How AI Could Solve Education (Instead of Breaking It)

31:03I had a hard time with it as a young guy. And I would have loved nothing more than to have had access to the tools available today to help walk me through it. And the truth is, the failure to grasp that concept pretty quickly and easily left me behind for the rest of a year. And it probably prevented me from taking the next step with another math class. Like, you never catch up when you have one of those struggle subjects. Like, I had things I was very good at, but that was one where it was difficult. And there just wasn't the time to put into helping me understand it. And I never had that epiphany moment that comes when you're trying to learn something new.

31:53But hopefully AI can give that to a lot of people. Yeah, this brings up a school that is doing an interesting experiment, which I want to get to in a second. There's just one other thing I wanted to touch on, which is actually like, so there's been studies on homework, right? Homework is another one of those things that's supposed to help you. Because I think I largely agree with what you're saying in that, like, look, if you're in the workforce and let's say your job is to accomplish a goal, as long as that goal is done and it's done well, the workforce more or less doesn't care how you did it.

32:27I mean, maybe some company policies care. But if your end goal is to get people to read something, as long as it's worthy of their time to read it and people read it and it's good and they like it and they share it, then it's successful. That's right. And the end result is what matters in the word voice. But in school, the end result is a little bit different. The end result is that do you understand or can you prove to the school that you understand? And I think there's two concepts that we kind of need to wrap our heads around with AI. The first one is like maybe proving it is not the most important part.

33:05Maybe actually learning it is the most important part. So how do you get kids to the point where you're no longer trying to prove that you know it, but you're actually learning it? And I think there's a lot of different ideas that are interesting. Now, where homework comes into play is homework has actually been proven since the rise of the Internet to not be as successful or helpful as it has been in the past at helping you learn concepts. In fact, by 2017, it was found that homework only helped 45 % of students, and that's down from 86 % in 2008 because a lot of people will just copy the answers online.

33:42I'm sure it now helps like 0 % of students because of ChatGPT. Like, I'm sure this is a no longer useful way to spend your time at all. I'm sorry with his algebra homework, wondering about those, you know, you always got the answers to the odd numbered problems and not the even. Chachibut would have been handling the even ones for me probably as a young punk. Yeah. So, like, it's just it's not I don't think it's an effective tool anymore. You could argue that writing essays at home is not an effective tool anymore. So then so then what do you do in that environment? And that's maybe another topic.

34:19We'll get into that. But to your point about, okay, like, let's say, like, you're a student like yourself, or I also struggled with math for a long time, who falls behind? How do you focus, like, okay, if we're saying the focus needs to be on learning, not just proving learning, how do you actually increase learning? And one school that is testing this in an interesting way is Alpha School. Now, this is not an endorsement of Alpha School, but I thought this was a very, very interesting topic that we wrote about recently. So they use this kind of, what's the right way to put it? Kooky? It's a little kooky.

34:57A little process? Yeah. Called the two-hour learning model. And I want to pull up their website because I think it's kind of interesting. To be clear, I think it's a little kooky in a good way. I, you know, I think I'm glad to see people trying something new. All right. So this is two hour learning. OK, so this is a system that's associated with Alpha School and a lot of other private schools. I believe they started in Texas in 2016. And basically the system says, what if children could crush academics in two hours? Two times faster. Right. That's the that's the pitch. OK. Very interesting. So how exactly do they do this?

35:39Well, basically, they need two things to learn. Academics at the right level and pace and motivation with the right rewards. This is an interesting process. So here's what this actually looks like. So they do two hours of accelerated mastery in the morning where students get cracking on two hours of AI powered academic sessions, mastering materials two times faster. Adults support their independent learning by providing motivation, motivation and emotional support while holding them to high standards. And then in the afternoon, they get the reward, which is just life skills to pursue passions and interests.

36:16So we wrote about this. Some of the things that kids said they do. They'll do wilderness camps in the afternoon. They'll do entrepreneurship classes. Some kids flew all the way out to Chicago to record a podcast with one of their heroes. I mean, this is K through 8 we're talking about here. And some of these kids are spending this time with guides or you could consider them camp counselors to basically learn life stuff. Financial literacy is another one. That's so important. I also love that they've included passions and interests because, you know, I hated school as a kid. I just hated it. I just didn't feel like I had the time to bother or care.

37:00But as an adult, I really developed a passion for learning. And that's not because of what I was being told I needed to learn at the school level, not even at the school level, at the government level is where that's. really coming down from in a lot of cases. And, but what I'll say is that my love for learning came when I had some independence to, or realized I just could go learn stuff on my own. And, uh, I really feel like, you know, if, if what they want to do is, uh, is learn about the history of video games, structure that, help them learn, teach them what it's like to do a deep dive into a new subject and kind of learn around it and the various aspects.

37:50I like what they're doing. I mean, maybe it's a little squishy. I don't know. I'm not sure. But I admire the heck out of anyone giving something new a try today because, you know, I'm afraid AI is going to really complicate the way we've been doing things. And that's not the fault of teachers, to be clear. I would argue it's not even the fault of local districts. A lot of that comes from how they're funded and the fact that they have these requirements they have to meet in order to get the money it takes to operate a school. Like when you look at small towns, for example, like I grew up in a small town with less than 2 ,000 people.

38:32School's the biggest business in the town. You know, it employs the most people, the most jobs with insurance. it was absolutely the biggest chunk of property it was absolutely the biggest business in the community and never thought that before but yeah I think that's true for my high school definitely yeah it's a it's a big thing and you know and it takes a lot of money to operate that and a lot of that money comes in especially in small towns it's coming from outside of the district. And that means you, you know, you have to hit these benchmarks. There are certain things you have to do. You have to perform an X rate on testing.

39:13You have to have an attendance rate of X that that's down to the hour, you know, and, uh, um, and, and I think that that approach has complicated it and that we're at a point where some sharp minds really need to come together and start looking at what that's going to look like. Yeah. As far as two-hour learning is concerned, they use something that's called mastery learning. So to your point, there's a lot of money that comes into the schools that says you have to do things a certain amount of way. And even the K-8, the actual grade system, It says everyone who's this age is in this grade and they have to learn these topics at this time.

40:05Yeah. It's very it's time based. And what this what the mastery system does is it's not it's no longer time based. It's based on did you master the concept or not? Outcome based. I love it. Outcome based. And it only you only move on like everyone's on their own individual track, which would not be possible with teachers. teachers, like to your point, like, like regular teachers, human teachers could not that you wouldn't have you have to have a one to one ratio of teachers to students to be able to keep track of all this individual learning. Well, the AI system keeps track of your individual track as the student.

40:41And you and you only move on to the next subject once you've proven that you've mastered again proven, but once the system thinks that you've mastered the current concept, then once you've mastered that then you move on so you know they say students learn twice as fast um some of the results of this um the top 20 percent of students show 6.5 x growth so not just two percent two x is like the average right some are learning 6.5 times faster wow um the classes the scores are allegedly in the top one to two percent nationally um and you know this is This is not, this is according to the school. This hasn't been independently verified.

41:27We're going to see, you know, this started in 2016. So we're going to start seeing these kids go out into the workforce or into the larger education system, higher education, I mean, and in the workforce. We'll see how that actually plays out. Like, do these kids perform better? You know, are they as equipped as other kids? Are they more equipped than other kids? But it's interesting. It's an interesting concept. It is. And high five for trying something. I know I've said that a couple of times but there's always such a temptation as humans, as professionals to stick to exactly what we know and how we've always done a thing as a writer I see it every single day and there's this temptation to be like well this is how we've done it and that means this is the best way and the fact is, it might not be you know but you don't know until you try you know and you figure out where new technologies new ideas new everything kind of fit into the puzzle you have and whether that's even a scenario that's going to work it might be that you know you need to do something entirely different but it's not always the case and I think the real innovators and the quality educators are probably these people who are are thinking outside of the box and looking for ways to use the tool to be better at their job the same way we do with what you and I do every day.

42:56I think, I think if you're using it as a way to get better, I think that's going to be the key. Yeah. Teachers are using AI too, by the way. It's not just, it's not just to check students from cheating. In fact, a new Gallup poll came out this year that said six in ten teachers use AI in the past school year. Wow. That they use it to save about six hours a week. That's awesome. On average. And 37 % use it for lesson planning. 80 % say it saves time on worksheets and assessments. And 64 % report better quality when modifying student materials. So 64 % actually think they do better at creating their students with AI.

43:37That's awesome. I think that's a great thing. You know, the truth is teachers are overworked. They got a lot of students. There's a lot of work that goes into it. And I mean, while it's absolutely important that, you know, they're reading these things and see with their own eyes how students are performing. If they were taking some of that same time and able to put that into building out engaging lesson plans or, you know, things that would excite these kids about learning, I think it probably benefits everybody in the long run. Yeah. And to be fair, this is not evenly distributed, right? Because only 32 % of teachers use AI weekly and 40 % aren't using it at all.

44:21So you got to ask yourself, are we going to start to see some of these teachers pull ahead in terms of their level of quality education they can give? And will those teachers be the ones who are sticking with the human-only skill set or will the teachers be the ones who are adapting? That's a good question. I mean, time is only going to tell, right? We're not going to know that until we start to see how this plays out. But it is going to be an interesting case study, basically. It really is. Something I think is going to be big in education is the idea of an individualized tutor. I think, you know, that you can have a tutor to help you through something difficult.

45:02And I talked about this just a little bit earlier, but I want to expand on it a little. Yeah, please do. I think that the ability to have access to a tool that can help you with a problem, even when you're frustrated, that can, and I'll go back to factoring as an example, that can teach me that in one way and I can just say I still don't get it. And it can try another way and then I can say I still don't get it. And then it can try another approach and it can just do that and come up with different examples because it knows you pretty well now. And I think for it to be able to know you, it's also then able to use examples that relate to you.

45:42And that's been a big problem with standardized testing for years is the idea that, well, these questions, you know, tend to lean toward these groups and they're not relatable to these groups. And like, you know, I can fix that. That's a very solvable problem. You know, you could absolutely have individualized questions based on your own background and such to make sure you understand and make. Take some of that frustration out of learning and even gamify it a little bit. Yeah, I mean, I had one of my earliest memories of playing video games on a computer and learning was playing the game Math Blasters.

46:22Again, dating myself on like an old school PC, you know. Yeah. And in that, you know, you're a little Martian and there's a or you're a little spaceman and a Martian is like shooting at you and you have to answer the math problem correctly in order to, you know, defeat him. I mean, that's like the most basic form of AI education. But think about how I mean, it's not AI because it's deterministic functions and whatnot. But if you take that concept and you expand on it and you say, like, okay, every kid is working at their own pace on their own, you know, subjects and with a tutor that knows them and is learning.

47:01And maybe you turn that into games. Maybe not. Maybe we've proved that that's not the right way to go about it. But maybe we do. Maybe it's something entirely different. Yeah, that could be a really interesting system. And it could be really beneficial. If you're having a hard time understanding to kill a mockingbird. You can use an AI tool and be like, listen, I'm not getting it. Talk me through this. Or I need to know these characters. Let's do some flashcards or word association or something that can be interactive activity. Generate the courtroom and put me in the scene and let me be like watching this as it's happening or be a part of the scene and interacting with it.

47:41We're going to get to that point. And who's to say that that's not more instructive than reading it in a book? If you're actually part of something. Yeah. You could interview Atticus Finch. You could absolutely right now interview Atticus Finch and be like, what was it like handling this case? And it will respond to you as Atticus Finch with all of the context and knowledge we have of him far beyond the book, going into college dissertations and theses and literary breakdowns over the years. I think that's such a cool thing that not enough people do. uh, is to, to go talk to historical figures.

48:21It does that pretty well. And with a reasonable level of accuracy, I mean, it depends on the model's training data. I mean, you know, you accept that it's probably not perfect. You accept that it's really not Albert Einstein, you know. That's a key thing because, um, the more that, I think the more that we're able to do these types of things, um, at scale or, or quickly, um, the more that you can create specialized models specifically for these purposes, right? Right now, we're all using these generalized, you know, OpenAI models or Anthropic models or Google models, but who's to say that we couldn't make better small models that are, like, perfectly trained on Atticus Finch or To Kill a Mockingbird?

48:59Like, you could create, like, and we can get into the news of the week, which is OpenAI's new announcement today, but you could go and create specialized models for pretty much every subject you want to learn, actually. And should. And that would be... And should. And the only reason we haven't done that is because we were still learning how to grow these things in the wild and figure them out. And now we're at a point where I think we probably could and should start making a lot more smaller specialized models to do these types of things. Yeah, I keep picturing like the ability to have an interactive back and forth video chat with Albert Einstein on the general theory of relativity and have him call you by your name and answer questions and talk back and forth and maybe, you know, laugh with you, explain it in kind of his style, sense of humor.

49:47imagine what an awesome learning experience that would be, as opposed to just watching a video or reading a book. I mean, those are still important elements, absolutely. You should look at the historical footage, you should read historical records and such, but I think that being able to do that opens up that level of education to an awful lot of people who struggle with it today, with learning disabilities like ADHD and the myriad of learning disabilities that exist. But to be able to put it in a way that engages people who struggle to focus. Yeah, I think to your point, I think it's not about like a one size solution.

50:36I think it's like we have more solutions than we've ever had, actually. Before we used to only have oral histories of how we taught each other. Then we learned how to write things down and then we had books and, you know, then we got video and audio books and all these sorts of things. Like now we have computer programs and software and all this sort of stuff. Like we have so much resources at our fingertips that we should be able to leverage it and use it no matter what the system is. Yeah. Or however you want to learn, you know. Exactly. And it's affordable. It's as affordable as affordable.

51:15For now. For now. I think at the individual level, there's going to always be something that would absolutely do that job well at a price that's bearable. I think the cost of large corporate scaling with AI is going to get expensive. but I think at the individual level, maybe not so much. Fingers crossed. Oh, definitely. Yeah, definitely if there's these, like we're talking about, like hyper-specialized learning models that are specific for, you know, XYZ topic, I think, yeah, you're right. There will remain to be a use case. Do you want to talk about the OpenAI news? I think we should talk about the OpenAI news.

52:02So, OpenAI just announced today that they've made a new initiative with 400 ,000 teachers to shape the future of AI in school. So this is the blog post here on the screen. They're working with the American Federation of Teachers to launch the National Academy for AI Instruction. Basically what this means is they're going to partner with this American Federation of Teachers to train 400 ,000 K-12 educators over five years. So that's one in every 10 U.S. teachers. They're going to provide$8 million in direct funding and$2 million in computing access and technical support. So that's probably to help actually run the models and whatnot.

52:47Yeah. And they've got a flagship facility in New York City, and they're going to expand nationally by 2030. That's awesome. So there's an interesting quote from Sam Altman in there that was pretty relatable to me talking about growing up in St. Louis because that's where I'm at as well. And he says, Growing up in St. Louis, my high school computer lab is where I first got curious about AI, mostly thanks to an incredible teacher who pushed me to experiment, even if I broke a few things along the way. Educators make the difference and they should lead this next shift with AI and we're here to support them.

53:23And I think that's, you know, that really backs up this idea we were talking about that, you know, teachers are going to play a big role in preparing tomorrow's work, as they always do, to be clear, in preparing tomorrow's workers for what this is going to look like. And I'm excited to see AI companies investing there. I'd also like to really quickly just thank them for dropping this article 30 minutes before we started recording this episode with no knowledge it was coming because it made so much sense to talk about here. Yeah, it's perfect. the other quote underneath is kind of interesting too you want to read that one?

54:08yeah this one's from Randy Weingarten president of the AFT and he says that AI holds tremendous promise but huge challenges which I think is kind of what we've been talking about today and it's our job as educators to make sure AI serves our students and society not the other way around the direct connection between a teacher and their kids can never be replaced by new technologies But if we learn how to harness it, set common sense guardrails and put teachers in the driver's seat, teaching and learning can be enhanced. And I love. I agree with that. Yeah, yeah, I agree with that. And I think it's a very pragmatic approach.

54:48And I like that the idea is that part of this is putting free and accessible tools in the hands of teachers to go learn with, to figure out what education is going to look like. One of the difficulties with AI compared to other technologies, like, you know, personal computers took 20 years to get to become widespread or more. I mean, like these were creeping up when I talked about Commodore 64. I mean, like that was early 80s. And you're working up from there to the point where they were in a classroom. Like I remember P90s being in a classroom and that being the first time we had like what I would call a true lab, you know, with with computers that were at least semi close to what you would use in the real world.

55:32You know, traditionally schools had pretty backdated technology and it's gotten a lot better over the years. But I love these efforts to support getting tools in their hands because they're the people who are going to have to figure out what this looks like for tomorrow. And there are kids graduating today without AI skills. Like, this year, kids are graduating without the skills they need to do work soon. Like, you're going to need to know how to do this to work at McDonald's at some point in our future. And it's not that far out. That's a bold claim. It is, but I think it is. And I don't mean that to minimize people who work at McDonald's.

56:11Um, but I think I say that because I think it's important to know that jobs that right now require only a high school education are going to need some amount of AI experience in the future, I think. Yeah, I can see that. I mean, especially if you're using the tools like to manage the computer systems or, you know, work with robots that are making you food. Like, yeah, I can definitely see that being a thing. Or if you work in an Amazon warehouse where you're around robots and, you know, might have needs. I mean, I'm sure most of that's done remotely, but it's not a big stretch of the imagination to imagine there being a scenario where you need to know how to help a customer get their order in talking to an AI unit or something.

56:58I think there'll be ways it's used we're not thinking of yet. If the AI takes the order and then the order is wrong, then you might have to go and dialogue with AI and be like, can you check this order and see where you made a mistake? And you have to be able to prompt it in order to get the right information out. I can see that. Or it's telling kitchen workers what to pull ahead based on the amount of demand they're seeing right now. And they know that at five o 'clock it goes up every day. So it's telling you, go ahead and pull out this many burgers, pull these out, because we know that amount is going to get used.

57:31You know, I mean, there are a variety of ways that I think we'll see this come into play that you're almost not thinking about unless you're hyper involved in a specific industry yet. Yeah, that's true. Well, I think this is an interesting place to talk about, you know, what are the takeaways here? You know, we've kind of broadly touched on a lot of things. I would also like to at some point touch on, you know, the sort of counterfactual or the counterpoint of AI in education. But let's first cover, okay, so what are the actual ways that you can use AI? So let's go back to the whole mastery concept, right?

58:12How, you know, I don't know how two-hour learning does it, but how do you actually learn something? Like, how do you create the neurons to spark in your brain to put two and two together to actually fundamentally learn something and make it stick, right? Yeah. That's sort of the tricky question here. Like, there's proving learning and then there's actually learning. So what is actual learning? Yeah, and that's such a good question. I don't even know if there is a clear answer. I'm certain somewhere out there a neurologist has a great answer to this or a sociologist may. But I think there are definitely some tips we can give to help you better do that on your own and to use AI in a responsible way, like Grant's mentioning here, to help you get that benefit from it as opposed to maybe, you know, dulling your intellect a little bit if that turns out to be a genuine concern as we see that, you know, spread out and get tested in larger sample sizes over larger periods of time.

59:27And I think my first recommendation is that you should absolutely struggle first. you know don't don't immediately dive into ai when you're trying to learn about something i mean obviously that depends on the subject like if you want to know how to cook a baked potato in a microwave absolutely just ask but yeah if if what you want to do is have a decent understanding of a concept that that kind of confuses you or you you don't trust your judgment on it wrestle with the problem first whether that's for five minutes or 30 minutes or or however long you're comfortable with. It doesn't matter. But I think it's better to use the bot as your sparring partner than maybe your ghostwriter, if that makes sense, for lack of a better phrase.

1:00:13It does. It does make sense. And so for context, we wrote about all of this stuff on the Neuron, and we have a nice long, would you say, 40-page blog post? Not actually, but it's quite long, where it's me wrestling with the problem of AI and education. And there's a couple takeaways that I had. One of them, it was related to Corey's point right there, which is, you know, first you should probably have like some sort of framework for, okay, is this something that I actually want to learn? Or is this something I just need to get done right now very quickly? Yeah. Like what do I already know? Yeah.

1:00:45But if you're a student in school, it'll be very tempting for you to say, I have the schoolwork that's done, that needs to get done. I just need to get this done. That is the opposite point of school. Okay. I know you want to get through it so you can go do other stuff, but you need to actually try to learn. And the reason I say that is because I've been reading this book called Make It Stick, where, to Corey, your point about neurologists basically trying to, you know, what does the science say about how we actually learn? And in school, a lot of times the thought is like, oh, you need to read it multiple times and underline it and, you know, practice memorizing it.

1:01:21And that is what actually leads to learning. And it's actually not the case based on the studies and the science that have been done. In fact, some of the ways that is actually the best way to learn stuff is not what you think it would be. It's actually quite counterintuitive. So we can get into a couple of those. But to answer this question, struggling with it makes a lot of sense because there's this thing called desirable difficulties. and desirable difficulty is basically you want something in a range where here's where you are now and you know what you don't know and to learn the next level of something you need to know it needs to be just hard enough that you can build on what you already know in order to get there and put it together without overreaching in some cases exactly and so the the the metaphor that worked for me.

1:02:15I think this might have been from a different book entirely, Atomic Habits, but basically think about playing basketball versus LeBron James. That's not going to be fun because he's going to kick your ass, right? He's going to be way too good. He's going to be way too good at the game. You're not going to have fun. He's not going to have fun. In the same way, think about playing basketball against a five-year-old kid. That's also not going to be fun for you because you're way better than the kid and you're just going to thrash him, right? So you want things to be in a desirable level of difficulty where it's like I'm going up against a partner who's just as good as me or right there on the same line and I can kind of struggle and it feels like it's possible for me to get to the next level.

1:02:57So that's why you want to struggle with it and try and solve it for yourself first. Makes me think of the thousands of hours in my life I've spent cussing at a guitar, trying to learn some song or some lick that I can hear and I can't materialize real well. Like there's sometimes there's just something you're struggling with and you're stuck on and you can get a mental block there. But that struggle is, you know, it's fun because you know what's at the other end of it. Yeah, we talked about this in the blog post that Simon Sinek has a good interview on, I forget the podcast, but we'll link to it, where he basically says the struggle is the point.

1:03:40He's a better writer because of the struggle that it took him to write the book. He is better because he took all this time and mental energy trying to put it together. So that's not to say that you can't use AI to help you learn, but that's to say you need to try and do as much as you can yourself before you pull in a tutor or a teacher. It would just be the same way if you were in school. Like you should try and actually put it together and get the answer before you raise your hand and ask a question. I mean, kind of common sense, but it's just applying that to learning. You raise your hand because you don't get it.

1:04:15And try to identify what exactly is the part you don't get. Like here's the part I do get, but then you said this. So what am I missing in the middle? Which to your point, a tutor who knows you well, you know, an AI tutor would be able to help explain that to you in a way that's personal to you that is relatable back to you. We're not there yet, but there are some prompts that you can use to help you kind of create this with the tools that are available now. So Ethan Mollick, who we shouted out earlier, he shared these creative common prompts from Wharton. And Let me go ahead and pull one up here.

1:04:53I'll bet I shout Ethan Mollick out twice a day to someone. Everything he shares is so valuable. Yeah, he's really great. Okay, so here's one. So this is quite literally called a prompt text. This is a tutor prompt. Sorry, this is quite literally called the tutor. And there's a prompt text here. So this is what you'd put into ChatGPT. You would say the goal. This is a tutoring exercise in which you play the role of an AI tutor. and you will help a student learn more about a topic of their choice. So it goes on a little bit more there. Then it has a persona. In this scenario, you play an AI tutor, kind of reiterating what was said above.

1:05:32Then there's a narrative. The student is introduced to the AI tutor who asks a set of initial questions to understand what the student wants to learn, the student's learning level and prior knowledge about the topic. So it sort of prompts the AI to ask you questions to try and figure out what you're trying to learn and what your level of understanding is. And then it goes through a series of steps. So, you know, gather information, you introduce yourself, you ask the students how, you know, the following questions. Then it tells you a list of tasks to do and what not to do. And then it gives specific instructions about how to tutor the student adapting to their responses.

1:06:12And this is all tailored and made by people who are professional full-time teachers. So they've thought this through on how this will actually help you learn instead of just saying, give me the answer to this problem. Yeah, and that's super valuable. That's such an intriguing way to look at it. And I always lean back. We always talk about, Grant and I joke regularly about how we're going to bring advanced voice mode on one day and do a real good interview. We need to do that. And we are going to do that. We just haven't yet. But we haven't even discussed the value that could bring to education in itself.

1:06:53Like, not specifically advanced voice mode, but the ability to talk it back and forth. Because, like, some people are better readers. Some people are better spoken. Some people need to put their hands on a thing. And I think by having both of those avenues, man, that's going to open some doors. It should. Oh, I've been thinking about that for years. is like the fact that text-based chatting, like the chat interface, is the primary way we deal with AI is like the first problem. Like some people need like artists, for example. They don't want to put in text to get art out. Like that's why they hate AI image generators because they are artists.

1:07:30They want to be able to create. They want to be able to put their hands on it, get their hands dirty, move stuff around. Apply that to learning, right? Same goes for videos. Same goes for people who would rather talk and are more audio-visual learners. We have the tools to make all of this stuff. It just has to be intentionally applied, I think. Yeah, I think you're spot on. I think you really hit the nail on the head there. Something else I think worth mentioning here, another one of our kind of key takeaways, is that I think pace is more important than seat time. We've got to get away from the seat time mentality that finances education.

1:08:11I think that creates a lot of problems. I'd love to see us pairing AI-generated quizzes with mastery checkpoints somehow to allow students to advance at their own pace instead of some arbitrary date. I mean, maybe the deal is like, listen, you're going to go learn stuff till you're 18. And that's just what it is. and it's going to be all kinds of stuff. It's going to be things you want to learn, things you need to learn, things that you maybe don't want to learn, but maybe you should. But, you know, maybe we can make some of the boring stuff a little better. I mean, like, economics was rough, man.

1:08:55That was not a desirable difficulty for you? That was an undesirable difficulty? It was interesting, but it was just dry. it was really dry and it was really sometimes it's the teacher sometimes you don't jive with the right teacher and student relationship you just struggle to communicate just like with any people in the world and I think this can help when that's the case in the long run like I don't think this is a thing that's going to happen tomorrow but it's hopefully around the corner yeah I mean to your point about time based I guess like seat based Could you define seat-based before I move on?

1:09:36What does seat-time mean? When I'm talking about seat-based, I'm meaning that kind of everything a school does centers often, and by necessity. It's not their fault. It's not their choice. But it has to center around this idea that you're going to be funded due to X amount of time in chairs plus test scores. You know? Mm. You know, there are other factors, but at the core, those are two that are very limiting, like, and trying to ensure, the idea behind it is good, you know, trying to ensure students at small schools are as prepared for college as students at big schools. But they simultaneously create a lot of problems.

1:10:19And when I think about time in chairs, I'm talking about, like, you know, they're funded at for even a day is based on, you know, number of hours. of students in a classroom. It's not based on GPA. It's that and generally standardized test scores. And I have probably oversimplified that. And I am certain there is an education PhD or teacher that's going to eat me alive in the comments. And I sincerely apologize in advance. I mean it to be supportive of you. Yeah, yeah, that's fair. And just for the record, neither of us are teachers. So obviously we're talking about something that we've researched and thought a lot about.

1:11:05We've been to the education system, but we are not teachers. So take that disclaimer with you. But yeah, okay, that makes sense. So I guess then my comment is, so to your point about it being more about mastery checkpoints and more about actually learning. Yeah. Alpha schools, one method for how they do that, where they're using, you know, learning is fixed, you know, at 100 % learned. It's not based on time. So that's one way that you could possibly do that. Another is, to your point, schools all over have a bunch of different problems or challenges, let's say, where you're located, you know, how much tax revenue is in your district, how funded or underfunded your school is.

1:11:53All of that plays a big part and limits the, you know, capabilities or the access of the students who go there and the teachers who go there by extension or work by extension. So having a tool that sort of brings the baseline level of, like, attention that a teacher can pay to a student, to all students everywhere across as many schools as possible would be good because now you're not limited by having one good teacher who you really connect with, you can now have at least a system, might not be the same as a teacher, but it's an augmented system that can connect to you and meet you where you are and help you.

1:12:31And I think in that way, you can get closer to do a mastery system because you can have some sort of AI or programmatic way of assessing a person's level of learning and mastery. Whereas before, Or testing was the best way to do that. Yeah. I think testing should be a test of learning, not a proof of learning, if that makes sense. Yes. And I don't even pretend to know what that needs to look like. Yeah. That's a challenge for other people who have the skill set, like the guy who gave the quote for OpenAI. That's a challenge for the teachers to figure out, but this is where we're at, essentially.

1:13:15Yeah. I think so, too. I think so. And, you know, this, as we've discussed before, these aren't tools that are going to go away. These aren't tools that are going to get worse with time. Like, it's highly doubtful that there's a scenario where LLMs begin performing worse. Like, you know, you've seen the curves. The curves are sharp. And they keep going up. And we keep discussing when that's going to stop happening. And so far, they just keep going more vertical. And we'll see this month with GPT-5. We'll see. Yeah. Hopefully by next week, we're talking about GPT-5. So if we're lucky, we'll see.

1:13:57So I mentioned Make It Stick. I do want to talk about this because as we're presenting these ideas to everyone who's listening, who is thinking about education, who maybe wants to use AI to learn or even in their work, I think that this is important to help us use AI in a way that actually works with the way that we learn. Because I think we need to design these systems to go off of the science that we have. And I'm sure there will be more studies that are done in the future. We'll learn even more. But this is something that's really important to me. And I want to share this with everyone and see if it's helpful.

1:14:31So like I mentioned, there's a book called Make It Stick. We'll include a link in the description. The full title is The Science of Successful Learning. And there's kind of four main tactics that I've taken away from reading this book that are helpful and, like I said, counterintuitive to learning with AI. So the first one is retrieval practice. So this is actually testing. So actively trying to recall information from memory is, as you might guess, much harder than simply rereading it. But it actually strengthens your brain better. So, for instance, the group that used AI in the MIT study, they didn't have to retrieve their information at all.

1:15:10They didn't have to retrieve any information. They didn't go to a website, read it, and then have to remember it and put it in the essay. They just had the AI get the information for them. They didn't remember it. They didn't retrieve it. So that didn't stick. Then there's also a tactic called spacing. Spacing is where you actually spread out your learning over time, which forces you to forget and then therefore re-retrieve. the information, which is more effective than cramming, which is trying to learn everything all at once the night before something is due. Yeah, sure. You might short-term memory memorize it.

1:15:45I did that a lot. In fact, famously one time in college for a psychology class that I took, I crammed the night before, slept for 20 minutes, went, aced the test. So whatever happened, but I don't remember a word of it, I'm sure. So it's much better to actually read something and space out your learning. So like, actually like, you know, take a break and go try to learn something. Come back, you know, a week later and try and revisit it. Related to that is a third concept, which is called interleaving, which is mixing up different subjects or problems, which can feel confusing. Like, wait, I'm trying to learn this topic over here.

1:16:25and then I'm also trying to learn this related topic, but now I'm getting them confused. But it actually trains the brain to recognize patterns and hopefully select the right strategy for how to connect those ideas. And so this is actually really important. So not only are you trying to retrieve information and spacing it out, so doing it over time, you're trying to learn multiple things at one time that are connected and basically helping you connect all this stuff and try and build it into the world model that you already have in your head, right? And number four is generation. So after actively generating the answer yourself, even if you get it wrong, helps you learn the material better than simply if you're given the correct answer.

1:17:08So if you're sitting at the keyboard and you're trying to recall something from a week ago, it's better for you to write what you like your best guess and try and get it right. Even if you get part of it wrong or all of it wrong than it is to go back and reread the answer and then try and memorize it that way. Interesting. Sort of a directional pathways kind of thing or something. That's fascinating. I need to read that. Oh, it's a great book. I highly recommend everyone read it. I'll shout out the authors in the description. Sorry, I didn't have, I didn't remember their names. There's three of them.

1:17:46So we'll put that in the show notes. But the way that I suggest that you use this with AI is so when you're actively trying to learn something with AI, you should first like use the prompt as your attempts to figure it out yourself. Like actually, like you said, struggle for 20 to 30 minutes. Actually sit there, define what you know, define what you don't know, actively try and struggle and wrestle with the problem and do the hard part first of generating it from scratch, like try to actually solve your problem before you go to AI. And then you pull in the AI with all of your, basically your notes.

1:18:30And you know, you could do like two or three things at once that you're trying to learn at once. And take that to the AI and then work backwards from that way. Wow. That's a good call. That's a really good call. I use it in prompts a lot as a like, hey, I'm trying to better understand this. I know this happens. I know this occurred. I know this, but I feel like I'm missing a step here. What is missing from what I've given you, or am I entirely wrong? I have a lot of prompts that look like that. You can use the Socratic method, right? You can say, don't give me the solution, but ask me questions that get me closer to it.

1:19:07Or here's my argument, vigorously challenge it, expose its weakest points. I'm working on this problem, A, B, and C. What are the principles that connect them all? Like, just really, like, actually try and learn and don't just have it give you an answer. You know, I think this is a good spot to kind of take the next step toward the end today. You know, I think one of the big takeaways here is that if all we get are new rules, students are always going to find a fresh way to dodge them, to cheat. It's called the whack-a-mole effect. It's a real thing. I think moving forward, it's going to be important to design assignments that make honest AI use a part of the school day.

1:19:50Give these kids what they need to go into a new world they're not quite ready for yet, not sure how to wrap up. But overall, I think as teachers learn how to teach with AI, hopefully cheating will drop and real learning will go up, and it'll hopefully be a win for a lot of people. I mean, I know that's very pie-in-the-sky view, but, you know, my hope is that, you know, this ends in a way that means teachers are more successful, students are more successful. You know, and I think if those two things happen, AI will have done us a good service. How about round-robin time before we call it a day?

1:20:35I have a question. Are you ready? You ready for this? Let's do it. Yeah, let's do it. What class do you wish an AI tutor could have rescued you from in school and why? I really wish that I had gotten further along in math than I did. You know, like you mentioned factoring earlier. So because now as an adult, I'm fascinated by math and I love using it. So and physics, too. so I wish that there was an attitude not to make the class go away but to engage with me on those topics when earlier when I was a younger man I was more interested in creative things and not school hard sciences and I wish that I had something that could connect the dots between why you would want to care and learn about these things more when you had all the time in the world that was a struggle for me too that's fair I think for me the class I would love to have been rescued from is probably, honestly, it's college algebra.

1:21:41College algebra was hard, and it took up a lot of time. I survived, I made it through, made a good grade, but it was particularly difficult. I was a non-tread student who had a few years off before college, so coming back in and taking college algebra was, it was a bear it was just tough and um i think that's probably the class i would absolutely have loved to have had a good ai tutor and like we had tutor resources available but the ability to learn it this way i think would have made it stick better and maybe kept me from stopping there yeah that's that's fair you know i think like all this talk about ai in the classroom it makes me think like they're like if if it's used well and it's helpful for students um you know what then what can the teachers do better now and i think like there's a lot of there's this concept of the zone of proximal development and that like like basically there's a learning sweet swap between what the students can do alone versus what they can do with teachers yeah and so the giving teachers the ability to read the room and adjust on the fly they can notice confusion they can provide different examples like a lot of ai can do that but the ai might not recognize at least not yet that the per that the student is struggling so like for you and me in like a classroom setting like if we had a individual personal tutor who was watching us and seeing like hey you're struggling with this concept like let me focus in on where you're having the problem instead of a teacher who has to focus on all 30 kids at one time, that's really powerful.

1:23:25And that can save you from... WTF is Mrs. Alvin talking about, you know. Yeah. And it can help teachers do that better too, because now instead of having one TA who's maybe helping the kids in class, you have like 30 TAs. You have a TA for every kid. So the teachers can even spend even more time and use the human intuition to help kids when they're struggling with the AI tutor. I just think it's a, I think when done well, it can be a very virtuous cycle. I think so too. That's a good way to put it. On that note, everyone, thanks so much for joining us today here on the Neuron Podcast. We had a ton of fun and I'm confident that left to our own devices, Grant and I could ramble on about this for at least three more hours.

1:24:08given that we have no authority on the subject of education, but we are absolutely giving it our best and what we hope, you know, is kind of something really good that could come out of this in the long run. But we lack in authority. We make up in passion. That's right. And please, if you're watching, wherever you're watching this on Spotify, YouTube, Apple, please like, comment, share, subscribe. It really helps us out a lot and helps us continue bringing these to you. we thank you for listening and look forward to next time farewell for now humans