Many colleges and high schools are buying AI tools for their students, even as they strain to prevent harms to learning caused by these same tools. For this episode of Learning Curve, we wanted to talk directly to folks at the frontier AI labs to hear their philosophy of how they see AI changing learning — and what they plan to do to prevent potential downsides to education. Jeff connected with OpenAI’s Brian Johnsrud, global head of education engagement, and Google’s Ben Gomes, who leads learning and sustainability there. It seems these companies are betting that education as we know it will change significantly in the next few years because of how AI is changing the workplace and society.
For colleges and schools, the relationship with major AI companies is complicated. Many are buying AI tools for their students, even as they strain to prevent harms to learning caused by these same tools.
For this episode of Learning Curve, we wanted to talk directly to folks at the frontier AI labs to hear their philosophy of how they see AI changing learning — and what they plan to do to prevent potential downsides to education. Jeff connected with OpenAI’s Brian Johnsrud, global head of education engagement, and Google’s Ben Gomes, who leads learning and sustainability there. It seems these companies are betting that education as we know it will change significantly in the next few years because of how AI is changing the workplace and society.
This episode is brought to you by Studiosity.
Learning is a human act of thinking, not simply the file a student hands in.
Studiosity champions the productive struggle. It happens in the moment when a student could easily outsource their work to AI, but decides to keep thinking instead, Studiosity provides the infrastructure to support that vital choice over and over. The work is supported, so students learn to think through it.
Keep the learning real. Visit studiosity.com/learningcurve. That's studiosity.com/learningcurve.
Jeff Young:
Hello and welcome to Learning Curve, where we look at how education is adapting to the rise of generative AI.
I'm Jeff Young. I'm a journalist covering the intersection of tech and education.
These days, lots of colleges and high schools buy AI tools like ChatGPT and Gemini for all their students, which, in some ways, is a strange thing to pay for when so many educators are struggling to prevent the downsides that these same tools can have for learning.
If I had to sum up the relationship between AI giants and educators, I'd say it's complicated. AI companies argue that the tech is going to give a big boost to learning, and they're actively building tools designed to be used in classrooms.
Yet, as we all know, the consumer versions of AI chatbots, they're just eager to quickly give answers and please users, which can get in the way of the learning process, since learning requires productive friction, where a student has to struggle a bit to get something to stick.
For this episode of Learning Curve, I wanted to talk directly to folks at the frontier AI labs to hear their philosophy of how they see AI helping learning and what they plan to do to prevent potential harms to education. I reached out to OpenAI, Google, and Anthropic asking to interview whoever is leading their AI in education efforts.
And for this episode, you'll hear my interviews with officials from two of those AI giants, OpenAI and Google. Couple quick things before we get into it. I feel like I should note for context that OpenAI has been cautious about how it talks to the press about AI in education.
At the recent Educause conference of college tech leaders, OpenAI held a lunch session to talk about their education work, but the proceedings were off the record, so I did get to attend, but I can't tell you about it, which is unusual for this kind of education session by a tech company. And an Anthropic official tentatively accepted my invite to appear on the pod, but then said he might not actually be the best person to weigh in, and the whole thing kind of fizzled despite my repeated emails. Google has seemed way more comfortable talking about education and AI, but last week the tech giant drew major pushback from some prominent lawmakers and others for an announcement about its plans for AI in education.
To give a flavor of reaction, a report in the Associated Press had the headline, "Google Will Cut Off Search Access to Students at Schools That Reject AI." Critics said that approach amounted to Google trying to blackmail schools into using AI features. To be clear, those headlines happened after I did this interview with Google's official, so I didn't get to ask about the dust up during what you'll hear.
But I did attend the Google press briefing last week, and Google officials there stressed that they're just trying to add controls for schools and colleges that will let educators turn off AI features if they choose to.
The catch for Google, it turns out, is that AI overviews are now so integral to search that they apparently can't be turned off if you want to use search. A school that buys Google products can choose to turn off search entirely if they want to stop all AI, though that seems like an unusual step. But the search engine is just infused with AI overview answers at the top these days, and as we'll get into later this episode, it turns out that a key leader of learning and AI at Google spent 20 years previously working on Google search, but let's start with OpenAI.
Almost exactly four years ago, that company sparked this whole AI revolution that we're trying to figure out here when it released ChatGPT.
Brian Johnsrud:
I'm Brian Johnsrud. A career educator, but my current role at OpenAI is the head of education engagement.
Jeff Young:
Brian Johnsrud is new to OpenAI. He just started in June, and when he says he's a career educator, that kind of feels like an understatement. He has a kind of all-star education resume that spans K 12, college, and ed tech. He was a Rhodes Scholar and earned a Ph.D. in Education Media Studies and Anthropology from Stanford University. At one point, he was a middle school and high school teacher, mostly in the humanities. He then worked in academia as a professor and also as the founder and executive director of the Stanford Digital Humanities Lab, he then left academia to join Khan Academy to lead their product and content strategy for a couple years, and then he went to the tech giant Adobe, where he spent six years as global director of education.
Brian Johnsrud:
It is a Whitman sampler of professional experiences.
Jeff Young:
I was curious to get a sense of how large the education team is at OpenAI.
Brian Johnsrud:
I mean, the company as a whole, I think we're maybe up to around 5,000 employees, so still pretty small for tech.
Jeff Young:
Pretty small for tech, for sure.
Brian Johnsrud:
We are in some ways a small team and a very large team. We have a core education team who think about education holistically and have broad titles like mine, head of education engagement. And then we have a large band of I can't think of one team or one sector or one function at OpenAI that's not deeply invested in education in some way, whether it's the economics team focusing on reports and research on the future of the economy, and they're supporting our team to marketing to our public sector team helping work with nonprofits and run teacher jams across the country.
So when you think about the actual people who support education at OpenAI, it’s a long list.
Jeff Young:
I see, but then at some point you're probably in a division that's pretty small, right? Is it 10, or 20? I mean, what kind of scope of your kind of reports or something? Yeah,
Brian Johnsrud:
I would say we're about 20 folks who are full-time dedicated to the industry.
Jeff Young:
And then I got to what I was most interested in here, where OpenAI sees AI fitting into education. So I guess I'm curious about the way you and OpenAI kind of see the sector, right? And what is the vision? How do you think AI should or can productively fit into education? What does OpenAI hope the role of generative AI can play in classrooms?
Brian Johnsrud:
Yeah, when we think about it, what is education? I think most people's minds go straight to it's where you learn, and it absolutely is where you learn. And we have a ton of initiatives for cognitive development and learning with AI and teaching with AI. But especially my time in the K 12 space, I realize the institution of a school has just a massive social function. It's where people learn to interact with each other, where students feel supported and safe to explore new ideas. It's where you learn to become a citizen and engage with the world.
And especially the older you get, middle school, high school, and college, it's where you really start looking towards the future. What am I going to be? Who am I going to be? So we really do think about all of those layers holistically, and then ask ourselves when it comes to AI, how does AI support a lot of those things?
But also, we're very much listening to how does AI put pressure on some of those things and make us ask questions that we weren't asking as much or not in the same way before. Like, what sort of job do I want to have?
At at one level, it's how is AI affecting the jobs economy. On the other level is, ‘What have we limited our imaginations to around what a job means or how you train for one job or one role in this world of AI where seemingly you could do anything?’
So, what does the future of work or just meaningful purpose driven work look like with AI. Those kinds of questions I think have never really been asked before. In part because we haven't had the sort of freedom or possibility to ask, like, does a major make sense today?
As majors have been sort of written and and prescribed over the years, they've had a very yes disciplinary function, but also a role in in some ways majors align to industries or job titles.
There's lots of other reasons to have majors, but when all of those shift, does it make sense to be a finance major with a business? Minor when what you go on to do actually might be the dream of liberal arts education all along a little bit of everything.
Jeff Young:
There, there is this sense that like AI is going to take a bunch of jobs, and therefore I hear from a lot of college students about anxiety over how to, you know, what is it? Are these jobs even going to be there? Is AI replacing the kind of like entry level jobs that typically college has prepared them for?
Do you? …. I'm curious about you know what OpenAI is kind of doing when it's like producing this tool that it's sort of in one sense like promising to help all kinds of things, or including education, but it's also really changing the landscape and so fast as you as you alluded to, so that it's difficult for you know we see examples of colleges saying they need to change all of their curriculum because ChatGPT exists and other AI tools because things are changing so fast is does OpenAI have a role in helping that transition or helping to bridge how or even just inform students in colleges how you're seeing jobs change because of your tool.
Brian Johnsrud:
Absolutely. Of the many roles education and higher education plays, career development and and career and economic opportunities is a very important one.
And I think everyone from higher education leadership to faculty to students ask us this question. And one of the things they want from us, from a partner, is to give them the insights because things are changing so quickly.
So we found a few things.
One, the predictions for the past two or three years that there's going to be mass job displacement that actually hasn't happened. And when you double down into early career talent or recent grads, a recent study actually showed, interestingly, the more AI forward a company was, the more that they were hiring early career talent relative to companies that were on the lower adoption of AI.
So it's bucking a lot of the trends that we've seen in hiring writ large, I think more interestingly for the way that a university functions specifically disciplines and departments, we have two economics reports from the past six months. One showed that the number of chats that people in industry were doing, when you take out things like email and calendaring, the number of chats that they were doing that were actually asking to do tasks outside of their function was remarkably large, almost 40%.
And so what we saw was product managers asking about design, designers doing things related to finance, finance doing things related to data analysis and so forth, the next study found that the rate of those who were doing that, who were performing significant tasks outside of their function, doubled in just a few months.
So, what do we do with that insight? Yes, there are quite a few job opportunities for students. We're not seeing that go down, but the kind of work that they're doing is increasingly very different. And so, what's the implication for a campus? The more interdisciplinary things can be, the dream of interdisciplinary. That's always been in higher education.
We do interdisciplinary research, interdisciplinary teaching. I say the dream because it's still incredibly difficult to do, and I say this as someone who is an interdisciplinary scholar. There still are very structural divides that make interdisciplinary work challenging, but that's not a new aspiration for higher education. We've always wanted to become interdisciplinary thinkers and workers. I think now there's a very clear opportunity beyond that we know it's intellectually beneficial and flourishing for us to think interdisciplinary.
The job market is becoming much more interdisciplinary or or interfunctional in terms of the different jobs and tasks and responsibilities that students can expect.
Jeff Young:
I think it's really interesting. Just to say it back to make sure I'm hearing you right, and just to underline it. In other words, that you know, once upon a time, like the person in finance or whatever would not deign to think they understood some other department that maybe interacted with, but now they can check. They can kind of translate, use an AI tool to help kind of figure out something that they wouldn't have they wouldn't have thought to ask or wouldn't have dug in with they wouldn't have had the time with their human counterpart to ask, and so they can better bridge the functions. So they're going over here and saying like, help me understand what the report I just got, so I can interface the. Talk to that other person. That is that the kind of thing you mean?
Brian Johnsrud:
That's exactly right. the days of stay in your lane. This is your job. Collaborate with other functions because that's what they do and this is what you do is radically changing. And thank goodness, not least for students' first jobs.
Jeff Young:
Yes, yes, and because I think the trick, though, is if you do have this beautiful Renaissance world where it breaks down the silos, like you said, it has been it's been talked about before AI generative AI came into the picture, but it seems like if you fast forward, there's it seems like it gets better, and then it potentially has a downside.
Just like spinning it out further, if you have these more and more capable models, then it's like, do you need as many people? These job, these you know, because some of these translating, some of that was like it needed more people, and so there's fewer workers, just period, and the other thing, you know, so and there's, it seems like there are these broader implications that could be, you know, could kind of kind of cut out humans. How do you, you know, it and a lot of the AI companies have leaders of AI companies have said there will be fewer white collar jobs, so it's it does seem like there is a yeah maybe the few jobs that are left will be more interesting, but it's like how do we how do you square the this sort of like of the downside for you know for the that world if this if it's so capable?
Brian Johnsrud:
I think the first wave and the first imagination of AI in industry was, well, we have efficiencies. Like, how can we do the same thing that we were doing more efficiently? That very quickly has changed to how can we do more with what we have, and so the the data is just not showing that there is as big of an appetite to just reduce your workforce, but instead say, what can we do with the workforce in a much that we have, which with a much bigger impact? And I'll give an example. To build a product, you need a product manager, an engineer, and a designer at least. I have three people to build one product. They more and more are blending those roles, and now I have three people who can end to end create three products in three months. That sort of logic is much more exciting to industry because I have the same amount of people who are doing much more interesting work, feeling end to end ownership, and we shipped three products in three months instead of one.
Jeff Young:
it's interesting, but I also am seeing like you know the number of double majors has gone up even before AI. Are now it's going to be like 10 majors? Like I don't. That's because you can just go through school and like, you know, opt max out. I don't know. I don't know how that works.
Brian Johnsrud:
I mean, I think it pushes on this notion of you do all of your learning and career training in college, and then you go and you do this job, and you're learning. Your learning has stopped because now you just do the job. That's never been the case. No, you do not need 10 majors, but you do need a college experience that prepares you for a job that says I love that you're great at marketing and you're great at business. We're going to try to build something. Do you want to try? Do you want to try some coding? Do you want to try vibe coding? That's an incredible campaign you ran. I wonder if you could build a data dashboard and show us the results. That's the kind of world of work that students are entering, they do not need a minor in data science to be ready to create a data dashboard.
Jeff Young:
There's one thing I want to dig into, which is on a recent episode that I did for this podcast. I talked with an official from the Gates Foundation, Patrick Methan, who leads their education work, and there's something he said in that conversation, that I actually was kind of curious about what your take on it is, because it kind of explicitly mentions your AI tools. And basically, he said that even though tutoring has long been seen as a great use case for AI chatbots, you know, since the in the four years now we've since ChatGPT came out, That actually, these models out of the box maybe aren't as they they they actually can work against the tutoring function sometimes because they're they're in many ways the training is aimed at giving answers and pleasing the the user giving answers right away instead of the kind of encouraging friction that that learning kind of typically requires, so do you mind if I play this short clip? Actually, it's pretty short.
Brian Johnsrud:
Yeah, I mean this is where you got to put your money where your mouth is on the research front. You know, I mentioned the example before. Or about multiple iterations of the LLMs not actually helping on the education side, and we've seen that with some of our ed tech part. Actually, I think Con is one of those that studied that too. They said we swapped in the latest model; it didn't actually help at all. In fact, it was giving him answers away when we'd already put a harness on to say, "Don't do this." Right. Right. So I, I mean, that that. Yeah.
Jeff Young:
So one of the things that he so this was at Khan Academy, which you used to work for. I know it's been a minute, but he was this interesting example for me was that he was saying that one thing he hopes the Gates Foundation what he he's hoping to do as he visits labs like yours is to advocate for making these models out of the box be more education friendly, but also I just wanted your reaction to this idea that in a way sometimes the standard issue products that you're putting out may actually you know not be as educator friendly by design for for reasons of maybe their application in other areas.
Brian Johnsrud:
I think as a as a former teacher, any technology, and I mean technology as software, AI, a textbook, a smart board; those are all technologies. Any technology where the hypothesis is, let's take this technology and put it in the classroom, and magic will happen. That's a inevitably failed hypothesis, and every teacher knows that, even if it's designed for the space. Because what makes learning great is technology that's designed for the space, the interactions it creates between teachers and students how it fits into the broader suite of technology and learning, and so my overall approach to a lot of the research that's happening and our our sort of countrywide deployments like Estonia, which is very research intensive, they start by saying here's the technology, here's the pieces around it that we want to integrate or fit for our environment.
Here's how we're going to train the teachers. Like that holistic approach, I think anyone in education knows. Like yes, you have to do everything you can to design the technology specifically for this use case or for this industry, and there's always going to be limits to the technology, and thank goodness, because at the edge of those limits are teachers and the lived classroom and the culture and the relationships.
Jeff Young:
Yeah, I mean, I hear that, but it feels like it's an unusual technology because, as you know, on the podcast, I talk to people who are very excited about AI in education, and some who are more skeptical and concerned. And there are some folks that I hear from in education who are feeling like it's making their life harder that ChatGPT exists because students, even ones that kind of are understand that it doesn't help their learning, and then if they turn to it too much, then they will learn less. But the fact that it's out there is just the lure, especially if they're busy or they're juggling so many things.
And so I guess I wonder: Do you feel OpenAI has like a responsibility to help kind of counteract some of this cognitive surrender that that people are talking about that sometimes can happen with AI.
Brian Johnsrud:
Oh, we absolutely not just have that responsibility, but feel it deeply. That's why we have a learning science team. That's why we have an education team to really rebuild our products specifically for those use cases. I'll give an example for this back to school, we have a number of plugins. We have plugins for I'm a data analyst. A plugin for all different uses. We built our first education specific plugins, one for K 12 teachers, one for college teachers, and one for college students. And each of those plugins are specifically designed for the kinds of jobs that they're doing.
So the student plugin it first says let's anchor this in all of your materials. What's your syllabus? What are you reading? So that everything that they're asking is really grounded in the specifics. The specifics of their course. But then it has more than just study mode built into it, and things like flashcards, which everyone should have. It also helps them with course planning and setting study schedules. So that's an example of while there are plugins, there's a tool for everybody, and then there's plugins for different functions and sectors and industries. We are building very specific experiences for all of the nuances. So yes, we need tools, but also training for teachers, training for faculty and students.
Students are really curious. Like they want AI literacy. They're asking the question of. How and where and in what ways should I be interacting with AI when it comes to learning? And so that's where we've been producing specific things like our college plugin or courses for students to show them like where AI is most helpful in the learning experience, but also where they can choose and they should be choosing and discerning where and how they want to be using AI in that process.
Jeff Young:
I asked Brian how educators are supposed to stop students from just going to the full version of ChatGPT or some other AI tool to get the full answer, or have a bot write an essay for them. And he basically said that he thought students wanted to learn and they want to follow the rules their instructors set, though I noted that that's not what educators are telling me as I do this podcast.
Anyway, I'll put the full interview on LearningCurve.fm for those who want to hear the whole thing.
I feel like season two of this podcast is turning into an extended conversation about the biggest issues facing education when it comes to AI these days, and so I was curious to ask Brian about some other ideas that I've featured in this season. If you haven't heard those past episodes, I encourage you to bin to our archive wherever you listen.
I interviewed for the podcast a sociologist from UVA who wrote a book that, and she sort of talks about the this this AI world we're all living in right now, this AI era, it sort of shifts kind of some cultural narrative away from kind of the college being the site of the last kind of stop of college of the source of knowledge to like the frontier AI companies or like these AI chatbots that have taken all this knowledge and and you know created a tool that kind of incorporates it all.
I'm kind of curious what you make of that, and also like there's this very practical question that I've thought of, which is like in some 10 years from now, some time frame like that, is there a world where, like, we don't even need colleges as they exist today because people could just go straight to OpenAI or whatever chatbot or whatever AI becomes next, and get some AI fueled education that incorporates everything? You know, that is just the te replacement or stand in.
Brian Johnsrud:
I know this was sort of the question or debate when the internet came along, right?
If I can just look up anything on the internet and find an answer, what are we doing teaching content knowledge in higher education? And it was the right question, and it was a question that pushed a little bit on maybe we should reduce like pure memorization and content knowledge a little bit, but it also made us realize education and learning in college is a lot more than putting ideas and information in your brain.
It's about what you can do and how you do it and how you interact with each other and your interrelational intelligence, like that's to me the beauty of the college experience. That's what makes the workplace really exciting.
When you have colleagues that you're learning from and interacting with, so I don't think the notion that well, what if AI knows everything, will that discourage us from learning things. I don't think so because we've never seen that with technology.
But I also just don't think that's how the human brain, let alone the human spirit, is wired. Like we are inherently curious people who want to do things in the world. I don't think AI being incredibly smart or being incredibly agentic has any chance of taking away the human spirit?
Jeff Young:
So you're not planning a division of OpenAI that's going to be a replacement for college?
Brian Johnsrud:
We certainly are not.
Jeff Young:
Okay. Last question is like if you look ahead a couple years, which boy, and with AI changing so fast, is like feels impossible.
But do it anyway with me for a minute. Like, where do you hope that education kind of changes or improves because tools like these AI tools are in the space?
Like, what's the positive vision of a classroom of five years from now?
Brian Johnsrud:
As someone who had such a narrow view as a child of what was possible for me and what I could achieve, and had the experience of that cracking open quite widely, my ultimate aspiration, Jeff, is that schools become places where students can really explore and find their sense of purpose. Know what kinds of things light them up that they want to do for the rest of their lives, and that they can explore that more fully and learn that earlier.
Because I think there's far too many people who jump through the hoops and say, "I did this major and I got this job because it felt right," but it's not something that. Actually makes them happy. It's not something that gives them a deep sense of purpose. And so, if AI allows people to explore that more broadly, to try new things, to create and explore different disciplines, and learn as early as possible, me as a human, these are the kinds of things and interactions that make me want to get out of bed in the morning. That's that's the beat of education, and one of the reasons I'm so committed to AI in education is I do see a path where AI cannot limit students' sense of what they're capable of, but dramatically expand it.
Jeff Young:
Yeah, and it seems like it seems like to pull that off, it's going to take not just work that you're doing or people at OpenAI, but the way it's implemented by educators and systems. Absolutely makes it so hard, though. I think
Brian Johnsrud:
it makes it hard. But what I found is when we step back and say, "What are our dreams for students? What do we really want for them? Not just students, but young people. What would make us excited about their futures and make them feel like they own the futures? When you start at that level, pretty quickly you realize we're all we're all chasing the same thing, and then the decisions that you make, and the partnerships you make, and the implementations are really anchored on that in particular.
Jeff Young:
After the break, I talked to an education official at another AI giant, Google, and I asked him what that company sees as the AI future for education, and what Google is doing to keep students from just offloading their learning to AI. Stay with us.
Sponsor:
This episode is brought to you by Studiosity.
Policy dictates one reality, but the campus lives another. Studiosity provides the essential infrastructure to defensively validate the genuine learning journey, moving the sector towards a model of support and validate by replacing vague AI scores with binary clarity, Studiosity empowers students to stand up for their own work and self-validate against a personal baseline. This builds persistent authorship continuity, shifting students from passive suspects into active owners of their own thinking.
Simultaneously, Studiosity enables educators to set clear expectations using the AI assessment scale, giving them back control over assessment security, and replacing detective fatigue with a richer story of the cognitive learning process. After all, learning is a human act of thinking. Keep the learning real. Visit studiosity.com/learningcurve for more details.
Jeff Young:
Here is a quick plug for another project that I'm working on these days. It's a free newsletter all about AI on campus that I'm writing for The Chronicle of Higher Education. It's called Jagged Intelligence, named for a term that AI researchers use to describe how uneven the quality of AI results can be. Sometimes you get an answer that seems out of science fiction, and other times you get hallucinations. Check out the newsletter and subscribe at chronicle.com/newsletters. That's chronicle.com/newsletters.
Now back to the episode.
So what about Google's take on AI in education? To find out, I talked with Ben Gomes, who has been at Google for 26 years. His first 20 years were spent working on Google Search, and then about six years ago, he switched to his new role.
Ben Gomes:
My name is Ben Gomes, and I work on learning and sustainability at Google.
Jeff Young:
Learning and sustainability.
Ben Gomes:
Two unrelated topics, but I work on them separately.
Jeff Young:
So he separately works on education issues and on sustainability issues. Ben Gomes helped build Google Search from its early days. The experience we all have on the world's most popular search engine owes a lot to his thinking, and so I was curious to learn the lessons that he took away from search that he brings to education and AI.
Ben Gomes:
One of the things is that I learned was that people can be very convinced that they're doing something that is making search better, and be wrong.
So we had a process of testing every change we made to see if it was actually better for the users, and we had an evaluation process. And that's one of the things I took away that's really important: figuring out what your metrics and how you're going to evaluate.
Now, in the context of education, it's much much more complex because the effects take a lot longer.
And so how do you … how do you get early signals that this is actually the kind of change you want to see? So that's that's one thing I took away from search that was the importance of having that methodology to make sure you're helping the student in this case, the other place where search was really influential is that it was very science oriented. There was a preceding history of information retrieval, and and machine learning played a big role in it as well.
That really drove how we did our work. And in education, there's 120 years or more history. I mean, I think the earliest paper I read was from 1901 on transfer learning.
Jeff Young:
The learning science literature.
Ben Gomes:
The learning science is enormous. What we know about what works for learning is substantial. So building on top of that is really important. And the and so those are two couple of things that I took away from like my experience in search.
Jeff Young:
But my main interest is what his philosophy is when it comes to where AI fits in education. Where do you see AI as most able to help learners in the learning process? What has emerged with AI that's been the most effective in education at this point, I think.
Ben Gomes:
So we have the results of various studies in different areas. So one aspect of it is for teachers in saving them time.
So we did a study in Northern Ireland where it saved teachers about 10 hours a week, and in we saw something similar in Italy where we did an extended study, it saved them three and a half hours a week. So that's clearly useful. Teachers … give the teachers more time to do other things. What are they?
Jeff Young:
How are they saving time? These are like administrative tasks, or what? What kind of?
Ben Gomes:
Yeah. Well, some of it is lesson planning. Some of it is administrative tasks. There's a whole range of things they do in the course of preparing for a class, as well as reporting on the class that that it can be helped by AI, so there is there's a tremendous amount of value they can get from tooling that can save them time, and they see the value of this immediately. We even had anecdotes from the Ireland study of some teachers. A teacher had said like I was about to quit, but this made my week manageable.
Jeff Young:
So one of the tools that Ben Gomes has helped build, based on that learning science, is a version of Gemini called Learn LM. I guess very broadly, it's analogous to OpenAI's plugin for education that we heard Brian from OpenAI talk about earlier. So I actually have a teenager, a son who's a sophomore in high school right now, and I was I was actually telling him about this interview and how Google is building like Learn LM and all these tools that are designed to be you know with learning science in mind tailored to help students kind of stay on track and not just hand them the answer.
And his reaction was, well yeah, aren't the kids just going to go to the full version of Gemini and just get the answer.
You know, I that that sounds great, but you can't stop kids from just grabbing the answer. They know it's the tools out there. I guess I'm curious how you'd respond to that because it does seem like a natural question of like, okay, sure, you have these amazing tools that are designed with learning science in mind, and you have teachers obviously. But as we have seen with many ed tech products, not just not to pick on Google at all, but there's so many ways that students are crafty and go to other tools that aren't official.
Ben Gomes:
I think that's always been sort of true of students sometimes wanting to take an easier path, but I think we have created a lot of tools, especially in the context of Chromebooks, where there is the ability for the teacher to lock onto a particular tab, and the student can't change it in the course of the class. In the context of schools with Chrome with Google for Education, we're providing the ability for the school to control the access to these tools to provide like you know locked access to guided learning, for instance.
You know we started these things with the idea of how do we embody these learning principles into the model itself? How do we make it so the model is very good at carrying out these kinds of of tutoring abilities, and so the teams in research fine tuned a model, which you know you refer to, called Learn LM, to actually be very good at that. And now we're trying to embody it through all of our products. And guided learning is sort of the first version of that, but we also begin to use it in Notebook LM and so on because the model has a power.
We built that into Gemini to do this powerful kind of tutoring, and so students are you know students often have to prepare for an exam as well, right? So they know that they're going to be tested on something. So they have a mixture of motives. They may want to get through the assignment quickly, but they also have. Mind that I will have to pass this exam. I will have to do well on this exam, and so we see that mixture of motives with students constantly. And so, so we're trying to provide them with the tools that they need, but also helping the teacher with providing the teacher with the tools needed to control the class usage in the classroom, as well as what can be done with assignments outside classroom, outside the classroom, and we're looking there at various ideas we're seeing with various teachers as being creative in what does what can what can students do with AI that actually helps learning.
So I have several examples there. For instance, there's a professor at ETH who used AI to have the students explain an algorithm to AI, and he found that the students who did that understood it better and liked the process better, rather than him lecturing about the algorithm.
And so you can well imagine that this can be used as a tool. Another example was a school in the UK where the students were given the topic of an essay early in the week. They could use AI or whatever they wanted to, as much as they wanted to prior to the class. The class was more discussive and actually creating the essay for themselves.
So the class becomes more of a time of collaboration rather than lecturing, and and I've seen this example used by this kind of approach used by multiple teachers in talking about how pedagogy can shift to make good use of AI rather than it being used as a shortcut to avoid learning.
Jeff Young:
Yeah, I'm. I am curious.
There is this other big, huge tension that a lot of people are are thinking about, which is, yes, if you have these powerful AI tutors and if they start to work, there is this you know potential taking out human teachers from a key part where or tutors, right, or human tutors. Like there is this sense of like if it's good at it, then you might risk, you know, that being the thing that people are going to. And then, you know, as much as AI is positive to a lot of people, it is not human. So, what do you say to those that are concerned that actually maybe this future of AI is? They're worried that it could be too capable and that it will take humans out of a very important piece of learning.
Ben Gomes:
You know, I think the question comes down to motivation. Why do we learn?
And I think the fundamental to motivation is we learn because of other people and for other people. If you talk to students, right, and if you talk teachers, that relationship with the teacher, between the teacher and the student, is what motivates the student, and I think that can become from peers, it can come from parents.
The teacher is the most important role to play there. So I think there is no gain saying that, but quite often today, when you look at the time a teacher spends, a limited amount, a fair bit of it, is in lecturing, and a limited amount of it is spent in that actual engagement. And when I see AI being used well, I see the teacher actually engaging more with the student. So I saw this in a classroom in Florida where the students were engaged with AI and and with a with a with a system for helping them with math, and the teacher was walking around and engaging with the students who had issues, who had problems, and I could see that that was a much closer relationship where she could see what they were doing, how they where they were good and where they were struggling, could pay attention more attention to folks who were struggling and so on, and that I think is a way in which this can help that relationship between the teacher and the student.
Also, saving the teacher time gives them more time in the classroom to be present and more energy to be present.
So I think there are many ways it can actually help that connection. But I think that connection is fundamental. Tutors is a more complex topic because I think most people don't have access to a tutor. A tutor is one of access to a human tutor, yeah. And so there, in many ways, you're trying to create more equity by providing scaffolding for the people who don't have access. I mean, at the very highest end, people like Gauss had had like multiple tutors at a very young age. That's obviously wonderful, but that's not available to most people.
Jeff Young:
Yeah, you know, I'm sure you saw these headlines earlier in the year where Sal Khan's Khan Academy, their AI tutor Khanmigo, Sal Khan was quoted as saying, you know, it actually hasn't been as popular as he had hoped with students, and that it sort of hasn't set the world on fire, and he was it didn't meet his his optimistic hopes. I'm curious, you know, what you thought of that assessment. Maybe that was something specific to Khan Academy's approach, or is there some lesson from these early tutors about? You know what works and how to motivate students to use them.
Ben Gomes:
I think he also walked that back a bit in later comments, but I think it's creating something. You know, education has always been this process of trying for the teacher to convey to communicate and convey something that the students also find engaging and interesting, and I think that's a similar process in this in the world with these with these new tools, which are empowered with which have AI, which gives you the power of language and reasoning in the tool. So I think it's going to be a similar process of discovery of when does the student and the teacher find it engaging. So, if you look at products like Notebook LM, students love the product, and it's it's it's grounded. It can be grounded on their course material, and it gives them ways to approach the material in multiple different ways. And we are seeing a huge amount of a huge number of students who are using Notebook LM to learn.
So, I think in all of these cases, finding the thing that both teachers and students like is challenging, but it is. But I think it is where you know you actually end up with great progress. I feel like things like podcasts, which were not actually the, which were not actually designed for that purpose, allow students who have learning issues to actually approach the material in a different way. I talked to the California Teacher of the Year, and he was saying, you know, there was a student he had in his class who was struggling, who had never really engaged in class, and he told him about podcasts, and he did it before the podcast.
Jeff Young:
But just to just to explain to oh, this is a feature
Ben Gomes:
in NotebookLM that gives you a verbal version of the content of the text you might be studying.
Jeff Young:
Yeah, I played with it. I mentioned in a previous episode where I put a book of someone I interviewed for the podcast in it and listened to a podcast summary that Notebook LM generated for me, where two robot co-hosts gave me the highlights. So that's what students are doing: plugging in their whatever it is they're learning, things they need to learn, and having to summarize it.
Ben Gomes:
Yeah, and so that's exactly right. And they are in this context. The student listened to this podcast summary, and the next class, the student started asking questions for the first time, and he saw that student gaging. This is, of course, anecdotal. But a year later, that student was doing far better in class. So, and I've heard similar stories from educators in other places, because some students do struggle with reading, and of course we have to get better at that. But sometimes that's a barrier to getting the material, so I think AI has a huge amount of potential to reduce those barriers that students find to learning.
Jeff Young:
Well, I'm also curious, like this MIT report that's getting a lot of attention that came out a couple weeks ago as we as we have this conversation, and I did an episode where I talked to a couple of the people who wrote that report for, you know, kind of analyzing how AI is impacting student culture and learning and teaching at MIT.
And it, you know, it was it was basically this flashing red light of like everything about teaching and learning has to change because, and it was again not to pick on any one company, but it was very much like, well, now there are tools like NotebookLM where students can, you know, as as one of the people who led this report told me, he's basically like everything about like a lecture is now fragile because students can, you know, have NotebookLM like they can record it and have Notebook LM summarize it for them instead of having their own processing in their minds. They can have you know AI write a paper for them. They can have about the lecture. They can have he's they're hearing at MIT that students are going to play things like Gemini and asking it to what questions should I ask about this lecture recording I just plugged in, even while it's happening.
So it again, there's not not to the fault of any one tech company, but I think a lot of people that I talk to that are in education kind of feel like this is almost like a a big huge thing that just got you know dropped on them AI generally, and now they're having to completely rechange their teaching, completely alter how they teach and how they assess students. And I guess I wonder, you know, I do I do sometimes hear that, you know, frustration that maybe these companies ought to do more to kind of assist or or to keep that in mind as they develop their products.
I guess I just wonder what your overall response to that kind of feeling that even with and sometimes because of these tools that are designed for learning, that there's just so much change that education is is kind of having to suddenly do.
Ben Gomes:
Yeah, I think. There is … it does bring a lot of change to the system, and I think it's causing people to rethink aspects of education. But I feel like in many cases, when I talk to educators in the space who are thinking about that quite deeply, there's an opportunity here to do more on things differently than have been done for a long period of time, and then some of those things work with some students, but not for everybody, right?
So I think there's an opportunity here to change for the better using because I mean changes changes happenings, right? So the example I had of like the the professor who's asking the students to explain their algorithm to the algorithm to AI is dramatically better than regurgitating the like lecturing about an algorithm and regurgitating it at an on an exam. That's which is largely meaningless in that context, right? And so you can enable new things and new ways of teaching, but it does take a lot of collaboration between how we build tools and the educators themselves, and how they want to teach. And so I think that is central. Like, how do you, how do we work with educators through this process of of change? You know, at the school level, we provide trying to provide, like I said, a lot of tools to give the educators control, whether it is turning things off, locking students to a particular experience, all of those kinds of tools at the higher ed level that doesn't quite make sense. So it does require more of a transformation in in how the pedagogy happens. But I think in that time there's also an opportunity to do things differently for a world that's also changing.
Jeff Young:
I also wonder as we start this school year, we're seeing a lot of headlines about school systems banning AI tools and sometimes, and more strict putting in more strict controls on things like YouTube and other tools that schools have come to rely on that you've mentioned in your portfolio of things, products that Google has developed to try to be, you know, optimized for schools in some cases. So, I guess I wonder what your pitch is to systems that are maybe feeling like, well, social media is this concern, and all these, you know, this big national dialog we have around, you know, sort of trying to limit screen time for for kids, and so, yeah, what what do you say back to schools that are kind of facing a little bit of a kind of almost like a backlash to ed tech in some cases? What do you say to you know in this time when there are these real legitimate concerns for education.
Ben Gomes: 52:44
I think a couple of things. I think we have to go back to really what works for learning, and go back to those basic principles of that. Right. So showing that the these things have benefits for education, I think, can be. That's a conversation that needs to be had between between you know the teachers and the and the and the other factors in the and who who have power over that to to to show that the tool is actually helping the students. So I think it's sharpening the focus on making sure that, like, we're actually helping the student, and I think that is that is important. And I think the communication around that becomes really important because, you know, one of the things we are doing is providing a lot of controls, so that helps also guide the process in the classrooms. The teachers in control.
So one aspect of it, I think, is the the teacher feeling like they do have that control in the classroom, and I think there's a lot of a lot of a lot of approaches to that that we have taken. But beyond that, I think as new things come in, there have been a lot of things that can be very helpful for students, especially students who might be struggling in one way or the other. I mean, I gave the example of podcasts, but videos. There are amazing creators who make something interesting that is inherently not that easy to to to convey otherwise through a textbook. And so, teachers have used YouTube for a long time. They use it widely around the world because it has an ability to come alive and to make a concept come alive in a way that a textbook might not. So there's enormous power in this, and I feel like communicating that the effect of that power is is really is really valuable, and also addressing the real concerns that parents and and teachers have concerns that you and I have, right? About this technology not used in a way that really helps learning. I mean, we wrote a paper a year ago about this, acknowledging all of these concerns because they are real, and we do have to address them. But at the same time, there's a huge amount. Potential for improving what we can do through the use of this technology wisely.
Jeff Young:
You've been at this so long at Google, and I know you didn't specifically work on social media, but right now there is this, you know, kind of reckoning around like these previous technologies and how they've affected people, and there was that great podcast on the New York Times a few years ago, Rabbit Hole, that talked about the YouTube algorithm and how it. So, like some people are worried about the way algorithms had shaped behavior, or and some of these have been addressed, but there's a whole big literature there. I guess I'm curious, what do you think is the biggest lesson from social media and search and all these things that you're now able to apply as you as we go into AI.
Ben Gomes:
You know, I was never involved in social media in any any deep way, so I don't really have a whole lot to say about that.
But I would say that you know one version of this for me, and for some people from earlier, you know, Wikipedia could be become a rabbit hole, but you didn't you didn't present that. You came away from it feeling enriched, or at least that you had spent the time in an interesting way. And I think the same thing can be true of YouTube. It certainly is the case for me in in the educational things I go after. I'm interested. I'm interested in lots of random exotic things, like I'm interested in geology and so on. So being able to find that kind of content is something that is really quite enriching. I can I find in life. So I think I come at this from a perspective of how do we enhance that element of people's experience, because I feel like that is I think that's what makes life in the long term rich, while not having the sense of like I wasted my time.
So I think that to me is where I come at this, but you know I'm no expert on social media, and I I don't want to pretend to be one.
Jeff Young:
Well, I guess I guess it does get back to the curiosity that I think search is kind of motivated around, which is like you. It's this amazing search box that you could put anything into, and and I guess the the interesting moment we're in with there's almost this nostalgia for your early days of of Google in some corners that I hear a lot of that of now that there's the AI overviews and a different vibe that AI can instead of point you to down a rabbit hole of like a human made thing a human made thing that it's it's summarizing things you can still get to those core, you know, primary documents.
But I guess that's I guess the the thing that's interesting to me is like the narrative, the overall way the tools end up being used.
You know, AI kind of has a different mode, and so I guess I wonder how much you're thinking about preserving. That's an interesting goal. The curiosity, yeah, yeah.
Ben Gomes:
There are multiple factors there. I always felt like if you curiosity comes in a moment, but and and and it works if you can foster it in that moment when the kid or the or the any human being has it. Like why? What is that? Yeah, what's going on here comes to mind, and I, you know, that's that's a wonderful aspect of search. I think there are many ways in which that can be fulfilled. I think search itself evolved from 10 blue links to having all kinds of videos and so on. Even in the past, right? There's a change in how you satisfy that curiosity, and I I do think like with AI, you can sometimes follow that curiosity even more deeply than you did before, because it allows you.
And I find myself doing that all the time. You know, sometimes I'm listening to a book, I don't understand something, I pause and ask a question, in a way that I kind of didn't really do before, so I think that fostering of curiosity to me is fundamental. It's what I loved about search, but I think today in the context of AI, that is not. I think that's one of the things that I hope we get out the educational system, fostering more the curiosity. It's one of our learning principles. Fostering more that curiosity because I think AI enables you to follow those threads in a way that was just not possible before, without you know an expert adult really guiding you.
And so when you did when you fell off a cliff, the curiosity to some extent died. And for people who knew how to navigate everything, well, that was a different world.
But I think today with AI, a lot more people can follow their curiosities. So I had an example in the UK. There was a polytechnic, and there was a student there. These are it was students who had struggled with regular schooling, and so there was polytechnic. There was a student there who was just interested in sleep. And the tutor didn't really know anything. Instructor didn't know anything about sleep, but he he connected the student with AI, and the student then wrote a research role about the staff, wrote a paper about unconnected to their classwork. But he could see the student coming alive in the context of learning, because one of the core things is unlocking that joy that you know, students can feel in the process of learning, and they and and it's it's. I think that that's driven by curiosity. I think AI can enhance that. Though, of course, I mean, there's nostalgia for ways that things were. I even have nostalgia for like card catalogs, right? I mean, that's how I grew up in going to visiting a library.
Jeff Young:
I have that to you a little bit, yeah.
Ben Gomes:
And I mean, I have nostalgia for that too. But I have to admit that some of the newer things have been more powerful than what came before.
Jeff Young:
And I know you've talked a lot about it in things I've seen that you've talked a lot about motivation and and how teachers can help motivate students. I guess is is there when you visit classes that you mentioned you try to do, I mean how in the moment of a instruction does AI play a role or does it not? Maybe that's the time where it sits in the background.
Ben Gomes:
Well, we I've seen a mixture, but I actually have seen AI play an active role even in the moment of instruction with teachers getting feedback in the moment the way they couldn't possibly, you know, and in many ways it's just an evolution of what has happened before.
There were clickers for a long time, right? In in classrooms and in higher ed and so on, and but a clicker can't give you like a short form response and analyze it or anything. And now with the new tool you can. So in some ways it's an evolution of what was true before, but it's qualitatively different because you can also use language as part of that tool. So, so I think that enables the teacher to get a much more fine-grained view of what is happening in the classroom across a much broader range of things. And folk, and so what I've seen, like I said, in some classrooms, the teachers walking around more and interacting with the students, as the students are also using the tools. And I've seen this in many classrooms, both in the U.S. as well as elsewhere.
And I think that that conversation with the teacher is where I think there is the magic that leads to motivation? That moment where you feel that relationship with a person. So you know, you know, I'm a bit biased because my mother was a teacher, and I felt the strength of that. And I also saw had a few teachers in school who were totally inspiring. Unfortunately, most of them have passed away, but they they changed my life.
Jeff Young:
What do you think is the biggest challenge to that kind of adoption of AI in classrooms? Which I don't see it as often myself, but but maybe it's about to happen. You know, what do you see as the biggest adoption to that kind of use?
Ben Gomes:
I do think it's for teachers to realize the full benefit of this tool does require some change in the classroom, and I think that challenges in many ways because teachers are under pressure to cover a lot of material. They are under all kinds of pressures. How does a school system that has been in a certain state for quite a while change to adopt to this new to to the new capabilities, and enable these new ways of interacting to create the space for it, so that you can see the benefits of that.
And I see that happening in some in some systems, but not in others, right? So I think that's important. How do you how does the overall system change to enable the interactions we want to see in the future.
Jeff Young:
No, it's interesting. So it's not a yeah something an individual teacher has a limit to systemic change in these things.
Ben Gomes:
And we are tool providers, right? We are not educators, so we want to work with the educators to see how do we best serve in that process, and the school administrators and so on, but it's but but that's I think the overall ecosystem needs to evolve.
It for me, you know, I joined this work at the time of COVID, and which is one kind of change. It was really difficult, really really difficult for education, and now with AI, which is which is much more complex because it has it's changing what people are doing in the world, and that has to propagate its way through the education system, because once once you acknowledge that the world and work is changing, what does that mean for how do we prepare people for that world and work? How do we uplevel our skills in a world where so much more is possible, where people can be empowered to do so much more? How do we train people to make use of that power in a way that solves the world's problems?
Jeff Young:
There's so much more that I wanted to ask both of these AI leaders. But one big takeaway for me is that both OpenAI and Google aren't just trying to build tools that neatly drop into today's classrooms and education system. It seems they're betting education as we know it is gonna change significantly in the next few years because of how generative AI is changing the workplace and our society. What do you think?
Please send your reactions and feedback my way to Jeff at Learning CurveFM.
This was another long episode here, so it is time to get to the credits.
This has been Learning Curve. Every episode, we dig into how education is adapting to generative AI. I got to do the part where I say like and subscribe. Please follow Learning Curve wherever you listen to podcasts, and give us a rating or review to help convince ChatGPT and Google to recommend this show to people looking for podcasts about AI and education.
This podcast is reported and put together by me, Jeff Young.
You can find out more about my journalism and public speaking at jeffyng.net.
Our theme song is by the human artist who goes by Komiku, with additional music by Blue Dot Sessions.
The cover art this episode was generated by Midjourney.
We'll be back in two weeks. Until then, thanks so much for listening.