Greg Toppo: The promise and perils of AI in schools

Hey everybody, I’m Alex Kotran. I’m the CEO of aiEDU and we’re here at aiEDU Studios DC edition. Very excited to have Greg Topo, a reporter for the 74.

One of the leading publications covering the education space, someone that I’ve talked to formally and informally over the past year, two years. And Greg has been both a confidant and sort of a co-learner as we try to make sense of this space and sort of how technology and specifically artificial intelligence is changing education or not. And so I’m very excited to to flip the script and and put the mic in front of you and give you a chance to talk about you know, from the vantage point of a journalist who’s been who’s talked to a really wide range of of folks in our space, what you’re learning, questions you’re asking.

But before we go into sort of the the day-to-day and you know this sort of this AI moment, this AI zeitgeist, I’d love to zoom out a little bit and hear why don’t you start with just sort of you know what you’re doing at the 74 and then you know how did you come to cover education? Sure. Sure.

Thanks for having me. And just as a quick aside, most journalists hate being interviewed by the way so you know I’m I’m good with it. I think we’re we’re gonna have a good time here.

I cover a couple of things at the 74. It feels like it keeps changing, but I cover basically the intersection of AI and education. And I also have sort of a I don’t know if you’d call it a like a side hustle at the 74, but one of the things we are very interested in is the future of high school.

And so one of the topics that I spent a lot of time on is writing about visiting and thinking about and writing about innovative high schools. And the two haven’t quite merged yet. I expect we’re going to be seeing more and more in the near future.

I mean there’s a lot of reporters covering AI. A lot of good ones. There’s a lot of good ones and a lot more today than there were three or four years ago.

Yeah. Not that many covering AI plus education that intersection. How did you get to be doing specifically sort of this like sort of view of the future of school and and innovation?

Yeah. Well, I mean I was interested in it for a number of years. I co-wrote a book in 2021 that actually looked at AI and education specifically, interestingly enough what it was going to do to high school.

We were very interested in in just sort of projecting forward what all these technological marvels, if you will, were going to the effect they were going to have on education. I had a co-author my co-author and I he was an educator. He still is an educator.

He and I wrote kind of a I I thought it was a fun, really interesting book. We basically did a what if a high school principal fell asleep in 2020 and woke up in 2040 and then went and visited his old school, what would it look like? You know after all these technological changes we’ve been seeing so sort of a Rip Van Winkle story interspersed with some reporting on what’s happening actually now and different models of high school.

So big picture I’ve been thinking about this for a number of years. I’ve been covering AI you know daily for two years and change at the 74. And it really began where you know I think most of us started covering it which was November 2022 when ChatGPT sort of burst on the scene and people started you know messing with it and freaking out about it and wondering what was going to happen and you know decrying the death of the five paragraph essay etc etc. that was when we as an institution started really paying attention to it and when I basically started saying you know okay this is something we need to cover full-time.

Yeah. To me because because we were chatting earlier you you mentioned that you had even did you do a story on GBT2 or had you just sort of in the book actually one of the things that we we looked at was and it was actually not specifically related to education. It was just in 2019 I want to say they were playing with AI.

Open AAI was playing with the ways you could essentially make GPT2 into kind of a an approximation of a journalist. And and actually the chapter in the book where we talk about it is where we talk about what’s this going to do to journalism? And the the example we gave which of course you could imagine even more clearly now was that it basically just they gave it a couple of words from a you know a real existing I can’t remember it was his guardian story or something like that and it just basically just went to town and wrote the whole story.

I remember reading the story. Yeah. And you know, it was lots of crazy stuff.

And it was even I think one of the one of the things they said at Lucon was imagine a reanimated JFK, you know, just patently crazy stuff. Yeah. I think GPD2 was as a time when there was, I think, only a couple hundred people in the world who had really had the chance to look under the hood of these language models.

But for me the the zeitgeist moment that ChatGPT ushered in because I’ve been in the AI space since 20 2014. Okay. AI has been around I mean you can actually go quite far back and some people argue the touring computer but you know machine learning reached you know technological maturity in industry in the 2010s.

So AI was not a new conversation. There were plenty of people pretending a lot of the stuff that is now reality today. It was sort of lived in the realm of futurism.

Mhm. I think the challenge for artificial intelligence in part was it was a concurrent technology with this suite of emerging tech you know crypto bit you know blockchain virtual reality much of which you know 3D printing the metaverse all this NFTTS and and while there’s actually arguments that a lot of those technologies are still nent and will you know bloom you know in later years and will transform society in people have predicted they were clearly a bit ahead of their time if we’re being gratuitous, right? And some of them are probably just like patently, you know, were overhyped and overmbellished.

And so I think there was this fatigue from okay, the futurists are at it again talking about some, you know, crazy future. And you know, it was just it was it was I think people were right understandably cynical and incredulous at some of the claims that were being made. Chat GBT put everything into their hands.

It made it tangible in a way that it wasn’t before. And I think in the abstract people understood that there was an algorithm curating the content that they view. It’s very hard to really wrap your head around, you know, what that means.

You’re not really using it. You don’t really understand, you know, what the weights are or even what that means. When you have a tool that’s just as and the user interface was really powerful, right?

You ask a question, you press enter and you get this output. And these outputs were, you know, astounding. I think today we’re sort of we are our our bar continues to to increase.

But I think most people will be able to tell you like exactly the first thing that they put into whether it was chat GBT or Gemini or something else. Where were you and what were you doing at the time? Yeah.

Yeah. Did you have you been in education your whole your whole life? I actually Sure.

We were all in education or you know the first part of our everybody’s lives generally. Exactly. I spent a long a long time in education.

I when I graduated from college, I I actually got right into teaching. Actually was in a teacher prep program the Monday morning after I graduated from college. And stayed in education, stayed in elementary and secondary education for about the next eight or nine years.

Mostly, I have to say, if I’m being honest, wanting to get the hell out of it. As soon as I got in, not quite sure how. I knew I wanted to write.

Knew I wanted to do something different. And about, you know, like I said, eight or nine years in, I finally just said, "Okay, I I need to change what I’m doing." And so started freelancing, doing lots of magazine pieces, and finally got my first staff job at a newspaper in New Mexico at the great great paper, still exists, the Santa Fe, New Mexican. And really just just loved it and had the time of my life there.

Spent three years there and then went to Baltimore to work at the AP. Spent two years at the AP in Baltimore, two more years at the AP in Washington. Was there just in time for things like the Clinton impeachment and ended up covering some of the Monica Owinsky stuff.

We we actually I as luck would have it I was the only person in the Baltimore bureau who lived a couple of blocks from the courthouse where the Linda Trip wiretap case was taking place. So I ended up covering that and on and on and on and on. Spent and after the AP in Washington spent 15 years at USA Today and was the national education writer there.

During that time wrote a book in 2015 that looked at the possibilities in learning games and really got kind of like a graduate level education all my own you know just self almost entirely like self-taught if you will about the kind of how that works how that sector works gamified learning Yeah. Yeah. And really had the time of my life.

By then, I was really wanting to do something different. Was kind of sick of the, you know, NCLB race to the top standards testing on and on and on. And just wanted to just explore something new and different and weird.

And that was really that was probably the most fun I’ve had in my career. And I’d love to actually maybe at some point talk about the parallels between what I saw then, this is now going on a decade ago, and what I’m seeing now with AI. Yeah.

So, you you found your way out. You’re covering national politics. We were just talking to a reporter yesterday from one of the big publications and people in education.

I think sometimes there’s a lot of stories of folks becoming cynical. Mhm. Education is a tough place for people who are well-intentioned trying to do work and have an impact and there’s just all of this systemic a lot of systemic challenges.

But national politics is its own cynicism factory. So you went from one very you know place where a lot of jaded people to back to education. But at the 74 you I mean your reporting is interesting because it’s a lot more it’s longer form.

You’re doing much more of a deep dive. You’re not like I haven’t seen too many process stories from you and is that is that because of your remitt at the 74? Is that something that you’re trying to be more intentional about in terms of I mean especially right now there’s so much to be to to cover.

Yeah. You know and AI is is not a a narrow space anymore. I mean how do you think about sort of the your beat as it were?

Like what is this the what is the frame or sort of the aperture that you’re using? Yeah, that’s a great question and and and I think in a way the DNA of the 74 million is that you know on the one hand we’re a national publication so we don’t have to be writing about the Cleveland school board or the or you know what’s happening at the you know the state board of education in New Hampshire. So we’re we’re not worrying about the those little processes unless they’re interesting to us.

So, so if there’s something happening on a local level that we want to dip into, we can dip into it. The perfect example to me is the the story I wrote a couple weeks ago about this case in Massachusetts with the you know, the student who was accused of students who were accused of cheating on a on an history project using AI. In a way a local story was in a you know, federal court in Boston.

But we found it interesting enough where we just said, you know what, we need to write about this. And yeah, I remember you reaching out to me for, you know, my thoughts on that story. And, you know, shortly thereafter, and it was around the time that your piece had come out.

I mean, it started to really gain national traction. Yeah. I don’t know if it’s actually the first, but it certainly appears to be the first, at least the first example of this that has been been talked about and covered nationally.

Mh. Do you know was this actually the first time someone has gone to court challenging an accusation of cheating because they use chat GBT? Yeah.

Yeah. Here’s a here’s the thing you need to know. We will never say first ever or or or never before or anything like any any superlative like that unless we can absolutely prove it because somebody always comes along and says, "No, no, no. you know, I was the first ever and I filed a lawsuit in 1988 and so, you know, we basically, you know, among the first or the known, you know, okay, first.

So, we will always hedge our bets. That said, I as far as we can tell, it’s it’s and it’s an interesting case, too. It’s an interesting case.

We’ll we’ll link to your reporting on this. I also wrote a Substack that sort of as a follow-up, you know, very much inspired by our conversations. The last and I and I I’ll admit I haven’t actually checked in on whether there have been any updates.

The last thing that I have seen was a preliminary injunction, right? That was denied by the judge. And I remember when we had talked about this case, you know, one of the things that you and I both sort of agreed on was we just don’t know enough.

Like we don’t like there’s we have one side of the story which is the student is you know claiming that they were unfairly penalized. And you know for for cheating when in reality from the students perspective this was purely just a misunderstanding and you know you know miscommunication or lack of sufficient communication from the school about you know what constituted cheating and I think there was a lot of reactionary outcry that was like ah this is completely unfair the student you know shouldn’t be penalized for the school just not being up to date or up to speed not having a good policy not having a good policy But when I read the preliminary injunction, the judge actually quoted specific excerpts from the school academic integrity policy where they say, you know, you’re not supposed to use, you know, technology. And they and sure they don’t necessarily specifically name artificial intelligence, but so there was even even the the outofdate academic integrity policy, I think arguably did cover it.

But then the other other sort of like you know detail I think was lost in some of the the national coverage is the school had actually had a whole class where they talked about appropriate and inappropriate uses of chat GBT right it was an English class I think it was an English class this in this in the and just for the record the kid had been punished in a history class. Yeah. And so so what we basically what the case kind of evolved into was to me very clearly the student cheated.

And this was basically a semantic argument about whether you know an a policy in an English class is transferable to a social studies class. And you know I’m not a student of the law and I don’t know exactly you know we don’t I don’t I don’t believe that there’s been a ruling yet in the case. The judge seems pretty aligned with the school based on his commentary in the denial of the injunction.

But you know I think I actually think that school in Massachusetts is an example is an exemplar of what most schools should be doing. And sure, maybe there, you know, I think there’s probably a little bit of constructive feedback in terms of overcommunicating in a mo in this moment is always good. And if you’re and and it probably does make sense to like, you know, clearly update your policies schoolwide and be really intentional about letting kids know.

But I also think that as a student, you know, there is a there’s there’s these gray areas even before Chad GBT. One of the one of the points that I make when people ask well you know Chad GBT shouldn’t be cheating because a student is just you know using a tool you know one of the questions I’ll ask and I’ll be curious for your take on this is imagine a student’s writing an essay and it’s an essay on cold war history and their parent is a professor of cold war history know where you’re going with this by and while they’re writing the essay their parent is sitting right next to them right and you know maybe not literally writing it but helping them write every single thing like sort of like you know provide them with all this feedback. You know not just reviewing it and providing like red lines like actually co-writing it with a student, right?

You know or there there there’s a sculpture project and their dad’s a sculptor. Yeah. Yeah.

I mean is that cheating? I mean I I like if you were the teacher in that situation. No, I think it’s a great it’s it’s a great great question and I think the the to me the kind of the short answer to that the short response is you know transparency says you should indicate that you know I got help from my mom who’s a cold war historian or you know here are the contributions that that she you know that she was able to offer.

And I, you know, it’s interesting because I think a good teacher, maybe this goes back to, you know, kind of like basic principles, like a good teacher would probably know like, you know, Alex’s mom is a Cold War historian, so he’s going to write a killer paper on you know, the Cuban missile crisis. And I think in in a way maybe that’s beyond the pale for most teachers. Maybe they don’t need to know that much about their students, but you know, it’s probably a good smart teacher would know that.

It’s kind of the same thing. I think you can ex you should be able to expect from your students a kind of transparency about their work and their sources and and you know the advantages they’ve got because the student didn’t site didn’t appropriately site their use of chat GBT and that was one of the things that they had covered in English class was you know first of all if you’re going to use chatgbt you know you have to I forget exactly how they framed it but it was like the the the spirit was, you know, you can’t just type in one sentence and press enter. Like the goal is that you’re, you know, and if you’re going to use it, like you have to site it in the same way you you’d site an encyclopedia or some other sort of like online technology based information product.

And so I think that there is but but there’s generally not I think the point with the parent is there’s not a right answer, but there is sort of the goal of education, the goal of an assignment is ultimately a student, you know, sort of this this like sort of productive struggle. Yeah. And I think and I think this is where I’m very I’m I’m very biased towards teachers in the debate about whether chat GBT is cheating because the answer is it’s complicated but ultimately I think teachers actually have an instinct and to your point you know teachers know their students and like they generally have a sense for whether something is like you know kind of wildly out of scope of what what that student was able to do.

And I think the challenge is the parents, students are looking for sort of a blanket answer to a question that I think society and industry has not even really quite answered because we have not figured out. I have I have not seen any company that has like really honed in on what does it look like for humans and language models to sort of co-create together. And so in this moment where we’re actually figuring this out together, I think there needs to be some latitude.

And I think you’re right and and you know I mean another way of thinking about productive struggle is you know effortful thinking right you know this is what school is about right this is you this is basically practice for your brain you know solving these problems. And I think if if we’re relying on a technology just to to get the student to the end to the end of the assignment, then we’re kind of we’ve lost the narrative. And to to assume that kids aren’t going to do that is naive.

Another reaction is, well, if students are cheating, it’s simply because the assignment isn’t it’s out of date. And I think there was an assumption that this was just sort of a standard book report, right? But again, in the in the judge’s denial of the preliminary injunction, so more was revealed about sort of the nature of the assignment.

And my understanding is the student got to choose a topic, right? They were working in groups. So there’s a lot of components that we would actually generally advise teachers like how do you design assignments and and work that is more meaningful to students.

I mean there was a lot of agency provided to students and it was also multifaceted and interestingly the teacher the reason the student got such a bad score on that assignment was because the it wasn’t even the final assignment that was turned in. It was the the research and the outline because I think the teacher’s mindset was it there were a few different ways that they could present what they learned. Mhm.

And so rather than grading everybody on just sort of like whether their PowerPoint was good or whether their like you know report or their write up was good, they were focusing on the process of gathering the knowledge, right? And so that that’s again like something that we’ve you know that’s advice that we often give teachers is, you know, you have to focus less on the work product and more on the process, right? So could the assignment have been honed?

Perhaps. But again, I I I find this an example of you know, even when teachers in schools are doing everything right. You know, kids are crafty and you know, I think there’s you know, I guess you you should get maybe some points for trying.

I think this was a little bit ham-handed. Yeah, I I guess it was ham-handed. I guess I guess there’s also a concern that teachers that that and I think this is actually reality.

I think some teachers are incredulous at the idea that students would cheat. And so rather than interrogating, this is actually Victor Lee’s research out of Stanford and and Victor’s research, which I want to talk about by the way, found that incidents of cheating postJBT have not increased. Mhm.

And the the the hypothesis is that students are always going to use sort of like the latest tools to cheat. Like when I was in high school, we had our lunchrooms were these round tables. And so, you know, before our AP European history class with Mrs. Janini.

We kind of had a pact where like at least one person would make would have the homework done and everybody would sort of pass the the sheet around and sort of like copy the answers. And you know that was sort of an old school that was like pre social media and and you know now students have Tik Tok and and other copy and paste your early copy and paste. So I so I think part of the argument is that you know teachers are always going to the worry is that teachers are taking it personally that it’s in front to them that students would cheat when researchers like Victor Lee would say you know teachers need to create learning experiences that students you know in sort of inherently understand and connect with their own world and that are meaningful right I I love the word I love the word I I I’ve seen in number of instances around this the word artifacts.

Because I think that’s a to me it’s a powerful word. It’s like what are the artifacts you want as a teacher? Is it just if you know in Alex’s history class you just want you know, eight questions answered and if everybody gets the right, you know, if everybody copies and pastes the right answer then then the work is done.

Is that an is that a satisfactory artifact or do you want a different kind of kind of artifact? Do you want something where you know it individualism is essential? And I so I I I think we’re we’re coming up against that in in terms of like what is what what do teachers feel is like the satisfactory proof of learning and I mean in a way like I I hate to say it this is a really interesting conversation for teachers to have and for parents to have and for students to have.

So maybe like all the maybe this technologies really like forced forced that upon us. So let’s do it. Let’s let’s have that conversation.

We certainly weren’t having this conversation 10 15 years ago. That’s right. So maybe so in a way progress progress is upon us.

Yeah. And you cover the future of is it high school or the future of school? Future of high school.

Future of high school. I I’ve been a part of a lot of conversations with folks and there’s different there’s different sort of monikers for this. There’s school transformation reimagining education, reimagining schools, schools of the future.

And those conversations have been going on and you can attest to this for for decades. Yeah. Centuries.

For centuries. And you know, as a nonprofit, as the the as the head of a nonprofit, I spend more time than I care to admit, you know, focusing on fundraising. And one of the phenomena in the education space, in K-12 specifically, has been a shrinking of the the number of philanthropies, the amount of money that philanthropies are giving to education.

And in part because they’re sort of throwing up their arms. They’re just like, we’ve been trying and we’re just not seeing the results. It’s just too hard.

And just there’s it’s very multi-dimensional. You’re trying to, you know, address math scores, but, you know, the schools that you’re working in are dealing with like, you know, literally a hundred different other sort of connected issues. Yeah.

And I think philanthropy has been been having trouble sort of like pushing that forward. I think, you know, at a certain point people are just exasperated. Not just the funders, but the the stakeholders and the parents and all the folks that are trying that are that you need on board for this.

Mhm. And I’m curious if you see this, but from my vantage point, the powerful thing about artificial intelligence perhaps even more so than what it can do in the classroom, and we can talk about sort of like what how AI is going to transform the classroom, but to me, the bigger the bigger part of this moment is it has actually reinvigorated interest in this conversation about the future of school. And suddenly it feels relevant, tangible and actionable and you know conferences are selling out.

Mhm. And I think the challenge is how do we make sure that the conversation stays on this question of what is the future of school without moving too quickly to the answer of well it’s just AI tutors and hyperpersonalized learning and just like throwing you know products because because basically I think the space has been inundated so quickly with generative AI products that you know while that and I’m curious for your take on this right is like like AI products certainly a component perhaps of this conversation like how much of the conversation with the future of school is truly about like the AI tools that we’re going to use as opposed to some of this more maybe philosophical or almost like the spirit this the conversation about like what is the spirit of what we’re actually trying to achieve. It’s a really good question.

So I have many thoughts about this and I think in at at basically at at I think at the the very base of this you’re right there is a there is a lot of interest in like what this is going to do like I think people’s imaginations in 2022 were really fired by this to your point and now they’re wondering okay how is this going to apply to my kids school like what are they g what could they possibly do with this that will make a difference I will say that it’s probably a safe bet to say it’s not going to make much of a difference just in terms of you know like you basic achievement in NAPE scores and we’re already seeing you know just last week we saw the you know lot of terrible disappointing very bad no good NAPE scores. Can you just for our audience explain NAPE? Sure.

A national assessment of educational progress. It’s a test given in in various forms every you know two years or so. The it it basically tracks fourth and eighth graders reading and math skills and as we’re seeing you know now 5 years out out of the beginning of the pandemic they’re just not really recovering in a lot of ways.

In a couple of places they are Louisiana I thought was a really interesting outlier where their skills of those kids and I can’t remember which group it was but really defied sort of the expectations but mostly you know we’re in a sort of a like a very kind of grim holding pattern and it’s been like that since actually before the pandemic since like 2019 I want to say certainly the pandemic didn’t help. So, but I I so I don’t think on the one hand it’s going to really have much of an effect these new tools, these new ways of thinking on that because feels like very little if anything has any effect on that. And I’m not I mean I’m not quite sure systemically why other than that you know I mean there are lots of people who would say you know well you know we keep we keep judging the the the standard for what it means to be basic and advanced.

And so, you know, we’re always chasing this this standard that keeps incrementally increasing. But I I I think to me the the the really interesting question is is like what do we expect in terms of these tools and like how much of that are we expecting to be studentf facing? Right.

I I would have thought at the beginning of all this two years ago, two plus years ago, that it was going to be like all student facing like we’re going to just get, you know, tools that students are going to be using, you know, you know, 24/7. And there it’s just going to be just this sort of amazing revolution of kids just interacting with the technology and just doing different things and different kinds of learning and it’s just going to be, you know, we can’t predict it. I think to me what I’m seeing which I didn’t predict and that’s just my lack of imagination is to me the most interesting piece of this is not the studentf facing stuff it’s the teacher facing stuff it’s the it’s the applications that use AI to help teachers do you know the kind of the basic tasks of life you know the sorting and the and the the you kind of in some sense the drudgery in some sense figuring out what students should be doing at a given time and that can be as basic as like you know predicting who’s going to be absent or you know I mean you’re you’re familiar with all these things to me those are the things I think we should really be paying more attention to and in a way it’s interesting because I talked to someone who a couple couple weeks ago who said, you know, AI is weird enough right now and is so unpredictable that I think until we like figure out more about it and how it works and and really what it does, I don’t think we should put it in front of students at all.

Like let the adults mess with it. But students shouldn’t have access to this. Anyway, I don’t know how you feel about that, but this is a well there’s this assumption that we are the AI education project and so often times I actually will often now start meetings with like this is what this meeting is not.

We are not here to try to put AI in front of all of your students or or rather we’re not here to accelerate that process. There’s I think a lot of questions that need to be answered in terms of how to do that appropriately, safely, but also in a way that compliments rather than replacing education. And I want to talk about like an anecdote specifically where that like mean replacing teachers or replacing when when you say compliment replacing the the that compliments teachers.

Okay. Yeah. But also that compliments the learning process, right?

I think you know the there is a world where you could imagine students and in fact if you go to China they actually I remember this there’s this picture that hit me really viscerally it was students preparing for one of the big national exams in China and they had this giant like I don’t think it was even a gymnasium I think it was a warehouse and there was like 3,000 students in cubicles okay with headphones on getting personalized learning and it just hit me that even if the scores are higher you know, the idea of just a student sitting down in front of a computer all day, you know, digesting information. There’s something really wrong with that. And I I I see sort of flavors of that with this sort of obsession with AI tutoring.

But it’s complicated, right? Because actually AI tutoring is still really exciting and interesting as an afterchool support. As a support for neurode divergent students or students in special education more broadly if you’re blind if you’re an English language learner if you’re deaf right the this technology is sort of unequivocally going to add incredible value and in sort of like giving you access to learning that was otherwise you know much harder to access.

So, so it’s very hard to say blanket, you know, AI bands. I think AI bands, I think, for that reason are sometimes looked down upon to me that if I was a superintendent, it would be more like we’re going to sort of just like pump the brakes. And you you described it really well is it’s let’s we need to understand this technology.

We can’t just rely on the what the vendors are telling us. And outside of the vendors, I mean, I’ve talked to most of the smart people who are really obsessed with this question. And most of the smartest people will admit that we don’t have the answers.

Yeah. And most of the vendors, by the way, in 12 months time are going to be moving on to something else if they have any capital left. Yeah.

Right. Right. And and so I think the but at the on the other the challenges if if let’s say we’re superintendents, we’re trying to design you know the this this high school of the future.

Maybe maybe we should do let’s put those hats on. Mhm. We are let’s go through some of the decisions we have to make.

So decision one we we don’t have any new budget. We have technology line item. We’ve been spending it now on things like smartboards Chromebooks for students.

We’re keeping an eye on the is it ENET e the government there’s a government subsidy for technology access. Oh -ate. Yeah.

We’re keeping an eye on e-ate funding. We’re worried that that might expire. So we don’t have too many extra dollars to spend, but maybe some stuff we can move around.

And we the first decision is do we we need to we need to send some sort of a letter to parents and students explaining whether or not students are allowed to use ChatGPT. We one step short of like writing a integrity policy, but but what’s sort of step one? Do we blanket ban while we figure it out or do we encourage students to use it but ask them to be transparent about how they use it?

What’s I’ll let you go first because I think you let everybody download DeepSseek onto their Chromebook and call it a day. No I I I this is a really good question. And for our audience, Deepseek is the the language model that was released by a company in China two weeks ago.

Now, technically in December, there’s an earlier version. Lots of privacy concerns there. Yeah.

So, and it was built for, you know, cents on the dollar, comparison to, reportedly, although there’s actually some new reporting out. We, we’ll come back to that. Yeah, we can come back to that.

You know, my feeling is I don’t think you send out a a memo to everybody to the to the parents and the kids. I think the first to me your first step is you give teachers some time to play around with it. And you you really give them the opportunity to find a comfort level with it.

Find what is the most you know likely way they might employ it. Let them figure out like what the risks are. And I’m not saying like security because I don’t think most teachers you know know enough about that.

You can read the 74 million to find out that because we cover it really well. But yeah, I think you I think you start with that. I think you start with teachers just just playing with it.

And and I really do mean playing and we can you know I I want to talk at some point about the parallels that I’m seeing with the reporting that I did 10 12 15 years ago when games and learning because I see a lot of similarities with that and with with AI. But I think I think you do I think you let teachers mess around with it and see what pops out and maybe as part of that you require them or strongly suggest that they like get together you know once a week once a month and just let each other know you know what the you know what the coolest aspect of this is. Okay, I’m in line with that.

So, we don’t ban, in fact, we just avoid that whole the all the controversy around that by just not addressing it because anyways, the kids are going to access it out of school. Yeah, they’re going to access it on their phones. The a ban only really makes sense if we feel confident that we can enforce it, right?

And then the other thing that a ban does is it orients teachers to this defensive mode where rather than focusing on like learning the tool, they spend more time focusing on enforcement. I think my modification to what you’re describing is and this this is this is a plug. This is the work that I eded.

You know we we have cohorts of teachers and we facilitate those professional learning communities so there so that there’s some structure. Yeah. I don’t think you need to work with nonprofits like us to to create those sandboxes, but I think it does help to have, you know, some subject matter experts who can sort of like sort of helpfully guide teachers towards you know, some of the best sort of like some proven ways to actually sort of like like some some scaffolds for their exploration.

Yeah. But the other reason I think that’s important is you know ultimately what is what is likely happening in our district is teachers and maybe even with the way we discovered you know Chad GBT I don’t know that we were covering you know follow my guess is as a superintendent we’re not reading wired or you know MIT tech review right maybe we saw a headline about chat GBT but we probably be before we actually got the chance to use it my guess is a teacher went to the lunchroom and was just like look at Sally’s paper. It looks like it was literally written by a college professor and I asked her, you know, I confronted her about it and she showed me this crazy tool called chat GBT like check this out and but it’s with that frame of like students are using this to cheat and but I but I think this idea of I think as as a superintendent or a co- superintendent my answer to those teachers who are like well how do we deal with this?

It’s like you you’re trying to design a bank that is, you know, impervious to being robbed. It really helps to consult with a bank robber who is, you know, who is really thinking about how to do this. You you have to get into the mind of somebody who is trying to cheat.

And the way to do that is you need to use it yourself. There is no sort of cheat code, pun intended, or or or shortcut or rule that we can put in. You’re probably going to have to, if not definitely going to have to adapt some of your assignments, and you’re not going to know how to adapt those assignments to be more resistant to cheating if you don’t have a sense for how the tool could be used.

Okay, I like that. Yeah, I I’ll go for that. Find find the kids who are who are doing the most mischief with it and maybe even you know have a have some sort of like a regular round table or something like a white hat.

We’re actually doing that at one of at the a couple of upcoming conferences that we’re doing sessions that we’re we’re having teachers I think we’re literally calling them like cheatons where we’re going to have some like you know sort of like traditional assignments and we’re like we’re going to spend the next hour literally cheating on these assignments and we’re going to show you how how you could do it sort of get into the mind of of a student and we might even have some students get involved in that. Love it. How are there you know as this Y superintendent who is who was around for the time of you know gamified learning what are those par you mentioned you saw some parallels there like how like can you just talk about that?

Yeah. I mean, you know, just like you’re sort of suggesting here, I mean, the the the place where 10, 15 years ago, games and learning were the most interesting and where things were getting done that I found really exciting were places where an individual teacher had discovered something on on their own or had discovered something with a group of friends or went to a conference and was like, "Oh my god, you got to play this game." Or met some developer or or became, you know, a couple of them actually became like, you know, game developers themselves, just literally teaching themselves how to build this stuff. And I I loved the excitement around that.

I I and I also loved what kinds of conversations were happening because of it because a lot of instances when teachers discovered really good learning games say in something especially something like math which is really easily adapted to a game. They were the teachers were asking totally totally different kinds of questions than they had been before then. They were really in a in in a way kind of like turning their curriculum inside out because this one game or this couple of group of games had just basically asked them to say to themselves like what is algebra like like why do we have algebra and what the heck is it and there’s one game in particular which I that I wrote about in my book was Dragon Box.

I don’t know if you’re familiar with that. So, this is by a French developer. And it basically asked students to kind of like almost like invent algebra from the ground up.

And almost like discover like what x is and what x + 1 equals y is. So it really just kind of like I said, sort of turned everything inside out. When you have students doing that, when you have teachers getting excited about that, it’s totally different kind of conversation than why can’t my kids do algebra?

Why aren’t they interested in algebra? And I and I see like poss I see the possibilities for that kind of thing with with something like AI. If students can like you know come at a math problem from a totally different point of view or you know do different kinds of problem solving then I think know then I think we’re really in in like a different place may and you’ve probably you know experienced some of these things.

I mean, you’ve probably seen teachers, you know, looking at some of these tools and getting, you know, pretty excited. I mean, I the the easy way to do it is just to ask ChatGPT to or the easy and stupid way to do it is just to ask ChatGPT to, you know, give me 10 algebra problems, right? But there are other ways to to to recruit these tools to do, I think, much more fascinating much more fascinating things.

Yeah, the so I’m going to sort of draw out from what you’re saying that stands in contrast to you know presumably there were lots of companies approaching districts and saying here is a you know gamified learning product going to improve your math scores you should put in front of every single teacher. Yeah. And my guess is that those districts, as is the often the case with new curriculum, they encountered a lot of push back from teachers.

It worked in some classes, didn’t work in others. Yeah. And and so as we the challenge is this is like you it’d be really nice if the answer to this question of like how does AI going to transform learning if it’s just like well you just give teacher, you know, just have teachers use AI to do XYZ.

They’re going to have more time. They’re going to personalize or use use it to personalize learning XYZ ways. And then your goal is just get that in front of every single teacher, train them on how to use it exactly in that way.

And you know, something that’s repeatable, prescriptive, and for, you know, districts like Prince George’s County where we’re just up the street from here, where we’re doing work, you know, they have 10,000 teachers. Yeah. I I suspect that the answer to the the best outcomes in terms of how AI will transform education is similar to what you’re describing where it is it is sort of you you create the conditions for teachers to sort of hyper customize and explore and and and use it as a as a force multiplier, right?

With their own spin. And I guess a challenge for for superintendent for schools or for school leaders is you know for PGCPS they have 10,000 teachers you know you start with the sandbox that learning community of 50 teachers but you really want this to get to every classroom and in fact you have if you have a mind for you know reaching the the kids who are at risk or most at risk of falling behind they’re very often in some of the lowest performing schools that really struggle to retain the brightest teachers. Is there anything whether it’s from your from your time in sort of like gamified learning, but is like how can this sort of get into the bloodstream?

Like what are the conditions for I guess the way to describe this would be it’s almost like what what is your understanding of some of the best ways to build and enhance like the teacher workforce and upskill them in the ways that they need to explore and understand and sort of like iterate on the technology in their own way. I have a couple thoughts. By the way, and I’ll just preface this by saying I think a lot of people in the game games and learning space would take exception to your use of the word gamified.

They they would say that gamification is is kind of a lesser form of of what they’re doing. They’re actually using games in a at a at a higher level than just sort of, you know, getting people to sort of behavior be behave a certain way. They they would say that game gamification is a little too low level.

We I don’t want to spend much of our time talking about that, but I I it’s interesting you you use a phrase a second ago get into the bloodstream. And I I guess I would maybe without your permission like maybe explore that a little bit like you know so you’re talking about like you know like a virus getting into the bloodstream something being viral. I mean why why are things viral?

You know in in the sort of social sense they’re viral because somebody’s really excited about it or it’s new or different or weird or just you know getting something done in a way that we’ve never seen before. And so people get excited about it. I’m not sure where you go from there.

But I I I do see the the like almost like the necessity of getting people excited and maybe I mean maybe it’s like just one big natural experiment you know like where the people who are excited about something you know test it out and try it out and you know have certain amount of success and the vast majority don’t have access to it and their either their results aren’t as good or their maybe their teacher satisfaction isn’t as high or you know I I I I am not familiar enough with sort of the way systems like these work to be able to say yeah this is all you need to do is just you know well I think even the deepest experts in teacher professional learning will would struggle to answer this question right there is no there there there is no quick fix is the is the challenge I I this this idea of like the the components of virality it has to catch people’s attention. Yeah. In education, you know, there’s it if something works presumably, you know, that is the type of thing that teachers are going to be talking about.

If something doesn’t work or for whatever reason they’re perceiving that it doesn’t work, then they’re going to they’re going to resist. And also they’re just in many way, you know, especially your teacher of like 10 plus years, they’re on cruise control. They’ve been teaching this class.

It really is going to take something profound for them to change course. It’s a lot of work. You know, you kind of have your your script, your lectures down pat, you have the homework assignments.

So, it’s it’s it’s not trivial even if something is pretty good. And and I guess so, so I do see the I do see the the the lever is this idea of if you find the right teachers, they actually can be catalysts. I I think I bought into that.

I’d be curious, you know, we don’t have to answer this now, but I think one of our wonderings, if you come across anything like this in terms of, you know, analoges that we can find, where, you know, really successful, whether it’s curricula or like teaching practices, project based learning, we we’re big proponents of that. Our most of our curriculum is PBL or inquiry based learning. A lot of the schools that we work with have struggled mightily to get project- based learning into the classroom.

And I think the answer is in part is because they just have project based learning is sometimes perceived to be at odds with like the standards and all the content knowledge and the rigor that teachers are expected. And so their teachers are getting these conflicting messages where they’re being told you need to have more project- based learning and then they’re also their salaries in some cases are literally determined by how their students are scoring on these like very rigid standards. Yeah.

Okay, being a superintendent is really hard. So I think this is our our chance to take off those hats. Thank Thank God.

Okay. So, I don’t have to be a superintendent anymore. Okay, good.

No, not anymore. Put your journalist hat back on. There’s so much happening right now in AI.

What are the stories that you’re just keeping an eye on? I’m especially interested if you’re willing anything that you maybe haven’t covered yet, but that you’re, you know, just trying to learn more about that we could sort of like dive into. Sure.

I’m actually going to be and this is not a self-promotional plug, but what the heck? At South by Southwest, I’m going to be doing a a panel with folks from Curriculum Associates and Khan Academy about u AI and assessment. And that’s one of the things that I think was really interesting about AI’s potential to mess with assessment and improve it and make it I mean the the word I would use would be sort of more invisible.

I I wrote a story a couple months ago about Cur Curriculum Associates buying this Irish startup that has developed speech recognition for kids and it seems like a minor thing but I think it’s kind of huge in a way because you know it’s one thing for you know an eight-year-old or nineyear-old to be able to ask Alexa you know to you play a song or turn on the lights or whatever you But imagine a three or four year old being able to do that. You know, I actually have a personal experience with that. I’ve got a three and a half year old grandson who has figured out how to very clearly ask Alexa to do stuff.

But I think that’s rare. And so they have developed essentially a way to to help with AI help computers understand the utterances of very small children. And this is, you know, it doesn’t take much imagination to think like, okay, if I’m a preschool teacher or kindergarten teacher, you know, this really revolutionizes what I can do with instruction because I can sit a an entire class down and just have them read into a, you know, a tablet or a microphone or what have you and you know, on their own, by the way.

And not, you know, not even in the presence of a teacher, just have them, you know, open up a passage and read it. And have the AI, you know, assess how well this kid is reading. That in a way that’s one of the examples of, to me, I was going back to, you know, something we were talking about earlier, like that’s not really studentf facing AI in a sense.

It’s it’s it’s AI for the teacher’s sake. And I think that’s a really exciting aspect of this. And I think it, you know, if you want to talk about, you know, force multipliers, that I think that could be really cool.

And I I’m excited about like the ways in which technology can play with assessments in other ways. You know, maybe make them absolutely invisible. Maybe you know, take let’s say work a student is doing already in in class and essentially make the work into the assessment, right?

Where you don’t have to stop and take a test on, you know, a chapter you just read because the system is already swallowed up all the responses you’ve just you’ve just offered over the course of a week or a month or what or a unit or what have you. You know, this, you know, getting back to the games piece, I mean, this is what games do, right? You know, games assess you constantly all the time from the, you know, minute one to the end of the game, right?

That’s one of the things people love about games, right? They’re constantly being assessed, but in a way that is invisible. And I think I think with AI, we have the potential to do that in a way we’d never have before.

Again, it’s not student facing. It’s totally teacher facing, right? But it it it does something we couldn’t do before.

Yeah. I’m I’m You’re skeptical or Well, I’m not skeptical. I I I just imagine the perfect iteration of this where you’d probably get the best outcomes from the perspective of like accurately assessing students would be to embed AI and obviously the students Chromebooks every single homework assignment you know if they have access to some kind of a chatbot or a tutor that data is all digested.

There’s probably like the classrooms recorded maybe with video as well. And let’s just assume, let’s take for granted that that we’ve absolutely solved the privacy aspect, right? Like the data is anonymized and so I’m I’m not my skepticism doesn’t come from a space of like, oh, like data privacy.

But in this in this world presumably we wouldn’t just use it for a individual class. The students you could you could use the same approach to evaluate the student over the course of their entire year and then over the course of their entire academic career and perhaps we don’t need college exam or you know perhaps we don’t need SATs or college applications colleges but in that world it’s it it’s kind of like the sorting hat you know it’s like you don’t you don’t you just you’re told here are the three colleges that you can choose from. Yeah.

And I I think you’re I I understand what you’re saying. It feels like you’re getting ahead of yourself. So I I’m absolutely getting ahead of myself.

I don’t know if that’s 10 years away, 20 years away. Yeah. I do think that it’s it’s it’s it feels inevitable and it’s it’s sort of like we just have to make a choice as society that that’s the direction we’re comfortable going down.

And if not, I suppose then the question is simply how do we build in a sufficient amount of student agency? And it’s it’s and I think we’ve we’ve done similar things in the space of like using AI to screen like you know resumes. You know I think at least you know in many places like even though you theoretically like we like AIU we don’t use any AI to screen our resumes even though we clearly could because we just made this decision.

Sort of like a normative decision about not wanting to surrender that process to to the AI. I just I just wonder if like are we sleepwalking into sort of this world where you know students are sort of just I mean there’s the the Iron Rand book where everybody’s sort of just told what their job is and I just Yeah. So your is is your take that’s just so far out it’s Yeah.

I you went very dark very fast with what I was saying. I guess I I I was just envisioning and I think a lot of people are envisioning just a just a a more seamless transition from work and thinking to assessment product. So I guess is there because maybe then maybe the answer is there’s some amount of this like because I assessments are clearly need to be changed right.

So like the the alternative the the flip side of this is the current assessment regime does not work and if we’re or it works in a certain way and that’s how it works and some of it’s great and some of it’s terrible. Maybe assessments it’s not that they do not work but there there’s a lot of room for improvement in assessments. And there seems to be a middle ground where we use technology to improve and enhance and address some of the gaps that exist in the education system.

And maybe the we just need to be sort of forward thinking about where do we need to sort of like draw the line. And yeah, no, I I would say that I guess I guess maybe the thing I’m proposing is this is a little different. Maybe I’m not describing it very well.

My feeling is that the best assessment is one that you don’t even know just happened. That you are assessed on something because you’ve been working on it let’s say all day or all week and at a certain point the system has enough information about what you know and how well you know it that we’re done we’re done here very very much again like a game right you know there’s that great quote by you know Jim G who’s a sort of game scholar he says you know if I get through every level of Halo. This is going back, you know, 10 15 years.

If I if if I get through everywhere level of Halo, you don’t need to give me like a Halo final exam. Like, I got through every level. I’m done, right?

I mastered the game, right? There’s no like final test. And and I think that’s I think what a lot of people are envision are envisioning.

Now whether that’s got you know ne necessarily you know an Rand implications to it because you know the student isn’t choosing to sit for an exam. That’s that’s another really interesting question but and I guess I’m not quite there yet. I’m I’m just thinking of like what is the what is the process we were able to just knock out and what can we replace it with?

Yeah. Okay. I think you’ve convinced me that I am this sort of catastrophizing is maybe value valuable for futurists and people really but actually I think I could also imagine the flip side where 10 15 years from now people are like there was a time when you spent all this time in in school and you did all this work and none of that mattered.

All that mattered was just the score you got on this one test on one day. And if you happen to be sick or if you happen to be having a bad day, didn’t matter. Or if it was a bad test.

And and oh, by the way, some of these tests are not very good. Yeah. And so yeah, I I think I think there’s a lot of there’s a lot of power.

There’s the the other piece about the sort of recognition like like sort of natural language processing. Mh. One of the things that people have been saying about, well, assessment, assessments need to change, homework needs to change, you know, we need to do more oral assessments, right?

And teachers will often kind of be a little bit they’ll scoff at that because they’re like, I just don’t have the time. Yeah. It’s I just like like homework is efficient because people are doing it simultaneously as opposed to like one student at a time.

So this is perhaps an example of a problem that AI can actually a problem that AI perhaps created but that it can also help to solve. I don’t generally think that about every single problem AI creates but I think this is this might be one of them. It is a bit challenging when have you have you seen this has been I think it’s one specific person but there’s I think probably more than one on social media.

Applying for like management consulting interviews. And he has like his setup where he’s like chat JBT live transcription of all the questions that he’s getting and then it’s like answering the question and he’s sort of like reading it oh aloud and sort of like literally doing the interview like acing these interviews with the help of his sort of like chat GBT module. That’s great.

That’s great. But yeah, that’s amazing. Yeah, you know, I I guess I am of the opinion that and I and I’m sure there’s research on this, maybe you can even cite it now, but that there’s a lot of wasted time in school and I think the product of that is disengagement on the part of kids and dissatisfaction on the part of teachers.

So, if there’s a way to to close that gap or not gap, but there’s a way to to to to get rid of some of that wasted time and you know, and use it for something much more interesting. I’m not going to say productive, but right then then I think that’s that’s a really interesting possibility. Yeah.

Teacher retention is increasingly I don’t know if it’s getting worse, but it’s certainly really bad right now. Yeah. You wrote a story a couple weeks ago about a survey that come out.

It was serving teacher schools of education and I forget these specific numbers but it was you know startlingly few colleges of education have any kind of a formal curriculum or approach to talking to teachers about language models or generative AI let alone actually teaching them any anything. And I was curious if any follow-ups on that have you No, I haven’t. But but but to me that you know on on the one hand I would say it’s it’s not a surprise because I think a lot of teachers colleges a lot of teacher education programs are just you know they’re they’re they’re kind of trapped in the past.

I don’t think they’re they’re they’ve got much incentive to do something new and different and and and innovative and some of them probably with good reason because they’re being asked to turn out a you know a certain kind of product. I I guess stepping back a little bit from that I would say not sure if it’s the job of a teachers college to teach a you know a 23 year old about artificial intelligence. I mean we’re all learning about it on our in our spare time and you know and you know over breakfast and you know like it’s to me it’s it’s kind of a in a way it’s sort of like a responsibility that we all have just to you know just to like we follow politics and the weather and you know climate and culture and music and all the rest.

Feels like this is just another thing we should all be kind of learning about at a very basic level. My kids are interested in it. They’re they’re by no means like technically savvy.

They’re just they just think it’s an important thing to learn about. So I don’t know. I I’m not I don’t want to say I give teachers colleges a pass.

I I just think it’s it’s just something it’s in the air. So seems reasonable to expect that that young people will just figure this stuff out at a certain basic level. Not you know saying you like they don’t need to learn you know all all the ins and outs but they they certainly need to know how it’s used and what it is and you know the difference between the models etc etc etc. Yeah, this is we we recently over the past like two years we started shifting from talking about AI literacy to AI readiness.

And and that insight I think is sort of part of the the the reason behind that and is is this idea that you know when I close my eyes and and try to look into the future yeah I see a world where AI is just it’s so ubiquitous and seamless and invisible it just seems it it’ll feel very like quaint that we would have to teach people how to use it in the same way that my guess is that teacher colleges you know 20 years ago didn’t have a class called like the internet, right? How to use your how to use your iPhone. How to use your No, I think that’s exactly right.

Mobile is even a better example because that was you know the early internet was actually kind of hard to use and so there probably would have been beneficial for people to know but yeah nobody there are no classes called how to use your phone and and teachers colleges actually have a lot of important work to do which is how do you build those durable skills which is when we talk about AI readiness sure AI literacy is a part of it we’re assuming that that’s actually going to be maybe one of the easiest parts is like teaching people about what AI is and how to use it the more important piece is like building agency crit critical thinking and those sort of those soft and hard skills that you know I think most teachers whether they realize it or not they are they like they have the the the knowledge and skills to get kids ready for the future. I just don’t know that you know I 21st century skills has kind of been it was it was it’s been it’s been talked about so much that I think sort of lost its its punch. I I my my big sort of like scheme is this is our chance to sort of reinvigorate 21st century skills and sort of almost like reframe it as as you know AI is really just another way of talking about 21st century skills.

Yeah. I mean I think in a way to me I don’t know what as as I think about like how I use AI in like my daily life you know I’m by no means like a technophile you know I’m interested in it and I like to kind of think about it and explore it but you know in terms of like being a journalist like to me the most I mean once you dig in and start looking at these different models you or ChatGPT or any of the other ones Google the Google one fascinating, you know, that came out a couple months ago. You know, that allows you to create these crazy little 10-minute podcasts.

Oh, notebook LM. Notebook LM. We had a lot of fun with that.

Not only in my house, but at at work, we were just we were playing with just feeding our news stories into it and just seeing what it created. But like so once you once you get past that the sort of basic literacy with that and knowing what these things do and what they don’t do and how one is different than the other then to me the big challenge and I’d like be interested in hearing your take on this is keeping and keeping them at arms length. I think that’s really maybe the most important skill of all.

And what I mean by that is like I would never use one of these models certainly to write a story. Okay? That’s like verboten.

Okay? And I I would never feel comfortable doing that, right? I wouldn’t even feel comfortable like taking like a rep.

Let’s say, you know, some government agency, you know, sends out a PDF of a report. I wouldn’t even feel comfortable asking Claude to like summarize this report. Yeah.

Because like I I you know I who the hell knows what Claude’s process is in doing this in a way like you want to you almost want to do you know like a kind of like a test of it. You know, you know, you want to take something that you know really well inside and out and ask these models to summarize it and then just see, you know, what it’s what it’s going to tell you. So I I I really feel like there’s there’s almost no way in which like relying on these models for something with any stakes is a good idea for me.

Yeah. No, I others may others may differ. I would but I I feel like at this point, you know, February 2025, this may change in March or April, but I feel like, you know, the place of something like AI is very low stakes territory, right?

I just want to I’m curious about something. I want to see something. I want to get the the you know, the AI’s take on something.

But I’m not going to rely on it in any way, shape, or form. And I I I don’t know if that where that comports with your conception of this stuff, but I you know, it’s just it’s almost like a as a journalist, I think, have a kind of an like an allergy to relying on this stuff for for for something that really matters. Yeah, it well, it does.

The challenge is that it’s it’s really good in some in some cases. And you are a journalist with integrity who has respect for the craft, but I think I don’t know the study there’s I think this is very hard to actually put a put a number on. But there’s a large portion of the content now on the internet is AI generated.

And if you go and read any sort of like the just like sort of clickbaity SEO optimized blog posts, like a lot of that is AI generated. I know ad agencies are using AI to write things like ad copy. Like if you’re doing programmatic ads, you now can not just do an ABC test.

You can do ABCD EFG all the way like 300 hyper 10,000 hyperpersonalized. But then there’s like certain very specific applications where you know quality is not necessarily measured. There isn’t necessarily like sort of an integrity aspect and you’re not necessarily even as concerned about accuracy.

Yeah. So setting those aside I worry more about what you just described. So, if you’re reading a report and you’re going to write about it, the way to really be able to write intelligently about something and add value beyond just all the the noise because we’re in a a world now where there’s just like everybody has something to say about everything and it’s mostly service level.

To actually add value, you have to really like deeply understand the thing that you’re writing or talking about. And I just do not think there is any substitute. I was talking to one of my board members.

She’s she’s a corporate lawyer. She has her own consultancy now and so she advises you know really big you know blue chip clients and and she gets tapped to speak at these they’re like they’re esoteric you know legal conferences so she was like I think doing a conference on you know financial services you know policy in in Brussels and like oh I’ve got this like keynote tomorrow or this presentation I have to make all these slides I have to like down digest this like 70page regulation or something and I was like well perfect use case for AI. Why don’t you use AI to summarize it?

I hadn’t even gotten to the place where I was still sort of taking for granted that AI was good at summarization. I have since found that it actually hallucinates maybe 20% of the time whenever I ask it to summarize. And so setting aside the fact that that’s just incorrect and you should never use it to summarize if you’re trying to like accurately reflect what’s in a report.

But let’s just assume that even if it was 100% accurate, her point was my like process for preparing to go and speak and the reason I’m tapped to speak is not because I have I’m reading off of like a a series of bullet points or talking points that I memorized. I actually am going up and like speaking as an authority figure. And the only way that I’m going to be able to like speak authoritatively and intelligently about this topic, like the process of actually reading those 70 pages and like writing out everything that I want to say, like that’s how I get to a place where I’m ready to go and speak off the cuff.

And my experience with writing has been ab I just wrote a something on DeepSeek on Substack and I used actually I I’ll admit I used Chad quite a lot and the way I used it was I downloaded I watched a lot of there’s a lot of you good YouTubers who had sort of like done like much deeper longer like a lot of the coverage was like you know a thousand words it’s like hard to get into into the weeds. So I went and watched like hourong YouTube videos, got the transcripts, plugged that all in the chat GBT, had to sort of synthesize it. And then created my outline and then sort of like had chat GBT sort of expand out.

In particular like sort of like the technical explanations of, you know, how Deepseek was how we how we potentially how Deep Seek was trained. But you know, CJ and I worked on this together and I think what we found is that it would it would get you something that’s maybe like, you know, a C++, maybe a B minus. But to get something that’s actually worth publishing, we basically rewrote almost every word.

Yeah. Now it was helpful to have a little bit of a scaffold, but you know, at the end of the day, I almost I’m still unclear about whether the juice is worth the squeeze because I couldn’t have written that without watching all those videos myself. And I think that if you when you talk to folks, especially people who are really deep in the AI space, this is something that’s really common is like most people are not using these language models.

The exception I think is like coding or data science. I think there’s some applications where it’s just actually legitimately gamechanging. But I think for the humanities, I don’t know about lie.

I I’ve only I only know the story from the the lawyer that got disbarred or or penalized for some for submitting a I don’t know if it was a injunction or something, some some document that was like just wildly hallucinated and right. Making making up cases or what have you. Yeah.

So my guess is most big law firms are not allowing their lawyers to use. I’m hope hoping then they’re not. What else are you covering?

Let’s see. And if not covering, what else are you just like tracking? Yeah, we’re I mean, you know, as you’d expect, we’re you know, we’re we’re covering the new administration and, you know, seeing what’s happening there with, you know, their work with executive orders and and whatnot, the fate of the department of education.

We’re really still interested in the future of high school. I’m always on the hunt for suggestions for the next school to visit. Just was was just a couple weeks ago was up in Albany.

Visited a really really fascinating school called Tech Valley High School. One of my favorite aspects of it is that, when you walk through Tech Valley High School, you see almost no tech at all. It turns out that this school was built was originated, at a time when people were very very bullish on the idea of like incorporating schools into sort of like tech hubs.

And so there’s this area in Albany called Tech Valley, and the school sort of got got its name from that, but it’s very, it’s it’s almost, you know, a humanities centered high school. That just is saddled with the name Tech Valley. A really cool high school, really interesting, you know, individualization, lots of lots of project- based learning, lots and lots of real focus on students ability to get up and just like speak about topics.

Really interesting place and I think it’s it’s a model that I that I’m really seeing more and more which is schools that are focusing on giving their students the ability to sort of be experts if you will if if there’s someone listening that you know is the head of a school and they want to reach out to you what I mean we’ll we’ll share your contact information or your email address 74mmillion.org org. What what advice would you give to a building principal who is trying to pitch their school? Like what are the types of things you’re looking for?

I’m sure you get a lot. I mean, there’s We actually don’t get a lot of pitches. Interestingly enough, we’ve been writing these stories for for a while now, and we I I do not get emails from people saying, "Come visit my school." Maybe maybe I should be happy with that.

But no, it would be great to to hear from people. I think the problem is that it’s hard without without really spending some time with it, without talking to people to know if they’re really doing something innovative. Yeah.

I suppose more powerful for, you know, for you to hear about a school, not directly from someone who’s like sort of like promoting it. No, I think that’s right. I think that’s right.

Maybe just some parting advice for folks. I mean, I think one of the common threads of this conversation is, you know, you clearly have this curiosity. It’s been part of your entire career and it’s it hasn’t always been AI.

It was really sort of much more broad about sort of like the future of education. But you are now really, I think, one of like a very small group of of journalists who are true, I don’t know if you call yourself an expert, but as close as you can get to someone who like just deeply understands the the intersection of these spaces. What yeah what advice would you give to our audience in terms of for folks who are they’re early on in the learning journey they get it.

Yeah we need to sort of like learn more they maybe they want to use chat GBT and experiment but are there any like specific like place like outlets that you’re following you know I don’t know is it like Reddit like how can they sort of just like stay keep their finger on the pulse? I mean my problem is I read everything. So, what are your like what’s the best what’s the what’s actually interestingly enough you know one of the podcasts that I like is the Times podcast about tech the hard fork.

Okay. Love those guys. Kevin Kevin Roose I I think his coverage specifically is really really really well done.

And he’s covered a lot of the you education safety aspects of it. You know, I I cover or excuse me, I I I I read and follow, you know, the the skeptics, if you will. You know the Ben Riley, he’s got a cog cognitive resonance who you I I’ll I will actually if you don’t mind my flogging you guys’ incredible podcast that you did with the disagreement which I’m hoping by the time this runs it will be public.

You guys and I even said it at the time I think on Twitter X or wherever the heck it is that the conversation you guys had back in the fall was like the best conversation I’ve heard so far about some of these issues specifically things like tutoring and AI. I thought you guys were just remarkable in in the way that you both agreed and disagreed at a very high level with lots and lots of respect for one another and but also were able to very carefully like you know pry the the threads from where you where you disagreed and not you scream at each other. I I I really enjoyed that.

So, I would recommend that highly. I follow Dan Meyers’s blog. He’s got a very specific focus on math.

And I think that’s a really important thing to think about when we’re talking about this world. I’ve, you know, I’ve always been a big fan of Audrey Waters, who, never met a piece of technology she, liked. And it’s interesting because like somebody like Audrey for people who may not know very much about her, but you know, she’s been thinking and writing about this stuff for years and years.

And I think one of the the sort of the big themes she has really been focused on for as long as I’ve known her is this idea that Silicon Valley is not your friend. That there’s a lot of dark thinking coming out of this place. And I just speaking from you know the some of the stuff we’re seeing with our new administration.

I mean I without getting like political about it I mean I think you can see the selfserving aspect of a lot of what these guys are doing. And I don’t think that’s something we we saw very clearly until now. I I think it’s surprising they got it passed for as long as they did because I don’t think people would look at you know car companies, oil companies, you know, electricity companies like all these, you know, mega, you know, megalith you know, capitalist in institutions, they their incentive structures are highly oriented towards one only one thing.

Yeah. And they will push in whatever direction that they can, right? So Yeah.

No, I think you but I but I guess you’re I guess I think it’s interesting that you’re you’re naming some some skeptics because I actually think that is we have a lot to learn from the skeptics because there’s plenty of optimists and there’s a lot of legitimate optimists who are super smart but then there’s also a lot of optimists who are shilling right like that they are they are self-interested whether it’s they’re either at companies or invested in those companies invested in the future of AI being successful and and I think to kind of sift through all cuz like I my the the the origin of my question what I want to leave with audience is like yes there’s a lot of noise you don’t have to read every single thing you know not everybody is a journalist covering AI and education but I think it is important to find some smart people that you trust and when you’re trying to figure out is this like legit like is there a there there you don’t want to be relying on you know the you you don’t necessarily be on the bandwagon while you’re evaluating that and especially for decision makers who are education systems you know I think they’re dealing with this a lot right now a lot of AI conferences are you know 95% of the people who are there and the sponsors are all you know aggressively pro AI and AI products and in reality I think it’s much more of I don’t know if it’s 50/50 exactly but it’s it’s definitely not 95% good and 5% we just have to figure it out you know it’s a much more of a it’s much more complex. And so the information ecosystem is not I don’t I don’t think it’s very balanced. I think it’s oriented very heavily.

And then you have like the and a lot of the skeptics or a lot of the the pessimists are not useful because they’re like the existential risk folks who are worried about, you know, the singularity and I I find that a very interesting philosophical discussion. I just don’t think there’s anything actionable that certainly an education leader can do besides pay attention and keep keep your eye out for for AGI. And I think I I mean to be fair, I mean a lot of skeptics I think have a different view of education than maybe you or I might have.

I mean they they have a they see it as a different kind of enterprise. And some some wouldn’t even be willing to you know entertain you know the functions of a teacher you know being you know augmented in certain ways by technology. So you know I I want to be want to be clear on that.

I I I I guess where I come come into it is I’m really interested to to read the and to talk to the, you know, the the makers, the vendors, the the people who run a lot of these outfits just to see what they think they’re doing, like what, you know, where they what their place is and all this. And some of them have really interesting ways of thinking about this stuff. And really interesting ways of thinking like, you know, here’s our niche or here’s all we want to do.

Well, you I I guess you can’t plug yourself, but I I will. I think, while it’s not everybody can necessarily talk to, you know, the founders of of companies like this, I think, you know, reporting like yours, you’re not blindly excited. Mhm.

But you’re intensely curious and I think you’re willing to I think this orientation towards like the future of high school is like an interesting place to come at this from as a journalist because you know your in your incentive is for schools to transform for the better. And so there’s a desire for things to be I I can sense a desire for for us realizing the potential that AI has. And yet I think you’ve been in this long enough that you you know that there’s there’s generally more to it than you know some of the the first claims I suppose that are made.

Yeah. No, I think that’s right. I think that’s right.

All right, Greg. Well, thank you so much for coming by. This has been fun.