The Creative Potential For AI in Education

In this episode of Make:cast, Dale Dougherty is joined by Ken Kahn, author of 'The Learner's Apprentice: AI and the Amplification of Human Creativity,' and Sylvia Martinez, co-author of 'Invent to Learn.' Ken discusses his experience in AI from his high school days to MIT's AI Lab and his journey into integrating AI into children's programming tools. Sylvia sheds light on the importance of empowering children with modern tools and technology. The discussion centers on creative uses of AI in education, including creating web apps and historical dialogues using chatbots, personalized learning, and the impact of AI in the classroom. This insightful conversation aims to inspire educators, parents, and students to adopt AI creatively and constructively.

00:39 Ken Kahn's Journey into AI
02:15 Sylvia Martinez's Background and Publishing Journey
03:17 AI's Role in Education
03:49 Ken's Book and Its Audience
06:44 Practical Applications of AI in the Classroom
08:08 Interactive Learning with AI
19:54 Personalized Learning with AI
22:30 Creative Uses of AI in Education
28:19 Conclusion and Final Thoughts

To learn more about the book, visit https://cmkpress.com/product/learners-apprentice/

https://make.co/make-cast/

Transcript

The Learner's Apprentice: AI and the Amplification of Human Creativity by Ken Kahn

Dale Dougherty: Hi, I'm Dale Dougherty and welcome to Makecast. I am joined today by author Ken Kahn who has written a new book called " The Learner's Apprentice: AI and the Amplification of Human Creativity." I'm really anxious to talk to Ken.

And I'm also joined by Sylvia Martinez, who among her other roles is publishing the book. But she's also well known as an author herself and a maker educator and other things out there. Welcome Ken and Sylvia. Ken, you are joining us from England. Is that right?

Ken Kahn: Yeah, that's right. I moved here about 20 years ago, so enjoying it. It's a nice place.

Dale Dougherty: Were you born in America?

Ken Kahn: Yeah, I grew up in Chicago and lived in California for about 20 years.

Dale Dougherty: And what's your academic background?

Ken Kahn: Yeah, so I got interested in AI even in high school and always wanted to do more with AI. I ended up at the MIT AI Lab for graduate school where I was working with Minsky and Papert.

At the time I was lucky enough to be able to do both AI research as well as join Seymour Papert's Logo group and work with school children. I've been very interested in programming languages for most of my career. So I was always trying to invent new programming languages for children. And the way we got to the point where I am today is after many years of building programming tools for children, and then even programming. tools for children that make it easy for the children to add AI into whatever apps they're building, I started getting interested in how far could you get with ChatGPT or now with Claude or Gemini or any of the others by having a long conversation in which you go back and forth saying, here's a little idea and it makes something and you give it some feedback and you go back and forth.

So that's been,

Dale Dougherty: right. We'll come back to that sort of conversational AI, which I think is fascinating, which obviously has its roots in Socratic dialogues and other things. Sylvia, would you introduce yourself, please? 

Sylvia Martinez: Sure. My name is Sylvia Martinez. Probably this audience knows me best for being the co author of Invent to Learn: Making, Tinkering and Engineering in the Classroom. And that book actually started me on a publishing career because teachers, as Dale, have fantastic ideas that they're doing in the classroom and they kept coming and saying, I could write a book, I could write a book, and I would say if I could write a book, you could write a book, certainly, my background's engineering, what do I know about writing books?

And so we started this publishing company, and Ken's book is the 16th book we've published; a lot of them on making, and when he came to us with this idea of this, these creative ways you could use chatbots, it fits so well into this idea of making, of giving kids the power to make things that are modern and wonderful and connect to their own interests and hopefully also connect to what happens in school.

So you know, it just fit right into this big idea of empowering kids with the tools and technology of the modern world.

Dale Dougherty: That's great. So we're going to be talking about AI's role in education, but especially from a practical point of view, like what can you do with it? I'm excited, Ken, that you have roots in Pappert and Minsky and all those people at MIT. And this is another wave, if you will of some of those fundamental ideas that started there about how kids could use computers and we haven't necessarily seen all of that come to fruition in the way we imagined it then. 

Maybe AI is giving us another shot to change the learning experiences of kids. Tell us why you had the idea for the book, Ken.

Ken Kahn: Yeah, sure. I've observed how many children get very excited about learning computer programming and making little logo or scratch programs and seeing them run, but they often pretty quickly level out, and they often have bigger ambitions but don't really have the technical skills or the design skills to be able to build anything very ambitious, what I discovered two or three years ago, even with the earlier versions of things like ChatGPT is that you could build really quite ambitious computer programs, simulations, games whatever, that take advantage of so many features, especially if you create things for the browser and you don't have to have much technical skill.

You just have to have good critical thinking, good communication skills. Some good design skills to create something that's much more ambitious than what was possible before.

Dale Dougherty: Who would you say your book is aimed at? Educators? 

Ken Kahn: Yeah it's funny I, normally when I'm thinking of the learners themselves, the students, but Sylvia convinced me early on that if I aim the book at teachers, then the parents and the students, the children themselves come along for free. But if you don't address some of the needs of teachers, you've got a much smaller audience.

And I'm really, excited about the idea of enabling, children to really be creative and create new kinds of things using chatbots. 

Dale Dougherty: Sylvia, I think you're dead on. There's a secondary audience of parents and tertiary audience of kids themselves. But teachers themselves are amplifiers.

It's, a way to, reach lots of people that may not have the motivation or interest initially to connect into this.

Sylvia Martinez: And, I think we've all seen how this topic has exploded on, social media and in articles and in books, but a lot of the uses are very utilitarian, how to make your grade book more efficient, how to write lesson plans, how to grade kids papers, how kids can use AI to write assignments, do assignments that are just old fashioned, work that they've always done.

And when Ken came to and both Gary and I, saw this early on came to us with this idea, it was so relevant for this practical aspect, what kids could do in the classroom, but also that the way he was doing it solved a lot of the problems that teachers were having with chatbots, the security issues, the sharing issues, having it write code that works in the browser is a tiny little twist that unlocks a lot of opportunity for use in schools.

Dale Dougherty: Yeah. That's great. Why don't you talk about that a bit more, Ken? 

Ken Kahn: So maybe I'll start with the first example from the book, which is just a very toy example, but one that illustrates a little bit about the process. With a very simple start where you just say, "please, can you make me a web app that generates random words?" chatbot, quite quickly and almost always reliably creates an app with a little button. Every time you click it, a different word. And then I said to it, could you have the app speak the word? And it knows all about the APIs within the browser to do text to speech and so on, and very quickly, it updates it. . If I wanted to, I could say, "Oh I wanted to have an interface where the user could decide how many syllables and how many words should be generated at once."

And, most of the time it does it fine, but you also have to acquire some troubleshooting skills. But most of the troubleshooting skills is to basically communicate what's going wrong and it'll try to fix it. And you could be very creative. I, as a playful thing, I said to it after it was working, I said " could you have it speak the Gettysburg Address where every third word is a random made up word?"

And did it fine. In the book, there's, a range of, scientific simulations, ambitious games, mathematical explorations. There's quite a lot. And there's, as we'll talk about maybe in a minute not just creating apps, but having very rich kinds of conversations or simulated panels or debates or adventures and stuff.

Dale Dougherty: Talk about that. I thought that was pretty interesting, that would maybe catch the eye of humanities teachers

Ken Kahn: Yes,

Dale Dougherty: Here's a historical character, let's have a conversation with them, or here's some period in time, and we want to talk to people who live in that period about their experiences.

Ken Kahn: Right, so there's at least two different styles of that kind of thing, one of which is where you simply say oh, I'll give you a science one, but in a minute I'll give you a more humanities one, where you say "please simulate a conversation between Aristotle and Galileo, add to the, every time they have an exchange, So it starts off, and they're very polite to each other, and they're talking about, what's at the center of the solar system and things, and then I start to ask him about how objects fall, and Aristotle tells his incorrect theory, and Galileo politely tries to correct him, and then, We talk more about these things back and forth.

And then I said to, in the middle of the conversation, I just simply said, "ChatGPT, could you bring Isaac Newton into the conversation? And then he ends up explaining how gravity isn't just something on the earth, but it explains how, why the planets orbit around the sun. So that's one example, but give, I'll give a more humanities example, which was where I said "I wanted to have a text based adventure where I'm visiting ancient Athens."

And it starts up with a description. You've just come to there and you could see the Parthenon on top of the Acropolis and over there's the Agora. And, what do you want to do? And you might say, Oh, go over to the Agora. And then it says you see a group of philosophers talking and you say, Oh, I want to join them.

And then it says, and one of them is Socrates and you can, do you want to ask him anything? And, you say. Start having a little philosophical discussion with Socrates and then if you want, you could go off to, the theater and talk to the playwrights or go up to the Acropolis and learn about their religion.

So it's, but it's a very immersive open ended kind of adventure that unfolds as you're having it. It's not one that's been pre programmed. 

You make the point that AI is not at the point where it can create, graphical video games automatically in the same way, but it can do text based adventure games. As a matter of fact,

Dale Dougherty: choices and, move on to,

Sylvia Martinez: It may be coming 

Dale Dougherty: soon maybe

Ken Kahn: Yes. I did experiment with asking it during the adventure. "Could you make illustrations every so often?" And it made one I made, I had an adventure witnessing assassination of Julius Caesar and it made three or four very nice illustrations where this high school student is visiting and there's the people and, he could talk to the either, the politicians or the man on the street or it was a fun adventure, but it was also enhanced by the fact that it could generate three or four illustrations to go with the adventure.

Sylvia Martinez: And, another example in the book is having a text based adventure in another language. If you were learning French, you could ask it, could you do a text based adventure? Keep the language simple. The chatbot will level up and level down and help you with learning a language in real time, with an interactive game. It's quite amazing what the chatbots know: everything from how to make a button to how to make the browser speak, and again, this is all in the browser, so it's school safe, school friendly, no new software, no downloads. If your kids are using a browser, you can do this. 

Dale Dougherty: We tend to talk about ChatGPT a lot, but there's obviously a number of different, systems out there. You will get different answers from different systems. And it's interesting to try them to see how they might work differently, operate differently.

Ken Kahn: Yeah, that's right. And also even the same day, the same one will respond differently and which one is better at programming, which one has a nicer interface. They keep leapfrogging each other every few months, but it's definitely the case that the top three or four are all more than competent to do all the examples in the book. Claude, ChatGPT, Gemini, I haven't really used grok munch, but I believe it's equally smart.

Dale Dougherty: So let's talk about from a teacher's point of view, standard high school teacher. Let's take that, let's not worry about the subject matter so much, my resources to date have been websites and textbooks probably, and other things you might think of.

What I find fascinating is, this is not like a teacher sitting up at the projecting a chatbot session. They could do that, but the power comes from the student interacting and feeling like they're part of that creative process to come up with, if it's not Athens, it's Rome.

If it's not Rome, it's Carthage or, whatever it is that they're, studying or interested in. How does a teacher make that transition in a sense to giving the power to the kids to create?

Ken Kahn: Yeah. So I think part of it is to, I'm very big believer in sort of project oriented learning and giving the students a lot of freedom to choose a project they want. But one thing that's nice about working with chatbots is it could be a very tiny project that's 10 or 15 minutes, but still leads to something interesting, or it could be, one that takes a few days of effort, or maybe, spread out over maybe weeks.

And my view of how teachers should view what the students do, is that it's the product, the illustrated story, the web app, the adventure, the conversation. Those are an important part, but maybe equally important is that I think the students should report about the whole process and reflect on the process.

Because first off, there's this risk right now that the chatbot will do too much of the work and the student will hand in something that's very impressive, but didn't really do much thinking or learning. I think you could avoid that if you're asking the student not just to provide the product, but also describe the process of creating it and maybe provide the logs of the conversation and something in the extreme. 

Dale Dougherty: Yeah. So I think that is one of the interesting things, how does a teacher assess a student's engagement? And, what are the kinds of directions they're giving to kids? One of the directions is, as you point out, is you know a prompt is, it can be simple to start with and you iterate.

That's one of the key concepts is just iterate over the thing. Don't just say I told it this and I got that end of story, right? You may improve your prompt and you may get a different result and so on and so forth. Having ideas in your head to begin with is important.

It's not entertainment. It's not just show me something that makes me laugh. 

Ken Kahn: Yeah, that's right. Much of the material of the book, I was trying very hard to role play students of different ages and background. In some cases, even the what the goal is becomes a co design rather than the student having a clear idea of what they wanted. One example, you could start with very simple prompts, and much of the book is about simple prompts.

But you could also set up, especially a teacher could set up a context where it's more like a paragraph. prompt that says, what age of the students, what kind of language to use, what tone to use, ask them if they've got an idea and if they don't, ask them about their interests and make a few suggestions and, you can tell it not to do too much to keep asking the student questions. An example of that was one case where I role played somebody that I felt very much like I was when I was 12 years old, and, I didn't have an idea of what to do, and then it said what are your interests, and I said I really like astronomy, and it says how about, an astronomy fact thing, or a simple simulation, or whatever, and I said, oh, a simulation sounds good, and then it built a very simple simulation with just, a sun and one planet going around in a circle.

And then I said, I've read about exoplanets. Could we make it, simulate the exoplanet system called TRAPPIST 1 that I'd read about? And it took, in this case a fair amount of back and forth, but we ended up with, you can see all the seven planets orbiting around the sun. It's a 3D view of it. The sun is actually the right red color, and it's lighting up the planets as they move around. There was a really funny thing that happened halfway through. At one point when I finally, said, use this data from this NASA site, all of a sudden the planets stopped moving. I said, the planets aren't moving, and you'll find chatbots are very polite, and they think, oh, I'm so sorry, I'll try to fix it, and then it didn't get any better, and I said, it's still not moving and, but then after a while, it said maybe the problem has to do with the time factor. And when we changed the time factor hugely, like 10, 000 times, everything was working beautifully.

So what had happened was, it assumed that if it takes the real planet nine days to go around the sun, in the simulation it should take nine days to go around the sun once. And it was probably moving one pixel every five or ten minutes, and I thought it wasn't moving at all.

Dale Dougherty: Yeah, interesting. Prompt construction, in a way, is what you're talking about. Now, it made me think of show your prompts are going to be a phrase at some point of what did you do to create that? Not, almost less interest in what the product was, but than how got there.

Sylvia Martinez: One of the big ideas is giving kids access to these tools, not just on AI week or, hour of AI or something, but giving them access to tools so that they can think of how they might be useful for things that they care about, when it's your work, when it's your idea And you say, Oh, maybe I can make a little widget that does that.

And if I can make a little widget, I can do something else. Now, today, there are a lot of schools that have either banned AI, or it's too expensive, or they don't have access. But this is going to change. So the teachers who are experimenting on this ragged edge, the jagged edge of how we use AI in the classroom, I think giving them ideas for how students can use AI to create simulations that help them understand the biology concepts, that can create models of what they're doing in math class, that can create games that support language learning. All of these things are going to be taken by teachers and students, and they're going to experiment and create thousands of new examples.

That's our hope for the book is that these little ideas are seeds that we're planting out there in these, every once these few schools that have access and that those seeds will grow and be examples for people as more and more students get access. You can imagine that a lot of people have this very dystopian view of students just taking the teacher's assignment, putting it into AI, generating a five paragraph essay about the causes of the civil war, turn it into a teacher.

The teacher feeds it into AI and grades it. And it's like. Who's learning? This is not what we're talking about. This is not a vision for school that is fun and engaging. 

Dale Dougherty: That's a kind of arms race, isn't it?

Sylvia Martinez: Oh

Dale Dougherty: doesn't lead to any good outcomes. It's just, but let's talk about personalized learning, because I think there's, you mentioned that in the book a bit, and it's one of the paradigms that has been out there for a while. But, it was an ed tech goal in so many ways, but it really wasn't that smart. It wasn't that individualized. A. I. Checks all those boxes. I think it's a question, how do you use it? But I'm intrigued by, I think it's a, you were talking about is ways to I almost know who the student is, and what you like. It's another thing, typical education doesn't do a lot of inquiry into who are you? What do you do? What do you like? And it would be interesting to have this kind of persona that you could query about what is Ken like? What is Sylvia like? And what do you want to do? And what have you done recently? And learn from that. So just talk about it personalized learning a bit.

Ken Kahn: Sure. So the simplest form of it is again that you just start with a sentence or two to tell the chatbot the context that you're interacting with a student and try to find out about their interests and their background knowledge and maybe what language level or complexity to use.

Then it'll start off with just a nice conversation and it's very good at being encouraging and friendly and so that, and then it could proceed, I've seen it -- very different response if you say you're 17 years old, or if you say you're 10 years old, or if you say that you're really interested in poetry, or if you're really interested in science, it will give a very different kind of, get a different kind of interaction. Technically it's quite feasible for it to remember these things between different sessions. But at least right now, you could always go back to the same session and keep adding to it, making something new. Because within the session, it does remember what's happened before, the student's answers to questions. But there is this limitation right now that if you start a new project with a new session, it generally won't remember anything or if you're enabling a memory feature, it'll just remember a few sentences about you, but it's definitely the future where it's going to know more and more about each user and personalize,

Sylvia Martinez: And also for accessibility and personalization these chatbots will speak. The spoken interface is getting much, much better. I'm using them on mobile devices. Any kind of accessibility that, screen readers, things like that, the chatbots all work with those. There's all these kinds of personalization things that the chatbots offer and are only going to get better at.

Dale Dougherty: Yeah. Plus you know, we were using the product as sort of conversation, but they can also create images, videos, music, other things that are also really interesting outputs and forms of expression. And I think that's one of the things that you touch on in the book a lot. I can imagine school administrators coming up with a hundred reasons why they could use AI. And one of them wouldn't be, we could get our kids to create, plays and songs and, videos and all these fun things. I really love to see that space brought out more, that idea that kids can use this as a creative companion and do things that even sometimes they don't have the skill level yet to imagine how to do it.

Sylvia Martinez: And for different kids, different things are fun. One kid might want to write a song and the next kid might want to make a Roman numeral calculator. 

Dale Dougherty: Yeah,

Sylvia Martinez: I was a different kind of kid when I was in school. I wouldn't have wanted to make a song. I would have wanted to do some geometry thing.

 The chatbot can sit there and go, "Hey, that's a great idea. Let's do that." The chat bot is really this sort of patient interested conversationalist. And, it's not that we want kids to get into deep relationships with chatbots, obviously not. Which is why I think the teacher is the central person in this conversation of how we get from here to there. The teacher creates the community of the classroom where kids can be interested in different things, where kids can try different things without worrying about, oh, that didn't work, so let's try again.

In a community of learners, you can experiment, try new things, try different things and it's all great.

Dale Dougherty: And, might also add that a parent could play the similar role in a family, and saying, hey, I want to work with my kids on this, not just turn them loose. 

Ken Kahn: I wanted to follow up on, the thing about songs and images and stuff, because there's one example I'm proud of; also I find it a very promising, exciting example where it starts off, you just ask the chatbot, " Could you create the lyrics to a kind of lighthearted song about the Euclid's proof that there's an infinite number of prime numbers."

so connecting it up with subject matter and mathematics, and it writes a really good, song that captures the intuitions of the proof. And then you take the lyrics to another generative AI program and say, generate a song with these lyrics that's a maybe the style of a musical for children.

Dale: Then you go back to a chatbot and say, okay, "for each stanza, what would be a good illustration for each one?" And then you've got a file with a song, and you've got maybe five And then you say, we'll put it all together in an app that will play the song and slowly change the images.

And you get something that's, combining all these different media to something that at the same time happens to actually connect pretty well with some advanced mathematical ideas. So the possibilities are really wide open.

Sylvia Martinez: And when Ken says that these examples are in the book, he does mean they're in the book. There's hundreds of examples in the book, not the entire conversation. Those are all online. So every, little conversations that's excerpted in the book and with screenshots of the simulations that were built is fully online in these logs that are connected to the content of the book.

Someone wants to dive deeper into how exactly Ken built this simulation of ants following pheromone trails. That's, two pages in the book, but it's all online, including the code and the conversation and the links to the apps. So all of it is available for teachers Or students or parents to really dive deep.

Dale Dougherty: It's the old view source and the web browser kind of thing, that you can see the work. You can learn from, studying other prompts. And I think that's one of the real interesting trajectories for students is, like asking a good question is an art and, it isn't just a few words. You may mean a lot more than what you're asking.

And so a prompt has to be a lot more specific, but you can iterate to get that. I've run across people who shared, three or four written pages of a prompt, and it's, Oh, wow. I was thinking it was a sentence. No, this is like an essay in itself.

Sylvia Martinez: So I've done something where I've done this iteration and then I chatbot "what's the prompt?" that will create the same thing the next time I want to do this and it like. Very helpfully goes blah, blah, blah. Here you go. And then I cut and paste it.

Dale Dougherty: I think they also, at least some of their setup that way it will show you how they interpret your thing. And that kind of gives you feedback on, maybe how they filled in the blanks, 

Sylvia Martinez: That's another part of this personalization. You can read the code, you can read the reasoning, you can ask it more questions. You can tell it to shut up. This is completely personalized to you and the chatbot will learn in its session, what you like and what you want.

Dale Dougherty: Good. I really thank you for spending time with me. I'm grateful for the book in part because I think these tools need to get in the hands of students and we have so many challenges, especially in public education today.

Kids aren't necessarily getting the kind of education they need to adapt and be participants in the future, and this is going to be the future. And there's no reason they have to wait till college to start doing this. 

Kids are already using it, but they're not necessarily getting guidance in what's possible. And I think that's what's really important. Not just off on your own. 

Sylvia Martinez: We certainly hope this helps teachers make the case for creative, constructive use of AI in the classroom

Dale Dougherty: Absolutely. I didn't get to the tinkering chapter and some other things that are in the book. There's a lot more in it than we covered in this conversation. That's why it's a book. But Ken, thank you for writing it. Thank you for joining me today. And Sylvia, pleasure to see you.

Ken Kahn: Thank you. I enjoyed this. 

Sylvia Martinez: Thanks.

Dale.