Event Replay: California Teacher of the Year Uses AI to Make Good Teaching Even Better

Welcome, everyone. I'm Natalie Cone, head of the OpenAI Forum, the expert community hosting our conversation this evening. I see so many familiar faces. Grant, so good to see you from Miami-Dade Community College.

Tina, welcome from UCLA, a professor. Such a wonderful delight to be hosting you all. In the forum, we spotlight discussions that reveal how AI is helping people tackle hard problems and we share cutting-edge research that deepens our understanding of this unprecedented technology societal impact.

AI is an innovation on the scale of electricity. It's transforming how we live, work, and connect. OpenAI's mission is to ensure that as AI advances, it benefits everyone.

We build AI to help people solve the toughest challenges because solving hard problems creates the greatest benefits, driving scientific discoveries, improving health care and education, and boosting productivity across the world.

The most important characteristic of our community is that we give the experts the platform to teach us how best to leverage open AI technology in their domain.

Tonight, we're hosting a gem in his field, Casey Cuny. Casey's approach proves that AI literacy is not about replacing good teaching, it's about enhancing it.

With more than two decades of classroom and leadership experience, Casey frames AI as both a challenge and an opportunity for educators. His core message is simple yet powerful. Good teaching is still good teaching. AI just changes the tools.

Casey Cuny is the 2024 California Teacher of the Year.

Cuny teaches 10th grade Honors English and Senior Mythology and Folklore at Valencia High School in Santa Clarita, California. He serves as English Department Chair and the school's Instructional Coach.

He also teaches at National University in the Masters of Education program, working with teachers from all over the country. He was recently awarded the 2025 National University Academic Excellence Delphi Award.

He's also currently working with ASU to develop their new California Teacher Credential Program. ASU, by the way, is one of the very first universities to grant Chat GPT to everybody on campus, from faculty, teachers, and students.

He drives to create an inclusive environment for his students and to champion their growth as individuals. He uses the synergy of social, emotional, and academic intelligence to push students to levels of rigor not possible without his concerted approach. Please help me welcome our very special guest, Casey to the forum stage.

Wow, thank you, Natalie. That was an incredible introduction. I really, really appreciate that. So I was gonna introduce myself, but you just did it. So that was better than I could have done. So thank you. So let's dive in.

Thank you all for joining this. And I know we have so many people in here, some people that I know, that we have state teachers of the year from all over the country and the territories in here, and as well as a lot of business leaders. And the AI understanding and expertise is, we have a wide, wide range in here.

So I'm gonna try to touch on all of it. I'm probably gonna start with some of the more basic stuff that I'm doing and teacher workflow, but then I really wanna get up to kind of some of the more complex stuff and some student facing stuff that I'm using.

lately and that I've been stealing from other great teachers as well. So let's dive in.

So yeah, Natalie introduced me. I don't need to go through much more. There's me in my classroom. One day I decided to flip it and I was Mr. CUNY GPT and they had to ask me questions.

So yeah, I've been teaching for 23 years. I'm passionate about education. I'm passionate about helping my students succeed and about helping teachers. So I'll kind of skip over my introduction.

But this is my family. This is the love of my life and my best friend in the middle there, Dawn. And these are our two boys. Dylan on the right just graduated from Seton Hall University with a business degree in sports management. And then Landon on the left there is a sophomore at San Diego State.

And I'll probably be talking a little bit about their AI usage throughout this as I've been using them as use cases over these last three, almost four years.

So this is my family, and we live here in Santa Clarita, California, which is Los Angeles County, just a little bit north of LA in a suburb here.

And yes, I teach at Valencia High School, which is a big comprehensive high school. I started my career originally at Van Nuys High in Van Nuys, California, and eventually moved up here to Santa Clarita and now teaching my community, which has been an incredible, incredible thing.

So I thought, what better way to introduce me than to show you what my students say about me? So for the last five years, I've been asking my students, what three words describe this class? And then this is five years worth of words.

So I'm very proud that fun is number one. I'm also very proud that number two is educational, followed by challenging, and then entertaining and enjoyable. And to me, this is what a class should be, right? I like to hold very high expectations, but I like to make it a lot of fun as we strive for.

those challenges. And so this is what my students say about me. There's some very funny ones in there. The word rad. I tried to I tried to bring back rad a couple years ago. I was like, you guys should bring that back. It's things are so rad, and it didn't stick.

When I talk about AI, throughout the presentation, I'm talking about generative AI, because I think the term AI gets thrown around a lot. And there are all these AI apps that are out there. And so I've talked to people like, Oh, yeah, I use AI all the time. And I use blank, and it's like an app. And so when I say AI, I'm talking about generative AI. And specifically tonight, really talking about chat GPT.

So if you do, if you have your chat GPT account, please log in, have it handy, I might be throwing some stuff at you to kind of test out.

So I want to talk about what my thesis for this entire talk is, is really two things that as educators, we must become AI literate.

don't see it as an option. I think when internet first came out, you know, I'm old enough that I was teaching when that happened. There were some people that were never needed to really get on the internet. It was gonna be okay. I don't feel that way anymore with AI. It's moving so quickly and I have massive concerns, of course, I think along with everybody, but the reality is I have 36 kids sitting in my class tomorrow morning 830 a.m. that I have to get ready for the world that they will walk into. So I've got to become literate myself so that I know how to help them.

And then number two, good teaching will always be good teaching. AI is incredible. It's this new amazing tool. It's this new incredible thing, but the science on teaching and the science on good pedagogy have not changed and I believe that we can actually use AI to amplify and streamline and really enhance all that stuff that we know works in the classroom.

So what do I mean by AI literacy? Well, I think when it first came out, a lot of us thought it was like a better Google, right? And I think you actually see that in the new data that just was dropped on the usage of AI over the last three years, how so many people still use it to just look up a piece of information. And I think a lot of us thought of it that way initially, especially. But this is completely different. There is a learning curve. It takes time to learn how to use and practice.

And so I'm gonna try to talk about a bunch of use cases tonight. I will show a little bit about teacher work type stuff, but I am gonna try to move pretty quickly to the more advanced stuff and talk more about learning and about student facing stuff. But in terms of a use case, you know, I think we've all had that experience where, right, where we we get everyone back to school and we get this the back-to-school sign-in sheet. I don't know, does anyone still do this? I still do this. And so I've got all my parents with their email addresses on the sign-in sheet and then after all the parents leave and I finally get home late at night I realize oh wow now I have to type all those emails into some sort of you know gmail group or whatnot.

Well let me show you a little hack for this just to illustrate how we can as teachers dramatically improve our workflow. So I'm gonna take a picture of this. I'm gonna take a picture of this handwritten sign-in sheet okay and then I'm gonna drop it over here on my computer and I'm gonna head over to chat GPT and I'm gonna drop this in to chat GPT. There it is and I'm gonna drop in. I kind of pre-typed my so that I wouldn't get so nervous trying to type in front of you all but this is my prompt.

Extract these names and email addresses and create a table for me to download. Do your best as I'll be able to double check the names

This is interesting. I found that if you if you don't say do your best, it's OK. It will sometimes come back and say, oh, I don't know, I might not be able to get the names exactly right. Getting handwriting. But if you tell it, it's OK. Do your best. I'll double check. It doesn't have a problem. So let's see how it does here. And I want this as a Excel spreadsheet essentially or something I can download as a table and it's analyzing it right now. We've got the first parent to sign up was Bucky Barnes. We've got Peter Parker. He's kind of young to be a parent, but OK. And here we go. So I've got my sign-in sheet. Looks like it may have gotten. Oh, yeah, there we go. It's got all of them. So I've got all I've got this. I can now download this as an Excel spreadsheet. But hold on. I want to do something more. And there's the also the link right here to download it. But I want to email these parents. So I'm going to say. Great.

Now give me a comma separated list of the emails so that I can copy and paste that right into my email. All right, there we go. Winter Soldier at Gmail. I've got all my emails. I can copy that. And then now I'm gonna give it my syllabus for Honors 10.

And I'm gonna ask it to read my syllabus and write me a welcome email to my parents using the ideas and themes from my syllabus, express that I believe in their students and strive to give them a great year. Here we go. There's my welcome email. Now, am I gonna just copy and paste this exactly? No, of course not.

And I think that's why I hear some people say, oh, I asked AI to use it this one time and I hated what it gave me. And well, it's technology and we'll talk about that in a little bit, but I'll probably use this as a very good first draft. And then I'm gonna, and it is, it looks really good. It can make it.

order for me if I wanted it could do it a little bit more parent friendly I could actually change the tone if I wanted but I could you know that looks pretty good to me
I'm gonna copy and paste that. I'm gonna copy these emails right here and that took about a minute and a half that took about a minute and a half from that now
when I'm in my classroom I always ask my students to give me a mind-blown moment at any moment that their minds are blown and I wish I could see you all because I hope that maybe there was some of that that just went on around the room in terms of like I didn't know that you could take a picture and yeah you can take a picture of anything and ask it to extract any of the text or you gotta take a picture of your living room and say help me redecorate this entire living room so yeah I can take pictures so we can use it as teachers to unpack here's the the metaphor here and do so much around our classroom in terms of that work that we have to do as teachers all that stuff that takes us away from students and AI is ground, it's a game changer when it comes to that stuff. In fact, I want to show you a little something here. We're going to go over to Chachuiti. I'm going to start a new convo, and I'm going to start something while we're talking. I'm going to go into deep research.

I don't know if you've seen deep research, but you click the little plus sign, you go to deep research, and it says I'm in deep research now. I'm going to take my learning outcomes for a unit I'm about to do on growth mindset and I'm going to drop these in here. Here are my student learning outcomes for an upcoming unit on growth mindset for a 10th grade honors English class.

Run a research report. Find me relevant sources, TED talks, videos, podcasts, anything. Then I gave it my learning outcomes and I want to get off of that. There we go. Because you can link it to your Gmail, you could link it to your drive, you could do a deep research on all this stuff in your drive. But here we go. Now it's going to ask me a couple of questions. Deep research.

What's the length of this unit? So I'm going to go, let's say, three weeks. Shift-Enter, a little hack on Chat GPT. If you don't want to accidentally hit Enter and then send that and go, oh, I didn't mean to hit Enter, it's Shift-Enter.

Two, are there any required or preferred anchor text or authors? Not necessarily. Maybe Angela Duckworth would be great. And maybe Bjork and Bjork. From UCLA. Okay.

Number three, technology resources. They've got Chromebooks. And would you prefer assessments mixed? And I'm going to say no, just to get this started. All right.

Now it's going to go, I'm going to leave this and go back to the presentation because this will take about 15 minutes. It's going to start researching. It's going to go look stuff up. So we'll come back to this.

But I do want to illustrate how you can use deep research. We might even start another one over here on, no, I'll come back to that. We'll come back to that. It's starting to research. All right. So I think it's going to blow your minds when you see what that can do.

So really quick on, in terms of using AI, a lot of people don't realize this, but you should refresh your conversations often. In fact, after 20 to 25 prompts, the AI degrades. Now this was in Chat GPT-4. So with five, it's probably a little bit better, but you should start a new conversation anytime you are changing the topic. Anytime you're changing the conversation and going to ask about something different, start a new conversation. When in doubt, start a new conversation.

Because what happens is every time you continue that conversation, it's spinning the entire thing back into the AI, into the LLM. And so after a while, once you get to 25 to 30, it becomes so much that the AI will start to shortcut in order to be faster.

for you and to give you great, fast responses, it may not read everything that came before it at that point and you'll start to get some bad answers. So, and be very specific. Don't say, help me with this email. Say, rewrite this email to sound more confident and concise. Vocabulary matters.

This sentence, this app helps students with writing. If I say, make this sentence better, I will get very different output than if I say, optimize this sentence. If you give it an old lesson plan and say, make this better versus optimize this lesson plan for a 10th grade class, you get very different output. And I've had a lot of friends of mine not realize this.

And I think we have to realize is that this is an artificial intelligence. So it isn't real, obviously it's artificial, but it is an intelligence. So the more intelligent you are with it, the better responses and the better product you will get.

So right here, when we said make it better, it improved it. It changed helps to supports, improving their writing skills.

OK, it added something. If I say optimize, I like enhances. I think that is better and provides targeted feedback and guidance. Now, I might decide I don't need to go all the way to optimize, right?

But that's why I'm doing an AI word of the week with my students now, words like evaluate. So don't ask AI, what do you think of my idea? Is my idea a good idea? Because it'll just say, oh, it's a great idea, right? It should be evaluate my idea. Give me pros and cons. What are alternate solutions to this problem?

I like the word robust. Give me a robust plan for this. Or the word curate, which, by the way, I learned from AI. I asked AI, can you get me a bunch of sources for this topic? And it wrote back and it said, would you like me to curate a list for you? And I was like, yeah, curate that. And now I'm using the word curate in my regular vocabulary all the time.

I think my good buddy Jeff is finding me.

and knowing how often I use the word curate.

But for those people out there, I know there's people that think that AI is gonna make us, it's like the movie WALL-E, and we're just gonna be dumb and floating around. And I just don't see that. Because when you talk to AI, it often will reframe what you're saying in more precise vocabulary, which then you're reading and taking in and learning.

And I'm finding that my vocabulary is actually improving by using AI. But what about this? What about the fact that it can be wrong? And what about the fact that it can hallucinate? Hallucinating meaning make stuff up, completely make stuff up. And the answer is yes, it can. Now I asked ChachuPT to make me an image of a hallucinating AI, and that's actually pretty freaky. Because it can though.

But you know what though, when I was back in high school, and not to throw any of my amazing high school teachers under the bus, but I had several high school teachers who hallucinated quite often. So who were wrong about things.

that I later discovered. So yes, it can be wrong. No one's saying it can't be wrong. But I think the reality is oftentimes it's about how we prompt it.

You know, I read a really interesting article. It said the article was saying how AI was not that great because they asked it, it wasn't even a complete sentence. It said, write me a lesson plan for science, boom. I guess that's a complete sentence, but, and it was not happy with the output.

Now I looked at that and said, well, that was the prompt. I mean, the prompt should have been what's the grade level, what standards do you want to focus on? What are your learning outcomes? And if you give it that kind of prompt, then you are going to get much better output.

And you know what? It is going to be wrong from time to time. You can ask it to double check its sources. And if you find yourself, and I've saw somewhere online, they called this the death spiral, where you can get so frustrated in some conversation you're having with Chachapiti because it's wrong. And you're going, no, that's not right. And you can get frustrated, stop right there and start again.

a new conversation and just copy your prompt back in and start over. Because sometimes it can just get a little screwed up, maybe because of the history of that conversation, and it can start messing up. So start a new conversation and it usually corrects it. So it can be wrong. You have to double check and you have to constantly look just like we would with Google. I'm sure we've all had experiences where you try to find something by searching through Google and you couldn't find what you're looking for and you've gone all the way to the second page and you still can't find what you're looking for. So it's technology. It's going to mess up here and there, but it is getting better and better. And often it's how you prompt it. Google created the age of search, but Chachapiti has brought us the age of ask and discuss. And this is different. And our minds are still in the age of search. You know, I think oftentimes when I go to Chachapiti, especially in the early part of it, I was still searching. I wasn't discussing. And so it is a behavior.

behavioral shift in a lot of ways that we have to think about in terms of how we're gonna use this technology to really maximize its benefit. And that is asking and discussing. We have to think of ChatGPT and AI as a thought partner, as a thought partner that is a co-professional that's at your side that can help you out.

And this is when I get excited though about what AI could do, because I think about the kids that we all have who are sitting at home, who do not have supports, who do not have a parent to look over their shoulder and help them with that homework, and who are struggling as a result. But AI could become an engine for equity if we teach kids how to use it.

I initially, actually, when AI first came out, I started calling it an engine for equity because I saw the benefits and the potential of where this could go. I've almost pulled back from it though, because I do worry that if we don't teach kids AI literacy.

and don't teach kids how to use it, that this will do the opposite and actually exacerbate existing inequities on a scale like we've never seen. So I feel it is almost a moral imperative to teach my kids how to use this. And it's so important to realize that they don't know how to use it. Because using AI requires a different kind of thinking. Oh, there we go, so we have to help our kids. Because good teaching is still good teaching.

And I wanted to talk about that now, about learning with AI. So I decided back in January as an experiment to try to learn something really difficult that I didn't know at all with only AI. Because I wanted to experience what it was like to try to learn with AI, so I could kind of unpack this and see how I could use this with my students.

So I asked Chachapiti to give me a 30-day course to teach me Python coding. Well, by day 17, I was writing programs.

and it has absolutely blown my mind. And now I'm building an app. But what I discovered really in the process of doing that was how engaged I had to be, how engaged of a learner I had to be. I had to constantly question. I had to always be aware of my own thinking about what it was that I didn't understand so that I could ask the right question to fill that gap in. It was much more engaging than I would have ever predicted until I had tried it out for myself.

Because learning, and do we really prepare kids in that way? Right? I think a lot of kids go to school and a lot of it's sit and get. You know, a lot of it, it might be modeled, something's modeled, we're pouring information into their heads, and then they show on a test whether they learned it. Now, years ago, I stopped asking the question, are there any questions? Because to me, that's a yes or no question. And it implies that if you have to raise your hand,

hand and say, yes, that there's something wrong with you. And so nobody raises their hand. And then the teacher goes, okay, great. Everyone understood it. And then they look at the test two days later and go, nobody understood it. So I stopped doing that years ago. And I use a nonverbal and I say, what are your questions? Let's go. Let me hear your questions. What are your questions? But even then it is like getting blood from the stone. It is so difficult.

I once had a total game of chicken with my seniors where I said, we will not go to lunch until somebody asks me a question. I need a question. And they were looking at me like I was crazy, which I probably am a little bit, but it went right to the end. It went right down to the bell. And finally a kid was like, fine, fine. I'll ask you a question.

And then it takes some training for me to get them to ask questions. And I'm sure everybody in here knows about neuroplasticity, about the power of habit, about growth mindset and the way the brain works. And it.

if we're not using neural pathways often, they're weak or underdeveloped. And so I do worry in a sense about the world that these kids are gonna walk into that's going to require that metacognition and that inquiry when we're not really preparing them for it, when we're not really getting kids asking difficult, thought-provoking questions often in class. Because when we're learning with AI, you have to constantly be aware of your understanding and your knowledge so that you can figure out the gaps where you're not quite getting.

So then you can turn that into a sharp, productive question in order to get to the information. But here's the crazy thing. If you start to master this, it's not just learning with AI, it's that you could literally learn anything with AI. And so this is the cycle that I see. It's a constant thinking.

So I know that, again, there are people that think that AI is making you dumb, and there was the MIT study about kids who used it to write their essay.

it was very little cognitive processing. And I read that study and I thought it was really interesting and good. And what it taught me was, don't let kids use AI to write their essays. You know, it didn't teach me that I shouldn't put kids on AI. It showed me a way not to use AI, which makes sense to me.

That if you're asking kids the AI just to come up with quotes for you and write paragraphs for you, well then, yeah, you're not really thinking. And we'll talk about guardrails in a minute. But when we're getting kids engaged with generative AI, we're asking them to maybe dive into a concept or learn something. They have to constantly be noticing their own thinking, defining the gap generated in question.

And then not only that, evaluating the AI's response to either, is this correct? Is this a hallucination? But more than that, okay, but what does that mean then? And like, what do I now need to ask it to go further? What are the implications of this? And kids don't know how to use it. I tried this as an experiment recently and I did not put it in the grade.

so no one gets concerned here. But I had done an assignment for years, this is totally gonna date me, but it's called a web quest, and I was still using this web quest. And last year when I gave the web quest in my mythology class, they were done in about a minute and a half, and everyone had nice big paragraph answers to the questions.

So I realized, well this isn't gonna work anymore. I can't give them a couple websites and say go explore and then answer these questions because they'll just ask ChatGPT. So I came up with this idea, I gave them this big prompt, you're gonna be a tutor to this concept, and I gave them some documents that I knew were reliable sources to upload with their prompts so that it was gonna be a little bit more focused.

And I announced to my seniors, you guys are gonna throw this prompt in, you're gonna learn about this, and guess what, we're gonna do a quiz in 30 minutes. And there's nothing like shaking seniors out of their stupor, you know? Like, wait, wait, what? Wait, what Q&A? We're doing a quiz on this and I'm

Like, yeah, you have like the smartest thing in the world in front of you. You could ask it anything you want. You're gonna learn about this. We're gonna take the quiz. Well, it went horribly awry. Only about 25% of the class, 25% of the class aced it. I mean, really aced it, 100%, which was really mind-blowing to me actually, minus one or 100% on this ancient Greek thing called the Ellucian mysteries, which they'd never heard about. And they aced it on a 20-question quiz after just talking to AI. But 75% of the room bombed it.

And when I was walking around the room, I would watch kids staring at their Chromebook, just static, and I'd look over their shoulder and I'd say, what's going on, what's going on? And they'd say, well, I don't get this. And they would just be sitting there. And I'd go, okay, well, what is it you're not getting? And I would coach them through it, right?

Oh, I don't understand this. Okay, what question can we ask? Oh, I could ask that, you know? They don't know how to use it just like we don't know how to use it. And we have to help them learn how to use it, I think. I think it's imperative. And I asked, we had a great class discussion at the.

and I asked a girl who got 100%, what did you do to get 100%? And she said, oh, well, I asked it to quiz me. I asked it to give me an outline. I asked it to make me flashcards.

And then this boy on the other end of the room called out, you can ask it to do that, which was just classic.

So again, I think about my students. I care about my students, and I don't want them to walk into the world disadvantaged because they don't know how to use this technology. And it works in all high levels of learning and thinking.

It can be used to streamline formative assessment like never before. We can increase the feedback cycle and the precision of the feedback cycle like we've never been able to do.

Differentiation, I mean, don't even get me started the ways that you can use it for differentiation and UDL. Now, if anyone right now is looking at that and saying UDL, what is either?

A, what's a UDL? Or B, I always hear about UDL in teacher meetings or whatnot, but I really don't know what it is. So this is your first chance to work on your AI literacy. I'm gonna drop this into the private chat. And Caitlin, if you could drop that in the regular chat.

I actually made you guys a prompt to head over, actually put it up here, to learn about UDL. And this is your first task, if you were just getting used to AI. Or if you are an expert on AI, but you've never heard of UDL, you need to learn about it. It's incredibly powerful. And everyone coming out of the teacher education programs around the United States, around the world, really, are learning this right now.

So this prompt will actually tell you all about UDL. Not only that, it'll ask you specifically about what kind of stuff that you're teaching or some kind of issues that you're having.

classroom and it'll show you how UDL could be used practically. And by the way, this is a side little hack. I stole this from Nate on Substack. These first two sentences here, you are a, and then our shared mission is to.

Those are key. I know early on in AI, everyone was saying act like a, act like a teacher. I've found when you do act like a, that sometimes the A on AI takes on like a caricature, almost like a theatric representation of that, what you're telling it to act like. So I find it better to just to say, you are a co-teacher and expert on UDL. Our shared mission is to, and I find that is really effective. It communicates that there are noble pursuits at work here. This is really important and we share this.

And what I've found is that the AI will actually give you much more detailed plans.

if you use that phrase, our shared mission. So a little hack on your prompt. Now, this is a powerful way that you can use ChatGPT by sharing prompts. You could put in a prompt for your students, and then up in the right-hand corner, you click that little share button, it'll give you a link, and then you could drop that into, you know, your Schoolology or Google Classroom, whatever, and then kids can just continue the conversation.

Now, one thing, so you do know this, is you cannot, if you, let's say you upload a study guide on your own, and then you create a conversation, and then you try to share that, the uploaded study guide does not go with the conversation, just so you know. So the context, that was what they call it, that you put into the conversation, whatever you uploaded, a picture, a document, that doesn't go with the share, just so you know. But like this example, your writing tutor, our shared mission is to help me understand how to integrate quotes. This was something I did with my students, and they had a lot of.

And I'm not using this all the time. This was just a fun warmup one day before we actually dove into some instruction on this. But it's a great way to share prompts that you know are effective.

And then oftentimes at the end of their conversation, I'll ask them, their last prompt is, summarize how I did, give me some guidance for next steps on what I should do next, and give me a full overview of our conversation. And then I asked them to copy and paste that into a Google doc and turn that into me. And then that's what I read in terms of, you know, some level of assessment.

Another way that we can use AI so powerfully is for retrieval practice. Now I call it study buddy, because I'm dealing with kids, right? And I find that with students and sometimes with teacher friends of mine, giving it a nice little catchy title works better than using the academic terms. But this is retrieval practice. And the data on retrieval practice,

research on retrieval practice is so overwhelming. There's a bunch of studies here, and these are all in the references, and hopefully we'll share this presentation with you later. But there was this one, the McDermott one, is just, they're kind of the foremost experts on this.

So when quizzed to study, so we're quizzing to study, we're seeing sometimes twice as good on the final product, on the final exam, versus rereading, highlighting, making outlines, just quizzing yourself on the material is better than all of that. And then this one study, this third bullet, was fascinating.

They found that when you took a two-week break and then needed to look at the material before a test, studying or retrieval practice to learn was by far the most effective. In fact, it got more effective compared to other measures, other methods, with that two-week gap. Now Bjork and Bjork from UCLA are calling this desirable difficulty.

And the idea is that when we allow information to we kind of forget it a little bit And then we have to work really hard to retrieve that information. We're strengthening the neural pathway We're retreat. We're strengthening the retrieval strength So that later on a test we're able to pull that information back out and this is so easy to do a chat GPT I actually have a prompt that I'll share with you guys at the end That is my quiz prompt That works perfectly for this and I'll actually just drop this in so you can see what it looks like Okay So this is my little quiz prompt. I give this to students on a Google Doc I say copy and paste the whole thing into chat GPT, you know, you're on my tutor This was for a Greek myth that we had read my mythology class I want a seven question multiple-choice quiz that was doing this in class But they could say 25 questions or whatever one question at a time

Time is key. You have to say that, otherwise it'll give the kid all 20 questions at once. So one question at a time, and then give me a little mini lesson. And then I added this, I like, and then I tell the students, pick one of these, delete the rest. And this gets to that metacognition piece.

It's so funny, because I'll ask kids, well, how do you like to learn? And what do you think they usually say? I don't know, right? And I actually have a rule in my class that I don't know is never an acceptable answer. I tell them, I just wanna guess. I just wanna guess. Doesn't matter if you're wrong. But this is important for learning.

I know that I like analogies, and I like sports analogies. That's why I probably put them in there with my bias about how I like to learn. But I had them tailor it to how they like to learn. And then I added a cool little thing. If I ever enter expert mode, then increase the difficulty of the questions. So I'm differentiating here.

And I do little mini lessons if they get it wrong. And then I send the kids on this.

and they're sitting there quizzing each other on this, and I've seen huge gains in end of unit tests by having kids engage in this. In fact, I had a mother email me because her daughter scored 100% on midterm for AP Chem. And the mom emailed me and said, I just wanted to thank you because she was really struggling in this class. And so when she got the 100%, I couldn't believe it, honestly. And I said, how did you score 100%? And she said, well, Mr. Cuny taught us how we can upload a study guide to ChatGPT and then quiz ourselves. And she said, that's how I studied for the AP Chem test. I uploaded my study guide. I went back and forth studying. And then what I didn't understand, I went and looked up in the book and I studied some more and she aced it.

My own son at San Diego State for his MIS class, he made a custom GPT where he uploaded all of his professor's slides and his own notes. And he told the custom GPT, I've got a midterm. I need you to help me study, quiz me. He went back and forth. He said for about an hour.

He felt really good about it. He walks over to his suite mate, who's sitting there on the bed, holding his textbook and a highlighter, and he looked up at my son Landon and said, bro, I'm so confused. I've been at this for hours.

And when I heard this anecdote, I couldn't help but think that that was like before the light bulb and the light bulb. I mean, this is such a dramatic shift in how we can help kids learn.

And then Landon goes in to take that midterm. It's a testing environment. There's no computers around. And he scored a 96 and the class average was a 68. And he's using it to learn. And I think that's the trick. We've got to show kids how to use it to learn instead of just using it to cheat. So the study buddy, use the study buddy. Upload texts, ask it to quiz you.

I'm going to come back to personas because I want to quickly show you, because we do want to get to a Q&A here soon, about these custom GPTs. These are pretty fascinating.

At Valencia High School, we build custom GPTs for most of our AP classes.

This is Penny, who's our AP econ tutor. I'm gonna pull over to chat GPT. They're over here on the left for me.

And so we've got Eisenberg, who's our AP chem tutor. We've got Abe.

Now we asked the US history students, we gave them a choice on which logo they wanted and they voted for South Park Abe. So that's South Park Abe.

But we can, the kids are using these to study. They're going home and they're talking to these and asking questions.

And just to show you kind of under the hood what this looks like. When you make a custom GPT, which by the way, many of you are going to be getting GPT Plus, which unlocks the ability to make custom GPTs.

And these are powerful because you can design this.

And as I scroll down here, I've got the knowledge base.

We loaded this with all the AP released FRQs.

cues, chemistry textbook, documents from the teacher. You can customize these conversation starters that show up as bottom boxes. And then these are the instructions. And when I show you this, this will freak you out because you'll think I could never write all of that. This is gonna give kids real world examples, visual aids, you name it, right?

Well, I didn't write all of this. I started writing it. I think I made it to like the fourth sentence. And then I discovered this, the create button. And over here on the create button, you have basically a custom AI that is an expert at building GPTs. So at a certain point, my big epiphany was I went in here and I said, analyze the GPT, how can we improve this and create some effective and innovative ways to help students? When I did that,

it started analyzing the GPT and it came back with a full thing. It was incredible. And at the end, see, it's already saying, do you want to do some gamification? And, and then at the end, it's like, do you want me to, which ones do you want me to implement? And then it just does it. So it's kind of incredible, but these are really cool.

Another one to show you is my Winston Smith custom GPT. I teach 1984 to my honors 10 students. So I made a little, a custom Winston Smith. I uploaded all these sources that we read in class around the book. And so let's see what Winston says. Does two, what is two plus two equal, Winston? Two plus two equals five because the party says so. The answer is not mathematical. It's a surrender. I love this. But then he points out that in a letter, Orwell, to a friend, which is one of the sources that's in here, he cites from that. The two and two could become five if the fear wished it. So again, not bad.

Not because it is, but because it must be. And I could keep going. Were you ever really in love? I hope I didn't put it. But I'm not going to go into that.

Julia and him, that's a whole other side story. But these are a lot of fun. And here's the thing too. We as teachers strive every day to get kids engaged, right? When I put them on Winston, it was crazy. I couldn't even get their attention to put the computers away before the bell rang. They were so into it. They were like, oh, look it, I asked them about that time in the woods, which that had me a little worried.

But yeah, so it's incredible engagement. And we all are trying to get kids engaged. And so they were learning, they were engaged. We then, of course, followed this with a conversation.

And this is so important. I think you have to think of AI, again, as a tool that you are inserting into your framework of teaching. So I like to think of it sometimes as human start, AI feedback, human finish.

Or, so we had already read the book, now I'm going to put them in AI for all this engagement and diving, kind of unpacking some ideas, and then we're going to finish with the Socratic Seminar with me.

So it's not like I'm only using AI, right? Or maybe it's going to be an AI brainstorm to start, human drafting, AI feedback, human finish. You know, I think it's important to think of it in those terms, but when could I use AI in here that could really help with this? And when should I not be using AI?

For the record, I'm using more paper and pen than I have since the pandemic. I feel like when the pandemic was going on, we were all in distance learning, I was all online and everything stayed in Google Docs and slides. And now I'm not.

I'm going back. I'm never going to ask my kids to write an eight-page paper with handwriting, though. I think that's kind of crazy. I hear about the blue books coming back. But if I'm going to have them draft a thesis statement, I'm going to have them do that with pen and paper in class.

so that I guarantee that that thesis statement, right, is coming from them. Taking notes, they're handwriting notes in my class again. I mean, there is a lot of research on that, too, that that's powerful.

So if you don't have GPT+, you don't have to go, oh, I can't do these custom GPTs. There is a kind of a workaround, and that's the persona.

So you can give kids a prompt, right? You could share the conversation with them, or you could make the prompt, put it in Google Doc, tell them to copy and paste it in, however you wanna do it.

And then it says, you act like a chemistry tutor, act like a culinary arts teacher. Give it a purpose, and I have a, by the way, a template for this in the Google Drive that I'm gonna share with you at the end, that you can just fill it in and give it to your students.

And then they can then engage with that. And then I taught my students that over here on the left, you know, you have your saved conversations. So they can just click this and.

name it and call it, you know, culinary art tutor, or, you know, debatatron, or whatever you're using it for the persona. And then they can always revisit that and they could use that to study and other things going forward.

You know, speaking of, I wonder if our growth mindset study is done. Let's go take a look.

It is. So it did 124 searches. It took six minutes. It found 16 sources. And here we go. Look at this. It's got some articles, Carol Dweck's article, false growth mindset. I already knew about that one, so that didn't help me much. But this interview in the Atlantic, I did not know about that. That could be really helpful.

And then I'm gonna keep going down. Angela Duckworth on Grit, her interview with ASCD, that could be a really good source. Look at this. I didn't do any work in terms of researching. I've now got TED Talks.

And I've probably got some podcasts down here. The Power of Belief, Mindset and Success. Oh, yeah, I saw Khan's talk. So, and here's some podcasts. He found me some podcasts that I could go find and get some kids. Oh, Dave Yeager. I'm gonna talk about him here at the end. Freakonomics. So, this is fantastic. I mean, look, I could keep going and keep going and keep going. Not only that, it took my learning outcomes and it's now giving me assessment ideas that are connected to the sources it just gave me. So, in terms of planning, deep research is a game changer. And I find that very few people are using deep research. It is an absolute game changer. Now, have I done some deep researches where I didn't really love all the sources? Yes. But usually that's only one out of 10. Usually it is absolutely incredible. And I wanna show you one more way we can use deep research as I have about five minutes left. So, we'll start in.

Convo, and this I think will blow your minds. Deep research. I'll have to run.

Summarize the ten most cited peer-reviewed studies over the last five years on the impact of feedback timing on student writing. I want citations and I want one-sentence takeaways. Yes.

I'm just gonna do this really quick. Show the feedback. Yes.

And I'm just gonna say yes and just go. Let's let that work and then I'm gonna go back to the presentation. This is deep research, you guys. It's incredible.

It's incredible. I mean, yes, I did use it to find the exact best blender to buy my son for his dorm room and maybe that was not the best use of deep research, but it is kind of incredible what it will discover.

So a couple more things here and then we'll wrap it up. Custom GPT's. Oh, final step feedback.

I love this. So when I had my students make that infographic

class or I've had a photography teacher friend do this where after they had their photos done and they felt like they were done or my students thought their infographic was done. Upload it to chat gpt and then I give it I give them a prompt act like a graphic designer or you are a graphic designer our shared mission is to improve this with canva or whatever photoshop or whatever they're using and this is this has been very powerful.

I had a student come back to me the next day and she's she was in student government and she said you know all the stuff I learned while improving my poster in your class we then all we I showed everyone in my uh my group uh how to use that for all the posters we just made for student government. So again when they're when we're using it in this way we're teaching kids skills we're also teaching them the skill of using AI which will generalize well beyond our classrooms because learning with AI requires curiosity it requires questioning we've got to give kids questioning stems we've got to scaffold that because they

they don't know how to ask good questions. We need to help them with those verbs. And literacy is gonna be more important than ever.

As you saw, all of this text, I once recently told the kids, we're gonna get into Chad GPT, and my seniors, again, one of them groaned. Oh, and I was like, whoa, I thought you guys would wanna get in GPT. And he said, it's a lot of reading. And it is a lot of reading.

And so it does worry me, again, that if we don't focus on foundational skills like reading and vocabulary, and help kids learn to use this to maybe speed up the learning of those things, especially for our English language learners, boy, they're just gonna fall farther and farther behind.

And that curiosity piece, I'm gonna come back to in a second, because we need to expect more.

I think that's the final thing that I'm taking from AI, is that we need to expect more from our students.

I had a student, they're making this infographic in a Viking mythology. They read a Viking myth.

and they're making an infographic in Canva about three life lessons from the myth and they have to siphon the myth and stuff. It was just a little project I wanted them to do after reading the myth.

And one kid said, can I use AI? I was like, yes, you better. And we hadn't talked about it yet, but then we talked about it. And what I told them was, here's the rubric. The rubric is expecting a lot because the rubric is expecting that it's not only you that's working on it, but a tech support person for Canva is at your side. A graphic designer is at your side and an expert on Norse mythology is at your side.

And the four of you, these are all personas in Chatterjee BT, mind you, should be able to produce something that is absolutely spectacular. And that's where I'm holding the bar now. We need to expect more, but we cannot just expect more without this next piece. It is all about relationships.

And that's where I was gonna bring up the Dave Yeager book. He calls this the-

mentor mindset, and I've been calling it rigor through relationships for years, and that it's all about relationships. And I feel like in the age of AI, this is more important than ever.

How do we get kids not to cheat on AI and want to care about their own learning? We have to build a relationship with them so that when we tell them we care about their own learning, they might actually respond. Are they teenagers and still going to cheat from time to time?

Of course. But we have to start with that relationship. And I think AI can actually free us up as teachers. We spend more time fostering those relationships. It's all about the relationships.

And I would urge you to read these three books. I think these three books are so relevant. Well, the one in the middle is actually a study from Stanford, challenge success, about the effects of homework and what is good homework and what is not good homework, because I think this is going to be more important than ever right now.

And then Dave Yeager's book, 10 to 25, is a must-read. I mean, absolutely must-read on how to motivate kids. And then Grading for Excellence.

if you haven't already read it, because I feel like everyone's read it, so powerful because he answers a lot of the problems with AI because we're gonna need more authentic assessments happening in the classroom and less points given to homework, which by the way, we never knew who was doing that at home anyway.

All right, this is what you were all waiting for, right? Well, what about the cheating? Oh, it's so bad, you guys. It's the worst I've ever seen. I'm sure you're seeing it too. For me, it's a lot of steps. Show your work. I'm breaking down an essay into parts. I'm doing a lot of pen and paper as first steps of drafting before they get to the computer. I have tons of guardrails up. Everything written is happening in the classroom that's gonna be typed.

I will talk about some guardrails in a minute, but I'm definitely having to rethink homework. And I already had started rethinking homework years ago where I really wasn't giving points for homework much anymore. Everything was happening in terms of assessments, like authentic assessments that were tied to standards, but we're gonna have to rethink homework.

you send home you have to assume will be AI'd. And so maybe we need to change homework to say go home and ask AI five questions about what we learned about in class today. You're gonna copy and paste your summary into Google Doc. That's your homework. And then come to class ready to talk about your final takeaway. And also I think we need to think about cheating. You know we were sending stuff home before and we had some kids that had parents looking over their shoulder doing the work for them. I knew I had parents writing essays here and there. We had kids utilizing the tutoring centers, the local libraries, which was great. But was that cheating? So now that they have AI at their desktop or at their phone, on their phone, is that cheating? Or is it actually, could it be, productive help? I think it's important that we communicate to students. I'm okay if you use AI for this, not for this. There's some research out that communicating that to students is bringing down some of the cheating. These are some of the guardrails I use.

Guardian all the time. I never used to use it. Now I use it on everything. I'm looking at version histories every once in a while, but I, and I'm trying to find a way though to not have to be a detective, right? So I'm trying to find a way to do a lot of stuff in the class, a lot of stuff, paper and pencil, a lot of discussion, a lot of authentic assessments.

I do think we will get to a point with verbal assessments. I don't want to talk about that because what is this going to look like? What is the future of our classrooms going to look like? I think it's important to play this out in our minds and think about where this is going.

I envision my students walking in the door, my AI taking attendance. So I don't have to deal with that anymore. And, and I tell them all, you were supposed to read chapter four of 1984 last night. So put on your headset and you're going to have a five minute interview with our, with my AI assistant here about what you learned from the book or if you read it or not.

At the end of that interview, the AI is going to probably be in an earpiece in my ear saying, great, 80% read it, 20% did it. But even of the 80% that did read it,

they totally missed the symbol of the bird in the woods. So we need to get on that. And I'm gonna go, yeah, you're right. Let's address that right now. So I'm addressing misconceptions in the moment.

And then we're in the middle of, we're starting in 1984 and I say to the AI, oh, you know what, generate an image for me so we can help the students visualize this. And boom, it shows up on the screen of Oceania. I'm then able to get an analysis of the transcripts. So this is where I see it going.

And by the way, we're not that far off. This is exactly what I did with this. So this was the prompt I gave it. You're my teaching assistant. Help me visualize the book in 1984, generate an image. And there it was, and I showed it to the students.

In fact, just the other day, literally yesterday, I'm reading James Joyce's short story, Araby, which I absolutely love. And we got to a part in the book and I was trying to break down this text for them. And I was like, look at all the verbs. And we were.

down this sentence and and how visual this sentence was and then I just threw this sentence into chat GPT while this was displayed in front of the class and said visualize this from the short story James Joyce's Araby and this is what it created and there were gasps from my students around the room and one student told me when she was leaving that this actually helped her so much to just kind of visualize the story for the rest of the story she pictured the main character running down that cobblestone Street so already it's becoming that powerful assistant

I was talking to an engineer about the future of AI in education and after I described all of this to him he said oh so it's like power steering and I loved that because what an engineer thing to say right but I also love that because the teacher will still be in the driver's seat but this is going to enable us to do things with good teaching with PBL informative assessment and UDL

we're all gonna become experts in now, in ways that we never could before. And to help kids at a scale we never could before, if it's used appropriately. So, we could have Abe Lincoln in the classroom, kids talking to Abe Lincoln, we got another tutor over on the side, and the teacher is still there, and the teacher is front and center. And I think it's so important that we recognize that.

Did you all have that teacher in high school when you turned in work, and you got it back and it had a check mark, and you thought, they didn't even read it, right? Did you work very hard for that teacher going forward? No, and so I really, really worry right now that some teachers are gonna sit at their desk, and they're gonna push the AI button, that's just gonna grade work and throw out some feedback, and that will be so destructive. That's the opposite of building relationships. The feedback cycle is an integral part of the relationship piece between mentor and mentee, between teacher and student.

So we've got to use it appropriately. And I think when used appropriately, it's going to make our lives easier as teachers. It's going to improve the learning.

Gosh, I want to show you guys so much more, but I'm running out of time. Because the other thing is, we don't know what the future is going to look like, which is why we've got to become literate. Why we've got to learn what this AI can do, not only so we can prevent the cheating, but also so we know how to use it.

And one thing I think is a lesson is, I love this quote. Take a look at this quote really quick. Isn't it as if that quote could have been written about AI right now? What I was just talking about, about that aspect of metacognition, that state of doubt, and then that systemic inquiry, which is thinking.

This was written in 1910 by the father of American progressive education system, John Dewey. So I know that some people are predicting

in the demise of education with the arrival of AI. I see a very different future. And in many ways, I see that future because of looking into the past and what has always worked in education. And I realize, I see now that we can actually fully realize the dreams of John Dewey, the dreams of the American Progressive Education System with this incredible new tool, if we all learn how to use it.

I'm gonna hold that up there for a second. This is a Google Drive filled with resources for you with these big custom prompts. And that was the last thing I was gonna show you. And then Natalie, I'll turn it back over to you. But I did wanna show one more thing here. You can create these huge prompts that turn Chachapiti into almost like a machine. So I've got this, I've got a supply and demand game. I've got a sentence tutor. I've got a linear equations.

coach in this in this folder that I'm sharing with all of you as well as I think it's an old presentation. I'll switch out the presentation and an elaboration conversation which I found to be incredibly effective. I've also got a claim evidence reasoning so let me pull this up. Oops there it is. So I'm gonna show you guys this just one more thing.

This is a very long prompt and I had Chachapiti help me write this. I didn't write this myself. Oops that's the wrong one. All right here we go copy paste. Okay so look at this prompt. You are a writing tutor. I'm just gonna show you what it does though because I tell the kids just to copy and paste this in.

This is to work on elaboration. So it'll come back and say, I'm gonna skip that, welcome to

To the Elabobot, first off, give me a topic. What do you want to write about?

I'll say Radiohead. I love Radiohead, the band Radiohead. So it's going to give me a claim about Radiohead. It has redefined alternative rock by blending experimental sounds. That's really good.

And then it's going to give me a piece of evidence. Albums like OK Computer and Kid A Push the Boundaries.

And then it's going to say, so here's the start of your paragraph. Now you elaborate.

And when I did this, all I could hear was typing. I had 36 kids all around the room typing about things they were interested in. I did this for two weeks, just for a 15 minute warmup, twice a week.

And then we took the district benchmark. And I don't want to, I hate to do this because I don't want to ever talk about test scores in this way, but I have to brag on my kids. They ended up scoring 23% higher than the school average. And I didn't do anything, okay? I was just.

them in here and you know what was happening? They were writing. And we've known for a long time that kids need to write more than we can grade, physically than we can grade. And this is enabling that for me. And I've got kids writing about topics that they're interested in using elaborative techniques that I have taught them because again I still am in the driver's seat as the teacher.

Well thank you all so much. I really appreciate you tuning in to this and joining and I hope that some of this stuff was interesting and at least got you thinking. Please reach out. You can find me at LinkedIn and whatnot. I would love to continue the conversation.

We're gonna pull some questions from the audience now.

So Tom Galley, project professor at the University of Tokyo. He's joining us from Tokyo. How should educational systems respond to AI at the city, state, and national level via curriculum design, student assessment, standardized testing, teacher certification, university admission?

et cetera. How do you think about these things?

I mean, yeah, I think that's a huge question, right? That's its own webinar. But I think what it comes down to is that we have to create a way, we have to focus on authentic assessment, right?

So if you're just going to ask kids to go do something at home and come back and turn it in, it's going to be AI'd, right?

So I see, for instance, ASU in building the credential program for ASU, they're doing some really innovative stuff. I have to give a shout out to the amazing Liz Frias, who's really doing incredible work with that.

But for instance, one of the assignments they had in there was that they gave, they're giving the teacher candidates a prompt that creates a debate between Dewey and Vygotsky. And they have to pick one side and debate Dewey or debate Vygotsky by the end.

And then the assessment on that is they have to make a video journal where they have to answer five of these deep kind of critical thinking questions where we kind of discover what they learned in that process.

I think we're going to have to think about.

about ways to do authentic assessment differently. I don't think it's that hard though. I think that these are just really educational principles in terms of curriculum design.

In terms of standardized testing, I don't think that has to change much. In fact, I think standardized testing is going to show who's using AI effectively and who's not.

You know, I use an analogy with my students. I said, if you hire a trainer and go to the gym, are you going to get stronger faster? And they say, well, yeah. And I go, yeah, but what if the trainer is doing the bicep curls for you? And they're like, oh, and I'm like, yeah. So at some point you're gonna have to take a test. You're gonna have to take the SAT. You're gonna have to take something, you know, down the road or you're gonna have to walk in and perform in that live interview. And if you cheated every single step of the way, like you're in big trouble.

So I think that we still, I think the standardized tests might've become more important in a way because we're gonna have to have some authentic assessments to really find out if kids are learning.

Yeah, definitely. And.

I'm not surprised at all that you're working with ASU, and they're so lucky to have you as a consultant. And they really jumped just headfirst into AI. I remember when they were first considering adopting chat GPT across the entire university, they hosted a competition for custom GPTs across faculty and admin and students. And I think that's pretty traditional for social scientists. Social scientists, you know that you need to, one, engage a reference group, and two, give everybody the opportunity to participate. And they just got campus-wide participation, Casey. So I can't wait to see what you develop with them. Please keep us in the loop for that.

Okay, so Sherry Zarabi, revenue operations consultant. She asks, I read the number of schools with AI policy has increased. So I guess a lot of people are really interested in policy, Casey.

So she's wondering, what do you see as the value of having one and does your school have one? And if so, what tips would you share with schools that don't?

Yes, I'm at the William S. Hart School District and we do have an AI policy. We have an AI task force. I'm on the AI task force with some incredible people. And I think that's the key.

I guess what I would say is, as a teacher, I don't care very much about policies. I care about my kids and I want to get my kids ready. So to be honest with you, but I know that they're important and I know that they're necessary. But I would say I would urge you to have teachers at the table. I worry that sometimes decisions are made like a lot of time. Decisions are made about education with no teachers there. And you need us at the table.

I got this incredible opportunity to go to the summit, the ASU GSB summit down in San Diego. I have to share this anecdote because I got the chance to talk to.

Arne Duncan, former Secretary of Ed. And I was talking with him and he said, what's your biggest surprise and takeaway about this huge education summit? And I said, well, that there are no teachers here. And he started laughing and he's like, wow, you know, you're right. And I was like, yeah, I don't, it's kind of shocking to me. That's crazy. It's crazy.

And you gave me some incredible praise and I really appreciate it. But one of the greatest things I got out of being this Teacher of the Year thing was that I met these other state Teachers of the Year, which are the most incredible people I've ever met. I'm going to cry just talking about them. They're just the most intelligent, hardworking, compassionate people I've ever interacted with. So bring us to the table. And I think that's going to be important for the policy.

And then I would say, you know, I know that districts get nervous about things, but we've got to move quickly. We've got to move quickly. AI is changing so rapidly that if we don't prepare our kids, you know, I was worried at the end of last year that we had kids that were going to be seniors this year that may have basically.

cheated on everything for their entire high school career. And that really worries me. The head in the sand approach is so not the way to go with this. We've got to get ahead of this and start teaching kids how to use it. There needs to be in-service trainings for teachers on how to use AI. That's got to happen immediately.

I told you as we were practicing tonight, I went downstairs and told my son, even, I just loved all of your use cases in the classroom. And I was like, son, you need to be watching this. I'm going to take all of that, Casey, because as a 15-year-old boy, he just requires a lot of encouragement and support. He has to do his homework downstairs at the dinner table with parents present. He needs us to be coaching him. And I thought we were doing a decent job using ChatGPT to do so. He does know about custom GPTs. But you brought in so many cool, innovative, fun activities that I think he's going to even...

appreciate homework more. I just can't, and as for bringing teachers to the table, thank you for introducing the forum to all of the teachers of the year, Casey. I will make sure that they all have the Chat GPT Plus accounts that they can experiment with. And we would love to keep this going as more of a series and host more teachers and learn, you know, how their methods for integrating AI to their classroom, just like you. And I'm sure people have different approaches, especially at different places in the country and across different disciplines. So we will stay in touch with those folks. And I'm really open to any of your referrals for hosting them because we love teachers. And you guys should be at the forefront of making decisions about AI policy in the school.

Okay, a few more good questions in here. Daniel Green asks, what's one feature you wish Chat GPT had that it doesn't today?

Oh wow, maybe I would love to see the context window open up so you could put 40 essays in at once. I would love to see more built-in flashcard type generators and things like that, which I think there is some of that coming, and I think study mode has some of that. I think if you know how to use it, you can get it there on your own, but it would be nice to see, in some ways, scaffolds in a way to create some of those outputs for teachers that would be easier to just click the button than having to come up with a page-long prompt to do it.

Yeah, definitely. Definitely. And maybe we can all work together with you and all of the amazing teachers that you know to create a prompt like cookbook, and we can publish it in the forum, and we can blast it to all the teachers in the country as a way to get started, because you have that Google Doc that you shared full of amazing long prompts.

So we have a good start, but yes. And I will take that feedback to the product team, Casey. See if we can do something about that.

Eric Luna, Creative Education, K through 12 at Apple, Professional Learning Specialist. He asks, well, first he says, awesome examples you shared. Thanks for that, Casey. When it comes to working on PL with colleagues, what challenges do you run into most often and how do you work through them? PL, I think he probably means professional development, right? PD, I would call it PD, but, so working with my colleagues in terms of AI. Well, I think the big challenge is that, there's a lot of very understandable fear and concerns and worries about what's going to come down the pipe on this. And I'm worried too. I saw a Stanford, there's a Stanford study that talked about 30% unemployment under age 24.

by 2030, which is in five years, which is my kids basically. And so there are, I have serious concerns about fraud and all these other things that are going to happen with AI. But then my mantra is always, yeah, but again, I've got 36 kids sitting in my class on Monday morning and I've got to teach them how to use this stuff or they're there as the kids would say, they're cooked. They're going to be cooked.

I mean, you know, they've got to learn how to use this. So I try to get, but I think it's important though, that we understand the, that the apprehensions are very valid and understandable. I think though, a lot of it is because the cheating is so bad that we're all seeing it. And it's hard as a teacher not to take it personally. When you have kids cheat, you're, it's hard not to go, why would they cheat for me? You know? And you take it personally.

And so it's so important to let that go and realize that if I was 16, oh my gosh, I would have been using this for everything. So we have to set up guardrails just like we would.

anything else, and I think learning to use it, though, is that necessary first step.

You know, I had a teacher, I was in their room, and they were saying, yeah, I'm not sending, I'm not giving the packet home, because they'll GPT it. And I looked over, and a kid was taking pictures of the packet. And I was like, what? Because the teacher didn't realize that you could upload pictures, you know?

So I think it's so important. I once had a kid turn this thing in, it was about Theseus, and I read it, and it said, Theseus is such a cool hero. So I call him up, and I go, you didn't write this. And he's like, what do you mean? And I'm like, you told Chad GPT to write this like a teenager. And he goes, oh my God, how'd you know that? Because I know how it works, and you're not gonna write like that.

So, you know, and the other thing is, there are no reliable AI detectors, so you really can't hold kids on that. So you have to create different kind of learning experiences.

I think, though, that there's gonna be some technology coming down the pipe, where like that, what I described, where you could have kids interview, like, or.

a verbal-oral assessment, whereas the teacher, I don't have to actually talk to all 36 for five minutes. The AI is gonna do it, but then I'm gonna get granular data completely broken down for me, and then what I need to teach next.

In fact, when I started teaching, I used to get these exit tickets, these index cards. And I would ask a question, and they'd write on the index card. And then at home that night, I'd be up late at night sorting index cards all over my living room, and making piles to then do the teaching the next day.

And I think that there will come a time when we look back at that and think it was crazy. Like, can you believe we actually had to sort out the exit tickets? Because now the AI is gonna do it. But I'm still the teacher with the brain, with the engineering, and the creativity, and the design to create the next instruction that's gonna happen based on that information, if that makes sense.

So I think, again, power steering. Absolutely. And a few notes on the future of work and these studies that are coming out about, you know-

entry-level jobs and young people struggling in the workforce. We host the Future of Work in the Forum a lot. So I've hosted a lot of conversations with economists that are all over this.

And two things to note, to start, is that AI literacy is at the very top of needed skills. And skills, when kids are coming out of high school and their undergraduate degree, they have to be AI literate. Like now, not even five years from now, now.

But another really beautiful data point is that social skills are becoming more relevant than they were the past 20 or 30 years, which means that there's just so much opportunity to develop that in a classroom, communal setting, and especially conflict resolution.

And that's becoming a really desired skillset. The chief economist of LinkedIn, Karen Kimbrough, just told us that in a talk. So.

I think there's a lot of opportunity there and while maybe there's this shift happening where work is beginning to change and so teachers like you Casey are really preparing students for the way in which things the work is starting to change and last I just wanted to mention for anybody that feels worried about their job is that if you are excellent at your job you will always have a job.

AI is going to take away the types of tasks that we don't love doing operationalize like really heavily analog tactical items like you showed us like taking a picture of the sign-in sheet and uploading it and creating a spreadsheet and organizing it and we can all really benefit from that but there's like so much opportunity for us to address new like to explore new.

opportunities when we take that work off of our shoulders. So I'm not an economist, but I just want to say that hosting people like you makes me actually excited for the future because I think we're going to have so much more opportunity to be creative. And my son will be looking for a job in less than five years as well, but I feel okay about it as long as he comes out of it literate with AI and has some creative and innovative educators along the way that are helping him to think critically because creativity is not going anywhere. Leadership's not going anywhere. The ability to manage people's not going anywhere. So there's so much opportunity, Casey.

And that critical thinking piece that you just mentioned is that I think if you go learn about the stock market, go learn about interior design, go learn about something you've always kind of wanted to learn about, but never have time, or you're not going to take a course, or you get a book.

and it's that big, and you're like, no, nevermind. Go learn about AI. And I think what you start to discover is to use AI well, you have to constantly be critical thinking. So I think we've got to train kids for that and help them, but it's a plus because we're gonna get kids thinking.

I mean, just to share one anecdote, I'm big on Socratic seminars and I've always done them, but I had a Socratic seminar and I said, I want you to have a conversation with GPT first, like two or three back and forths before you come out with your answer. Well, I had them write down their answer first, like their first draft. Then they had to talk to GPT three times. Then we had a conversation and it was the best Socratic seminar I've ever had on that topic. Their thinking was deep and nuanced. It was human start, AI in the middle, human finish, and it was powerful.

And so I think that, again, thinking about it as a framework and as a tool, it changes, I think, the threat of AI and you start to see it as this power.

Absolutely, absolutely. It's actually very exciting. Very exciting to think, again.

know, so many economists in the forum, because the future of work is on everybody's mind. And I just remember hosting David Autor, and he was describing to us how value gets assigned to different types of work. And for instance, my kind of work and your kind of work that's so heavily dependent on interpersonal communication skills and social skills, it's going to be even more value than it ever has been in the history of valuing work in the United States. And I think that that's actually very exciting. And it's something that those are skills that really begin in the classroom. So I think.

Well, I would love to see a nationwide push for education. I would love to see it as a priority. I mean, I'm not holding my breath. But I think that the root of all of our problems are, you could draw an arrow right back to lack of education or lack of training or lack of kind of ethos given to education so that the best and the brightest would maybe.

be considered going into it. And I think AI could actually free teachers up from a lot of the rote tasks we have to do so that we can focus on what we know works. And that's why my mantra is good teaching will always be good teaching.

Yes. And you have 36 teenagers in your class. That's insane. And hearing that, it gives me so much empathy.

Well, I have five classes of that. So that's a lot.

Oh, my god. How in the world are you supposed to even respond to the teachers? I mean, sometimes there are no responses. And I totally understand why. Because that's so overwhelming. You guys have a really hard job. I have nothing but respect for teachers.

Well, thank you.

OK, one more question.

Because in that case, you know, go ahead.

No, yeah, go ahead. What's the other question?

OK. Oh, man, we have so many questions. So I think I'm going to send you some follow-ups. And maybe when you have time, aside from your totally full capacity at work, you can respond.

to folks if you ever feel so inclined. Let's see, find one more good one. They're all good. This community is so amazing and I'm so glad that you get to be a part of it now, Casey.

Yufen Lin, assistant professor at Cal Poly Pomona, asks, how would you assess student learning using ChatGPT? Are you using the API to create automated grading or through custom GPTs?

Wow, that's a really good question. So yeah, I mean, I wasn't really gonna talk about that too much tonight, but I, so I am using the APIs. I've got a completely custom front end using Django and then I have a backend using FastAPI and I use API calls over to the ChatGPT assistants and everything from the initial analysis of an essay to then class wide and all that kind of stuff. So it's been pretty, it's been insanely powerful, especially with the exit tickets.

a lot of kids where they're just scanning a QR code on the way out the door, they're hitting that question and I get instant feedback on not just what, who got it right, who got it wrong, but like the nuance to why there were wrong answers. And not only the answer, but I trained it to try to analyze student thinking. Because I think, see, I think this is gonna change a lot.

Prior to this, you had people making apps that weren't educators. And so educators, we know that a good exit ticket is going to elicit evidence of student thinking, which is what Dylan Williams said all about formative assessment, elicit evidence of student thinking. So my AI assistant is trained to do just that. So then I get that information, I can then make next steps teaching.

So to answer that question, and I'd love to talk more about that if you wanna talk more specifically about the technology. Maybe we'll do that for the next event, Casey. That is almost like step two. It requires a bit more technical literacy.

I actually am gonna ask one more question because I think you're gonna, you're going to appreciate.

Appreciate this one, Casey. So this is from Igor Bigas. How do you see AI's potential to expand learning opportunities in under-resourced communities without relying solely on expensive infrastructure or privileged access? Yes, this is the best question. Thank you, Igor.

So like when I first did Elaboration Conversation, my biggest mind-blown moment was that prior to that moment, that would have required a team of engineers building a software to do that task of creating like a dynamic writing activity. And that would have never happened because there would never been money in that. They would never have seen that there'd been money in that. So it never would have been designed, right? But now I could just make this on my own and get all my kids writing.

I had this student, Ramil, who was new from Russia and his English teacher was, I was just tutoring him on the side and his English teacher had wanted them to do, write a lead-in to the quotes. And he couldn't understand this. And I kept trying to explain it to him and trying to figure out cognates, but I don't speak Russian.

and we went to Google Translate, it sucked. And this was like literally a year and a half ago. So Chachapiti was pretty new. So we went to Chachapiti and I asked it to explain to me this concept of introducing a quote, we're providing ethos, you know, an appeal to ethos to give that credibility. And I watch him read it and his eyes lit up and he looks at me and he goes, giving context.

And I thought that was so brilliant because I would have never thought that context was how that translates, but to him, he understood it now. So what I'm seeing with my English language learners is mind blowing. And so I'm with you. I think that's where we have the most that we could do with this.

And I have seen studies and I don't have them to cite for you right now, but I have seen stuff that when they look at AI helping people at work, AI helps the people that are doing the worst job the most. And I think that when we think about literacy levels or learning, if we can get kids to read though, that's gonna be the challenge. And they can use this. Oh, I think we could level this.

the playing field like never before, which gets me very excited. Casey, you're honestly amazing. This is probably the most inspiring conversation that we've ever hosted in the forum. And we have hosted more than a hundred events in the forum over two and a half years. You are very special and I'm so grateful that there are awesome teachers like you out in the world.

We've come to a close and Casey, I hope we get to have you back again. That was so wonderful. And maybe next time I can even just prepare you even better and we can set you up to hardwire so that your wifi can handle all the live demos and stuff. But honestly, it was absolutely perfect. So thank you so much for being here. And you're a member of our community now. And I hope this is just the beginning of our collaborations.

Well, thank you. I wanna give a shout out to Valencia High. Let's go Vikings and the Winnipeg District and National University.

A lot of people hear from that and the NASU and then the State Teachers of the Year who have become my new family in a lot of ways and love you guys. Yay. Good night, Casey. Good night. Oh, cut him off just a bit quick.

Everyone happy Thursday. Thank you so much for being here tonight. Before we leave, I just remind you of some upcoming events on the horizon. If you have aging parents or grandparents, we're hosting a really cool event that can support us in helping them. You can even invite them to the forum. It's about scams in the age of AI and how to prevent them first by just being able to recognize them. I'm actually incredibly stoked about this one because I literally just had to open up a new checking account for my mom because she has been shopping on the internet and has been targeted for all these AI scams and oh my God.

It's just overwhelming. So I'm really excited for this one. And I have learned that we are really welcoming an audience of senior citizens. So it'd be really rad if you have older parents or grandparents that like to join us for that. This one is for them.

We're also gonna be hosting one of the most foundational AI researchers on the planet who happens to work at OpenAI, Lukasz Kaiser, for learning powerful models from Transformers to Reasoners and beyond. Lukasz, for those of you who aren't familiar, he was a co-author for Attention Is All You Need, which is basically the paper that brought Transformers into the world.

And then Tina Austin, who's here tonight. She's one of our forum members and a teacher at UCLA. She is going to be attending this and also presenting. But we're hosting an in-person higher education guild on the OpenAI forum.

inaugural summit. That one is limited capacity because it's in person, but we have reserved a few spots you can request an invite. So if you're a faculty member at a university, a community college, if you're a graduate student, or if you're a leader that has really led the charge in AI adoption or creating innovative AI initiatives in your higher education institution, please request an invite. And then you can even drop me a line, make sure that I have my eyes on it.

I hope you guys learned a lot from Casey Cuny tonight. He's such an amazing educator. I know how hard it is to teach teenagers. I have one at home and I'm infinitely grateful for him sharing all of his practices tonight.

We'll see you soon, everyone. Happy Thursday and thanks for joining us in the OpenAI Forum.