and it has some specific guidelines. On the right is CanMingo. CanMingo isn't available in Canada yet but it's the Khan Academy's version of a tutor bot. So it'll be interesting in the next year. I would expect we're going to see a lot of custom bots that are not the whole of GPT, not the whole of these models, but slices of the models. And an interesting thought experiment would be what would a bot look like in your course? What would a bot look like in your context? How would you develop it? But let's move on to the discussion part and to kind of get you going, what I want you to do first is to take a moment and prompt it. So we've talked about three ways of using prompts, coaching, tutoring and simulations. I've given you a guidebook with different prompt templates. Select one of the prompt approaches, edit the prompt template and create your own prompt in chat GPT or Bing Chat. Look at the quality of the prompt and the output and then share your prompt and the output in the chat. And we can use that as a point of discussion. While we're doing that, please feel free to raise your virtual hand, ask a question, you know, share what your prompts looking like. But we wanted to kind of get you going, doing some prompting right away. So I'll get you to open GPT, open a GenAI tool of some sort, grab a prompt based on what we did and see what the output's like and kind of let us know. So I'll give you a, let's go five minutes on that. But again, if you have questions that have come up, please raise your virtual hand and jump in as we do that too or drop them in the chat as well. And I see John's put a couple of links to some tools in the chat as well. And again, once you have your prompt and your output, share them, just drop them into the chat. You can share them as a link if you want or you can share the overall prompt and output. So I see a question in the chat from Annie saying, I saw in the promptathon that you could use chat GPT to create learning games. What sort of games have you had success creating? I've had it successful doing question and answer games. So where it asked me one question, you know, a really interesting game for it is a prompting game. You can ask it to give you an output for a prompt and your challenge is to create a prompt to get the output and it will evaluate your prompt. So kind of a meta game. I've also done a logical fallacy game where I had it. I think I might have showed that in the promptathon where I had it play a logical fallacy game with me. It gave me a logical fallacy, a sentence with it in it. And I had to guess what was it ad hominem? What was the logical fallacy? And it would tell me if I was correct. So I played those couple of games with it. I'd like to hear from other people, other games they played as well. And another question, how can we use it well with graduate students when writing course papers and even theses? So I kind of open that question that's from Mary and up to the room as well. How can we use it generally to support writing? Any ideas around that from folks in the room? And I see that Judy's answered. It can be used to create a very basic generic outline. This is known to be able to help students overcome the initial barriers to writing. Students certainly need resources and voices to create the final product, the research paper. Thanks Judy. And I have some concerns about this. This is from Amit Upadthay. As we about this, how can we as an educators promote AI tutoring when we know of its tendency to hallucinate, especially with more complex concepts when there is not as much information online to begin with? Well, we can advise students to be aware of hallucinations. They might not be confident with the subject matter to differentiate between the correct answer. Absolutely. I think this is a big challenge. I think some of the custom bots may solve it as we get into more custom bots. They may have more accuracy. But how can we put a human into the learning process for students? Judy mentioned she's trying to create a jeopardy game. So maybe if folks had a chance to prompt chat GPT, would you be able to share your prompts now and your outputs? Just drop them in the chat. That would be great. Great. I tried an escape room game, but the output was too simplistic. Thanks, Annie. Any other prompts that you tried? Just give folks one more minute, and then we can move on to some couple broad questions. Great. So Brandon's prompt, you're a grade 10 student who is struggling to grasp conservation of energy, ask questions. To me, the teacher as a student would, to a teacher, one at a time, allow me to respond. I'm having issues with this. Chat GPT is answering its own questions. Interesting. And that can absolutely happen. Often I'll be telling it, please stop, try doing it this way, changing it around a little bit so that it doesn't start doing things like answering its own question. Worst comes to worst, you may even have to open a new tab for other points here. I just grabbed one of the prompts from your Google Doc to compare Bing and chats as Yens. After learning that Bing doesn't work in Chrome, I did realize that Bing's references look a lot more appropriate. That's a good point in Bing. If you use Bing AI, it does use references, and it has links to the references. One of the challenges, it's nice to be able to find the links. I think one of the challenges with that is some of the links are, a lot of them are to Wikipedia. Some of them are like to Quora and some dubious web sources. So I tried to start a Q&A back and forth, but my prompt said write a set of questions and it provided the answer before I could try. When I changed the prompt to suggest pose it worked, interesting. So a single word that made a big difference. And Eden shared her prompts, communication skills. Effective communication skills vital in medicine. She did a situation. You're a lead physician in a multidisciplinary team caring for a patient with complex health conditions. The team is having difficulty agreeing on the best treatment plan. And tensions are rising. Your goal is to ensure effective communication and find resolution. I love the specificity that we're getting into. And prompts can be long, or they can be something you do ongoing with the tool. So thanks for sharing that. I just want to now open up for a discussion. So how might generative AI be used in your discipline? What are some other ways you've used it already to support learning? And what limitations and constraints do you see in these tools? So if you want to turn on your camera now, that would be great. You don't have to, but why don't we kind of discuss the significance in your discipline? I'm sure we have a lot of expertise and thinking in the room. So please feel free. Turn on your camera. Raise your virtual hands. Jump in there if you'd like. And I'll keep fielding questions in the chat. Does anyone know of GPT-4? One moment. Does anyone know if GPT-4 is capable of generating correct citations? Rachel, a really good question. So I don't know if anyone else has done experimentation with this. I have. Yes and no. Sometimes yes. If prompted under certain, sorry, sometimes no. It's given me lots of false citations. But when I prompt it in education, I have got a lot of correct citations. And one of the ways that I've done that is using, it's called a COT prompt, a chain of thought. So once it creates an answer for me, what I'll do is say, tell me how you got to that answer. Tell me the references you called on. Or I'll even add a role into that. So I'll say, act as an experienced researcher in X. Analyze the results you that were given to you. And tell me the research that it was built on. And what I found is sometimes when I do that, I get very accurate references. It's easy for me in education because I can see whether the references are legitimate or not. Sometimes I won't get accurate references that way. And it's so interesting to see, again, that sometimes you're going to get something clear and sometimes you're not. And how do we work with students around that? So ways that you're using it in your disciplines, what constraints, what limitations are you seeing with it? Maybe another question is, would you be comfortable using it as a tutor or a learning coach in your area of practice? Once Analyze's mentioning that she finds it quite superficial, but with great potential as it improves. And I would challenge you to, I think we can move beyond superficial sometimes with a lot of prompting. It's not always going to be there, but I found it really depends on how we prompt it. But I do think it runs into these barriers. So it's an interesting experiment to see how deep we can go, like how much, if you queue it with very specific things, can it pull on better data? Ani, can I ask you which field you're in? Yeah, science education? Let's call that. OK. And where are you finding this superficiality? Well, just when I use it generally, and I ask a question, I just find it basically, it always kind of responds at the Wikipedia level. And so, which probably is enough for most people, but when I'm using it for my own work, it's like, no, need to dig a little deeper here. That's a really interesting point. How about the rest of you? Do you find this superficiality with it, or have you been able to go a little deeper? So I'm just looking in the chat. I work in a specialized area of medicine, and I find sorting through the hallucination, which versus what was actually correct can be time consuming. I think that's quite interesting is how long we're spending to try to wade through this and to try to find the valuable information. An example that I'm not sure I shared this in a previous workshop. Yeah, sorry, Mara mentions it goes deeper with follow-up questions. And sometimes it can in different areas. An example I had was I paddleboard, and recently I went to my hometown, which is called Outland BC up in northern British Columbia. And I asked it, can I paddleboard? Where should I paddleboard? And it said, well, you can go across the lake. And if you've been up to Outland, it's the lake 70 kilometers long and really wide and glacial, and you will not do well going across the lake. So it's pulling this generic information. But then I was able to say, I want to take my paddleboard from Torres Channel to First Island, and suddenly it called on totally different information. And it said, you can't do that. If you tried that, you would need to do xx and x. You'd want to cross here. So it's like it was able to access different data points. So it's interesting to play with different words, different phrases that will get at those different data points. Or as Jennifer Rice mentioned, perhaps in certain areas of medicine, certain areas, it's not going to be able to move beyond that data or move beyond that generic Wikipedia type answers. I find that disconcerting how much you have to play with it to get good information. It's a very good point. Wonderful. So I'm just looking at our time now. It's almost 5-2. And what I want to do is share some further resources with you. And thank you for the conversation in the chat. That was a really interesting conversation and lots of really good prompt shared. So let me kind of end with a couple of resources. One moment. Give me a sec here. I'm going to take my chat down. So resources. Again, acknowledging this is a really emergent space and a single workshop only touches the surface. At CTLT, we have a GenAI website where you can go and we've included resources there. We've included links to future sessions, as well as some faculty stories about using AI. Related to that on the bottom link there, it says share your experience. If you have an experience using it, we're trying to collect these stories right now, how you've used it in the class, what's working for you, what the challenges are. And finally, for this session, based on that guide, we created a web resource that goes over each of the different prompt approaches, as well as, so the different prompt approaches, as well as the different ways of using it. So please feel free to access that as you go. And that kind of brings us to the end. I wanted to thank everyone for taking some time to explore this and it will be really interesting in a year or so to kind of look back on this and see where things are then and where we were now and what the changes are. So thanks for coming to the session. We'll put the recording on the website as well and we'll share an email with you, with the recording as well as with the website resource.