Education for the Age of AI - A Conversation with Charles Fadel

Welcome to the Future Learning Design Podcast.

We have to look at that intelligence the same way we look at animal intelligence and say, it's not like us. It has different capabilities. All the AI analyses from very early on were all about creating a human mind. That was just the one way of looking at it. The human as the pinnacle rather than each one of these animals as the pinnacle of their own environments.

Hi everyone, I'm really happy to be back this week with a special episode on the status of AI in education. Once again, cutting through the hype with the fantastic Charles Fadel. My name's Tim Logan and the podcast is produced by Good Impact Labs. This is quite a deep dive into the topic, so if you're early in exploring it, I'd recommend having a listen to episodes 107 and 108 with a great selection of reflections on AI in education from young people, teachers, leaders, policymakers, and edtech entrepreneurs. But it was great to be able to get into some of the complexities with Charles, who's been working with AI for the last 30 years. Charles is a global education thought leader and futurist, author and inventor with several active affiliations. His work spans the education continuum of K-12 schools, higher education, and lifelong learning. He's the founder and chairman of the Center for Curriculum Redesign, and among many other accolades, he's the author of Education for the Age of AI with co-authors Alexis Black, Robbie Taylor, Janet Slezinski, and Katie Dunn. You can find more links to Charles's work and the extremely useful resources shared by the Center in the show notes.

Thank you, Charles. An absolute pleasure to meet you, having been at the IB conference last week and seen your fantastic keynote and having read your book and followed your work for a long time. So it's a great pleasure to be able to have this conversation. I've got 1,000 questions I want to dig into. So yeah, no, great pleasure to meet you.

The pleasure is all mine.

Fantastic. So maybe just to start with your most recent book, The Education for the Age of AI, is highly pertinent. Let's say everyone's obsessed with AI right now, more or less productive ways. I find the hype is very strong and very real, but you've been, as you mentioned, telling your background story, you've been involved in AI for a lot longer than many people have been obsessed by it in the last 2 years, given that you've been around maybe, what did you say, 30 years perhaps you've been involved in AI in all sorts of ways. So, but we're really interested to kind of hear your perspective on some of that. But maybe just setting the context for now, you talk in the book, at the beginning of the book about how you've, you wanted to update the Four Dimensional Education work because clearly in the last decade a lot has changed. And there was a great quote, if I may, where you say that you, because clearly, I mean, it's not news to anybody who listens to this podcast that education has been somewhat averse to change. And you say that the AI-powered wave will sweep away the reluctance to adapt. And in fact, there will be— education systems will reach a point of no return, an event horizon moment. Very interesting. So perhaps you could just set the context for where do you think we are in relation to the technology, but it's also in relation to this huge question of how do formal education systems adapt?

Sure. Well, you know, it's not about AI at first. If you look at the shape of the world, one can make the statement that all the problems of humanity that we're going through, you know, global warming, pollution, et cetera, et cetera, relate to very deep instinctual drives that were developed by, you know, through our zoology, through our evolution. And this is what Christian de Duve, Nobel Prize winner, 1974, wrote a book about and the genetic aspects of why we are short-term oriented, why we follow the leader, why we do a number of things that were very useful during our evolution, that are absolutely detrimental to our survival as a species now. So you could say that basically education has failed to recognize in the past 50 or so years that business as usual was not going to get us any better because humans have such a reductionist view because of all these other tendencies that, you know, De Duve talks about, that we don't pay attention to the long term until it's there. And by then the price to pay is 10 to 100 folds what it would have been to pay it. As the saying goes, an ounce of prevention is better than a pound of cure. And yet we systematically go into many pounds of cure.

Yeah.

So education has fallen behind into recognizing that because of human nature, let's call it. And these problems have accumulated to a very, very critical stage. That was the first driver. The second driver that started hitting us was related to technology. Starting with the internet at large, devices like the iPhone, but really the heart of it is social media. And social media democratized things beyond the reasonable in a sense, because anyone with any wacky opinion can find 100,000 people like them around the world. When you were constrained to a village, you were the odd person in the village. Obviously there are, there's good and bad to being the odd person in a village, but unfortunately, you know, you wanna create a religion based on human sacrifice, I'm sure you'll find followers. It is that crazy. So that's where we got seriously hit. And now you add onto that this wave of AI taking some cognitive roles that humans were comfortable with, feeling, well, you know what, machines have taken some of the psychomotor roles, but at least we have the cognitive and the affective roles that are still quote unquote human.

Yeah.

And then you start seeing AI chipping away at these things. And that's what the book is about. Mm-hmm. We talk about what exactly is being chipped away at, how much can we know, how much can the future hold about all this. But in essence is to say, look, there is no refuge. You cannot say, well, the effective domain is entirely us. For instance, your 5-year-old will love to be read by an AI for the 5,000th time with the same level of energy, if not more. More, whereas you as a parent will be exhausted and not so patient.

Yeah.

And AI— so AI can be tireless. AI can also be very courageous in the sense it doesn't know fear. So if you have to send it on a dangerous mission, it will do fine. It just doesn't have the concept of fear. So even the affective domain can be affected by AI's capabilities, not only the cognitive domain. Because of all of this, we're rattled appropriately. And we're trying to figure out, well, what are the limits? What is our next role here?

No, absolutely. And I, I mean, I really appreciate the fact that you've, that there is a, a long history in the book, cuz I think one of the things that interests me is the relationship that humans have always had with technology. And this is just the latest iteration of that. And of course it's moving at an incredible pace, but it is still a relationship with the technology that we have to reckon with, I think. And one, one of the things you talk about and you were talking about in the keynote was just how our brains have evolved the fabulous capacity to be lazy, to save energy. And I think at a very deep biological level, that energy kind of optimization or, you know, energy saving is a really important function. And so therefore, one of the things we do, obviously one of the things we do is generate cognitive biases and rules of thumb for, you know, to simplify and, and be efficient. But at the same time, what we also do is externalize into technologies that we can use, whether that be language shorthand that expresses ideas in a, in a more efficient way, or whether it's, you know, then it's later tools and technology. And that for me, there's something very deep and important there about the relationship. In fact, there's a feedback loop as well in how does the, not only how do we use the AI or the latest technology, but how does the technology shape us as well as we—

Absolutely.

Dialogic relationship, I think.

Absolutely. And this is something that anthropologists have been telling us for a long time, you know, the relationship between the tool and, and evol— and the human evolution or any evolution for that matter. The simple fact that we have a, an opposable thumb had a huge impact on our capabilities, right? And so you have this feedback loop between the ability to create a tool and the tool's ability to give us better nutrition, for instance. Mm-hmm. You know, we, we domesticate fire, we eat better, then our brains grow better and your babies grow more healthy and on and on and on. The, the difficulty now is how these changes with AI are happening far faster than evolution can adjust. Now, I don't know if epigenetics, uh, will be sufficient to adapt us fast enough. It may be that, uh, we're gonna have to accept at some point that any way to enhance our brains will be a good thing. The same way we use caffeine to enhance our brains. We may decide in a couple of decades when we're confronted by all of AI's capabilities that we have to supercharge ourselves with nutraceuticals or do, um, advanced CRISPR or whatever. I'm not saying we will, I'm just saying these are the sort of things we'll be confronted to.

Yeah. And do you, do you feel that it will have an impact, AI particularly, on even the way that we think about knowledge and the way that we think about even our, our existence? There's a comment you make actually in the book. About the way that AI, you feel, will help us understand the mind and intelligence better through the development of better ontologies, which I found a fascinating statement. So where do you feel that it can have that feedback loop or may, but in not necessarily that everybody might see as positive ways, right? Making us perhaps a bit less human and a bit more computational. I don't know.

Well, you think there are really two angles to the question. Uh, one, which is the anthropocentricity of our views, right? You know, We, we look down on things that are intelligent, animals that are intelligent, but they're far more adapted to their world than we would be. So their intelligence was a result of their adaptation to their surrounding, nothing more, nothing less. Uh, so whether you are a, an octopus, right? And you have, you know, your intelligence distributed in your arms and you're dealing with your environment and you're pretty smart, you know, even in human terms, or a cat or a dog, et cetera. Or marine mammal, and on and on. So we have this continuum of consciousness, continuum of intelligence development, uh, that adds up with more and more capabilities, right? Well, so AI comes and it's gonna have a number of capabilities, not fully like a human, very jagged, sometimes very poor, sometimes incredibly powerful and everything in between. And so we have to look at that intelligence the same way we look at animal intelligence and say, It's not like us. It has different capabilities and we haven't scratched the surface enough on the animal intelligence side. And that's something that I think will benefit from this approach. Because if you remember, all the AI analyses from very early on were all about creating a human mind. That was just the one way of looking at it. The human, right? As the pinnacle rather than each one of these animals as the pinnacle of their own environment. Yeah. So, so that's one angle. The second angle is, of course, neuroscientists and computer scientists have for a long time trying to look at each other's work. But in recent time, I think that computer scientists have gotten ahead of themselves by using anthropomorphic words like reasoning, like emergence, like intelligence, et cetera, without clarifying what they mean by that.

Yeah.

So it's been rather sloppy, honestly. Yeah. And some of the things we are trying to do as a center is to try to figure out What exactly are the definitions or consensus of definitions so we can tell whether or not we're gonna have AGI? If you don't know what the goal is, of course you're never gonna get there. And it's whomever hypes it the most that will claim arrival.

Yeah.

So that's one angle, which is really define the terms and understand how they may or may not apply. When we talk about reasoning, are we seeing reasoning or not? Well, okay. What qualifies as reasoning? Let's decompose that. Hence this issue about ontologies and taxonomy being so critical so that we can be much more specific in our definitions, and then we can see whether we, we get there.

Yeah, I know. Absolutely. And I think it, it's so important for educators dealing with young people as they navigate this world, right? I mean, just as a personal anecdote, I can think of many examples where I'm sitting with my son and he's not understanding whether the thing that he's engaging with is a human or a machine. And so, you know, whether it's just a very basic chatbot that's been around for a long time on a website. Right. And so there's a very blurring boundary, particularly because of the lack of clarity in some of the language about what's actually happening. And obviously part of the hype as well. And I think that feels really important to me as an educator because I'm thinking, you know, one of the things I want to help young people to do is to understand and read the world. And if the world is being presented to them in a way that's quite blurry in relation to this computation thing happening over here, and, you know, and then there's this very human carbon-based thing happening over here, and they're not the same, you know, I, I believe they are fundamentally not the same thing, whereas some people may disagree, I guess. But I think that's part of what we do as educators, right?

Exactly. And that's actually one of the problems of technologists who want to dress these things with human faces, et cetera, to further blur and go beyond the uncanny valley.

Yeah.

And I think it's a really big mistake because we really should constantly refer to these things as it, not as he or she. So the anthropomorphization in my opinion is dangerous and that's what we need to constantly pull ourselves back from. So for instance, I was in Madrid at a conference and they were showing this avatar, you know, basically it was a little cardboard thing with projection on. It was of course chosen as a pretty woman, young, and of course it was chosen as a talking avatar. And I said, no, I refuse to fall in that trap. I'm gonna call it an it. Yeah, stop it, people. You know, it seems cute to do so, but it's actually bad. So we need to keep that distance and keep reminding ourselves it's an it. Because it's an it, it's going to have a different behavior. We call animals in English, it. I'll call them he or she. Yeah, the same.

Absolutely. And I've heard Yuval Noah Harari say the same thing, that one of the most dangerous things— or actually, one of his recommendations was that any AI should have a disclaimer on it that says, you are engaging with a machine at this point. Now, this is not a human being.

Exactly. And so I know technologists are very, very enamored by making people believe it is almost like a human, basically triggering the ELIZA syndrome.

Yeah.

Where everyone, even in the '60s, thought they were talking to a real human just because of very crude response types. So imagine if the responses are much, much more sophisticated.

Well, exactly. I mean, and that's what we're seeing now, right? I mean, you know, some of the simulation, the digital twins, the vision, you know, I saw Reid Hoffman engaging with his own digital twin and it's, I mean, it's, it's quite incredible. The capability at certain levels and obviously based on some significant compute power. But it is quite amazing and it's quite hard to tell if you're not studying closely which one is the real read.

Well, it starts with images, right? But I'm pretty proud of the fact that there was a New York Times test just a few weeks ago and I succeeded with 10 out of 10 being correct which ones were humans, which ones—

Oh, nice.

Okay.

Yes, but I'm not sure that this advantage is going to last very long.

Yeah.

Uh, in another 2 years it'll be impossible to tell the difference. Right now I can still, based on some subtle cues.

Yeah.

But this is something that we've known for a long time. This was gonna happen. And you, you mentioned my past. There's something I've been watching for decades and you, you saw how these things progress by S-curves, a succession of S-curves. And we are now going through a very, very major one. And it will— it's starting to flatten. There's still no GPT-5. I think single types of algorithms running models are not gonna be everything, the same way that the brain has multiple types of cortical structures to deal with multiple situations. But over time, over the next 20, 30 years, we're going to get incredible capabilities that are hard to fathom, honestly. And so the question for education is, well, what do you do If you have no idea what's gonna come next, what do you educate for?

Yeah.

And I mean, and, and that's why it's a great updating of your Four Dimensional Education because a lot of that was already laid out there in really interesting ways and built around the broader, well, there's a, there's a few points I'd love to pull out, but obviously the big one is the enduring goal of education therefore being wisdom. Based on, on all of that. But one of the things I really appreciate about your work is that you are integrating and both-anding, let's say, rather than either-or, because I mean, everyone's a bit sick of all of the, the educational culture wars at this point. No one's ever going to win. And what I feel you've done with that model is presented a really clear statement that of course it's all of it. I mean, it's not, you can't just pick knowledge or you can't just pick skills or competencies or character as your pet project and ignore the rest. It has to be all of it as we grow whole human beings, right?

Right. So I've, you know, not being from education originally, I was floored to find all these, all these parochial fights to the death. You know, it's my view or your view thinking, wait a second, both of you are correct in some context and both of you are wrong in some context. But you know, because academia is rewarded by its publishing a specific philosophy, people live or die by that philosophy rather than accepting that their philosophy is always going to be incomplete. And so that's why you have this, this tension between academia rather than realizing, well, it's not only about didactic systems and it's not only about constructivist systems. Of course it's the blend of both.

Yeah.

One is better at condensing 3,500 years of human evolution into 12, and the other one is much better at real world motivation. So you have to combine the two.

Yeah, absolutely. You made it, if I, if I may, you made a really interesting comment, almost an offhand comment in the keynote that I heard you give recently, that maybe we can compress experience. And that's part of what education does. But the reason I'm, I wanted to ask you is because sometimes that's part of the nub of this, some of the kind of progressive traditional debates is something around me going through the experience of coming to know the thing that we're talking about is a deeper learning, let's say, than just being told by somebody who's already figured it all out. And you, yeah, so you made a comment that actually experience can be compressed. So that sense of compression, that's what knowledge is doing. So yeah, what—

so think, think of even didactic systems, I mean, are compression. This is what we're compressing 3,500 years of human evolution of human knowledge into 12. Making choices here and there, of course, about what matters, what doesn't matter, but still you don't go and recreate mathematics from, you know, Sumerians onwards, right? It will take you forever. You will never converge. It would be incredibly time inefficient, if not completely ridiculous. However, knowing how to apply this in a real context that's transferable, that is really important. It's very motivating. It's very grounding of the theory. And so that too is a compression because trying to do simulations ahead of time is what a project is about. When you're in school, you're doing a project. It's not going to be truly starting your, your mega company. I mean, for some it may be, but for most kids, it's not going to be that. It's going to be a simulation. You're going to fail at it and you're going to be failing forward. That's all good. So it's basically doing simulations and thereby compressing the time to real experience.

That's, yeah, no, that's really interesting because I spoke to Carol Sanford on the podcast recently who has done a huge amount of work in regenerative business and regenerative education. And she talks a lot about the value adding process and actually what's part of what we talked about there was that it's actually the simulations that is often quite a dangerous part of what we do as educators because we are trying to simulate real life. So that students don't participate in the value-adding process of engaging with the real, the world as it is, right? I mean, you've, you've invoked context and grounding a couple of times, and I think there's something incredibly important about helping young humans inhabit contexts, right? In a very real way. So I, I do share some of her reservations that the, that the simulation, while useful to compress experience because you haven't got time to really get into it or the system is constraining you. I do worry that the simulation is, again, is artificial in a, in a quite significant and important way.

An engineer would never say that, right? As an engineer, as an engineer, we will always do simulations. You're not going to build a bridge without a simulation first, right? Yeah. It's not as much simulations that are a bad thing, is that poor simulations are a bad thing, the same way that poor assessments are a bad thing. And so think of, of the following. Would you prefer to only read about the plight of refugees, or in addition to that, don virtual reality goggles and go through a refugee camp? If you cannot be taken for safety reasons, cost reasons, or whatever, to a real refugee camp.

Yeah.

Won't you want at least the simulation aspect? It's better than just reading about it. More emotionally connecting.

Sure.

And not everyone's going to have the, the latitude to go to a refugee camp.

Yeah.

And so a simulation is not a bad thing. Same thing, you're not going to be able to spread the pandemic to test your simulation. Better not.

Sure.

But it would be a good thing that you see a pandemic, you simulate its parameters, and then you apply that to near transfer another disease, far transfer to something like global warming. And obviously you're not going to be able to simulate the true reality, but as we say, all simulations are faulty, some are useful.

Yeah, no, sure, no, okay. I mean, definitely that makes sense given the constraints that we're operating in and the examples you've given, for sure. I guess I would just when it's an experiential thing, so that, you know, with the refugee example, absolutely, there may be real value in the multimodal engagement. I guess I would then say to go that next stage further, why not actually engage with real people and real people's lives if that's possible? It may not be possible, but—

Absolutely, but that's what the project should do. So when we say simulation, we use that word at large. Necessarily in a narrow sense. So we want a project to be real-world focused with real parameters, etc., etc. So it's not simulation as in completely devoid of reality. Yeah, the same way that simulating a bridge structure is not devoid of reality, it has real parameters embedded.

Yeah, no, absolutely. I, I guess the other differentiation I quite appreciate is, um, a difference between the systematic and the systemic, and I think there's a A bridge is very much a systematic, of course, you know, it's going to behave in a broadly similar way one would hope in a theoretical simulation as in reality.

Yep.

More systemic, you know, complex ecologies like messy human interactions.

Yeah.

There's something very important that I think we just need to be really clear about what it is we're talking about and not cross-fertilize too much to make the systemic into a systematic thing.

Totally agree. Those wicked problems are, well, wicked for a reason.

They are wicked for a reason. Yeah, it's true. Perhaps could you say just a little bit, just on the point about wisdom, which I, you know, I think is a, is an often, yeah, misunderstood concept, dusty old concept kept in libraries. But you, there's a great appendix to the book which I'd really recommend people have a look at, of all the research, which is fantastic. Could you just say a little bit about the meta-learning piece? Because I, I personally, I feel like integrating, because you've got the knowledge, the skills, the character around the learner, but then there's this role of this meta-learning on the outside. So, and I somehow that feels really important from a wisdom perspective to be able to see the things or integrate or understand the things that you are doing.

It is, it is extremely important because at the end of the day, a machine in some way, shape, or form is going to follow its training, its instructions.. It's not really gonna have, at least for a while, who knows until when, an ability to adapt. That's really what makes a big difference. And so what is meta-learning? Meta-learning is metacognition and meta-emotion, if you prefer mindfulness, but I prefer meta-emotion. It's less charged.

Me too.

Yeah. And so, so basically it's your ability to reflect on your actions and your emotions. This ability to self-reflect is actually a very inexpensive pedagogical practice, very inexpensive with extremely high ROI as a return on investment, because it allows you to stay ahead. Basically any task requiring adaptation and coordination, which is any job, is what an AI cannot do and cannot do for a long time. And this is where humans need to be. So that means paying attention to meta-learning, particularly adaptability, particularly self-directed learning is where it's at. So sure, we train you for 12 years. So you know how to continue adapting and learning on your own, because we're launching you like a little bird out of the nest. We're launching you and the rest you have to do it yourself. That will be training you to do that. So rather than training you only on content, So you go into university and painstakingly figure out the world afterwards. We're going to train you to figure out the world, you know, from day one and the mechanism to continue figuring it out by yourself. And that's where this conversation goes back into the project, the constructivist element.

Absolutely. Yeah, it definitely does. I think The other thing it does for me is, I don't know if you're aware of Mary Helen Immordino-Yang's work, but there's something there for me about this. Education is very, very goal-directed in a lot of its, the way that whether it's skills or knowledge focus, it's just, we do this for that outcomes, outputs everywhere. What she talks about in relation to transcendent thinking and the default mode network activation and the reflection and what does this mean to me, building meaning. How do I fit into this? How does it connect with me? All of those kind of bigger meta questions. Her research seems to be suggesting that they not only do they have really beneficial outcomes for young people from a competency perspective or from a readiness for the world perspective, but also from a wellbeing perspective. And she talks about life outcomes being enhanced over the longer term based on how much time, obviously not all the time, but, you know, an important proportion of the time they're spending in this transcendent zooming out kind of phase. So I don't know if— yeah, any reflections on that?

Well, so there's really two aspects that you have mentioned here. One, which is the zooming out, meaning meta-learning itself.

Yeah.

But the other aspect, which relates to identity, agency, and purpose, right? What's— as you were asking some questions, your questions were related to It's true. What am I here for? What motivates me? Why should I do this in the first place? Why not rely on the engine to do the work for me, et cetera? What motivates me? And that's why in the framework you saw the addition of this triangle up front about motivation, which contains identity, agency, and purpose, which are the mechanisms to foster that intrinsic motivation in an age where people are gonna be accessing very powerful machines. What keeps them in the game? What motivates them. And even that has to be nurtured, you know. You don't— you, unless you're Mozart, you know, you're not born knowing exactly what you're gonna do with life. And so you have to explore, figure out who you are, what you like to do, what you can be good at, etc. And you develop passions.

You don't—

you're not necessarily born with passions. Yeah, most people are not.

Yeah, absolutely. And, and that also speaks to what you say in the book about the growing need to personalize education. So that's, you know, an important section of the book is about then how do educators provide those opportunities for young people to connect with their own purpose, agency, identity, and motivation? But that's not gonna be the same for everyone, clearly. So therefore you need a level of personalization happening.

Exactly. But, but that level of personalization doesn't need to be as coarse as choosing a given disciplinary path. Like, you know, I want to be in math or in STEM at large or in humanities or whatever. It could be at a finer level where let's say all of us have to build a robot, but the agency we have is to decide which type of robot. And the purpose is going to be asked of us by our teacher saying, okay, Tim, you want to build a submarine robot? Fine. What for? And you're going to think for a while and think, well, I want to count plastic particles in the ocean because I see on the UN SDG that's an important thing to do, pollution. And same, another student will say, I'm gonna build a flying robot so that it can help crops.

Yeah.

Right. So this is how you can, at a finer grain level, develop a sense of agency and a sense of purpose even within boundaries. Creativity is always within some boundaries. And so you have to develop that as part of the game.

Yeah.

Yeah. I totally agree. And, and the other thing, I, I have some concerns about the whole kind of personalization move, particularly when you start bringing in intelligent tutoring systems and things like that, because there is a— clearly it's highly efficient in all sorts of ways. And there's the— you mentioned the two sigma problem as well in the keynote briefly, but there's a— there's some kind of an idea that wouldn't it be great if everybody had an intelligent tutoring system that personalized their learning absolutely to exactly where they are, really is so optimizing to help them to move quickly down a path in terms of particular things. Clearly that can play a role and it, based on Bloom and those things, it seems like it, it works pretty well. However, it feels to me that there are some, yeah, some, some concerns about what that then does. Again, talking about the way that technology then feeds back to what we expect of the world. Do we then inhabit the world expecting that it's gonna deliver to us exactly what we need at the right time, which I don't think the actual world does that very often.

Yeah, it may spoil you like, you know, parents spoiling a child, always feeding the right thing. Yeah. Now, of course you can add some randomization to it. You can do all sorts of things to make it wonky, but look, at the very, very base of it, it's Bloom's work, right? But Bloom's work has not been well replicated. And so first of all, if we're building all of this in thinking, if we build a tutor, therefore we solve it all. Well, we have to first of all really examine that carefully.

Yeah.

First of all, understand, is it really 2 sigma or 0.1 sigma? I mean, hey, I'll take 0.1 sigma, but at least let's know what it is. So there's that aspect. So number one. Number two, a lot of these systems presuppose that our progress is very linear, very simple, very structured.

Yeah.

That works in very rare disciplines like mathematics, which are very structured by their very nature. But, you know, discussing with you whether or not you should read Dostoevsky at this point because you've read Rousseau already, um, that's becoming like a recommendation engine on Netflix more than anything else. So it becomes much, much, much more complicated, and you could say just as fraught as a human recommendation. You know, if you have a teacher Yeah, they're gonna make that recommendation. They may be right, they may be wrong.

Yeah.

So there's a tension in me about this because on one hand you can imagine situations where without a good teacher or without a teacher at all, an ITS would be so much better than nothing.

Sure.

So, you know, even if you have a bad teacher, right? I mean, that happens. Not every teacher is great. Sorry. In any profession you have people who are not good.

Of course. Yeah.

So what you have that situation, or you have no teacher at all, or you have only a 15-year-old that can keep the classroom together, but then the rest is Italian tutoring. That's great. But it may work for some disciplines, not for others. It may work better for some students who love learning on a screen rather than with other humans.

Yeah.

On the other hand, some, the other 90% are going to be very social and want to learn socially. With a teacher, with other students, in which case the ITS is only a record keeper of sorts, which is not a bad thing again. But you know, there's this range of scenarios that we are only starting to scratch the surface of, and it's going to take a while to figure out what, where, and how. And that's even assuming that the base assumption of a significant sigma improvement—

yeah, absolutely, absolutely. And I think there's also an equity issue in there because it feels— I also have the worry that, you know, the, the context where they can't get highly trained teachers to young people. You know, they'll be shipping out technology and ITS systems to young people across Sub-Saharan Africa or wherever, because it's just, it's more difficult to get teachers, you know, at scale to young people. So there's an inequality, I think, potentially there. But the other thing I, I think is interesting is in a way, some of this is modeling on the kind of the old aristocratic tutoring, right? All the great polymaths of yesteryear had their own aristocratic tutor who was, you know, an absolute master and who would, would support them through. The equivalence of then saying that an intelligent tutoring system is more or less the same thing. I mean, it's back to our simulation conversation, right? Is it, can you simulate that? Because actually one of the things I often think about education is the things that are being learned by the young people are not the things that the teacher thinks that they're teaching. So by that I mean there's a ton of signaling, there's a ton of non-verbal interactions, there's contextual responsive information that young people are swimming in when they're in relation to— not to sound romantic and Rousseauian about it, but they are in relation with another human being. And that's, that feels from a learning perspective incredibly important.

It is, it is. And If you don't have it, what's the alternative?

Sure.

Yeah, no education whatsoever.

Yeah, absolutely. Yeah, yeah.

You're like, well, okay, this is better than nothing. Go— that goes back to the example about virtual reality goggles and the refugee camp. It's better than nothing. It's better than just reading the book.

Yeah. I can't leave you without asking you about one of my favorite topics, if I may, which is energy related to the climate crisis. And you, you had a question at the— this came up at the conference and I just wanted to come back to it with you because I think the elephant in the room of AI is how much power, energy, water, etc. is consuming. How much, you know, the kind of arms race for energy production from the big AI companies and, you know, the carbon emissions increase of Google and of OpenAI. And, you know, there's an environmental cost to this development of the technology, which we can obviously maybe have an agreement or a disagreement about whether that's worthwhile and valuable and worth the cost or not. But one of the things that I'm just very keen on is making sure people are aware of that. I mean, it is a, it's a phenomenal need for, and one of the, I think the big resource wars potentially of the future will be around how can we find enough energy to power our AIs, whether that's, you know, nation versus nation, we shall see. But you, you made a comment about the nuclear alternative to that. So I don't know, maybe I'd just love to hear your thoughts a bit more about—

or so first of all, if I'm gonna cut back on, on the source of wasted energy, it would be on cryptocurrencies. They generate absolutely nothing. It's the Ponziest of Ponzi.

Nice, I like it.

Number one. Number 2, there are plenty of models that are showing up now, smaller language models that are quite efficient. They may be running on your laptop, next generation, et cetera. And so that's gonna cut back on the data center side. Now, of course, it may just push the problem to the client, who knows, but perhaps not. Perhaps it'll be more efficient that way. Third, there are all sorts of new chips coming on, techniques, et cetera, that are dropping the power consumption. Fourth, I think it's a good use of power consumption, honestly. At least it's a likely productive use rather than cryptocurrency, for example, or even financial manipulations all day long. So there are plenty of ways we could cut back on power consumption with particularly the finance folks that are wasting all this energy and use it for more cognitive, more useful tasks. There is a judgment here about usefulness. Yes, I'm making it. Then finally, there's the strange possible unintended results of all of these data centers pushing for development of technologies that are cleaner.

Sure.

For example, fission and energy cleaner fission with containment, et cetera, et cetera, that we know how to do now. And certainly not follow the example of Germany that killed all of its nuclear power only to fire back coal plants.

Yeah.

And so that's, that's the sort of thing that I have high hopes for because there are plenty of small reactors out there. You know, think of submarine reactors, you know, ship reactors and so on. Those are being transformed by the Rolls-Royces of the world into small portable citywide reactors. And so those, those could power a data center, etc. So it's pushing on fission, is pushing on fusion because a lot of them are making big bets on fusion, and that's accelerating small fusion a lot faster than the giant efforts. So I really think it's a better use of energy and it's one that may foster more advances in renewable energies than if left to its own devices or to the coal and oil companies to develop.

Sure.

Yeah.

Okay. So yeah, interesting. So you're very much in the techno optimist camp, shall we say?

Well, from this narrow standpoint, otherwise I have to force myself to be a, let's say, humanistic optimist. But it really is not easy every day.

Yeah, no, it is true. There's a lot of reasons not to be, but it's, it, I find that it's just very interesting because there's a game theoretic thing going on with AI. There's a kind of a, you know, it will just accelerate and it's accelerating in this. I don't know if you know Daniel Schmachtenberger, but he talks about The fact that a very adaptive technology is obligate, like everybody will compete because if you don't compete to continue to develop it, and we're seeing that now, right? With AI. So there's, there's an acceleration there. And so in a way there's a, a doubling down on the technology will, the efficiency gains and the technology will kind of solve our challenges. It's a kind of, it's a difficult position, I think.

It's a difficult thing to really understand. I mean, one thing is clear, we've gone from the science phase to the engineering phase. We're smack into the engineering phase.

Yeah, sure.

I think we're doing better, but we're not gonna see a new S-curve until we develop completely new methods and algorithms. And those are being swamped away by the focus on large language models. So in a sense, I'm happy because it's gonna Things are gonna slow down. It's gonna give us a chance to catch up and gather our thoughts before the next wave.

Yeah.

I'm not unhappy about that, honestly.

Thank you. No, I really appreciate it. I mean, it's great fun to talk to you, but it's also, you know, and I'm conscious that we've gone off into quite complex territory. It's not the kind of 101 for teachers with AI, but I think there's a lot of that out there and that's fine. You know, people invite people to read the book, but no, I just, I so appreciate what you're doing and the fact that you're kind of raising an opportunity for a really sophisticated conversation about what this means and not an overly hyped conversation about what this means in education. So thank you so much.

My pleasure. My pleasure.

We hope you've enjoyed this episode. Please feel free to continue the dialogues with our guests with us on our blog or on social media, or within your own networks.