The Science and Art of Teaching Creativity in the Age of Gen AI: What Every Educator Needs to Understand

B-R-N, where educators share what works. What these systems have gotten really good at is style. They can imitate visual styles amazingly well, but the interesting one was, you know, once this new model came out, I started taking pictures of my family that I had and I said, "I will make them and Pixar style." And my daughter had a photograph, she's Indian, of course, Indian American, and with two of her roommates who are white. When I did the Pixar version of that, it made her African American, and that was fascinating. And I think these become really interesting points for us to have a conversation about how these systems work, what kinds of decisions it is making. When we look at them in educational context, it's really important for us to unpack. Hi guys, great to see you again. Hi, Rubim. Hi, everyone. I've been really curious lately about these image generators that are out there. And I've been playing with these image generators with my student teachers, who are getting ready to teach K-12 students. And they, to say the least, are very nervous about AI. And one of the things that we've noticed as we've been looking at these generators is that there's some questionable things going on within them. For example, one of the things that we've been generating are images of our campus at the University of Michigan. And what we notice is that it will generate an image that looks like a campus, but it's not our campus. And we'll even ask for more details, such as to put in particular locations or stadiums or buildings. And while those buildings will be added, they're not the actual building that you would see on campus. It's something that might look similar to it. So there's definitely some questions around accuracy. And then when we would ask for a student or students to be placed in the campus, we notice that the students tended to be white males as the first generation of it. And then as we would ask the generator to generate more students, we would have to be very specific about having students that were more diverse, that were of different genders. And so the image generator would generate those, but we would really have to push. So I was just really curious what others were noticing with these image generators. It seems like there's some potential, but there's also some things happening that might be concerning too. So I'd love to have a conversation about that. Sure, it's a great question, and Bumbley definitely need to get into. One of the things from my perspective is, first things first, we should always have been telling students, "Hey, just because the image says it's from somewhere or it's even a photo from somewhere, doesn't mean it's representative or that it's neutral or that it's anything in particular evidentiary unless you have another context that gives it that. I mean, again, I'm not exactly just born into the internet age, I come from an earlier era. And in my era, for instance, National Geographic was one of these standards for, "Oh, look, National Geographic." Well, if you look at the images in National Geographic, they had huge, depending on which era of National Geographic, it was a huge resource, but it also showed biases, it was also very selective, what it showed, it had certain ideas about what this culture versus. That culture should look like, so it was never a neutral source at all. And the same thing with Encyclopedias, I didn't have a Britannic, I had a La Rousse Encyclopedia growing up, and again, some of the illustrations in that thing, you look at them today and it's gone, "Oh my God, it's incredible." So I think one of the first things we have to do is, no matter where the image is being sourced, teach students about, well, visual literacy, and that what they get out of something like, whether it's a chat GPT or Gemini or whatever, will represent a subset of images that has been collected. With a given filter that will take what they say and render it as something that will represent something coming out, that may or may not be representative of that location. So what we need to get into is a real issue of media literacy, and also, frankly, when things have always encouraged, we will reach out across cultures, I used to, you know, in the early days when you could first, this is before the pandemic, before Zoom, etc, I always encourage people to do things like digital writing projects that have reached across cultures where people will talk about the reciprocal representation of each other as a way of exploring this. I think, you know, just to get the ball rolling on their front, so I know, Helen, what do you think about this? Yes, so I have been exploring images for a good while. In education, we often have to show images because it's very helpful. I present a lot, and it was always difficult. My family had that many pictures taken at them using computers and things, and I used to use all their children for all my presentations, you know, to again, focus on good digital citizenship and not just grab images up the web. So now it's very helpful in that the images have got so much better. I agree in Tyler Ruben, what you said, in particular, about media literacy, digital literacy as a whole. And I've been looking at images in particular, we went from suddenly this kind of hazy version of something that we recognise as a person, to weird hands, to then this better picture, and then suddenly now we've gone to a whole completely different level. In full disclosure, OpenAI gave me access to the new image feature that they have about a month before it came out, and asked me to kind of explore it. And I spent ages, like can I have a picture of a middle-aged woman, and I was checking for grey hair wrinkles, asking for pictures of students, and it was in an interesting case. Women and I have a very good image creator, but what I know to see even still now, if you ask for a student say, this student is an absolute beautiful looking student, you know, perfect skin, amazing buzzing eyes, and it's, you know, we're not all looking fabulous like these pictures show, and what I really liked about the Changju BT one is it showed a very true image, you know, like hair, stuck up, you know, like we'd see people. But again, like I said, I examined it very thoroughly, I asked for pictures of people wearing hair and aides, cleft lip, all sorts of things, and it was very good. There's certain things though that still are fun. What it can do now is it can show as an image of a wine blast with the wine topped to the top. It couldn't do that before. It would have always the three quarter height of the wine, and then if you ask for it over flowing, it just puts a piece of the top with the wine going over. So it's figured that time is still has issues with analog clocks, but otherwise I'm finding it really nice, and for using it in education, it's got so many opportunities, you can do so many things from it, you can create flyers for your class. You could create cartoon images of scenarios, I create various ones of those, that I'm using in my class now, you can create multiple images, and what I really liked, an opening I actually published, was things like learning theories, creating posters, just as like a quick memory buzz for students to remember what each one is. So as you can see, I'm very positive and I don't recognise that it's still a bit of a way to go to be absolutely perfect, but it feels like it's gone from okay to wow, I could not tell if that's a human, with a photograph or actually AI, so Caroline, what do you think? I've been trying things out a little bit, and so one of my classes is forensic linguistics and we were really interested in how AI deals with pronouns. So we were getting ChatGPT to write short stories and giving it unisex names and not giving the gender and then asking it to write a story about a teacher or a professor or a police officer, this kind of thing, which it did and it didn't use any gender at all, but then we asked for an image of it, and that's where the stereotypes started coming out, so again we were getting white male doctors, so that's how we noticed, there's a difference between what it actually writes and then the images it comes out with. One thing I, it is terribly polite generally, but it's got a bit better, when I try and get images of myself and I describe myself and then I say, you know, overweight or a bit chubby, 50 years old, it tends to go a lot younger and a lot slimmer, I really have to push it to get what I want coming out of it. Issues that I've got better, so if it's using text in images and words in images was always a massive problem, you know, it would spell things incorrectly, it didn't seem to really recognize language as such or words, now in chat GPT it seems to have nailed it, but not in other ones, so that is much better. But what I've been trying recently is like you were saying Helen about posters and things like that, so I've been trying to get it to create infographics and I was just trying this out with experimenting with examples from bushcraft, okay, so I get chat GPT to describe how to create a longfire in a forest, for example, and it would give a really great definition and the stages and describe how it could put that together. As soon as I asked for an infographic, all of that is lost, again, it's great in text, but it's not in the images, things weren't making sense, it was breaking the laws of physics, so I think that's certainly an area I'd like to see it improve in, if it does infographics which have lots of text in them, like a flow chart, something more like a flow chart than an infographic, that seems to work well, but anything with images in is really struggling with at the moment, so those are my most recent experiences, Ponya, what are your thoughts on this? So, images are interesting, right, because they are so literal, right, so I think it's not surprising Caroline, when you've talked about the examples of when you ask for a pros, non-gender sort of specific, but when it has to create an image, an image is like it has to make something, and whatever it does, we can then interrogate it, right, for its biases and things that way. One of the things I think that these tools have become way more powerful than they were before, so I completely agree with that, but here's the interesting thing, it's like if you ask it to do, I'm going to have some examples on my website, when you ask it to do scientific diagrams, simple things like the solar system, and things that way it creates wonderful garbage. My favorite example is the year knows throat system, it'll create a beautifully rendered diagram, except create weird labels and place middle year in the middle of your forehead, very surreal, very fun, but also very incorrect. So, I've found actually image generation to be really good way of introducing what LLMs are and how they work to students. You know, there's these ones where I did these experiments with like optical illusions, where I would take something that looked like an illusion, but really wasn't, for instance, the one where the lines, you know, where you have the veins on the lines and one looks longer than the other, and I would actually make one longer and it would say these two are the same length, and even the latest ones make the same mistake, and I think that's very interesting as to what it can get right and what it can't get right. They can imitate visual styles amazingly well, this whole controversy about studio Ghibli that we are seeing is just one example of that, where it nails the style very well, but content is still, like you said, with the infographics and so on, an issue. The interesting one was a bias. So, you know, once this new model came out, I started taking pictures of my family that I had and I said, "Oh, make them and Pixar style." And it was so fun to see those images and I was sharing them with my family and it was great. So, my daughter reaches out to me with some images that she has and she's like, "Papa, you have the paid version, can you make me some?" Like sure. She had one image. So here, when we did my family, we were all, you know, Indian four of us and it came our skin tone and all of that, it did it correctly. My daughter had a photograph, she's Indian, of course, Indian American, and with two of her roommates who are white. When I did the pixel version of that, it made her African American. And that was fascinating. And I kept trying to fix it and it wouldn't because it was almost like the contrast that it was seeing, it pushed one in one direction completely. And there was no way I could fix it. That, I had a conversation with a friend of mine and we went back and looked at another photograph that he had, where it had made me white. Where it was a picture of four of us or three of us and it had basically even out any differences between us. And I think that that's, to me, images become a really wonderful way of, of sort of a microscope into how these large language models work. And that's been sort of fascinating for me. So I have these little tests that I do whenever a new model comes out to see how well it does at certain tasks. And what I'm finding is it is style to a large part over content. So if you want to create an image of a child playing in a park, it will do a pretty good job of that. But if you start getting a little specific about what you want, maybe not. And I think that, that to me, is incredibly fascinating. The optical illusions want a hilarious, because I will have red, massive red circle, a tiny red circle with some arrangement of black dots around it. And it'll say those are both of the same size. And be very confident about it. So one of my colleagues, Melissa Ward, actually asked the, the reasoning model to say, can you, I think these are different in size. The reasoning model took 10 minutes to say, let me compute and calculate the width. And came out, yes, you are right. They are different. They differ by 0.5 pixel. Nobody knows what 0.5 of a pixel is, because a pixel is like a basic unit. But that's, so I think we need to be careful and intentional about how we talk about these things. I think they are incredibly powerful, but they do have issues that when we look at them in educational context, it's really important for us to unpack. You brought up something important, Pugna, which is understanding how the things work. And it's important for people to realize that we don't have yet in terms of most of the mainstream models. Something that really understands what any of the stuff in there is. You know, these are all still, and I strongly recommend if anybody is curious about this, Google just published the Google AI team just published a paper on their latest model, which is Imagine Free. And they're, they're very honest. They are, it is, whether it's not what the panic cannot do and how and why. But the key thing to remember is that these models, they, they, they are all diffusion models of ones or another still to this day. And that means that what you're doing is less a question of saying, hey, give me Wily Coyote in here. Rather, it has a database of show me 10,000 images that are tagged Wily Coyote and show me 10,000 others that are tagged road runner. And then after, well, just what they have left over is what's the minor differences that distinguish an image that says Wily Coyote from one that says road runner. Now, in the original, a diffusion models, you know, you go back to stable diffusion. The other ones have made a big splash when they first came out and so on. Everybody was working with images that were really low resolution. And frankly, just tagged with a few basic labels, some basic labels for style and so on. And not much more. So that was why you were giving them the images, just labels and saying, well, you know, boy, 15 years old standing on dual, just you have to keep it very simple with, you know, just even how you label the style and so on. Nowadays, what's happened is that as LLMs get more powerful and they get integrated with these models and you get integration of image recognition and you get bigger models that can handle larger resolution images with richer descriptions. You're shifting away from those types of very telegraphic descriptions to things where I tell people is trickling at the story about the image you're trying to generate. And what type of story works? You know, that would get deeper into this. But your point for you is very well taken. You need to get people to understand how it's being generated. Because I think sometimes when people say, see this, they attribute more intentionality to the model or they attribute less intentionality in some cases. I kept the view it more as well. Well, how would this happen described in the source images that were used to train it? And you have to understand there is no actual, for the most part, image in the model itself. There's no actual picture of your curioli and another picture of a world run. And there are, there are, my new differences between them. And I think a touchy bit is now doing some interesting things quick, grabbing images from other sources and integrating them with diffusion. But I'm not sure because they haven't discussed this. And there are some things that are strange in terms of how the latest generation of images is. But again, they certainly are important to discuss. And sorry, I'm going to go back to one of my favorite topics playing with the open source models with models that are generally in other countries using other sets of images, like the Chinese models. For instance, you start to get a very different picture. You start to say, well, oh, okay, how is this generating like this? It'll look at images generated like MVDS working on getting some physics into this. So it's not just, you know, the question of the wine glass and so on. It's well, it has no idea what a wine glass one of wine is. And you used to be able to generate images quite happily that you broke the bottom of a glass and they wine just stayed in their magic. But that's changing and video is working on getting more of a physics modeling here. So all sorts of interesting things are happening. And if you play with open source models with Libre models, you get to see some of the different things that, you know, I understand why chat GPT makes a particular type of model for its particular audience, but there's a very wide world out there of what you can do. I was just thinking about these at a lunch conversation. And we were talking about cartoon physics, just like you were saying, Ruben, that, you know, like you might have the coyote with a piano dropped and I mean, it makes an impression on the floor and then it comes out still absolutely whole and everything's good. But it's like you said with the wine bottle that it is basically focusing on what it think it should look like and forgetting the physics like you said behind it. Why I find interesting about the images though is it there seem to have been one of the harder things to do. If we look at generative AI text, it wasn't too bad and obviously it has things going wrong but it's improving very quickly. It went from what I say one to a hundred, it produced, you know, magical music using audio and so on and you know, things like that. But images, it's struggled for so long, you know, to have the right number of fingers to show more contextually correct things that I think would take a massive leap at the moment. One thing I do like though is, what ChatGee PT have done is a loud consistency. You know when you get an image, you ask for an image and you say oh that's not too bad. I just want this changing and you'd ask for something and bring up a complete brand new image and it was kind of no I don't want that. I just wanted this correcting. Now I don't have many people know this but if you actually on the image on ChatGee PT, if you click on the image then it gives you a choice and you can even highlight look just this bit I want changed and it will remain the consistency of that thing but then layer will allow you just to change something small so I kind of is getting to a better place but light is at room and I think still the physics needs to be improved but it is very very life like and I will say beyond education I have had some fun we're doing things like and I was just in South Africa and I took a picture of a war tug just on the side and I asked it to do a professional heart rate image of the war tug and it was fantastic you know the little tiny hairs I mean I wouldn't have known if it was real or not but it was just so perfect that I think it's we're getting to a nicer place which is good to see because I think images have been the slowest in progression. Yeah I was fascinated by playing around with the infographics and yeah the laws of physics all over the place with that and coming back to what Prenure was saying as well with regards to how we can use these images and what's produced to learn about the systems. I'm planning a lesson that's coming up soon with my students and we're again we're looking at forensic linguistics and we're trying to find we've looked at lots of cases in the past that were solved before AI and now we're going back to those cases and we're trying to find real life tools or specialized tools that could have been used in those cases and one of the things I'm asking my students to do is research the tools find out how they work what the input is, what kind of data it goes on, how it comes to the output and I'm getting them to sort of draw a little diagram or a flow chart or an infographic themselves and then I'm trying to get them to do it with chat GPT which is failing, absolutely failing which is a great activity for the students to do but then we're asking why. So the next activity is how did you go from you know this textual description of what you're doing with these specific tools, how is that not translating into the images. So that's something that's coming up that I'm going to be investigating and I find myself quite often now asking the systems why did you make this decision or you know why with the gendered one why did you choose a white male in this example, can you explain your thinking behind it and I'm finding that's a really useful activity for myself and also for my students to understand better how these systems are working. Carolyn I just want to jump in on that because you know I started this conversation a little bit about like my students and as they're preparing to work with you know younger students, primary students and secondary students they're struggling with these systems and one of the areas that they struggle with is this area around kind of using those higher cognitive thinking skills and creativity and their concern of course is that by having these image generators and other areas of AI like the creative podcast with AI type of thing that it's removing the creativity of students that they no longer have to think outside the box or that they have to think creatively they can just plug a little prompt in and then make a few adjustments and then they're good but I think one of the things that you brought up and Helen did too is that you know we need to part of the creativity with AI is digging deeper into what's going on behind the scenes right clicking on certain parts of the image and asking like why did you make this what's going on here or can't you do it this way or did you think about this so I think there is some room for creativity but it's a different maybe kind of creativity than we're used to and it's something that I think a lot of my students are not as comfortable with but I think maybe we're moving into a world where we need to kind of rethink what creativity is or how we you know engage with analysis and those types of skills when using AI so I think you both gave some really good examples of ways that that can happen. So if I can jump in so first of all I think Liz the point you made about creativity I think what'll happen is over time I mean right now those notebook LM podcasts sound terrible to me they were so cool six months ago right the voices sound the same and so on so I think we will redefine a way certain things as being creative creating a podcast from your pros is not going to be regarded as being very creative because the barrier to doing that has completely dropped right and so I think there's going to be a definition of some of these things as some things become easier to do and so I think that's going to be interesting but I want to come back to something LM you talked about where you talked about the fact that that images have been harder to for these large language models to sort of male rather than like text I think I have a difference like a different take on I don't think images have been harder to create I think we it's harder to fool us with images we are highly visual creatures so any anomaly in a visual thing it pops out at us we ask questions of it I think we are easier to be fooled by text so we I think these large language models the way they are creating images the way they are creating text is essentially the same sort of probabilistic sophisticated don't get me wrong it's incredibly sophisticated but it's still this probabilistic models I think we get fooled by text because language is such a human ability that it's still doing the same it's still making similar mistakes and so on but when it comes out as a visual it pops out when like a sixth finger is shown it really makes us uncomfortable and it's it's it's like you know we see it like literally I think that so my my my my thinking about this is which is why visuals become such a great way of probing how these systems work is because the discrepancies become so salient to us while we get lulled by text we are not as discriminating towards text as we are towards images so I think that these systems are getting better in both areas but they suffer from the same flaws in both I don't know what others think of this this idea but yeah they're what you think about audio because that will be more jarring in some ways if that went wrong yes and you can see that when if you like like I said I mean the the example of those notebook lm anytime I hear one of those voices now it makes me cringe music I think is different so I think that that this is tapping into certain foundational things about how our mind works and I think with images we are particularly sensitive compared to text for sure I mean this is just a speculative thought in my head I mean I that's how I've been sort of thinking about it yeah and just to go a different direction slightly as well we're thinking a lot about the generative AI is in creating these images such a chat gpt jennie that that seems to be the best at the moment and a few others then but what about things like napkin AI if people haven't had time to explore that yet napkin AI is one of my favorites as well at the moment for the fact that you can put text in or it can help you with text and it creates a like kind of graphic of a flow chat you know like if you're trying to think okay we've got five steps in using a tubster you know I want a visual of that it gives multiple of them it it's cookie cutter in that it's got so many choices but they are really nice choices I've just helped design because we knew it's not napkin AI for all the rest of the image generation like that and I think it's the only one out there that kind of just does that job but what what people thoughts on that now have you been able to have time to explore it no I've not seen that but I am going to be on this as soon as I land so thank you Helen Hubby this anyone else nobody used to use that one here no I haven't used napkin oh it's it's it's it's very exciting I do a lot presentations as we fly all over the world and whenever I show napkin the excitement I see if we think about AI we're asking like chat GPT Gemini we're asking them to do so many things it's like seeing GP napkin is the the specialist one for creating these flow graphs flow charts whatever you want to call not just flow charts but you know a list of things in multiple different ways and like I said I'd love to hear when we come past each of the next time in the airport the great city thoughts on it yeah you know Helen one of the things you are pointing out whether it's exactly something that to me is very important which is we need to make these things do things that are making things and I don't have a more refined turn for that sorry but you know one of the things is for example people actually show what do you do with this as I from you know reflect on the AI itself well lots of things for instance I work a lot on climate change and what changes in cities what happens if you get a small rainfall a big rainfall a flood etc how does that change well I have good images I have historical data but I can generate images that say take the city and now flood it right up to this level now show me what it would look like if suddenly we had to use boats instead of cars now for instance take a city like Jakarta which where the waters are rising the capital is going to be all from the corridor and they say okay show me wow it evolves if this is the assumption if that's the assumption so it allows me to do a lot of what ifs and that's where I work a lot of people how do you use it to visualize what ifs it also allows me to generate images that are not for the realistic and that's really important because sometimes for the realism can lead people to just a very narrow pan has a hola let's take a step back let's abstract the image a little bit so I could have a series of images for instance you know Notre Dame just reopened after the fire etc and I'd be okay let's assume that the waters rise in the sand what would this look like but I deliberately did not make it a for the realistic rendering of Notre Dame I play with different French comic book band this scene if you want to French term artists what they might look like and this is actually another particular artist I synthetically created by training the AI a new artist that nobody has ever heard of because it doesn't exist in the world outside the AI to generate these images to get people to think so tell me what you think now's this provoking you certain ideas for action how do you have you reflect upon this how do you put your own priorities in those so that's one example and the other thing I tried to do and this also I think helps a lot but soon since I tried to take it so it's not just on the computer screen so I have an image great let's print it out let's plaster it as posters a little bit of a place see what people tell us we have something that we can print out as 3d I've been doing a lot of work lately with 3d AI generation wonderful can we start using this can we make it for toys can we make it to explore ideas can we make it as provocation as critique as political statement in other words get students and really this is students of all ages trust me I've actually worked with pre-case students in getting them to think no I'm not going to ask them to please comment on current sociopolitical trends that doesn't quite work with them but it's only to get stories that matter to say hey holy tell this physically make me a play and we're gonna make the props using this we're making a scenery using that and some of it you're gonna draw using crayons and markers and some of it you're gonna cut up and you're gonna do a collage make a nice mess of this because great masses are a great way to think if you're thinking about how that mess is going to get you somewhere and that's to you want the things that's very attractive about what we can do with this generation off the rule that I love I love what you're saying here one of the rules I have because I create a lot of graphics for my website and other stuff one of the rules I have is that I never create people in them and I always try to go a little abstract for exactly those reasons because those are a more evocative and be prevent me from sort of the biases and things that way that are there but I love love some of the situations that we made there that's really really interesting so I know what I'm gonna be doing as soon as I get home I'm definitely gonna be on that kinei just from playing around very briefly right now it's creating exactly the kind of yeah like you're saying Helen it's charts and it's numbers and it's it's connecting ideas together and I'm certainly going to be using this with my students when we do that lesson where we're looking at how different AI systems work then I'm gonna see if they can use this and generate good images to explain how the systems work so thank you so much for sharing that my pleasure and I will just expect exploring all of them all because I just find the image generation in a very new place very exciting and we'll always continue to test it out because we're always looking for perfection huh we were always interestingly as educators we're always saying okay what does it not do but I I think we should also look at what it can do and how we can use it for so many things in education okay that's my thoughts because I do have my flight just about to take off in a minute so just doing this so Helen building on what it can do and and Ruben and some of your ideas what are the things that I have a lot of fun doing is that I love typography so I make these just for myself typographical designs and then I go to the AI and give it the design and ask it to write a piece of concrete poetry about it and then it'll create something and then we'll go back and fold that and now make it happier or make it like a Zenkhon or do this and that's been served the most fun sort of creative sort of exercises I've had and so not something I'd ever thought that I would spend time doing but it's very personally fulfilling and great fun and I can share some at some point if people are interested but it's actually quite as opened up an aspect of creativity in the visual space for me in ways that I had not thought possible so I think that's very exciting for me personally speak I don't know if it would mean much to anybody else but to me it's been great I would love to see that put your please if you could share that I would really love to take a look I'm working on a presentation about that so hopefully soon yeah you know this has been a great conversation and from my perspective you know there's a multiple ideas I think you know we've all talked about exploration so then there's one more thing I'll just throw it out sort of as a provocation which is that this also opens up the door for us to revisit in a really meaningful way some questions about who's an artist what's a art you know not in a banal okay it's three in the morning and we've all had one drink too many way right but actually in ways that matter that matter because somebody wants to say something about how they live where they live what's happening to them what's not happening to them it has to matter somebody is saying hey I want to express something and before this I didn't have the tools and now I do so how do you use them powerfully but also ethically how do you use you know where the satire come in I think one of the things that was fascinating with the studio Ghibli things was some absolutely retro images were created using it some other people use it neutrally but some people used it for very powerful satire just some very interesting things come again to the table that perhaps had gotten pushed aside have become a little bit more well that's an academic question and so on I think it's very interesting to look at the possibility to see this reenter a conversation in a powerful way across multiple places that aren't just you know traditional education I think so just throw that out as a provocation and a punja I'll also definitely second care I really really want to see this I'm fascinated to see what you can do but it's so that also sounds absolutely great yeah I will it's I mean Ruben just want to underscore what you said like what does it mean to be an artist you know and I do these typographical designs is that just for me I mean I spent hours on them really pepper show them anybody because it's for my pleasure and to have this partner kind of quote unquote looking over my shoulder and engaging in the conversation about you know sort of a verbal response to a visual design it's been it's just been a blast and and that's fascinating that that's an angle that I'd never predicted I would have ever predicted that I would be using AI for right well I would love to help close this out by just saying I thought this was a great conversation and I think that as my student teachers are grappling with AI and how to engage with it with their future students I think we are all living in this space where we know that AI is going to change the workplace and change different jobs we'll have new jobs because of AI and we'll also lose some some occupations as well and that's something I think for my student teachers to keep in mind as they are very nervous about using AI in any way in their classrooms on that they do need to prepare their students for the future and I think this conversation really lends itself well to some areas that they can focus on you know definitely they can focus on the biases but they can also focus on this idea of creativity through AI or me ways to generate old ideas or as Reuben said you know kind of the question of what's an artist and we know that's that's evolved over the years right and what it's considered art and so I think this lends itself to an opportunity for the next generation teachers to lead their future generations into new conversations as well as new ways of looking at cakes through these lenses so I just really appreciate the conversation she knew how to route this. To the flight gate, see you. Hi everyone. All right, see you guys. Hi. Catch you next time. Good night. If you enjoyed Air GPT you'll probably appreciate AI Cafe. To listen click the link in the show notes or look for it on Apple, Spotify and Audible.