On this episode, John Bailey, who advises on AI and innovation at a number of organizations, including the American Enterprise Institute, Chan Zuckerberg Initiative, and more, joins me and Diane Tavenner. We discuss AI’s potential to democratize access to expertise, weigh the costs and benefits of its efficiency-boosting applications, and consider how it will change skills required for the workforce of the future.

Michael Horn: Hi, everyone. Michael Horn here. What you’re about to hear is a conversation that Diane and I recorded with John Bailey as part of our series exploring the impact of AI on education, from the good to the bad. Here are two things that grabbed me about this episode that you’re about to hear. First, John made the point that this technology is really different from anything we’ve seen before. Specifically, how these large language models could, from the get-go, produce artifacts of work that would rival what an entry-level person in a variety of professions would create. And how we’re just scratching the surface of their capabilities. And most people don’t even realize that yet. So what could this mean for education? Second was John’s observation that just because we can do something faster doesn’t mean it’s being done better. Said differently, making the wrong work more efficient isn’t necessarily the right solution.

Now, when we finished up the interview, I had several reflections. But one I wanted to share with you now is this. John’s big framing is that through AI, everyone now has access to an expert in virtually every field. So if the internet democratized access to information, the analogy essentially is AI is democratizing access to expertise. But I’m curious if someone isn’t as skilled or knowledgeable or experienced as John, would they know what to do with or how to use such an expert at their fingertips? I’m excited to be in conversation with Diane for more sensemaking after we’ve talked with a number of people. And we’d love to hear your thoughts and reflections. So please, please share, whether over social media or by dropping us an email through my website at michaelbhorn.com. But for now, I hope you enjoy this conversation on Class Disrupted.

Diane Tavenner: This is Class Disrupted, season six, and the first. I know. Can you believe it? The first of our AI interviews. And we, in this case, we have the first best person, John Bailey, as our guest. Hey, Michael.

Michael Horn: Hey, Diane. Good to see you.

Diane Tavenner: It is always great to see you. There’s so many things we could talk about. But I’m really eager to jump in today to our topics. We’re going to go there right away. When we kicked off last season of this podcast, Class Disrupted, we said that one of the things that we really wanted to delve deeper into was our curiosity around AI. And it’s hard not to be curious about AI right now. In our most recent episode, we were pretty straightforward about kind of where each of us are at this point in time and our understanding and our perspectives. And we overviewed some of the kind of current debates that are taking place specifically around education and AI. And today we get to go deeper with someone who, I think you’ll agree with me, frankly, knows a lot more about AI than both of us.

Michael Horn: So I agree with that. I think it’s very fair. It’s one of the many reasons I’m excited for this conversation, because, as you said, it’s going to be the first of many where we bring folks on who, frankly, have very different views from each other around the impact of AI, sometimes from ourselves as well. And so to start this, we’re welcoming back someone to the show who’s been with us, I think, twice before. So this is like a three peat, if you will. So he’s clearly one of our favorites. None other than John Bailey.

John Bailey: It’s so, so good to be on. Congrats. Six seasons. That’s huge.

Michael Horn: Yeah, we’re still kicking, right?

Diane Tavenner: Thank you. And just in case anyone has missed John previously, quick, quick background here. John’s served in many, many posts in the state and federal government around education and domestic policy more generally. He’s a fellow at AEI. He holds numerous posts supporting different foundations. I could go on and on and on, but what some people might not know, John, is that you originally entered education as an expert on technology and ed. And, you know, we’ll hear that expertise coming through because you have gone deep in the world of AI and how it’s going to impact education, and so, welcome. We are so excited to have you back.

John Bailey: Oh, my gosh, I’m so excited to be here, and I just admire both of you and I’ve learned so much from you. So it’s so good to be on the show today.

John’s Journey to Education AI Work

Diane Tavenner: Well, before we get into a series of questions we have for you, we’d love to just start with how, I guess how. And maybe it’s a how/why did you go so deep into AI specifically? We know you have a lot of experience with sort of frontier models, and maybe you can describe that term for us as well as we sort of begin this conversation. But tell us how you jumped into the deep end and come to this conversation.

John Bailey: It’s such a good question. And it’s also like, my point of entry into this was interesting because, as you mentioned, I’ve been involved in a lot of technology and policy intersections for a number of years, including in education. And if I have to admit, like, I’ve been part of a lot of the hype of, like, we really think technology can personalize learning. And often that promise was just unmet. And I think there was, like, potential there, but it was really hard to actualize that potential. And so I just want to admit up front, like, I was part of that cycle for a number of years. And. And then what happened was when ChatGPT came out in December of 2022, everyone had sort of like a moment of ChatGPT, and for me, it wasn’t getting it to write a song or, you know, a rap song or. Or a press release. It was. I was sitting next to someone with a venture team and I said, what is, like, what is an email you would ask an associate to do to write a draft term sheet? And she gave me three sentences. I put it in ChatGPT and it spit back something that she said was a good first draft, good enough for her that she would actually run with it and edit it. And I was like, oh, this is very different. And then it just sort of started this process of seeing, like, what else could it do? And it just became insanely fun to kind of play with it. And then I was posting a lot of this on Twitter, and that caught the attention of some of the AI companies. And then they gave me early access. So I got to play with something called Code Interpreter for OpenAI, which was the ability of analyzing spreadsheets and data files, and then did some work with Google beta testing, Bard, and a handful of other things as well. And so I get to work with some of the companies now on safety and alignment testing, but also seeing kind of a little bit what’s over the horizon, Google Notebook LM I’ve been playing with for the better part of Over a year and giving them some feedback on it. So I think what’s happened though is that for me this feels very, very different from all the other technologies I’ve been exposed to at least over the last 20 years. And that has caught my excitement. I’ve rearranged my entire work portfolio to spend more time on this, just because it’s rare to see something that I think is going to be so transformative. I don’t think that’s going to be immediate. I think that’s going to play out over years and over decades. But also just the pace at which this technology is improving and new capabilities are being introduced is something like I’ve never experienced. In just the last two weeks of December, you saw so many announcements from OpenAI and Google that you can’t even wrap your heads around it. So better models that do deeper reasoning did not get a lot of attention. But OpenAI released Vision Understanding so now you can use your camera. And so I walked around a farmer’s market and it analyzed all the produce and the meats and it was giving me recipes on the fly.

Thanks for reading The Future of Education! This post is public so feel free to share it.

Diane Tavenner: Yeah, we were playing with it at the holiday dinner table. Yeah. And just like what, what’s on the table and what are ,you know, and, and I think the amazing thing was with my 82 year old mother in law who was like into it and so, and wanted us to get it on her phone so she could go show her friends.

John Bailey: Oh yeah, it’s. Yeah. I mean it just feels different. It feels like something I want to just dedicate a lot more time and attention to understanding it. Both the benefits, lots of risks, lots of challenges on it. But it just like I’ve seen, you know, my mom’s using it to your point, like it’s just an advanced voice and the style of. Is just great entertainment for kids too with telling stories and whatnot. So anyway, so that’s my journey into this space.

The Best Case Scenario for AI

Michael Horn: My kids have started to leapfrog me by just taking their search inquiries right to ChatGPT themselves and then get frustrated with some of the answers. Let’s dive in then John, because you’re getting to see a lot of these large language models clearly up close. You’re getting to experiment and help advise these companies that are at the leading edge in many cases. And I think what we want to do in these conversations, frankly is have both the advocates for and skeptics of AI and you clearly have a little bit of both from what you just said, make the case for both sides. You know, how’s it going to impact positively, how’s it going to impact negatively? So we can start to unpack the contours and figure out where the puck’s really going in classrooms and schools. And so I’d love you to start with this, which is to make the argument for how AI is going to positively impact education first. So leave aside your concerns and skepticisms for a moment and in your mind, like what’s the bull case, if you will, for AI?

John Bailey: One is, I think you have to do a lot, I’ve been wrestling with this a little bit. I think most of the other technologies up until this point have been about democratizing access to information. So that’s everything from the printing press to the computer, like CDs and with disks to then the Internet, the Internet democratized access to Wikipedia and you could get any information you want within your fingertips for almost no cost whatsoever. What I think is different about this technology is that it’s access to expertise and it’s driving the cost of accessing expertise almost to zero. And the way to think about that is that these general purpose technologies, you can give them sort of a role, a Persona to adopt. So they could be a curriculum expert, they could be a lesson planning expert, they could be a tutoring, and that’s all done using natural language, English language. And that unlocks this expertise that can take this vast amounts of information that’s in its training set or whatever specific types of information you give it, and it can apply that expertise towards different, you know, Michael, in your case, jobs to be done. And so for the first time, teachers have experts available at their fingertips, just typing to them the way they would type to a consultant. So give me a lesson plan. Here’s an IEP of a student, help me develop three lessons that I can use for that student that’s based on their learning challenges and the interests that they care about. So I think that’s going to unlock both, it’s going to be an enormous productivity tool for teachers potentially. I think it’s also going to be an amazing tutoring mechanism for a lot of students as well. Not just because they’ll be able to type to the student, but as we were just talking about, this advanced voice is very amazing in terms of the way it can be very empathetic and encouraging and sort of prompting and pushing students, it can analyze their voice. And then this vision understanding which was just sort of introduced. Google’s had this in a studio kind of lab format for a couple months now, but I think that’s going to just unlock, imagine a student to be able to do a project and presentation and having an AI system give them feedback and encouragement. That is like science fiction two years ago. And it feels like it’s very much within the realm of possibility. Maybe not right now, but you see the building blocks for where that could actually be assembled into a pretty powerful set of tools for both teachers as well as students.

Diane Tavenner: So John, when you, when you step back from everything you sort of just described of what’s possible in schools, teachers. Well you didn’t say schools. So among teachers and students, I sort of mental mapped a school on top of that concept. What part of that do you actually believe is going to be real, you know, for students and teachers and why. And maybe I think you’re probably going to put a timeline on it too is my guess based on what you’re saying.

John Bailey: Yeah, I think, I mean if other industries are a bit of a roadmap here, what you’re seeing in almost all the other sectors is that where AI is getting deployed first is a lot of back office functions. It’s in their IT shops. With coding, we don’t have that in education. But there are other, a lot of back office things where again the benefits can be pretty high and the risks of it being wrong are a little bit less than if like it’s engaging in a tutoring lesson with a student and hallucinating. That’s like high risk. Right. And so, you know, I suspect we’ll see a lot more sort of back office improving parent communications. I think we could see this, you know, beginning. There’s already been, you know, decades of legacy of trying to use AI or technology computer based scoring for assessments. I could imagine that. And then I think you’re going to see it roll out with a handful of tools for teachers. You’re seeing companies like that already with like brisk teaching. But also, I mean all these capabilities we were just talking about with Google, I mean they, if the moment they flick a switch and roll that out over Google classroom, that’s bringing AI into 60, 65% of classrooms and teachers around the country. And, so I think what you’re going to see is a lot of teacher productivity tools and then over the next, let me call it two to five years, a lot more sort of student facing things. As those technologies mature and as we build more robust products around it that have some of the safeguards that you want and need that ensure accuracy and quality as well as safety, I think for students as well. So I think there’ll be a lot of potential, but I think we’ll roll it out to students over a longer period of time. Meanwhile, like the teacher productivity, you know, enhancements for this could be pretty huge immediately.

The Risks

Michael Horn: It’s interesting to think about building off that Google classroom platform and just the access. Right. That solves in terms of distribution that perhaps historical products have struggled with in schools and gaining access to teachers and students. Let’s turn to the other side for a moment, John, and just like, where is AI not going to help things with teachers, students, schools, learning, you know, what’s sort of the, the place that people are dreaming up right now that AI is going to do something and you’re like, I just don’t buy it.

John Bailey: Oh, it’s interesting. Don’t buy that’s a different I, where I was going to go. I worry a little bit of, just because something done faster doesn’t mean it’s done better. And I know like, if any of the white papers are like, teachers should always be in the loop and teachers should always use their judgment, but teachers are also human. And I think one of the aspects of human is that if you’re overworked and you’re tired, sometimes the fastest response is the one you go with just because you’re just, you’re trying to maximize your time. And that’s one of the reasons we see teachers using like not great instructional quality resources from Pinterest, you know, and from Teacher Pay Teachers and from some of these other websites. That is a problem that exists now that I worry AI will exasperate. You know, if you’re a teacher and say, give me a lesson plan on literacy or reading something of reading in the third grade, you have no idea if that’s based on the science of reading, if it’s based on, if it’s aligned to your curriculum, if it’s adding coherence. And so there, there could be a sense of this instead of really augmenting a teacher’s judgment it could lessen it. In the same way that I think we worry about this with students, that part of the way you learn is through struggle, and struggle comes with not writing a perfect first draft. It comes from the first draft, the second draft, and the iterations and revisions on top of it. And I worry that the moment like students have just have a button that can automatically improve a paper, a paragraph or a sentence, they’re atrophying a muscle that is really critically important for this going forward. And then lastly, you know, we’re in the midst of this national discourse and debate right now about social media and phones and is that leading to more social isolation, loneliness and mental health issues with young people and inject into this these AI tools that I think as much as people say this will never happen, the risk of an AI companion where you’re talking, literally talking to an AI that’s empathetic and warm and adopting Personas and that’s going to be easier than the friction of talking to real life people. And so I worry that there’s a scenario where this AI companions will start leading to exacerbating the social disconnectedness and divide. And that is something that if you look at kind of the headlines that we’ve already had a couple cases with some tragic situations with kids who have committed suicide, I don’t think it was because entirely of the AI, but the AI was a contributing factor in that. And that’s something I think if we want to get ahead of where we are in the social media debate now, that’s something we should be thinking about researching and adding some guardrails to as well.

Diane Tavenner: John, I’m wondering, as you’re sharing these perspectives, how you think about. I guess what’s coming up for me is I feel like the main structures of school and education are still in place. And I agree with you, like the efficiency plays are the first places people go and does AI sort of risk reinforcing the existing model of school and education because it will make it more efficient? So like if teachers were just like barely, barely holding on and now we can keep everything sort of the same but just give them this like boost of efficiency we can keep things the way that they were. And obviously I’m biased because, you know, I want to, yeah, change up the way, pull apart everything but I’m curious just how you think about that, especially as things will unfold over time and like the easy places to start and the asymmetry of adoption too, you know, I mean, not every teacher in America has even ever logged into ChatGPT before. And then there’s some that are like power users at this point.

John Bailey: Yeah, I mean a common theme for both of your works and including over the six years you’ve done the series too, has been, you know, we have this system and institutions within the system that are remarkably resistant to change. And I think what we’ve seen is like technology doesn’t change a system. The systems have to change to accommodate and harness and leverage the benefits of whatever technology or sort of new innovation has been introduced to it. So I’m a little skeptical there. I think you’re going to have capabilities of AI outpacing the institution’s ability to harness that. It’s going to take time to figure out what that looks like and what that means going forward. I do, I come back though to this idea of like it’s access to expertise and I wonder if that mental model starts unlocking things as well, that if you’re a school principal, all of a sudden you have a parent communication marketing expert just by asking it to be that Persona and then giving it some tasks to do. And if you’re a teacher, it means all of a sudden every teacher in America can have a teaching assistant like a TA that is available to help on a variety of different tasks. And going back to what Michael’s point was saying with like Google Classroom, imagine if you’re a teacher, you’re in Google Classroom and you have your TA that’s able to look at student folders and just answer questions. You have. Like, I see like John and Michael really struggling in algebra what are some ways I could put them in a small group and give them an assignment that would resonate with both of their interests and help them scaffold into the next lesson? That was impossible to do before. Like that those three sentences could easily do that. And, and that’s why I think you’re going to see this idea of assistance very much kind of entering not just the education narrative but also the, the more sort of broader corporate landscape as well. Where you see that also by the way, is, is a little bit in how OpenAI is thinking about the pricing for this. There is an OpenAI model. Most people probably didn’t see it. The most robust, smartest and the one that has the most reasoning and they’re charging $200 a month for that. And most people are like oh my gosh, like I would never pay $200 a month for software. And that’s because it’s the wrong way to think about this as a software. The way to think about it is will you easily spend that much on a consultant or in a part time staff person. So OpenAI is even adopting almost like a labor market pricing strategy or the expertise that they’re giving you. And so I think this is an amazing thing for schools to think about at time of tight budgets is, you know, again, if you want to maximize your teachers, how can this fill different types of labor market roles in the education system to enhance and support teachers in the limited staff, given budget tensions that are going to be coming out in the next couple of years here.

How AI Is Changing the Skills Landscape

Michael Horn: It’s interesting hearing you say that and draw that analogy, John, because actually Clay Christensen, before he passed away, one of the big interests he had was how do you scale coaching models in education, in health care, in lots of these sort of very social realms as the recipe, if you will, for sustained behavior change and success and things of that nature. Never got to really dig into it and write about it. But as I’m hearing you talk about this, it suggests that maybe a disruption of that might be afoot. I guess that’s the question I want to lean into though, as well, which is you named a few things that this could hurt. And so the flip side of it being a great coach is that it might take away social interaction. Or you talked about essay writing and that, you know, actually the learning is in the process of doing it in revision and sort of pushing the easy button, if you will. Right. Jumps you ahead to the product, but not necessarily the learning and the struggle from it. I guess what I’m curious about, and I’m going to borrow an analogy that Brewer Saxberg, former chief learning scientist, I think was his title at CZI Chan, Zuckerberg Initiative and you know, Kaplan and K12 and a variety of places. He talked a lot about how Aristotle back in the day worried a lot about as the written word became a thing, that people weren’t going to be able to memorize Homeric length epic poems anymore. Aristotle was absolutely right. And I don’t know that we regret the fact that most of us.

John Bailey: Speak for yourself, Michael.

Michael Horn: Two of the three here could do it, but I, so but the question I guess would be, you know, of these things that might hurt, which are really going to, are they still going to matter in the future or are there going to be other things that we, you know, other behaviors or things that are more relevant in the future? And how do you think about sort of that substitution versus ease versus actually like really, you know, frankly, I think when you talk about social interaction that could be, forget about disruptive, that could be quite destructive.

John Bailey: Yeah, no, it’s, it’s a great question. It’s a good point. It’s also this is an area where some of the best studies of this are happening in the labor market and looking at like, how is AI changing? There was just one study I was just reading today with Larry Summers and Deming from Harvard that are looking at, you know, AI, one of the things that they’re finding is AI is chipping away at some of the entry level jobs. It is for the same reason that, you know, you don’t like, if I’m in Congress, now all of a sudden, I don’t need an intern to just summarize legislation. I have something could summarize it for me better in five seconds. And that actually hurts that intern because they’re not developing the skills of reading legislation and analyzing and summarizing it. But it also means the other thing that they’re talking about in labor market sort of terminology is that it’s really raising the skills for those entry level jobs. Now you’re not expected to summarize, now you’re expected to do more and a higher level cognitive functions with it. That, that’s interesting. But I also mean that’s going to place a huge strain on our education system. Like if you’re looking at just the results of TIMSS and NAEP and where kids are, they’re not in that higher cognitive function in terms of being able to ask those questions or do those capabilities. And so in many ways I think if this is going to change the future of work and going to raise the level of what’s expected, that’s going to put more strain on our education system to make sure that we get kids that are capable of doing all those different things. I think about that with myself. Like I’m not like, there are many people who are Excel gurus, very good at analyzing data and they do P tests and other things that statistical things that are very important and I would not be able to do. And this was one of the first experiences with code interpreter, with OpenAI is that all of a sudden I had again an expert, a data analyst who could do that for me. But what that meant is that for work I can no longer say, well that’s not something I can do. Now I could do it because I had an analyst that could help me with it and that in some ways don’t tell my employers this, but like now that could like raise their expectations for me as well. But I have to get smart on the type of questions and the type of direction to give it in order to get the answers that I can use to synthesize into some sort of response. So anyway, I think this is going to be a very messy way. It’s going to change the labor markets, but it feels like it’s lowering the floor in many respects and access to these higher cognitive tasks, which in turn then raises expectations in a lot of different ways. And that’s very powerful. But it’s also, I think it probably a huge strain on our human capital systems. Did I answer your question?

Michael Horn: Yeah, I think it does. Before I think Diane has another set of questions. But before we go there, just one quick follow up, which is it strikes me that then you knowing that you can ask those sorts of questions and sort of having a sense of the contours, right of like what are relevant questions, what are. What is knowledge base that is out there, that I could ask this in meaningful ways and how to structure it. Like those are topics that I might not need to know all the mechanics of how to do it, but I need to know that they are questions that can be asked and, and the relevant place to ask them is that a…Where am I on or off on that?

John Bailey: Yeah, I think that’s right and also again, this is where AI is amazing. Like you could give it a spreadsheet and say what are 20 questions you can ask with this? Or give me 20 insights that you glean from it if you don’t know where to. Like I’ve started again, treating a lot of AI people will tell you not to do this, but if you treat it, if you treat it a little bit, almost as if you’re talking to a person, it does unlock a lot of capabilities. There’s risks of doing that. But also I just find sometimes like I want to do X, like give me the prompt in which to do that or I want to do Y. Like what are. Ask me all the questions you need to be able to answer that. And then it asks me 10 questions and then spits back an answer. I just helped someone with, she’s coming up with a name for her social impact advisory firm and so we created a little GPT and AI assistant that was a brand advisor and it asked her questions the way a brand advisor would and then it spit back 20 names and one of them she’s going with. And so that’s like incredible. But again, she had expertise that could ask questions and facilitate a conversation to unlock some of her thoughts and preferences and then spit back an answer from it.

The Interplay Between AI and Policy

Diane Tavenner: So much there especially given my current focus of sort of 15 to 25 year olds and who are going to be intensely impacted by, I think every, are already intensely, I think impacted by everything you’re talking about. I want to flip over to policy and I want to come at it from the angle of, you know, most people think about AI policy around safety and you know, what are we controlling and what are we, you know, protecting people from, et cetera. But let’s come from the other direction that you sort of introduced a little bit ago about the structure of education in schools. We’ve got some pretty interesting policy movement happening in education right now. We are seeing the rise of ESAs or educational savings accounts, which, you know, puts money in the hands of families to spend it where they want to spend it. We’re seeing a lot of states adopt sort of portraits of a graduate or graduate profile that are these more inclusive, holistic views of like what someone should be able to graduate knowing, doing, being able to do and an openness to how they actually get to that place and the different pathways. Talk to me about like those things going on sort of in the policy world and AI happening over here is that kind of the intersection where we could sort of start seeing some structural differences. And again like a more user centered approach to educate, you know, a student centered approach potentially. So I’m curious your thoughts there.

John Bailey: No, I think it could, I think it’s a yes. It’s a yes, but in some ways the yes is, you know, I think there’s a whole class of ways of using AI that is about navigating and navigating really complex systems. And ESAs are one of those. And I think, you know, I. One of the first GPTS I built on OpenAI to demo this was, like if you go to Arizona’s ESA, it’s like two websites, there’s a weird random Excel file of expenses and then PDFs that like a 78 page PDF. And again that was the best that team could do with limited resources and also with the limited technologies. And I just put that into a GPT and all of a sudden it was a bilingual parent friendly navigator. And if you said can I use funds for Sony PlayStation? It didn’t say no, you’re a terrible parent. It used warm empathetic letter answers to say like no, you can’t and here’s the reasons why, but here’s what you can do. And it was all conversational. And I think this friction of dealing with education systems and education policy could be immensely improved by using AI. Another example, I have a friend, she has kids in a school district and they send these terrible absentee reports and I say terrible. It’s like her daughter’s name is capitalized. So it’s like shouting. And then it’s like has missed six days of school. It’s very, it is reading, reading like a hostage like script. It’s like your daughter’s missed six days of school. It’s very important for her to go to school. We are here to help you. And then it does this weird bar chart at the bottom that’s like meaningless and like I just gave it to ChatGPT as an image and say make this better and give three questions a parent could ask their kid for why they might be absent. Amazing. It was like. And that I did in an Uber ride crossing the Key bridge in Washington D.C. like, you know, that’s an amazing set of powerful tools that can remove friction and help improve the system to make it work better for parents and for kids and also teachers and administrators too. So the but on all this is like, I think that’s going to be powerful and it’s going to make policy easier. I’m still, until we create more flexible ways for teachers to teach, for students to learn and students to engage in different types of learning experiences, I just think we’re going to end up boxing and limiting a lot of this technology capabilities. On the portraits of a graduate. I do think like again, an easy navigator on this is to take student work and student interests and student grades and say I’m not really sure where to go, like help me, Ask me the 10 questions I need to figure out. Should I pursue an apprenticeship program, a two year degree or a four year degree. It feels like again, we’re very close to being able to do something that, you know, it may not be perfect, but it’s much better than what the vast majority of students have access to right now. And if it helps them make a better decision in this process and pick a better path that’s based on their interests and their passions and their skills and their abilities. That’s great. Like we should do everything we can to help maximize that.

Diane Tavenner: Awesome. Maybe just to round out anything. What policy do you think we should be keeping our eyes on as we focus on education in relation to AI? What should we be worried about? What should we be thinking about? What should we be paying attention to? I know you spend a lot of time thinking about policy.

John Bailey: I do, yeah. A little bit. A little bit of policy. So one is that Congress is going to move very slow. We thankfully though, in this day and age of such polarization in so many of our politics, there are two remarkable bipartisan roadmaps. One from the Senate, Senator Young and Senator Schumer introduced. And then there was a House report that got reintroduced right before break that is also bipartisan, remarkably good. It’s 218 pages and they have a lot, I take great comfort in the fact that there’s a bipartisan, durable consensus. It’ll take time to enact that. That’s okay. It’ll take time. At least we have a little bit of a pathway on that. The thing I think for most of your listeners to really pay attention to is what’s happening at the state level. And there, I mean, just last year we saw close to 400 something bills that were introduced at the state level. Everything from dealing with deep fakes to copyright issues to regulating the models themselves. The most famous one was in California. And those don’t on the surface look like they have anything to do with education, but they do. If that California bill had passed, that limits in many respects the types of models that would be available for teachers and for students. There’s another bill, similarly in Texas right now that’s being debated. And so I think we need to pay more attention to what’s going on at the state level because that is going to either restrict or enable access to a bunch of these different types of tools in the models. I think, Diane, you had mentioned too in one of the previous questions, like most people haven’t used ChatGPT, and I think that’s exactly right. But I think what’s going to start happening is ChatGPT and Google Gemini are going to come to where people live already. And you’re seeing that with ChatGPT being integrated into Apple’s iPhone, that, you know, I think for the vast majority of people in the country, their first experience of ChatGPT is going to be through their iPhone. And I think for a whole other set, especially teachers, their first experience is going to be using one of the AI tools on Google. And that’s okay. But again, what’s going to either restrict or expand access to those different types of tools are going to be these laws that are either restricting or adding more scrutiny to the models themselves. And what I will say there is, I don’t think anyone’s cracked the code on how to best regulate this. Whatever policymakers think they have the models improve or they’ve done something that they didn’t think was possible. And for the longest time, policymakers are like, we have to restrict these powerful models and it’s based on computing with some astronomical number. And then on December 24, China announces something called Deep Seek that is pretty much as good as ChatGPT4 and Llama3. And they did it with far less computing power. And so that would slip in underneath as like an exception. And I think policymakers are really wrestling with the best way of thinking about this and restricting it. So anyway, I would do more of that. You’re going to see a lot of other attention to AI literacy. I tend to be. I think these literacy efforts are great, but I have lived through, we need tech literacy, we need media literacy for everyone. It has felt like it. This is by no means to disrespect folks that are approaching this that like every new technology gets attached to literacy component to it. It is not really clear we got much from tech literacy back in the 2000s or some of the other things. And so maybe there’s a way to make sure that we get right what we got wrong before. But I don’t think that’s going to be the quite the silver bullet that we need it to be.

Diane Tavenner: I think that’s right. This has been really such a good way to start. Michael, do you have anything else you want to.

Reading, Listening, or Watching

Michael Horn: No, let’s. Thanks, John. This has been a really tremendous overview of a number of currents that I know both of us have been making notes on the side as you’ve been talking and we’re going to want to dig in more. Maybe let’s pivot away from the topic that we’ve been delving in as we wrap up here and just, John, what have you been reading, listening to, watching outside of the AI education conversation? Hopefully AI is not dominating every single thing. Although I won’t be surprised if you give us some movie or fiction or something like that with AI coursed in its veins. So what’s on your list?

John Bailey: Oh my gosh, what is? Unfortunately, it is like, it’s not unfortunate. It’s just I have. I found myself waking up at like 5am like 2 years ago just thinking about this. So like all of a sudden you’re reading books on, you know, intelligence and human expertise and human psychology because you’re trying to understand like intelligence and what is, what makes something intelligent and that. So anyway, that’s nerdy stuff. The new Henry Kissinger book with Craig Mundy, the Genesis book has also been good. I’ve been reading David Brooks’s book How to Get to Know Someone, which I sort of have missed the first time it had come out. But I think also it has an AI play too because that’s trying to get to know the essence of someone and the humanity of someone. And so it’s been great kind of reading through that in light of kind of everything that’s happening kind of around then what am I watching? I don’t know. Some great series on Netflix, the Lioness. Yeah, it’s good. Oh, and all the Landman too which has also been quite good. Coming out of Yellowstone.

Diane Tavenner: Cool.

John Bailey: I don’t know.

Michael Horn: That’s good. I’m impressed with your range. Diane. What’s on your list?

Diane Tavenner: Well, my new exciting project for 2025 is we are planning a trip to Greece. And as Michael knows, when we sort of plan these trips, one of the big parts of it is spending like six months reading and learning and exploring before we go. And so I actually had a conversation with ChatGPT like you have advised John. When I flipped to just talking to it like a person changed everything to structure a reading and listening list and like all the things I’m going to do. So I have started in on that list that we co constructed and built together, which is pretty awesome. With the Greeks by Roderick Beaton. And this is on the nonfiction side. I have fiction too, but this one rose to the top because I really asked Chat to say I, I need you to find history that’s like engaging and that’s going to keep my attention and you know, give me all the, the way that I want history, the sort of the big swaths and so, so far so good.

Michael Horn: Very cool. Very cool.

John Bailey: One other thing, this summer when I did a vacation, I actually created a GPT with all, the travel itinerary, the PDF and everything else into it. And then it was awesome because I could just ask it questions, but it would give me, it would also speak phrases if I needed it to.

Michael Horn: Oh that’s next level, that’s very cool.

John Bailey: It was kind of, it was just kind of a fun little, little thing. But I’ll share the prompt with you later. Yeah, yeah.

Michael Horn: Because we used it for itinerary planning for, for all the different interests in our group, but did not jump to that level. John, that’s, that’s a good one. Mine has just been a book, so I feel boring compared to you both. I polished off Israel: A Guide to the Most Misunderstood Country on Earth by Noa Tishby, which has remained in my mind quite heavily. And so I highly recommend it. I thought it was quite good and quite humorous and quite engaging the way she wrote about it. So I enjoyed it. And that’s what I’ll, I’ll recommend for folks, and I think we’ll wrap there. But John, huge thanks for joining us again, kicking this off with a lot to chew on for all you listening right in with your questions, thoughts, things that are on your mind coming out of this conversation. We’ll look forward to the next one on Class Disrupted.

The Future of Education is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

Thank you for subscribing. Leave a comment or share this episode.

EP 82: Ali Moresco Talks Advocacy and Her Lyme Journey