Coding without Borders: inventXYZ's Mission to Reach Every K-12 Student with Nikil Ragav

work. Nikil was awarded Penn's president's innovation prize to start inventXYZ NFL player Travis Kelce tweeted that he was sponsoring a STEM lab for underserved kids in Kansas City, and Nikil eventually collaborated with him moved to Kansas City to help design and

run the STEM lab. Now inventXYZ makes it painless for middle and high schools across the country to teach 100% of their students computer science and AI even if they don't have a dedicated teacher Nikil has also published a paper about teaching AI in K 12, with researchers from Johns Hopkins University, and we'll put a link to that on the show notes for this episode. Nikil Ragav Welcome to Edtech Insiders.

I didn't. And actually there was this post, just like yesterday, I was looking from a guy who goes to University of Waterloo, which is like the university in Canada. And he basically said the same thing that I've been feeling, which is, you know, Waterloo has all these amazing computer science and engineering graduates. But it's not because of the coursework, it's because they go out of their way to go and find side projects and build a portfolio and then find

internships and Co Ops. And that's like, kind of the problem I've been seeing from middle school all the way up to college is that like, you know, there's a lot of like computer science and engineering and stem like course content. But none of it prepares kids to do the right things to build real products. It's all like very heavily

theory. So like, my like specific story is, and in middle school, I wanted to create an Android app that like allowed me to define words just like on iPhone, where you can just click a word and press define. And for whatever reason, Android only made that feature two to three years ago. So my mom's a computer scientist, I thought, okay, like she can help me like she knows how to code. She gave me some book, and like, it sort of made sense. And then I was totally lost when actually

trying to make the app. And then in 10th grade, I took AP Computer Science A in Java. And I was like, this makes total sense. My teacher is great. She's explaining how everything is stored in memory, and all of that. And then I tried again in 11th grade. And I realized everything I learned in the 10th grade computer science class, and then the 11th grade, sort of like an advanced data structures class was almost totally useless to building an actual app because we didn't learn the

right things. We didn't learn how to fetch data from an external source, right? I was trying to get the definitions from Merriam Webster, I had no idea what would happen, like would my app crash if I didn't use all of the definition information? Like, how do I make sure it doesn't lie? Like I had no idea what the hell I was doing. And then I got to college and I thought, Okay, well, this is going to be different. It's like, you know, university is where people learn how to build

things. And the only reason I know how to build things is because I was on the Penn electric racing team. And we built an actual electric race car, but like, almost none of the classes are except for just like a couple of them how to build real products. And so that was super frustrating to me, and I wanted to kind of fix that and education.

it's too theoretical. So tell us how these experiences added up to lead to what you're doing with inventXYZ and how you sort of launched that through your pen boarding experience.

excited. And then I came out super disappointed, because like, they all had like essentially the same problems, either. It was actually just a big empty room with tables and chairs with wheels. Or they had some cool stuff. But the person who was like, initially there would leave after a year or two, and nobody else knew how to use any of this stuff. So just sit there MPs. So I initially thought, Okay, let me try to

solve that problem. And then towards the end of college, I ended up interviewing 54 superintendents, and assistant superintendents and innovation directors in a span of like three weeks, and realize that there's actually a little bit more of a nuanced problem in that, like, yes, the makerspace thing matters, but they also like actually need content in

the first place. And one of them said, like, I don't want this space to be used just like one hour a month, I want it to be used every hour of every day, give me hands on STEM project for social studies. And I was like, wait, I can do that. That makes sense. That's kind of

where the journey started. And then more recently, like there are now 10 states as of last month, that have passed a law requiring every student to learn computer science in order to graduate from high school, but like, Where the heck are the

teachers, right. And to what I was saying earlier, like, if we only have sort of like, part time teachers are people who like barely understand the stuff leading that class, then all kids are going to come out with is like learning some basic block coding, and maybe sorting some random numbers with a for loop, which is a massive far cry from what we actually need, you know, supply the high tech

workforce of the future. And also, it's totally not what kids need to build their own products and start their own entrepreneurial experiences, right. So really trying to solve that problem of like, let's solve that teacher shortage. And also, let's solve the quality and content gap.

money for it. And people who truly know how to code don't often become teachers, they often become coders, because that's a whole different kind of career path.

generation. But I find it very interesting, because I think people in your generation, very young realize what the pathway is you need entrepreneurial mindset, you need to be able to build products to test products, even the process that you just took, oh, I went and went to nine states. And then I interviewed all these directors. That's a product mindset. That's

user research. Yeah. And I think there's such a gap, you know, between what young people are thinking about and they know, they need to know to get these really lucrative jobs or very impactful jobs. And the education system just does not know that. So let's dig into what you're actually doing with

inventXYZ. So given your experience talking to all these people teaching computer science, how does inventXYZ build on that and tackle the challenge of sparking interest in technology and entrepreneurship and product among K-12? Students?

inventXYZ. And I had gone to like the National Superintendents Conference presented at the like school of the future segment and one of the superintendents said, Oh, I'm in Pennsylvania to I'm in and I said, Okay, great. I did not realize he was in a town of 13,000 people four and a half hours away from Philly. So, I was like, you know, I got a bias for action. So I screw it, I'm just gonna go drive over there and do it. And when I went there, you know, we had two

projects for kids. One was to basically build in code their own Pong game, and like algebra class. And then another was shooting a professional short film using proper lighting techniques in English class. Because a lot of times like English teachers have no idea what kids are actually interpreting the poems, right, they tell you what to interpret, but they don't actually know

what you're thinking. And when I went there, I realized, wait a second, these kids talk about essentially the same stuff as the kids I grew up with, right? Like, they have gaming, like GPUs, and graphics cards, they play like the same games, they have Instagram, they, you know, like, daily life is not that different. What is different are the opportunities post facto, and the expectations that they have for themselves, both in school and out of school for

like how far they can go. But like, you know, GitHub is a totally remote company, for example, a lot of gaming studios are like kind of semi remote now, like, the next best company could be in St. Mary's area, Pennsylvania, town of 13,000 people, and there could be like, I don't know, 10 kids, or like 10 adults at that point, making like three or 400k, you're in a town or a house only costs 100.

Okay, to begin with. Okay. So now back to your original question number two, basically, what inventXYZ actually does is there are these computer science mandates that every high schooler and in some cases, also, every middle schooler needs to learn computer science

to graduate. And so if you think about how that impacts schools, rural schools, in general, are going to just struggle to find anybody, most schools are going to have to hire more than one computer science teacher, because they just have that many students. And so even if they have one teacher, they're still going to struggle to find

additional staff. And as we talked about earlier, like, even if they have stuff that things that they're teaching are not actually solving the workforce problems, and like how to build real products, which is why like, companies lobbied for these things in the first place, right. So what we do is we create masterclass style projects featuring industry practitioners and industry

challenges. So for example, we have kids recreate historical empires using Unreal Engine, which is the software that fortnight was built in, it's being used in the Batman and the Mandalorian. Movies, architecture firms use it, we have kids analyze the DNA of a plant that's being attacked by a virus to figure out what gene could make it immune to that virus. And so this is a computational biology project, in partnership with a researcher

at Duke, right? So real scenarios, and then as students are doing the projects, we connect to them with virtual mentors who are current college and master students. And we try to make sure that the students we collect, have a variety of backgrounds and experiences, so that every single student has somebody in the mentor team that they can talk to and be like, I want to be like that person in 10 years. And they're all you know, sort of learning that engineering related field,

right. And we try to make sure that they all know how to build real products, like they have a portfolio like so they know, not just the basics, but like how to build real things. And then, the last feature of our thing is STEM project for social studies, right, all of our content integrates into existing courses that kids already have to take. So all of this means that a school district doesn't have to

hire any additional stuff. And they don't have to retrain their existing staff to become computer science and AI and electronics experts. Because our remote stuff takes care of the technical stuff. And it correlates with their core content areas, meaning they could run a project from on as a unit in their biology class or their history class or whatever.

English class. This is about social studies class, we have this, you probably have standards mapping, you know, how do they react to that?

subject area. So now, we're actually helping the teacher to like, make sure they cover enough content in some schools, especially where they're like a little bit more innovative looking for project based learning type experiences, they really liked the stuff. And I mean, in general, the feedback that we get is like, if you're going to put it in a core class, the standards alignment needs to

be really spot on. And it needs to cover at least a units worth of stuff if you're if the project takes almost a month, right? Otherwise, teachers We'll feel like they're losing time, which is a totally like, I think that is like sort of the right way to look at it. And it's definitely taken us some time to like find that balance of making sure we're covering things, but not in a trivial way. Right, it takes a while to figure out completely.

English and history. And then saying, you can learn these really important, you know, English and social studies concepts through this technology project. But the technology project is going to be really meaningful and going to give you a portfolio piece that you actually can use right to get a job. It's a really noble attempt. It sounds difficult, but it sounds incredibly exciting as well.

about. So, I mean, at the end of the day, the goal is for students to learn not necessarily for teachers to teach, right? Because the absorption is what they're measured on what they're measured on what the students are measured on. Yeah, and what the teachers and what the schools are measured on by the states, right?

the logo, right? And right there, right? Exactly, yes. And then secondly, you know, like, one of the schools that we worked with here in Kansas City, we did like a summer version of our projects there. And we did a project in Tableau, where students analyze a dataset that's interesting to them. So like, sneaker sales, or housing prices, or crime data, or Tiktok influencers. And one of the kids there was like, super getting it. He did like a custom formula, following our tutorials

and stuff. And then I was like asking him and interviewing him later. And he said, like, I asked him, wait, why are you in the summer class anyway? And he said, Oh, I failed two classes, algebra and English. And I was like, hold on, you're like one of the smartest kids here? How did that happen? And he said, well, because I couldn't force myself to pay attention and

care, right? So just flipping that key, like, you know, it's not usually gonna flip it for the entire year, but at least for the couple of critical units, if we can flip that and have students really care about what they're doing, what we're having them do is probably harder than anything they've done in their life in terms of, at least academics, right?

work together. So we ended up moving to Kansas City to help design that innovation lab. So my COO Annette, and then Joseph, whose are like electronics curriculum lead, were basically the teachers for a year and a half, working with the students directly developing new content, building tutorials, all of this stuff. And you know, we developed a lot of empathy for teachers at that time, because it's, I understand why it's hard, right? It's not something that's for the faint of heart.

For sure. Men, all the students, especially in some underserved communities come from, like, rough families, or they might have IEPs that are sort of undiagnosed, or they they might be diagnosed, but there's no way to like have special extra care for them. Because they're just

not enough people. So like, they're they're all these things that like we were trying to build a product that work for students with drastically different backgrounds from ours, which I think in the long run is super helpful because 40% American students are kind of in a similar boat, right? socially, economically and otherwise. And so we, we learned a lot there. And then, in the last year and a half or so we've been scaling to school districts across the country.

dedicated to Kansas City. He's also now beating Taylor Swift's, which has just made him one of the most famous people in the world, I believe, what strikes me so interesting about your approach to this, obviously, you've been thinking about product and coding, and this and entrepreneurialism, from a very young age, but you're also very opportunistic, and a lot of really clever ways. I mean, you're looking at all the different needs from these different constituencies, right?

This football player who wants to help his community, the teachers who want to be able to teach their subject, but not necessarily go through lots of computer science training, or that's not available to them, the school district leaders who have to legally teach computer science but don't have the staff to do it. So they need outsourced staff, even we haven't talked about this one

much yet. But these mentors that you've just mentioned, people who are current students learning computer science or AI, who are getting the chance to, you know, pay back to, you know, in high schools around the country, by supporting kids who are doing these amazing projects, I'm sure that has a little bit, maybe a lot of impact on their perspective to tell us about the experience of those folks,

TA. And there's also that like, the whole point about, you don't learn how to build real products in college, for example, we have students build real circuit boards here, and a lot of electrical engineers don't leave college knowing how to actually build an order a certain party.

100%. It's amazing. So you mentioned that Penn President innovation prize, and that's a really interesting, you know, aspect of your story as well, you've been, you know, right in the middle of Philadelphia, in the middle of Penn doing this work. you've navigated urban and

rural landscapes. I'd love to understand a little bit more about how all of these different experiences which I imagine going into college, thinking about computer science, you probably didn't see a lot of this coming, you know, how has this changed your perspective on education, technology on society? How do you see the world differently? Because of these experiences?

So they can just like, be me up on the screen and call me if they need help. But when stuff is going wrong, and you don't know what's going wrong, like it becomes really overwhelming fast. And so like the teacher didn't call me for like three or four days. And the issue that students were having was something that I could have fixed in like two or three minutes, right. So that like asynchronous, kind of like have the intermediary in there. Like

that doesn't work. And now with school districts across the country having like much more robust Internet, and like students also having much more robust internet at home, we've kind of changed our approach where like, we basically directly beam into each student's computer and they have headphones and stuff. And that's like, a lot easier. But it's also still really hard to debug

electronics over a camera. So now we have to develop processes for students to say how they've connected things in a repeatable way. That's, like easy to debug. Right? So like, that's another thing we've built. And then how do you make like, you get the same questions over and over again, how do you how do you do that more efficiently and faster

for students? So building a library of FAQ tutorials, and if students have the same question and kind of at the same time trying to route them to sort of the same mentor and small groups. The other thing I've like seen in sort of technology overall, like outside of how we're trying to use it to make our product a better experience

for students and teachers. is like, districts often talk about how or like you see reports that like Some teachers have something like 120 or 420, I don't remember what the exact number was, like number of distinct apps that they use. And like, that might be true. To me that like, doesn't sound crazy, because like, each one is just a website like, okay, like I go to

a URL. But for people who are not necessarily like comfortable in that, right, like, they may not have grown up in the same, like, there are only a few people who are roughly my age who are also teachers, I can see how that's like super

overwhelming, right? And how many of those tech products are actually providing educational value, versus providing like, some sort of form collection system, or like student roster management system, or, like, there's so much sort of like meta technology to manage the other technology, which I think is like, kind of missing the

point. Like, I felt this about, like, when I was going into college two, I chose electrical engineering, because I cared about computer science, but I also cared about the impact on the real world. Like, it doesn't matter if I'm making software that is only useful for other software, or for like, optimizing sort of minut things like it needs to actually have value to humans,

really, really working. And I think what you're talking about there is really interesting, this idea of this meta layer, that there's a lot of technology that's sort of there to help other tech or there for this system there for the schools use, which doesn't mean make it bad, but it does make it less relevant to the core mission of

the school of teaching. And it's an interesting lens to look at all of this through because I think districts all over the country are trying to figure out what their strategy is, what do they need to keep? What do they need to double down on? And what might they you know, let go or

let the contract run out? It's interesting to hear a perspective of somebody walking into this space and saying, Wow, you know, yeah, for you, hundreds of different tools doesn't sound overwhelming, you grew up in that environment for a teacher that's had each of these sort of land on them, and there's dedicated training for each one, and they have to use it for particular moments, it might be a whole lot, we are seeing teachers say that they're

pretty optimistic. I've been pleasantly surprised at some of the surveys about teachers use of technology, they're saying, hey, this, you know, it is working, students are getting a lot out of it. It's a lot. But like, you know, we think that there's something there's something really good, they're they're not saying get these things off of me. But at the same time, I think it's a really smart lens to say, you know, what are the core mission of the

educational institution? And how does tech serve that?

down. But one of them like the bar on the screen progresses twice as fast. So you kind of see that like, ratio piece for that kind of standard. Like, there are some really cool, like intuitive things that you can do. You can teach, for example, you can teach calculus, and specifically, like you can teach Multivariable Calculus without ever going into the formulas, right, because you can teach the intuition behind all of this stuff. Very, not complicatedly.

technology. It didn't sort of work for them as a business but they're starting to look into edtech and Ganguli from prisoners. We are We interviewed her for this podcast that will be coming out. Just believe next week. Oh, by the time audience listens to this, it'll be already in the past. She was amazing, really, really cool product. So yeah, I agree. There's some really amazing

stuff happening. And that's why we were you know, in this space, you mentioned and what you just said, we both mentioned sort of AI in passing throughout this conversation. You know, you're talking about electrical engineering, computer science, computer science projects that tie into core subjects like social studies, English math, but AI is obviously coming to all of us. We you know, we just had this conference about AI and

education. Tell us about some of your take about the AI I know, AI modules that you've already been doing. into inventXYZ, what are your thinking about how to get AI into this equation of Yeah, realistic projects happening in K 12? School?

And this kind of thing has happened in real life to bananas. Like, we have a monoculture of bananas, meaning all the plants are basically clones of each other. And they all

randomly generate mutations. And then from there, we give them like, essentially like, Okay, we use CRISPR, to like, put this in a lab and test all the things, how do we know which ones worked, which mutations actually mattered, and which ones didn't matter, because it's bad to keep in extra mutations, because they might have some other side

effect. So then to do that, they actually go through a data science process called plsr, which is basically about shifting your perspective to figure out which of these things are correlated with survival, and which mutations are also correlated with each other. Because sometimes you do, like one small grouping of mutations and your like, survival rate goes up, but then you add like one other mutation and your survival rate goes down. Right. So like, which cluster actually

is important. And that process using plsr, or like related techniques, like PCA, and so on, is a very realistic kind of data science that people do. And you know, AI is not just the GPS of the world, right? It's not just like text generators. And so like, you know, that's kind of like why it came into, into into significance. But actually, when we do like, presentations at conferences, or like trainings for teachers, I'll break down AI into essentially four different

categories. So one is like language models and or generative AI. And I include, like image generation pieces as well in there. But the point is that it has some kind of representation in like, some kind of 3d multiple, multiple dimensional space of like, where words belong, and like, how do we get from one word to another word, right, or like how concepts relate. The second type is signal processing. So that's like, analyzing audio and video, and trying to find correlations

within itself. Then third type is more of that traditional data science like we do in that project I just mentioned. And then the fourth type is reinforcement learning, which is like, similar to teaching a baby to walk, you give it some objectives and some rewards. And Reinforcement learning is by far the most complicated and like hardest to train. But you know, try to split that up, give some

specific examples. And then I actually have teachers, like, come up with examples, or I'll suggest examples to them and ask them like, okay, which versions of AI is that? In some cases, there might be more than one, right, but really trying to like, split apart the understanding. And like, the key thing about AI is you really need to understand matrices. But again, like I was talking about with calculus, I don't need you to like understand the formulas and how it turns into systems of

equations. That's not what I'm talking about. What I'm talking about is we can represent things in multiple dimensions and give them sort of a coordinate in space. And things that are closer together in space are related or like, we can shift the way we look at the space to want things that should be related closer together. That's the piece that we want kids to understand. And then doing that enables them to understand

deeper insights. Right? So for another one of our projects is for like government class, where we have students analyze the text of Supreme Court cases, using AI tools again, so they do sentiment analysis. They do like topic modeling and clustering. And one of the things is, bias in AI comes in multiple ways, right? So one is, you just like, don't think about the scenario that you're trying to analyze and you're using a model that's

just not fit for purpose. And so one of the most popular sentiment analysis models actually basically hard codes, like specific words to a sentiment and so when you like initially run this model Like every case comes out with like a negative sentiment. Why is that? And it turns out like police government are all hardcore to those negative, or like that might have made sense in one

context. It sure as hell does not make sense in the context we're using it for so really helping kids understand that like piece of the ethics and the context, right.

at all. And if so, it's usually at the highest level, like you learn Plessy versus Ferguson, you know, you learn a handful of Supreme Court cases, you certainly aren't analyzing the text of them as a corpus for an AI model for sentiment analysis, it strikes me is, you know, it's so your perspective on

education. And the future of it, I think, is really prescient for where we should all be thinking in this age of AI, it's just a very different world, you don't have to be limited to the same old subjects, you don't have to be limited to the same old types of projects, the capacity of students and teachers to do amazing things is just, it's

there. It's here already. Yeah. So it's really just the four quadrants or the four sort of types of AI that you just mentioned, feed into my next question, which is that you have written a paper with Johns Hopkins researchers about K 12

ai education. And you know, we're in this moment where President Biden has just put out an executive order ordering sort of every department in the government to start really thinking much harder and faster about how AI is going to work, including the Department of Education, and there, have a mandate to create this AI toolkit for K 12. About how to use it. Some of the things you're thinking about feel very

AI toolkit. Ready. Yeah. Tell us about this paper and how you've structured it and how you're structuring your program to sort of meet what you've discovered in it.

And like the answer is yes. But how do you use AI scenarios to enhance student's learning in terms of creativity and depth of knowledge. So we give a bunch of scenarios of like, what that actually looks like, and some sort of sample like one pager lesson plans. So like, some stuff that I haven't talked about, in our discussion so far is like one in physics, where this is a signal processing example. So like, I'll give you sort of the inspiration piece.

And then the actual example, right, like your Apple Watch, or like, if you have an Apple Watch, it can detect when you have like a heart attack, or it can detect when you're in a car crash, and like call the ambulance. How the heck does it do that? So it turns out like it has some sensors in there, the sensors measure like motion, and like, what does this motion mean? Like to the sensors, there's just like some lines on

a graph, right? And so with signal processing, you can correlate certain funky patterns of wiggles with specific movements. So in physics, we have an example in there that's like, using motion sensors to figure out like, the impact level of a force, right? Like, what is the change in acceleration that's happening? And what does that impulse look like? And then based on that, you can kind of categorize what kind of impact it is. And to do that you basically do some signal processing on the sensor

data. And that kind of thing is like, yes, that's how like, the Apple Watch works. But it's also like, for example, there are some motion capture suits and like animation, right? And they have like a bunch of sensors strapped to you. That's how also they figure out like, where your skeleton is, and how do you calculate that? And how do you counteract for like, drift and

some other issues, right. So like, we give some examples like that, we talk about how it's really important to use AI to help kids see unexpected connections, because that is kind of the basis of creativity, right? Like, if you think for example, that like inventXYZ solution is creative. It's creative, because we're making leaps that like are not traditionally made across the different constituents, and across the like program design.

But the only reason I can do that is because I have a lot of knowledge sort of memorized in the first place. And I like have a lot of knowledge about a variety of topics. And I can think about them and process them at another level and say, Wait a second, these two ideas are related. What if I like took this process from, I don't know, let's say agriculture, and then plopped it into filmmaking and

things like that, right. So like, Yeah, we talked about how we can create these AI modules, where students are learning how AI works and building their own models, so that they can better create a creative mind and better understand the core content that they're actually learning.

necessarily feel creative. But when you think about where we're at now, with AI, it is incredibly creative. And, you know, deciding where to point these AI tools and how to use one technique in an interdisciplinary way or, you know, in another context is everything. And kids love creativity. I mean, this is like, this is something that I think young people who don't necessarily have enough context knowledge to do some of this to your point, but also don't have

some of the limitations. They don't see anything as siloed. Right, they don't see all these disciplines as totally different, might be a source of incredible creativity when it comes to this stuff. I can't wait. Actually, I'd love to ask you about this. You obviously trying to put tools in the hands of students that sort of core to

everything you've been doing? Do you think we're close to the time when we're going to be able to put, you know, really powerful generative AI tools, let's say video creation tools in the hands of students in school? I mean, do you think that's six months away? Three years away? What do you

stuff. And that process is okay. But it's like a little bit error prone sometimes, and I'd like is

a little slow. So then, like, maybe last year, there was another process that was like essentially the same, but like via video, so you have a bunch of pictures, photogrammetry like to take a picture of our entire studio, I'd probably need at least 200 pictures, if not more than that, this new thing called a nerf like you can move your camera around, and just like record a quick video, and it'll like generate sort of the same thing with a similar level of

fidelity. And then there's another one now that came out like maybe a month ish ago called Gaussian splatting, who's even better, and it's also way faster, like it runs almost in real time. And you can do it with video, too. So for like Halloween, I dressed up as Neo from the Matrix. And one of the things that I want to test is basically doing like bullet time, with like just a phone or

a couple of phones, right? And then like, the phone might have only been in one location while we recorded this thing, but because of this Gaussian spotting algorithm, now it can like move the camera in 3d, and tools like that. There's like Luma AI, for example. And if like schools have CAD labs, or a lot of schools or interest into like, eSports. Now, so if they have computers that have the graphics cards, like you can run

these models locally. I mean, yes, if there's a little bit of a pain and tooling, and maybe that's something we need to build on our side to make a nice UI to use it. But like, kids can run this stuff and like do cool videos like that. Right? And it's not that far away. Now in terms of like privacy rights and stuff around like Gaussian swatting, I think it's fine. In terms of like deep fakes. Obviously, that's where like most people are concerned, I think there are lots of ways to

use deep fakes for good. One of the things that we're looking at is like a deepfake. Professor, that's like the student but like, 10 years older. Yeah. And then, like, empower it with our FAQ knowledge so that this student can ask questions at any time. I think we're a few years away from that, mainly because deep fakes don't currently really run in real time, all that well.

potential in schools. And you're saying this guessing splatting is a really good example of something that is a very powerful, very modern tool, that would be a lot of interest to kids doing creative projects, even in the service of, you know, they could do that for historical places, or they could do it for you know, the house in the Raven, it's a really exciting vision to give tools of that power and that modern nature to students both because they can be creative with it.

And because it teaches them how to use modern tooling for their lives, which is, you know, when they are 18, and graduate from high school, this world is going to be completely filled with tools of that nature. And once we came into the match,

like, it's real, right? Like those are like real starting salaries like that's how like not necessarily straight out of high school but straight about college for sure, like people who graduated with me, he's made 190 straight out of college plus like signing bonus, right? It's already a thing. And for those kids, they will graduate college and like 2030 2031, right? So we need to be preparing kids for 2030 2035. And it's not all of those jobs haven't been invented

yet. So I'm going to teach kids just soft skills. You can't learn soft skills without doing a hard thing in the first place. Like you can't learn teamwork without having something meaningful to work on a team with. You can't learn critical thinking or grit without having done something really hard. That requires critical thinking and requires multiple attempts. Right?

talking about today. But I guess the question is, in this moment, you know, we're here in 2023. It's only been a year since generative AI came along. And it's only been very recently that it's sort of universally acknowledged that AI has to have a place in education, you we have to teach some version of the skills, we have to use the tools, we have to allow teachers to use the tools for themselves

and for their students. How do you see the next few years getting you know, the far future is just too lucky to see we try. But how do you see the next few years evolving in terms of people coming closer to your grand and I think very cutting his vision of what you know, middle school in high school should look like

And so I think I think it's really important to reframe the activities we do in education, because he can be a super powerful tool, even if it's just using like an off the shelf chat GPT or cloud or Bard or whatever, right? Like, for example, you can do like a Socratic seminar, wherein one of the participants is a chatbot. Or you can have students write an essay using the chat tool, evaluate the essays from each other, and go, Wait a second.

It's kind of good language, but like, didn't really say anything, right? Or is this citation real? Oh, I got the author's right. But I don't know where this actual paper is, by the way I was. I'm into like workout science and like food science. So I like I asked it something about like, what's the optimal protein intake? And like, is there any study around how much protein one can use for muscle building in a single meal? And it like, cited some articles? And the researchers

were absolutely right. But the articles did not exist, right. So it's actually very important for students to see those points where it breaks and for teachers to see that, because that's a great learning opportunity. So I think, in the next few years, I think a lot of schools are going to start doing the like basic stuff of let's use a generative tool as like an activity in an English class, let's adopt something like Kira Potter or whatever that like helps generate some good explanations.

And I hope that will like raise the standard of what students learn while also lowering the amount of effort teachers have to put in, right, because if you can use these tools to like semi auto grade students responses, and I'm not saying like, entirely graded with AI, but at least if you can cluster like 150 papers down to like, essentially three categories of the way in which they frame their argument, now, grading becomes so much faster, right.

So I think there's definitely going to be space for teachers to start, like, just looking for tools that make their life easier, and probably a freemium model there. And I think there's going to be a very non freemium model for stuff like ours, where it's like a much more immersive, changing what students actually learn experience, and I think there's probably both of those things are gonna have legs.

It's being able to give students rapid feedback about what they're doing without having to wait a week or two for a teacher to give them we No, that's important. It's also about putting some of the sort of metacognition in the students hands, right. It's like, there's a great story in the conference from Ben Waitley. He was one of the founders of memorizer he was talking about his daughter had a, you know, an assignment and they told her she was allowed to use chat GPT. So she used it and

came out with a product. And he said, Okay, now great, but like, how well are you going to do on that? Like, how's the teacher going to? You know, is it good? And she was like, well, Chapter GPT helped me write it or wrote it for me, he was like, well, now use Chatri, to to figure out if it's good or not, and make it how to make it better. And it was like, a little light bulb

went off. And she went back and, you know, basically tried to create a whole model for that's up to you to grade her own paper using the type of rubric and thinking that the teacher might use and it's like, ah, you know, that's metacognition, and that is very hardcore learning. It's really interesting to think

about it that way. And then what the teacher can do, instead of having to, you know, cross out the misspellings, or say, this is an awkward phrasing, they can say, Okay, let me zoom out and actually look at this and not have to focus on these tiny things, instead, look at the entire submission as a full submission and actually think about it, at the height of my intelligence, rather than having to come down to these many errors that students are making that really are sort of not that

relevant to anything. Yeah, it's a really interesting world to imagine.

thing. And I came up with this idea of like clustering responses, like back then to like, with one of our first schools, we tried to do that. And I think you really need some like LLM or like a what, vector database or something to do it because we tried some of the more naive language processing techniques, and it just didn't work. Because he just only have like, 15 kids, right? Like, that's not enough,

can answer these things. But like, there's also a little bit of creativity involved of, you can't just like go check off all the bullets in a linear fashion, you kind of have to still make it cogent,

check the boxes, right. And I think, you know, there's a real equity piece to that, too, because I think a lot of students in private schools are sort of taught how to think like that early on, and a lot of students in public schools are underserved schools aren't, that's just sort of not how they're trained to think about about homework, you know, and it creates, I think, a real gap in how the graduates of different types of schools in America are able to do things in the rest of

the world, because they haven't been sort of trained to think of how is the teacher gonna interpret this? How do I do something that is gonna make them right, really understand and respect and you know, admire my work. And that is such an important thing to be able to do it when you're applying to college, for example, or applying to a job.

memorization is bad. It's just like, lack of context, lack of like, that metacognition piece just makes it feel very detached and like, not usable. Right. But you contrast that with like, a massive portion of Indians go to private schools and pay like, pretty hefty tuitions. And they're better about it. It's not like everybody repeats the

thing. But there's still like, not a lot of metacognitive pieces, like one of the complaints that I make is that when I go to India, like a lot of the engineering graduates are like design graduates or whatever the TV like HD television shows on India in Indian channels are sometimes stretched and squashed. Like they don't understand aspect ratio, and some of these like important things for like the

user interface. And it's like, man, you're not teaching like really basic attention to detail pieces and metacognitive pieces, and like having pride in your work. And you know, there are obviously people in India who have those skills, but it's like, you know, when people hear talking about like, we need to teach kids critical thinking skills, like I mean, that's true. Like this is something that America absolutely needs to do, but Like, again, you gotta teach it in context with the

modern tools. So like, yeah, interesting discussion and like, it's making me connect the dots in interesting way. Yeah.

But the idea that it only makes sense to teach them when you're doing something really meaningful, or potentially very difficult, makes a lot of sense. I mean, that is how the work world is you don't often have to work in a huge team to do something very simple that you could have all done, you know, one person could have done, that's not how the world works. When you're working in a huge team, you're doing something incredibly complex and very far reaching. And it makes sense to

model that in school. Yeah, definitely. I like it. Okay, so we have reached our final questions. What is the most exciting trend that you see in the education technology landscape right now that our listeners should keep an eye on? And ideally, this is something that you know, you have an eye towards which you have, obviously, a very good sense of where things are going, that other people may not see coming?

a teacher, right? Like, she can make more at Apple, she could start a startup, like whatever it is, but you want at least, maybe not a CMU grad, necessarily, but like a Missouri grad, or KU grad, like, you want those kinds of people in schools, so that you can raise students expectations of themselves, and what they can do. But if the salary is only, you know, 3040 grand, like, yes, there's a pension, but like, that is 30 years away, I'm not thinking about the pension.

Right. So that disconnect, and like, you know, 10% 15% raise is not solving the problem of getting the wrong people in schools, you need a structural shift. And like, there are some that are like, remote teacher, like we beaming them, somebody from Jersey into an underserved urban school, that's really having a problem staffing, like chemistry teachers something,

right, that sort of works. But the problem is still fundamentally that we have a shortage of qualified people to teach at the given salary range. And so either you need to like outsource people, or do something kind of like what we're doing, or you gotta like massively, totally changed the salary structure. And I think states are looking at it, some districts are looking at like doing innovative things like

changing that structure. I think that area is like something that is going to be something majorly innovated on in the next. I don't know, five, six years, probably

sort of a sacrifice, right? Like a martyrdom kind of opportunity cost is humongous. Yeah. But that's that's not how we should be thinking about this, we should be thinking about, especially in the global context of like China coming from dirt poor country in the 70s and 80s, to like, this amazing behemoth. They really invested in education, and especially in like STEM related things. And now there are like smartphones cannot be built in the US.

Because there are simply no factories that can build precision circuit boards in the way that is required for phones and laptops. And that's interesting. Yeah, but a lot of these components and suppliers and factories are there in China, right? Like, if we want to compete on the global stage of, regardless of what industry we're talking about, we really need to change the priority of like how we think about education.

school teacher? Well, that PhD in chemistry makes it that the salary is $150,000 because they are so well trained that they can be like maybe that's already exist somewhere but or if they have, you know, sold a startup, you know, to Google, then that is counts as experience that would raise their teaching salary and yeah, school can actually recruit that person and say, hey, look, we can pay you

more. And you might have a really amazing time teaching the next generation of kids and inspiring them with your experience, but we'll actually make it worth your while. And rather than you going to another tech company like, right, I wonder if that's a world that we will see,

right? So in education, kind of, as you're hinting at, there could be something where like, maybe there's just like a second class, because like, most of these salary schedules topped out at like, 80 grand or 90 grand or something, right? Like, even if I put that PhD person on the, like, 30 year run, it still wouldn't be competitive, right? So you almost need like a second category of we want these, like actual professionals who have like some level of experience and for those people will pay

them what they're worth. And maybe the state has a separate allocation for that. Like, that could be something that, like, I know, some of the states are like looking at, like raising the minimum threshold and stuff. But like, yeah, maybe that's the other approach. And it might actually be cheaper to do that way. Because you only need like, I don't know, six or seven teachers like that.

learning science studies. And it has like meta analyses, meaning multiple papers have like studied this concept and trying to figure out like, what the overall impact actually is. And you can it basically ranks every type of intervention based on what the standard deviation of improvement in students outcomes are. So we'll link that as well. It's, I think, one of the failings of a lot of professional development is teachers don't actually understand the latest recommendations and learning

science. And it's that's a tragedy, because like, this thing is happening over here and this like small cluster of education researcher PhDs, and it's not being translated to the actual practitioners, because they do a terrible job of communicating it. Right? Yeah, it's just totally distinct, anyway, that we can get more people actually understanding what has been tested, what actually might work, what sample sizes, because sometimes the sample sizes are still really

small. And the other also thing to note there is just because something has been studied once, it doesn't mean something that hasn't been studied is bad, right? Like, you need to have an idea, and then have a hypothesis to create data, you can't simply only look at data, because then you're only looking at things that we have already tried. But there are always going to be new things that we haven't tried that might be significantly

better than the old thing. So like, you need both, but you have to measure them properly.

You're clearly looking in all sorts of ways at the future of education, the future of AI the future of, you know, project based learning and the workforce. And it's really exciting to see what you're doing in K-12 schools with middle schoolers with high schoolers and even with these AI mentors, as well as teachers, thank you so much for being here with us on Edtech Insiders.