AI Gave Every Student a Supercomputer — Education Has to Catch Up

0:00I think we need to move away from preparing kids for jobs. because jobs are going to change. The way students are taught and what they're taught to optimize for is very at odds with where we're at right now with AI. The safest assumption we have to make in this moment is that kids are going to be using artificial intelligence at home. So the classroom really has to be a place where the deep learning is happening and where we're raising the bar on knowledge. When you outsource your cognitive work to an AI, you actually become cognitively weaker. This is potentially really detrimental to society more broadly, that we become so reliant on these technologies, we stop believing in our own ability to make decisions.

0:42In the age of advanced technologies, kids need to read more and we'll have to make school harder. The most important skills for the future have nothing to do with technology. What should kids be studying and learning today for a future that's going to be transformed by artificial intelligence? and how does the institution of education need to evolve to meet the moment. In the spirit of South by Southwest, I'm going to share a talk that I delivered on this topic, where I go into detail on some of these ideas from where we can expect this technology to take us to the skills kids in school today need to be cultivating and the system-wide change the institution of education is going to need to undergo to prepare kids properly for this world.

1:26I'm Sinead Bovell, and this is I've Got Questions. I'm super delighted to be moderating this conversation with the fabulous Sinead Bovell on AI and the future of education. She's got a massive fan base. It was hard to get to this stage, actually, with people kind of coming up to her and saying how much they admire her work. Sinead is a strategic foresight advisor. She advises C-suite executives and senior leadership across governments and global corporations. And she's an 11-time United Nations speaker. And she has delivered formal addresses to presidents, royal families and Fortune 500 leaders on topics from synthetic biology to artificial intelligence.

2:14And today she has advised, very relevant to this conversation, 16 ,000 educators, government officials and policymakers on redesigning education for the age of AI and emerging technologies. So, Sinead, tell us what it is to be a strategic foresight advisor and your lens on AI and the future of education. Yeah, education is the bedrock for a healthy democracy and for a functioning society. And it's not just essential for things like economic mobility and economic security, but fairness as well and for well-being. I believe there's no such thing as a state that over-invests in children's future.

3:03An investment in children is an investment in national interest. So you want to foster an informed and adaptive citizenry that cannot just safeguard the future but thrive in it, especially a future that's going to be as complex as the one children in school today are entering into, which will be shaped by quantum computing, genetic engineering, artificial intelligence, commuting back and forth to space. This is an incredibly complex world they'll be entering into. So the more that they can understand it, the more we can support them in that journey, the better equipped they are. And that's an investment in a country's economic security, in our collective health and well-being, and our overall security and national security interests.

3:47Goodness. So couldn't be a more pressing topic. And before we dive into some of the kind of final points, where would you say, taking a step back, like where are we at this moment with AI? Why? So if I were to say where we are, I mean, it's very, very early. Maybe it's 1992. The internet has dropped. Companies are experimenting. We know it's maybe going to be a big deal, but there's still a lot of doubt. Some people are also kind of playing around on it. But we have yet to fully comprehend the way it is going to fundamentally transform our world. The Googles, the Apples, the Amazons of the future, they have yet to be invented.

4:32But they're coming. And artificial intelligence, it's also a general purpose technology. So similar to something like electricity. Think of how pervasive electricity is. We don't even think about it at all. It's so foundational that it's moved into the background. So we will soon be streaming artificial intelligence the way we stream electricity. That is going to be a fundamentally different society to live in. And these general purpose technologies, they take time to get so entrenched in society. But you know when it's reached that point, because when people can't access general purpose technologies, whether that's at a country level, certain neighborhoods, we deem that wildly unethical.

5:15Who doesn't get fair access to the Internet? Who doesn't get access to electricity? That is the path artificial intelligence is on. So if artificial intelligence is going to be this general purpose technology and fade into the background, what does AI and education look like now? Like how should educators be considering AI and education given that it will be in the background, but today we're at this very early phase? Yeah, so I think that there are kind of three pillars that are related but distinct in terms of how we should be thinking about AI in education. The first pillar is safe adoption for kids and for learners.

6:02So this means equipping kids with the tools to navigate artificial intelligence because they're going to be on these tools at home regardless. this. They have supercomputers in their pockets, supercomputers on their iPads. So giving kids the skills to utilize these tools safely. So that's conversations like AI isn't your friend, right? Your chatbot isn't something that you tell secrets to. This is what we do or don't share with artificial intelligence. And this is also how you ask it good questions and validate its answers. So that's pillar one. Pillar two is how do we more urgently adjust what we are teaching in school or just the formula for what happens in the classroom versus what happens at home, knowing that kids are going to be leaning into these technologies at home to do homework and to complete assignments.

6:51The third pillar, and this is where I think we are rushing into, but this is actually the long-term game, is how do we fundamentally redesign the entire system of education for the age of artificial intelligence? But what seems to be happening in this moment is we are kind of merging all of those pillars in a sense of urgency. And this leads us to deploy AI in schools for the sake of feeling like we need to meet the moment by bringing AI into the classroom. And there are a lot of technologies that aren't ready. So I think we focus on pillar one, giving kids the tools to use these tools safely if they're going to be using them on their phones.

7:32We slightly adjust what we're teaching to account for cheating and homework. But it's much more at the departments of education, the ministries of education level to take this longer term lens and fundamentally redesign AI or school for the age of artificial intelligence. So I think we've heard this week a number of different ways that educators are experimenting with AI, different pilots, different ways of going about it. Like is now the time to be, sounds like we should be teaching it and educating about it. Is now the time to be experimenting with it in deeper ways? Yeah, yeah. So teaching it, it's more so AI is a hard skill.

8:14So that should be happening. Experimenting with it, yes, we need to be running these pilots. We need to be gathering the data as to what's working and what's not. But it has to be in very, very intentional ways. and not just assuming that we can just throw in an AI tutor somewhere arbitrarily, and that's going to be sufficient. And making sure that we're not running social experiments that jeopardize learning outcomes for the sake of just feeling like we need to quickly meet the moment. So yes, I think that these experiments, they're vital, they should be happening, but they need to be very, very intentional and very, very controlled.

8:51I know that you're working with Fortune 50 companies in this space and advising them on how to navigate AI and education. What are some of the data points and the advice that you've been giving them? Yeah, so it's been really, really interesting to look at some of the data that's coming through. And of course, we're still very, very early in the age of using artificial intelligence in education. But there's one clear trend that stands out. So I'm going to walk us through a study that I find particularly helpful. And this was done by the University of Wharton, the University of, I believe, Pennsylvania, and Budapest British International School.

9:32And it implemented artificial intelligence in math classes. So there were a few math classes in the high school, and they broke the class up into three groups. The control group, which is just your traditional doing homework problems with your textbook. The GPT base group. and this is the students that got uninhibited access to artificial intelligence, and then the GPT tutor group. So these are students that got access to an AI that has been designed to just guide them through problems, give them hints, but not the answers. All students got the base lesson for math together, and then they broke out into their respective groups and their respective tiers of AI access or not.

10:12So the study showed that when it came to the practice problems, The children that got uninhibited access to AI did 48 % better on the practice problems than the control group. The students that got access to the GPT tutor did 127 % better than the control group. But when it came time to actually test students without access to AI and do the final post-unit test, the kids that got the uninhibited access performed 17 % worse, so AI harmed the learning outcome. and the children that got access to the AI tutor performed at the same level as the children that didn't get access to any artificial intelligence.

10:54And so the conclusion of the study was that generative AI harms learning outcomes. But then there was a second study that happened at Harvard and of course we have to control for the fact that self-directed learning is a little bit different at a university level and clearly if you're getting into Harvard there's also some kind of higher order thinking that you're able to do. But that aside, it was a physics class, and they broke the physics class into two groups. The control group, which is the students that went to the traditional lecture with the professor. Then they broke out into peer groups and worked with one another and their peers and had instructor-led guidance on solving problems.

11:31The second group had no in-class lessons at all. The entire process was done with AI, But they specifically designed the AI to be self-paced, going with the student's needs, to provide immediate feedback, whether the student was on the right direction or off the right direction while they're doing the problems, to provide motivation and to really take all learning best practices and implement it into that system. And continue to adapt how it tested the child based on how they were evolving and doing the problems. When they did the general test after those two experiments, the kids that went the pathway of artificial intelligence performed twice as well as the peers that didn't get access to AI, and they were more motivated and more engaged.

12:19And so what we can learn from just those two kind of isolated studies is that you have to adapt the entire ecosystem, right? It's akin to inventing electricity, but only swapping out where the steam engine was and putting light switch there, not building out the entire assembly line and rethinking about how we design the system. That's step one. Step two, immediate feedback is absolutely vital in AI learning outcomes. Gone are the days if we're going to incorporate artificial intelligence where we wait for the unit test or midterms or the end of year exam to see where students are. Artificial intelligence needs to be able to extract the data in real time.

13:08This is how somebody is adapting or this is how they're falling behind. And AI needs to provide that feedback or else we learn visibility. We lose visibility into how well things are happening. Self-paced learning is also vital. So if you go back to the high school, everybody had an hour and a half to learn the math problem. So whether you were working with your textbook or working with AI, that hurt people that were doing the AI method and it helped people doing the traditional method. So kids need to learn at their own pace and the system needs to be able to adapt in real time. So those are just a few of the key takeaways when we think about AI in education.

13:43But that's why this is a longer term redesign and that redesign should not fall on teachers. And that's one thing that I think we need to really make clear, that this moment shouldn't fall on teachers. They already have way too much on their plate. This is something for government department heads and those designing curricula more broadly. And I think we're getting that wrong. Right. Yes.

14:13I had the opportunity to sit on a few panels yesterday and there was definitely, you know, a very understandably irate educators like we had COVID. We had, you know, we're in an underprivileged area. We've got so many pressures and now we have AI to learn and we have to do this two day course. And then the pilot ends and then we've got to go and do another two day course. And AI is the last thing that we need. So it sounds like there is a right way to do it. And there are definitely ways that can be harmful. And some of those right ways involve a lot of structural redesign of how students actually engage with AI.

14:55Yeah, this is like the institution of education we need to approach differently. It's akin to asking the accountant to redesign the concrete and the bricks. That's not what we should be doing. And I know that you, there was a, we had talked about a school that you had been tracking. So I think it could be helpful to share some of the insights there. Yeah, absolutely. People familiar with Alpha School in the room? A few people. What it is, it's a two-hour learning process where like all the hard skills, all the knowledge that you need to learn at school happens in a two-hour period with an AI tutor.

15:29and the experience is entirely personalized and adapted to where that student is at. So if you walk around the classroom, you'll see different students working on completely different math problems, let's say. And as it's understood what that student is interested in, the math problems become kind of contextualized within topics that they love. And so when critically and again sort of going back to one of the best practices or you know mandatories in AI being successful in schools there's real time feedback and performance on performance of that child. They can see their own performance and actually start to own that journey for themselves.

16:21And they get everyone into the 99th percentile, no matter where they've started in their journey. So I think that's a fascinating way to look at it. And I think that besides those two hours, so the point of getting all of that work done in those two hours, and some people are able to, some students are able to accomplish more, double in those two hours, and some on the higher performing end five times more. But the critical part is freeing those students to focus on life skills, on EQ skills, on kind of developing their own human ingenuity. And so I feel like that is, you know, from all the research that you've discussed, just an amazing example of what's happening today.

17:05Yeah, totally. And that's kind of the moment we're in, right? We're in pilots and experimentation and innovation, really a redesigned zoom out wide lens and some in some ways taking risks, but it should never hurt the learning outcome and it should never be a burden to teachers. And both of those things need to be true. Absolutely. So did want to kind of like dig in a little bit more into some of the current challenges with AI that, you know, educators, students, parents are experiencing today, which is around AI and cheating and using, you know, chat GPT to get to the answer right away. And what impact that might be having on the learner experience and kind of the point of being at school?

17:57Yes. Oh, absolutely. And I think that we're in a bit of a crisis in this moment when it comes to artificial intelligence and cheating. And we can talk about what happens when you kind of short circuit that thinking. But I think that the safest assumption we have to make in this moment is that kids are going to be using artificial intelligence at home. So whatever happens past 3 p.m., expect that to be powered by a supercomputer in some way. So we have to start there. That means we have to change what we are doing in the classroom. And so in some instances, that means maybe we flip what happens at home, happens in the classroom.

18:34But in other ways, maybe, for example, you teach history. You give children the research portion, and they can go home and do all the research with chat UBT that they want. But the higher order, critical thinking, deep learning and discussion, all of that happens in the classroom. So the classroom really has to be a place where the deep learning is happening, where the testing is happening and where we're raising the bar on knowledge. But we do have to assume everything passed for is likely going to be done, co-created by or outsourced to an AI system. And then again, the broader goal and the longer term goal is that we've entirely redesigned the curriculum to account for the fact that kids can lean into supercomputers.

19:20Because that is the actual goal in the end. Kids in school today are going to step out into a world with advanced robots, with supercomputers that are polymaths. We want them to know how to engage with these tools and these systems, how to utilize them, how to invent with them. And we'll have to redesign education to account for that. And we'll have to make school harder because you do get access to these supercomputers. And so in a kind of superficial way, maybe that means people are learning about quantum computing at seven years old because that learning is facilitated by a teacher and by a supercomputer.

19:54But that part is going to take time. So the more urgent kind of redesign is flipping what happens in school versus what happens at home. What happens if we don't do that? And I think it's quite obvious, right? We end up just short-circuiting the thinking. So there shouldn't be anything to cheat on because what happens at home isn't what we are evaluating. And that is, I think, the kind of the baseline that we need to move towards. And that's, I think, what we should be doing more urgently. So many places to go from everything that you just said there. But maybe a positive example of how outside of hard learning and maybe an AI tutor helping you learn the things that you need to learn at your own pace in the most individualized and sort of data rich manner.

20:44Well, how can you use AI outside of that core learning in a way that helps children become more human? Right. So like what does human flourishing kind of look like? And does AI have a role in that? I think we need to learn how they work and learn how to use them. But also, what should we as humans be focused on? Because we want to collaborate with AI. So what are the sorts of subjects and skills that are deeply human that we can focus on? And something I've been thinking about quite a bit recently is like different types of knowing. There's a cognitive scientist called John Viveki, and he plots out different types of knowing and kind of the most sort of like academic or kind of research based and fact and knowledge based learning is procedural knowing.

21:36And that's the kind of knowing that AI is really good at and is getting increasingly good at. But what AI doesn't have is lived experience and deep insights that change you as you experience them in the world, change your perception of the world and then change how you connect with others. And so I've been really interested to hear some of the talks this week about experiential learning environments. I learned about Thinkery, which is here in Austin, and it's kind of this interactive learning museum environment. And it's just really interesting to think about what are those deeply human skills that we can be focused on teaching students while they're learning what AI is and the best ways to collaborate with AI.

22:26Yeah, I think that that's vital. And I just wanted to quickly go back to the cheating. Another thing we'll probably need to do in the short term is introduce more pop quizzes and surprise tests. And they don't need to count towards grades, but to see where students are as we are in this kind of new territory. So a lot of times we don't know how much they're using AI, how much they're cheating with it, to insert more chances for assessment and be tracking that data. Because that's another thing, we don't have a lot of visibility into how this experiment is going. in terms of what AI can do and what AI can't do, and therefore what should we be teaching kids in particular and what skills should be fostering, my philosophy is we should never assume AI will never be able to do something.

23:10And the reality is we cannot predict the future, what jobs will be there, how advanced AI is going to get, and how quickly. That means we have to prepare kids for absolutely anything. Whichever way the future evolves, however quickly we get to the moon, start genetic engineering, kids can pivot, adapt, and think critically about the world around them. And most of those skills don't actually have anything to do with technology. They require deeper thinking. Critical thinking is absolutely vital. In the age of advanced technologies, kids need to read more, read for the sake of reading, and read in a way that they can come back to school or with their parents and discuss the ideas and have those ideas challenged.

24:00Kids need to play more in the age of advanced technologies. The future Steve Jobs of the world, they're not going to come from a corporate cubicle. They're going to come from people that have imagination, that can play freely, experiment, work collaboratively. Long-term thinking. So getting kids to think beyond the immediate horizon and beyond just this unit test in chemistry or in math. But how could this impact things in five, ten years to come? And even cross-disciplinary thinking. Kids in school today are likely to hold 17 jobs across five different industries. They won't be doing just one thing.

24:38So we have to get them to think, how does math connect to what I just learned in history, which might connect to what I do in philosophy or in English. So all of these new, and they're not even new skills, but it's just about centering these types of skills. Most of the important, the most important skills for the future are ones we can foster for free. And that's what I think we can sometimes miss in these moments where we feel like we have to lean into technology for the sake of it. But it's actually the other skills that we need to make sure we are doubling down on in the age of advanced technologies.

25:15And one kind of lived example that I do with my nieces and nephews constantly, since the age of about six or seven, I theoretically introduce them to technology. And this is what can happen in the classroom as well. I explain concepts in age-appropriate ways, like genetic engineering, and I ask them to interpret what that would mean for their world and their sense of ethics. So if we could theoretically make sure nobody gets sick in the world with these technologies, should we do that? But what if, to my nephew, it meant all that basketball practice you do, somebody didn't have to do because that same technology allows them to suddenly be really good at basketball?

25:54How should we think about that? They engage in the higher order thinking. They're exposed to the longer term concepts of technology without actually having to play around passively on an iPad. So these are the types of deep conversations and higher order thinking that can happen into the classroom that teachers are uniquely positioned to deliver and to facilitate. I mean, when you think about a teacher and they don't get enough credit for all of the things that they do, I mean, the curriculum is one small part of it. They are social workers. They are therapists. They know their children inside and out.

26:28So being able to go deep into these types of conversations, that's what we also need to be focusing on. And I know it sometimes feels counterintuitive because I'm a futurist and I spend most of my days in patents and technologies talking about robots, uploading brain interfaces. Yet the most important skills for the future have nothing to do with technology. Great. And I want to go back to something that you said. So technology for the sake of technology is absolutely not the right way to go about things. But learning for the sake of learning is. and some really interesting insights this week about how schools and test-based learning does not set students up to enjoy or take pride in just the sort of active learning and sort of encouraged and optimized to find the answer, get the answer right.

27:22And then even in a critical thinking class that cognitive scientist Christine Legare talked about yesterday, Even in that class where there is no right answer, what the students wanted is they wanted the rubric to get there. And what she said to them is like, well, you know, do you think you're going to when you have a job in the real world? Do you think that there is a way like, you know, you're not going to be asked to to discover the answer or where we should go or the right path, I should say, by being given the rubric. So it sort of seemed very, we're at this kind of acute moment where the way students are taught and what they're taught to optimize for is very at odds with where we're at right now with AI and the fact that it is designed to give you the answer.

28:17And I actually talked to a teacher who said, teaching 16 to 18 year olds, and she was like, OK, so what I do to try and kind of circumvent the use of AI in writing is I have my students write in class. And, you know, I do give them a bit of a rubric like this is, you know, a good structure for an essay. And then when it comes time to actually submit the essay, they go home and type it up. In some cases, in a few cases, that student had kind of ripped up their essay and basically just completely generated a new one in chat GPT. And what that said to me was, I mean, there was no time saved or cognitive load saved in doing that.