AI Learning and the Global South (an episode of the Great Minds on Learning podcast)

This podcast is supported by World of Learning Summit. This leading L and D event returns to Olympia, London on the 3rd to 4th of February 2026. Two days to learn, discover, interact and network with the best in the business. Register now. It's free@learnevents.com this podcast is supported by Synthesia. Training videos shouldn't take weeks to make Synthesia lets you turn text into professional multilingual training videos in minutes. It's fast, scalable and so simp well. Anyone can use it. You can write an email, you can make a video.

Hey there, my name is Alex and I'll be guiding you through this training session.

70% of Fortune 100 companies use Synthesia to modernize onboarding, simplify compliance and scale global training. Podcast listeners get an exclusive 30 day free trial. Go to synthesia IO greatminds. Don't miss it. This offer won't last.

Hello, I'm Kate Fitzgerald from the Learning Hack Team welcoming you to Great Minds on Learning. In this acclaimed series, Professor Donald Clark joins John Helmer to explore two and a half thousand years of learning theory, from the Greeks to the geeks. This episode was recorded live at the online Educa Conference in Berlin, a city that's always a pleasure to visit, especially in the run up to Christmas. We had a wonderful time there and the session on AI and the Global South Spanish sparked exactly the kind of lively, generous debate that we love at Great Minds on Learning. It's a vast subject, touching history, technology, language and power. But in our current moment of technological transition, what does AI mean for the Global South? Could it offer the hope of a more equitable future or the threat of even greater inequality?

Welcome everybody to this live recording of Great Minds on Learning. It's great pleasure to be back here once again, I think for my fourth year and you've been coming here since Cambrian era. I'm your host, John Helmer and I'm here with a very much alive and kicking great mind himself. Please welcome my esteemed co host Donald Clark. Our theme for this episode is AI in the Global South. It's a subject I'm embarrassed to admit I don't know a huge amount about. But you do, Donald, so I'm looking forward to learning more today. One thing that quickly became clear from my preparation for this episode, such that it was, is that the Global south is not so much a delineation of a specific geographical region as an idea. People used to talk about the third world, then they talked about the developing world, and then that too seemed to fall out of favor. So more often we hear the Global south now the Donald and I'm assuming these coinages are coterminous, they all mean about the same thing and I assume you'll put me right if I got that wrong. But tell us what in your experience has caused you to focus on this topic and how do you characterize the Global south in relation to the Global North?

Well, you're right to question the terminology there because the Global south really, you know, is a bit of an odd term. There's no exact term here and that's got not get too hung up on geography for a moment. I mean, the bottom line here is that all the money, wealth, banking, regulatory power, political power is actually no longer in the Global North. Even so, the power is really in two countries. We're talking about AI here and it's China and the U.S. i would argue that Europe is in danger of becoming part of the Global south very shortly if we don't waken up and get on with things. Nevertheless, there is a vast difference in terms of GDP of countries and by every measure on the big continents in the Global south, and that is especially Africa, with a vast and growing youthful population but very massive levels of inequality compared to the GDP in the North. And the same could be argued for some parts, but less so in South America. But you certainly wouldn't include, for example, New Zealand as part of the Global South. And then you've got India, is it or isn't it? But I think the big distinction here is just between in terms of AI is where the power, the money, the investment lies. And that's what we're going to be discussing. So let's not get too het up about this. The real reason for me getting really interested in this is historically I've been Africa quite a lot and I've been going there nearly every year since 1989, which is maybe before some people in this room were born. And I'll be there on Monday, in fact back. So I have a great interest in the place, just culturally, but also in terms of learning technologies. So I've witnessed Since the late 80s some terrible things in terms of learning technology. So in this one we're going to talk about phase one, which is what I call device fetish, which is Nicholas Negroponte, some of you may know that name. The following one is Sugata Mitra Sugata gave a keynote here in 2010 saying, don't worry all you people in the south, we'll put put a computer in a hole in the wall and all will be well. That didn't work out well because they are literally holes in the wall. So there was a phase one I witnessed that I was incredibly critical of at the time and wrote voluminously about. And then I think we have another layer in the debate here, which is contemporary, around the ethics side, John, which I'm really interested in, because I think the whole thing is massively overblown. I think there's a lot of white saviors in the north, bell ringing about people in the South. So I had a bit of an epiphany this year, and Marquano was there. I was in Tanzania, in Dar es Salaam, and myself and Callum had a workshop, a room of similar size that was jam packed full of people from all sorts of countries in Africa, and they were having none of the ethical debate. They didn't care at all. And these plenaries we get at these conferences, I think they're sort of laughing internally a little bit because they want to get on with it. They're practical, pragmat people. And I really came away from that going, my God, what are we playing at here? We have very strong opinions about the south, but are we asking those people what they really want from this AI thing? And when you do ask them, in my experience, it was a very pragmatic and positive response. And then I think there's a third level. Who are the players here? Now, this is a vast number of people, but I was really interested in people who were actually Africans. And I know I'm not John. There's the contradiction. Nevertheless, there are some key figures here. One in Google, the very senior guy who reports to the chief executive, and another really good academic, an expert on African languages at Harvard. And I really did pay attention to them because they know their stuff. So we want three layers of the onion here. Let's try and learn our lessons from the past a little bit and not make the same mistakes as we did by giving a standing ovation to Sugata Mitra, because that ain't going to work this time. And we can go into that in more detail, certainly. But I've been around a lot in those countries, from Egypt, I know really well, and then Senegal, Uganda, Rwanda, Tanzania, Namibia, South Africa. Myself, Callum, spent some time there. And my universal experience has been that people are really very, very interested and curious and positive about the technology in a way I don't feel. Sometimes when I come to these conferences, people are. And I go along the plenary, it's a technology conference, and everybody seems to hate technology. And I often say this, you know, imagine having a conference about baby clothes and everybody hated babies. It would be a bit weird, wouldn't it? But we seem to be in that never ending cycle of moral high horsing. And I don't think it's often ethics. I think there's a big difference between ethics, looking at the good and bad, and serious discussions on an ethical plane as opposed to moralizing or simply having a beef about something. Because I hear a lot of that in the ethics debate from people who don't have a degree in moral philosophy and don't know what a geontological theory might be, or any of the academic framework behind ethics, but have very strong opinions about it. So I've grown a bit weary on that stuff.

So apart from the attitudinal things that you've mentioned there, are there any areas in which the global south has an advantage over the global North?

Yeah. Well, anyone who's been there since the year 2000 really will know that what happened there was, could be summed up in one word, and that's the word mobile, because we had that huge leapfrogging. And by 2030, about 70% of Africans will be in the sort of smartphone world. But it's a very different profile from the Northern hemisphere where we have much more focus on laptops and home devices and so on. And then the second big thing that came out of Kenya was M? Pesa. And M? Pesa's sort of morphed and is very different now. But of course that was another huge example of a leapfrog. So people in the Southern hemisphere were doing financial transactions and finding jobs and 101 things on mobile phones long before anybody here was involved in that stuff. And we're only just struggling to do online banking here. So that thing took off originally in Kenya and then spread to Tanzania and across the whole of Africa. But the big problem has always been data. The interesting thing about the smartphone phenomenon is how cheap they've become and therefore ubiquitous. When you get down to a 25, 35, $50 smartphone by Chinese brand brands, of course, then you have a completely different profile in the population. And what makes AI different is when the Internet came along, that's when Mitra and all these Negroponta came along. We had to build the infrastructure and there were only cables down the side of Africa, so it was fine on the coast, nothing in the middle. Now that was a huge problem. The Internet took ages to get going in Africa and therefore we were shoving devices at them but they had hardly any connectivity, hardly any electricity, didn't have any money for maintenance, and it all fell AP. But this time is different. The AI revolution is different from the dot com bubble because we have the platform, the Internet is here, guys, we have the devices. Everybody's got a smartphone. We even have Starlink that gives no blind spot access at a reasonable price, $30 to $40 a month, which is going into SMEs and remote places in Africa because there are no blind spots. And that's about 8,700 10,000 satellites at the moment. But it will get much bigger, much quicker. So there are very different profiles. And one last thing which I've noticed over the last few years is some people call African, you know, the sort of the African savannah entrepreneurship. But in some countries like Rwanda, Kenya, South Africa, of course you have an increasing ecosystem of small companies who are quite high tech. And therefore those three things that leapfrogging the mobile profile and also a developing African AI industry is something we should celebrate and encourage, I think. And I think also the needs are very different. The reason I like AI is it's not like the technological escapism of the first 20 years of the 20th century where what did we do with ad based, which basically we're trying to tap into your brain by having ads behind Google. We've got only fans and porn, we've got ordering Deliveroo or Uber. Is that it, guys? Is that what all that technology is all about? All that sort of consumerism, that attention economy? Then along comes AI and said, the hell with that. I'm going to give you intelligence on tap and it might revolutionize your healthcare and your education system and your agricultural sector and tourism and a company conference like this and everybody's got to do their own AI. No wonder Africa's like this stuff.

That's exactly what they need.

They don't need onlyfans. And so I think it's a mistake for us in the global north to start poohing this technology when the people who are furthest away from education but have the greatest need for it and we have a chance here of solving some of those problems that we never had in the past.

Maybe, yeah, maybe we should stop obsessing about things like Italian brain damage and having a profile picture on LinkedIn. It looks like Studio Ghibli and cure cancer. Okay, so if you've got any questions in the. I've learned something already there, Donald. So that's very. Okay. If you have any questions in the audience about anything, we're saying today there will be time. Time for you to give them at the end. So long as we get our skates on. So let's dive in now and look at a couple of figures who we're going to deal with fairly briefly, starting with Nicholas Negroponte. Nicholas Negroponte is an American architect and digital technology pioneer, best known as co founder of the MIT Media Lab. Trained in architecture, he became a leading figure in early computer graphics and human computer interaction. Negroponte launched the One Laptop Per Child initiative to expand educational access in developing regions. And I remember seeing him present that at the Learning Technologies Conference in London, sister conference to this one, Plug the sponsor. So good and valuable work, Donald. Or was it?

Well, to be fair to Nicholas Negroponte, I think when he first hit the scene, he founded the media lab at MIT. That was quite a long time ago, in 1985 or so, and I read all the early stuff and his book Being Digital was read by everybody. And that was a shift from the notion of atoms to bits. A general, quite small book, in fact, that was very influential and that was quite admirable, I think. But then he got involved in the One Laptop Per Child. Anybody who's been in Africa will know that phrase, One laptop per child. Now, to cut to the quick, there was a big problem here because Nicholas had skin in the game. He actually owned a part of the company, which I don't think is good for an academic in terms of selling into Africa. And that came back to haunt him somewhat. But the real problem was that it didn't work. And there were innumerable trials in many countries in Africa. And when it came to the evaluation at the end of all this, it was a massive failure. And it was a massive failure because of what I've called device fetish. Now, this is not unusual, not only in Africa, but also in Europe and America. We have been shoving tablets into schools like crazy for decades. None of that worked either, by the way, and I'm a real tech guy, but I was dead against that. Does anybody really think that young kid should learn to code or write an essay on an iPad is just pedagogically stupid because it's basically an entertainment device. But we did that here, keep on taking the tablets, as it were. That was a failure. But in the global south, the One Laptop Per Child thing was a particular disaster because I think it was exaggerated by the MIT Media Lab in terms of its success, when the objective truth was that it was a failure. One other thing I would say, of course, you may not know this, but Negropontia has also come a cropper on the Epstein file. So I bet you're going to get that mentioned at this conference. But of course, from 2002 to 2018 19, they were taking a lot of money from Epstein, and the head of the MIT Media Lab had to resign, of course, because there was more than taking money. He had to resign for very clear reasons associated with islands. And I think I'm quite suspicious about the output from the Media Lab now. I think it's really going downhill here, but I think we have to be very careful with this notion that devices are the solution to problems in learning technology because they become consumer items that everybody has. We can perhaps focus on the software side, which is what AI is.

Firstly, I'm here to have a break from hearing about the Epstein files, and I'm a bit surprised that Donald Trump isn't here for the same reason. So, moving on. Sugartha Mitra, are we in safer grad with him? Sugarta Mitra is an Indian computer scientist and educational researcher, best known for his Hole in the Wall experiments, which demonstrated how children, particularly in underserved communities, can teach themselves to use computers with minimal supervision. Everyone got very excited about it. Donald, are we on safer ground with that initiative?

Well, again, I think people have good intentions, but it was exactly the same phenomenon here in the same period, which is this device fetish. So Sugata's first experiment was a computer shoved in a wall in Delhi in 1999. And then there were some massive claims which turned out to be exaggerated. But the real story is what happened in the rest of India and Africa. So I was deeply suspicious about the Hole in the Wall project because nothing in the psychology of learning told me, which is my subject. Nothing told me that it would work. It just seems stupid, this idea that you shove a computer in the hole in the wall outside the school and people will learn. Naturally, nothing in the psychology of learning says this is true. And so it turned out. So I did something quite unusual and myself. And also, I don't know if you saw. If you heard the. If you heard a last year pile, Aurora, she did the same in India. So I went to Africa and went to a Hole in the Wall site and Paolo went to India and we both found the same thing. We found holes in walls. There were no computers in there. And then we interviewed the teachers and the teachers hated the thing. It arrived on their lap. A truck turns up, drills a hole in your wall on the outside of your school, no teacher support. And in Fact. All the evidence we gathered also marked. Washour, who was at the University of California, he did a similar thing. All the evidence pointed towards this being a disaster for the schools and it was mostly dominated by young boys playing games. This was the testimonials we got directly from teachers. Remarkably consistent. This was the very opposite of the self published research that came out of the Highwell group. So having gone to see a hole in the wall in a township and heard this story and having had it corroborated by other people, it was worrying really. And then other things came out at the back of that, which is the solely the self organized learning theory, which I disagree as well. I'm not a social constructivist. But the idea that you just ask kids questions, give them time to discuss it and then have a little bit of reflection at the end and everybody picks up something is a bit of a disaster in my view. And that's what the Soli thing was. And that's all gone. Then there was the granny mentoring project, which is gone. If you go to the website, it's actually a Chinese, Chinese gambling site now. And then on top of that, we had the famous schools in the cloud. Anybody know what happened to the schools in the cloud? They've crashed to earth. They no longer exist. The 1 million TED Prize so these false dawns and false promises are a disaster for Africa and India. They don't work and it makes them turn, those people turn against the technology. So that phase, this is a historical phase, remember I'm talking about here. I think we've got over that. But we have to learn our lessons from that. That's a really important lesson for me, that we don't make the same mistake and start selling similar false dawns in those markets.

Okay, so just to reassure you, there will be some stuff that Donald approves of later in the talk. We're kind of doing the good, the bad and the ugly, but in reverse.

Can I say the one thing I should say? And there's the stand right outside and right hand side.

Oh yes.

Just on the good side. To balance this out, there was one initiative at the same time which I really love, and that's Moodle. So I know Martin Dugumas and we had an amazing time once in Rwanda where I got punched by a gorilla with Martin Dugumas. That's another anecdote for another time. But it wasn't his fault. It was my fault. But the interesting thing about Moodle was it wasn't selling devices, it was selling platforms that allowed people to loosely structure learning in an institution, which could be a group of schools or it could be a university. And I think Moodle, in a sense, was a signal to the future. It was open source. And Martin is an amazing guy. He fought off blackboard. Blackboard tried to take him to court because they claimed they had IP over the LMS market. And he fought that one. This is a guy who's running a sort of charitable enterprise at the time, not for profit, but he fought some great battles. And he had to be a sort of benevolent dictator, really, because I don't know if you've ever been involved in open source, but it's like herding cats. So I'm a huge admirer of Martin and the people involved in Moodle, especially in Africa.

Comes out the same kind of era as Wikipedia, I suppose, which for many of the same reasons, is kind of laudable.

I think you can describe these projects as either mosquito projects, to use an African metaphor. You got mosquito projects like Hole in the Wall that sort of get well funded, sprout up, and then die really quickly. Gone. And then you get turtle projects. I'm a turtle guy. I like turtle projects that go on for 10, 20, 30 years. And Wikipedia, Moodle, they're turtles. And I think we don't have enough turtles in this world, mainly because of the stupid funding mechanisms. Here's a bunch of cash. You spend it at the end, dead. So there's no sustainability in the model. And that's what we've seen repeatedly, decade after decade.

If you're looking to connect with the top minds in learning and talent, there's no better place than an iventive event. That's Iven Tiv. Since 2009, they've hosted nearly 300 interactive conversational events exclusively for senior leaders like Global Hitler Learning. I've been a sponsor myself and found these events invaluable. Less about sitting and listening, more about real, impactful conversations. It's an unmatched opportunity to collaborate with leading chief learning officers on the challenges they're tackling today. Join the conversation. Visit iventive.com learninghack that's iventive.com learninghack. So, turning now more to the core subject here of AI in the Global South. And Donald is launching his book about AI here at the conference, which I hope it'll go to tomorrow. Donald, AI has its critics in the Global south and a few here at the conference. Can you give us an overview of that criticism in the Global south, who the critics are, its dominant narratives, and so on. Before we move on to look at individual theorists.

Well, if you walk into a bookshop, so I live in Brighton in England and I did this just before I came to the conference. So I buy most of my books on Amazon, but I went into a bookshop and grabbed the first two or three books on AI I could find and read them. And it was very interesting. One was if you know about AI, there's quite a famous book called Empires of AI by Kate Crawford and it deals with the global south, but it is a complete and utter hatchet job of AI. And then I read another one called Code Dependent, which is a great title, terrible book. But what's interesting about Empire of AI, if you really look at the people who know about AI, they absolutely ripped it apart. So the claims in there, some of the claims are quite interesting on mineral exploitation, but she had a go specifically on the energy use and water use. And Rigk's study was in Chile. It turned out not to be true. Data centers, by the way, don't use water. It may come as a surprise to you. It's not like the book, like the book. You'd be amazed at how much water has to go in every page in doing a book. But we don't have big moral plenaries about how bad books are of course, but we do in AI. The truth of the matter is that most of it was made up. So she had exaggerated the math. It was actually a mistake on her side by three orders of magnitude, which is a big mistake in a times bestseller, by the way. And the same was true in the energy front. So when you quite often you get questions, isn't it terrible all this energy on AI donor. But when you turn around and say, well, how many kilowatts per hour do you think an average prompt takes? It's an interesting question. So we have a comparator and the answer is about 5 to 10 seconds of watching Netflix. In other words, it's not that big a deal. And the amount of energy taken to train a model of course gets ameliorated over the now billion users of the model in the case of ChatGPT, to name just one. And then co dependent was similar. She had to go at people who were doing data work in places like Nairobi and also in Uganda. And I think that's also exaggerated. When you actually speak to some of those people or speak to the people who are living there, they're actually quite pleased. For reasonably well paid jobs, you're $2.50 a few hundred dollars a month. It's all relative in terms of the wages in those places. And they're clean jobs. It gets young people because they're mostly young people on that first step in the ladder in it and helps the general ecosystem grow. So there are a lot of people in the global north sitting on 50 to €100,000 salaries. It's criticizing kids who are working on data and other things in the global south, but I think it's somewhat exaggerated. And again, the code dependent book, Madhati's book. But again I found that disappointing. Then we come to the big moral debates where we have a very famous figure, Tim McGebrew. If you're involved in ethics, you'll know her name. She's an Ethiopian. She was at Google. Sacked her or did they sack her? And she wrote a very famous book. I saw a very famous paper in 2021 that set the ethics world on fire. So anybody know the name of that paper? Stochastic Parrots. That's absolutely right. So in the dangers of Stochastic parrots and RL end too big, that was the premise of the thing was two things. First of all, should I just give up a quick. You can give it to my Rory.

Stewart bit into a quick explainer. Timnit Gebru is an Ethiopian American computer scientist and a leading voice in AI ethics. Co author of foundational work on the right risks and biases of large scale data sets. AI had a hand in this bio. She co founded the Black in AI initiative to support greater inclusion in the field. Gebru's research has shaped global understanding of algorithmic bias, environmental costs which you've mentioned, and the social consequences of AI systems. After leaving Google in 2020 she founded the Distributed AI Research Institute DAIR, an independent community rooted research organization focused on the harms of AI, especially for marginalized populations. So carry on Donald, give us a more in depth reaction to her work.

Yeah, so as my good friend mentioned at the back, it launched the phrase stochastic parrot and this was parrotied I suspect because I think most of the people who used the word hadn't read the paper. The paper was actually written, but the main author was Bender. So Gary was the second author in the paper. But the interesting thing was it was meant to be predictive. In other words, LLMs would get too big and the bigger they got, the worse they would get. It wouldn't get any benefits. Of course this is patently false. It's actually quite complicated, this argument. But that didn't happen. So the paper was completely and utterly wrong. But people kept on using the word stochastic parrot. And remember, AI doesn't parrot anything. It doesn't take anything like a traditional database seek, pluck it out and say it again. The word parrot is completely and utterly inappropriate, but it had a sort of catchy ring to it. But the other thing that people didn't read much into was the ethical arguments. And some of those were quite good actually to be on. They predicted some of the energy issues and so on, but again they were just grossly exaggerated. And one of the things that disappointed me about the paper and subsequent debate on ethics and AI is ethics is about good and bad. It's not just about chopping the legs off of AI, that's moralizing, that's just taking a position. If you're not also discussing whether it's a utilitarian theory perhaps of AI, whether you're balancing out the goods and the bads and the possible harms, or you could take a more deontological view of this and say I have a basic set of values. I think people do that all the time without realizing it. I was in Norway yesterday and there was a speaker on the ethics of AI telling me that we have to align AI with values. And my question was, whose values? Your values? Academia's values? Let's suppose we were to align AI with academia. That would be a massive and suicidal act of bias because academia is mostly left progressive. I am as well, less progressive, feeling about 90%, you can do it. The rest of the population in the democratic voting system in Europe have certainly not gone in that direction. So if you're adopting the values of a bunch of people in academia, that is massively biased in itself. And that's why the word of values is heavily weighted and loaded. And you've got to be super careful in using it in ethical debate. But to go back to Timnit Gebru, I think she did actually some quite good on data cards and the way you record the ethical issues around large language models. But she set the tone for a destructive and negative one sided form of ethical discourse, which I think has been a disappointment because we don't hear much about the upside, which I see every day in papers on healthcare and in education and in agriculture. So I think there's a lack of balance in this, which was caused by Timnit Gedru. On the other hand, I'm a real fan of Payal Arora who spoke at this conference last year. I don't know if you saw her on the plenary because she's written extensively in the subject as well, and maybe a good introduction for her would be useful term.

Paol Arora is a digital anthropologist and professor of inclusive AI culture at Utrecht University, known for her influential work on technology inequality and the everyday digital lives of people in the Global South. Her book the Next Billion Users challenged common assumptions about how low income and marginalized communities engage with online platforms, reframing them as creative, pleasure seeking users rather than passive beneficiaries. Aurora advises governments and international organizations on ethical and inclusive technology design, with a particular focus on digital access, gender equity and cultural context. Her work foregrounds the human realities behind global digital expansion, so it's well aligned with conference themes here. Donald, tell us about those human realities.

Yeah, well, Payal, remember she was my comrade in arms. She went to look at the Hole in the Wall project in India and I went to see one in Africa. So I had known her since then and she's a real breath of fresh air for me in terms of a balanced view of ethics and AI in relation to the Global south, which she writes about extensively. So she's written a lot about this. The Billion People book's very, very good. In other words, what she's saying is the people in the Global south, you might think they're interested in learning all the time in healthcare and technology. Actually it's a sort of oral culture and they like leisure and they love music. Have you ever had a flash, one of those flash sticks in Africa? You hand it over to somebody and within like an hour they've got music on it because you're out of the, I don't know, this massive divide. But there is certainly a different attitude towards technology that I've experienced and many people have experienced when they go to Africa that sort of shocks people from the the global North. She calls it Western paternalism and she thinks we're constantly trying to impose these sort of puritan northern European values on the Global south, which is a mistake. We should be asking them what they want and letting them build it from the bottom, which I agree with. So we've constantly got moral frameworks on responsible AI and we're producing oodles of documents which I doubt they read. Ordinary people with a smartphone in a township certainly don't read that sort of stuff. And she thinks that AI is a massive opportunity and is full of hope in terms of healthcare and education and these other things, but only if you build upon the culture that already exists. And I think she really cares about the big moral issues around equity especially, but also learning cultures and good social outcomes. But I think she is a real leading light on this because I think she understands that those cultures, we need to pay attention to what they want as opposed to what we want to do to them. And that's been the story so far. And actually, when that story turns toxic and you say don't use AI, believe we in the global north, we know what it's about. There are far too many risks. When it's ready, we'll give it to you. That's the sort of tone we've had, paternalistic tone we're giving out here through endless frameworks and policy statements and God knows what. Very often on AI that existed a year ago, it's almost stillborn from the press whenever it comes out. So I'm a big fan because of her previous work on the device fetish thing, which she agrees with me on, but also her work on ethics, because I think it's balanced and truly is ethics.

Can you give us a bit of a flavor of her ethics, her idea of ethics and how that differs from Timnik Gabrius?

Yeah, so she seems to me. I haven't really discussed the background philosophical schema she might have on this, but she's certainly not one of those people who have a deontological view that there is an absolute set of values for humanity and if only we align AI with those values, everything will be hunky dory. Just travel anywhere and you'll see how untrue this is. And therefore what we are doing is imposing basically a set of post Reformation European values. That's what we call values. It's not what the rest of the world calls values. And they're certainly very different in China. I think they're very different even in the US but certainly different in Africa, and that this is the fundamental mistake. So she's not a deontological person who believes in absolute values in any sense. And that's not to mean that she drifts into the swamp of relativism, because there are other theories of ethics here around utilitarianism and other theories. Of course, I get the feeling that she's much more practical and pragmatic in terms of morality and values and what we have to adhere to. And I'm also in that camp, I should add. So that's heartening, I think.

Yeah, you do hear people talking eerily about our values, our Western values. And there is always that kind of first person plural, we and us, which if you actually press people to define what they mean by that is it becomes very Difficult, I think. Yeah. So interesting.

It does.

Shall we move on to the. The next person?

Sure.

James Magnica is a Senior Vice President for research, technology and Society at Google Alphabet where he guides company wide thinking on social, economic and scientific implications of advanced technologies. Previously he chaired the McKinsey Global Institute leading major research programs on automation, AI productivity and globalization. Manyika has served on numerous international commissions including on the future of work and the digital economy and has advised governments and global institutions on technology, politics, policy. Born and raised in Zimbabwe, he brings a distinctive perspective on how digital transformation affects both advanced economies and the developing world. Donald, we have another Google alumnus here and someone whose work has taken him to the heart of government in the U.S. he was appointed by President Obama to the U.S. global Development Council. Probably isn't around anymore in the administration. Presumably this gives him a unique perspective on a. On our subject here.

Yeah, I think it does. I think what's happened has been really heartening. So James reports directly to the chief executive Google and has a really powerful position there as heading up the research labs and the sustainability projects. A really powerful figure. He was one of the times 100 in AI people. But he's Zimbabwean, he's African. So let's ask James what he wants. And James is very clear here. And the first project that Google initiated on his back and he's pushed and pushed and pushed is a thing called the Thousand Languages Project. That's not just African languages and I'm a big fan of using that, you know, this base, what we're going to build upon in the global South. I think languages are a good start because there's a big problem here. Anybody who's been to Africa often enough will know that amazingly unbelievable numbers of kids in schools are taught in English and French. It's like colonialism never went away. You go to a university, you get taught in English and French which is almost everybody's second language even in this day and age. Now There are about 7,100 languages in the whole world and about two and a half thousand of those languages are in Africa. It's a very language rich country. But 1,500 of those languages are oral only. They have no text. So how do we get AI to work if you have no data? Now here's a interesting weird thing. Colonialism historically meant that many of those languages were translated into the Bible. It's a really weird freaky thing. But the Bible turns out to be quite a good data set because it's got a very rich vocabulary and language sentence structures so they have that in some administrative files and so on. That proved to be quite useful. But the thousand language projects is a real thing for Google and they're pushing forward with African languages, low resource languages, very, very specifically. And then if you turn they have an open speech datasets thing, which means they're gathering data from mobile sources as well. Remember, going back to mobile phones are the heartbeat of Africa. So if you're going to gather data about languages that have no writing, then you can do using AI speech to text, you have the data, you can move on. And I think those open speech databases are particularly helpful for the people who are doing this in Africa. So he set up two big research centers in Accra and Ghana and also in Kenya. So they're spending money in Africa and letting Africans lead the way here, which I think is a super urban positive thing. And I'm not critical of the big tech companies here because I think they're doing what big tech companies and companies never did in the past and that's going to Africa and spending money. And Meta have done a similar thing with no language left behind. So it's a big project and that's very specifically looking at low resource languages. And in that case they have a thing called the Florist 200, which Florist 200 is a real set of databases which is looking at languages that don't have much data and trying desperately to get the data together to build models that will be efficacious in Africa. They've also got a benchmark and they're also using it in a very practical way to encourage people to use their local languages in Wikipedia, for example. And they've had some success with that. So I think all of these initiatives show that some of these companies actually have some heart, I think, and I think James in particular, this is about power. But you need somebody to affect change and if you need somebody to affect change in Google, there's nothing better than somebody who reports death to the CEO and actually goes to African. Africa is African, he's from Zimbabwe and actually does things in Africa. That's what we should be supporting and funding. So I'm a big fan of James.

First time I've heard the Bible referred to as a great data set. I got a wedding or watch a courtroom drama where somebody swears on the Bible. I said, I swear by this marvelous data set to tell the truth, nothing but the truth or anything about the truth. That's brilliant. That's my biggest learning from so far, I think. But I'm sorry, I have a facile mind. Professor John McGhany is just director of the African Language Program at Harvard University and a leading scholar of African linguistics. Again, language. His work focuses on the structure, diversity and evolution of African languages, as well as their cultural and intellectual histories. Meghani has contributed to major initiatives to document and preserve African languages and has championed their role in shaping Africa's digital futures. He's been involved in efforts to ensure linguistic diversity is represented in digital tools and resources, arguing that language is central to equitable participation in the global knowledge economy. This scholarship highlights Africa's linguistic richness as a foundation for innovation. Yeah, I'm really interested by this focus on languages, actually. And when you think of how culturally specific chatbots seem to be, you know, I've complained long and hard before about how chatgpt doesn't understand what coriander is and you have to call it cilantro before it will add it to your shopping list. But Donald, tell us about Professor Magani's work.

Well, I learned a lot from John McGanni. Specifically. He has an angle here which is why should we focus on low resource African languages? And he makes this point about the dominance of English especially, but also French in Africa, which are colonial products, of course. And his view is, and if we actually go to the next layer, and I was people's first language, Swahili, whatever local languages are, and there are many of them, then that affects education. This is an interesting inference. He thinks then you'll free up education in a way we never have in the past because people can learn in their local language or their first language. And so it would be a bit of a shock if everybody here suddenly had to go to school and learn in a second language. Imagine having how unacceptable that would be in the global North. But it's actually what Africans have to do very often. And I think this is an enlightened position here to get out of the trap of colonial wrongs, but also in terms of building from the bottom up and raising the standard of education in places in Africa because we have AI as an aid. Now that brings me to a second whole load of studies. And we don't have time to go in here, but we are getting some, some really good academic evidence now about AI being incredibly useful for teachers, especially in Africa, who are often badly paid, not turning up working in terrible conditions, but when they have the tools that allow them to make structured content, do the usual curriculum development stuff, help them in the schools, that's been particularly potent. But also learners in Subjects like mathematics and so on. I've been collecting these papers because they're far more interesting than the sort of stuff that's being done in America in the Alpha School, where kids get paid $40,000 a year to go. I'm far more. I'm not interested in those kids. They'll do fine. I am interested in the kids in Africa who perhaps are smart enough but don't get the chance to make a life for themselves because they are having to learn in a second language. Now, there's a second thing that happened here which I think is wonderful, which is a feature of AI, which is, okay, we have a language, it's an oral language. We have no written data, so you can't make a pot without clay to use another African sort of thing. But you can actually, with AI, it's quite odd in many ways. They have synthetic data. And actually the Bantu group of languages are 500 languages that are Bantu based. If you do the basic Bantu stuff, Swahili and so on, you can actually almost first. You can almost sort of first shot this stuff in the associated languages. So AI has this funny ability. There's a very interesting case of this. You can see this on YouTube when you first. This was Google. And so this is the chief executive. Google went to test the very first version of their LLM and he said, he's Indian, he put it to the test and he asked a question, Urdu. There was no Urdu data in the data sets. He knew that and it came back and answered them in Urdu. This is quite weird in many ways, but if you know a bit about language law models, you see how, as a guess, its predictability, sort of minting new words every time. This blew their minds. And actually it's come to pass now that you can actually do some quite useful adjunct language translation work without having a lot of data stored as text. And there's also, of course, if you record the stuff from mobile, which he's been involved in a lot as well, it was. With just a little bit of data, you could do wonderful things. This opens up a marvelous world in Africa, I think.

Yeah. And on that subject, is it. When the first generation of language learning models came onto the market, it was very text specific, but now we have kind of audio and video as well. Does that kind of move into an expanded range of media, make it easier to record oral languages?

That is a brilliant question because.

Thank you.

Well, it's a brilliant question because, I mean, it's a bit of cliche. Oh, yeah. Africa's an oral culture. You hear that all the time, but it sort of is. And so the mobile phone is a good example of that. You know, them leapfrogging, using phones. Now, if that's true, that you have a much more oral. I mean, myself and Mark debated in Tanzania, wasn't it? And no, it wasn't Tanzania. It was somewhere else, Senegal. And actually, when you give one of these debates, you'll see it tonight. Africans in the audience go apeshit. They're absolutely wonderful. They stand up and give Shakespearean speeches. They don't ask questions. It's great. It's absolutely, completely different cultural context. They're not fearful at all at challenging people and so on and so forth. But the wonderful thing about this in terms of the voice side is you have OpenAI. I don't know if you've used advanced voice and OpenAI. Now I'm Scottish, absolutely. A broad accent and I can ask you very specific pieces of Scottish vocabulary that only my son and wife might know and it comes back with perfect Scottish answers in a Scottish accent. This blew my mind when it first happened and that was in version one. It's super good now. It even swears at me. So I think, imagine, take that example and apply it in Africa where the voice data. The interesting thing is that 75% of data in Africa comes through mobile devices and increasingly huge amounts on voice because there's an oral culture. You can see why that would be so. And I think this is a great gift of AI that we've never had before. Translation and all sorts of things, obviously.

Do you think, taking that wider, do you think this might loosen the grip somewhat or of text over our culture? We have a return of orality. I mean, some people have talked about this in terms of media.

No, I think that's right. I think there's a great. There's a sort of moral panic. Paola talks about the moral panic in the global north, which they sort of throw southwards. And one of the big moral panics that a lot of middle class people have, I think is kids are not reading enough Harry Potter or whatever, they're not reading books. I think this is massively. Every generation thinks the kids coming up are stupid and dumb and de. Evolving. I've even heard people in the audience say evolution is going backwards. Really. I think you should go back to school and understand what evolution is. The truth of the matter is though, I think that we are seeing a shift into not an oral culture. If you look at the amount of time we spend looking at Video now has gone through the roof because you have all that stuff on social media. You have the whole TikTok thing, you have all the streaming services. I'll tell you a little story. My dad is 90 years old and he had Netflix. Didn't know how to use it, so I showed him how to watch a western called Godless. The next day he said, God, that was a long movie, wasn't it? He had watched seven episodes in a row till 3 o' clock in the morning. You know how it goes on automatically to the next one? I said, God, that's the longest movie I've ever seen in my life. I had no idea that it was coming up automatically. But people, none of us are immune from that. I'm sure all of us have sat there watching the 6, 7 episodes at some time in our life. We're watching more video, we're reading less. Certainly. I don't think that's a bad thing necessarily. I think the podcast culture, for example, is a wonderful thing and therefore I'm not too sure. I mean, I'm a bookish person and read a lot of books, but I'm not trying to force it down the throats of everybody else.

Yeah, yeah. These moral panics are what you're saying there about kind of old people always thinking that young people are dumb. That's not just anecdotal. I've interviewed academics who've done studies of this, looking at kind of magazine articles going back over the decades, and it is absolutely cyclical. Everything that people say about millennials or was saying about millennials a while ago, they're now saying it about Gen Z, they were saying almost identical things back in the 60s about teenagers. And we have these crazes where we're kind of worried about what is happening to always people in a younger age group than the speaker. And I do think, well, let's hear from them. I mean, what do they think about it?

Yeah, well, with the 30 under 30 group, speak to them. Callum's been doing some work with them. Speak to these kids and ask them about these issues. But I think it's interesting because going back to one of the people we mentioned, Payal goes on a great deal about this moral panic, inducing moral panic in the AI world and abstracting it and throwing it at other people. And what you're doing is between the lines saying, just don't use it. It's evil. And I think that's a big mistake because there are certainly some problems, but there are problems with every technology. There are problems with books and reading, let's take that as an example. We have to chop down trees. Do you know how much water it takes to make a book? But we don't have debates on that sort of stuff because we love books. AI however, we don't really like it very much, we don't really want it very much, therefore we'll have massive debates about its energy use and water consumption. We're so confused about this. I mean, I grew up in a world, you know, people who want to ban social media and devices and smartphones because people are using them too much. Or another argument, it's quite an interesting moral argument, is, oh, I'm sorry, Donald, some of the kids in the class don't have a laptop, therefore we can't use technology. Now, supposing you go back to when I was at school and say, oh, I'm sorry, Donald, we didn't have any books in our house. I came from a working class family, not a book in the house. But nobody turned around and said, I'm sorry, Donald, you don't have any books in your house, so we're not going to use books, books in school. That would have been a stupid and absurd thing to say. But when it comes to technology, we have a different set of rules around ethics. We could happily say, I'm sorry, Donald doesn't have a laptop, therefore nobody can use a laptop. Well, the solution to that problem was to get a laptop for Donald or get him a book, which is exactly what they did. So I think there are some very odd standards in the ethical debate around AI that are going to be destructive for the Global south because it's a form of, as Payal said, Western paternalism that we're abstracting onto other people.

So having kind of opened up quite a lot of issues there, maybe we need to start drawing them back together. Before we go to questions, what does all this tell us about the Global south, particularly an impact of technological change there?

Well, I think we should recognize that we're coming out of the era of technological escapism and fractious social media and all the sort of stuff that we've had for 20 years. I think we're moving into an area where we have intelligence on tap, which has massive potential. I mean, I presume everybody's used the technology and has had those aha moments. My God, you know, I'm going to the doctor, like I go to the doctor, but I consult ChatGPT always before I go there and it's got so accurate now. It's giving the treatment of the ointment that the doctor recommends to me when I go. So its utility is its strong point. We're moving from technological escapism to, I think, an era of intelligent productivity, which is a boon to people who are poor, because most of it's free at the moment. It works on a freemium model, which is great. Education isn't free in the north or in America. 50, $70,000 a year. So I think we're on the edge of something wonderful here. If we avoid the device fetish approach to it, if we avoid frightening people to death with puritanical ethics, which I think a lot of it is. And I think we see this almost. I remember coming off the plane in Johannesburg and the big science company had put up called African AI, African intelligence. And I think that's really behind everything I've said here. If we make that happen, the premise is that Africa and Africans are doing this, so give a huge amount of help at the bottom so they can build up and get proper infrastructure and ecosystems going around the growth of companies. But also in education, then we stand the chance of doing some wonderful things. If this technology is as good as we think it is, it will absolutely be revolutionary. In that context, you're using a way there.

As good as you think it is.

Yes, I think it is. I think it's pretty good.

We're talking about some very poor parts of the world here. Do you think that kind of relative, that inequality of money slooshing around in those economies, is that going to kind of inhibit progress towards AI? Do you really need the money to do it?

Well, it always does. But the interesting example of AI is the orders of magnitude in terms of the drop in price per token. This surprised everybody. So myself and Callum work on real projects, and I don't know what the orders of magnitude are, but. What would you guess at, Callum? Something, you know, it's coming from, let's say from launch to now on ChatGPT type tokens. Yeah, yeah. 100 times cheaper. Now, what other market does that happen in just a couple of years? Imagine a car becoming 100 times cheaper. Okay, so that solves a lot of the data problems. And of course, you also get the satellite things coming in, so you get the cost of landline tech stuff, which Africa Leapfrog. They're ready for that. So Starlink's waiting to go there. Everybody here hates musk. I don't. Again, I think that's an example of us projecting things. The only social media thing that wasn't put up in the plenary this morning was Twitter. Come on, grow up, guys. Grow up. It's a social media platform. It's one of the biggest in the world. Are we now attending conferences where we cancel social media platforms? It's ridiculous. So I think, I think we have a chance here because it's cheaper to solve some of those problems in Africa because actually most of it's free at the moment. All Those kids using ChatGPT in Africa are using the free versions, of course.

Okay, Minority report masks are twat. Questions. We have got time for a few questions. Fellow at the back there on the right, towards the back. Have we got the roving mic, please? So that this person picks up on the podcast and is immortalized in in our podcast.

Hi.

Thanks. Thank you very much. I agree with a lot of what you said and some of it I'm like, oh, I need to push back.

Sure.

These are kind of ill formed thoughts, but I just wanted to get your perspective on them, if that's okay.

Sure.

So maybe working backwards, starting with Musk, which put me in mind of doge, which puts me in mind of USAID and the pulling of funding. Right. And so what you have there is a model from the Global north, let's say, that is charity based. Okay. So they put no effort, as far as you understand it. And others in the room can correct me if I'm wrong, into building capacity, for example, around the development of pharmaceuticals in Sub Saharan Africa. Okay. So that now transpose that idea to LLMs. And you talked about Google Meta, et cetera. And two issues I think I have with it. One is around language. And you made that nice example of, I think the Google Thousand Languages project, which I think is more about translation. It's not about using multiple languages for training dates, data. So if you go from low resource languages to high resource languages, there's very few, you know, English, Arabic, etc. That you can build LLMs from. Right. So you have a cultural hegemony of English potentially because it's about 60% of the training data. So one of you could speak to that in relation to languages. And the second part is.

To do.

With the development of the centers themselves. Come back to the, to the health model and maybe people talk about decentralizing LLMs. So multiple small ones, which I underst. I think I'm not huge on the technology in this. Others can correct me if I'm wrong, but is a hugely technical challenge. So I guess we're overarching. What I'm saying is how do we shift Will the power structures and the innovation to the countries itself and maybe your experience or of doing any, or dealing with those kinds of issues. Thanks very much.

Thank you. Yeah, the first question is really interesting. So the Thousand languages project, to be clear what that is. So it would be a mistake to characterize it as just translation because the underlying technology is universal language model. So in other words, they're trying to build something that will work with all of those languages. This is devilishly difficult should add, but. So there's something there that's far more than just let's translate a few. And then the open language data sets that I mentioned, they already have 40 African languages, which is a fair number. I mean I know there's 1,500, but you've got to start somewhere. Nobody else did this, but they have 40 and another 50 I think are coming this year in 2026. This is a huge effort. I mean technically it's a massive effort. So there's much more to it than the translation. And let's get a few pages in Wikipedia up. I mean they really are doing what John McGann thinks, trying to get this into the educational system and the research center in Accra, for example, that's their goal, to do much more than translation. You've got to start with translation. Of course, it's the practical thing that people use. And then the second point on healthcare is interesting because I actually think that is much more greater efficacy than education because people die. You don't die because you don't get a GCSE in Latin. But they're already, if you look at the general literature. So I just finished a book on AI and productivity and I did a whole chapter on healthcare and I immersed myself in the progress it's made in that area. So it's the one sector where there's been progress from patient presents, triage, diagnosis, investigation, treatment, even well being, education, administrative savings. And I'm seeing that in my own country in the uk, you know, it's absolutely astonishing. I go to see I have a neurological problem and my consultant actually reads the summary of the meeting with me in the room. The letter which gets transcribed into an actual letter and gets sent to me. He does it on the spot now last year he had to wait two weeks while he got sent off to somebody to type it all in and so on and so forth. Massive savings are possible. Now imagine that across the continent of Africa and healthcare and every one of those because you're. I mean the interesting thing about healthcare is if AI is getting as good as your average GP in Europe because they have a 5% misdiagnosis rate. 1 in 20. So if you go and see a doctor, they will make a mistake. 1 in 20. It's been absolutely rock solid, solid for decades. Actually. AI is getting devilishly close to being better than that. So you'd be a fool going to your doctor if it gets down to 1%. Now imagine that across Africa for free where AI is better than a doctor for most illnesses. This is such a huge leap forward in terms of the problems they face in everyday life. And if you, I mean, you know, I remember when I was in Rwanda, you know, people were bleeding out with malaria all over the place because you can get blood to them, you can't refrigerate blood, you can't get to the north from the south. We forget this is a massive problem in terms of mortality rates, especially amongst children. So I think there are some great, great things happening there already and I think really interesting another one just to add this because it's a real personal favorite of mine, this topic, because the tools in AI are advancing research so quickly. You're getting real breakthroughs in things like drugs for malaria. So you'll see this autogenesis. In other words, AI gets so good, it improves itself. That's starting to happen now. So we're at the great. I mean, people forget this has already happened. Watch the documentary on Dennis Hassabis on the production of AlphaFold 200 million proteins in medicine. They solved the whole problem. It used to take a PhD four or five years on one protein, multiply that by 200 million, you're getting about a billion years of research done in an open source database available to researchers everywhere. And it's already bearing fruit in terms of tackling tropical diseases. So let's carry on with that track because there are people who need it right now. And let's not halt progress because we think the data's biased when. Well, have you looked at human data? So I'm quite passionate about the healthcare thing. I think that's the one thing I would spend real money on.

Can we squeeze another one in? Just one more. Anybody else? Yes, over here, down in the middle. Who wait mere seconds for the roving mic to reach you.

Yeah, thanks very much, Mark west from UNESCO. Donald, always a pleasure to hear their perspectives and, and, you know, these, these opinions. I just wanted to get a few things. One, we're at an education conference. You've spoken a lot about the uses of AI for healthcare. I was just wondering if you could also elaborate what you think some of these, you know, major breakthroughs are as regards education. And with a caveat maybe, that a lot of people speak about education, of course, of, you know, information transfer, knowledge transfer and other things, but also, you know, sort of putting on my UNESCO hat that education is about much more than that. It's about acculturation, it's about learning to live together. It's about having conversations with people who often will disagree with you, interact with, with people from different backgrounds. And the sort of obsequious nature of a lot of the interactions with AI are, according to many educators, are almost, you know, they're not necessarily educative in the way that they will sort of always agree or sort of send you down the track that you want to be. And good education, I mean, that's why I love coming to hear you speak, is about coming, encountering beliefs that, you know, you may not necessarily share, or we, you know, like to come to them. So on those education issues, be great to hear your kind of elaboration. And then just one other thing I heard a lot of, you know, we're the. The title of this talk is about AI for the Global south. And you've talked a lot about sort of Western or Northern paternalism as regards AI. I was wondering, is that accurate? Like, that's not how I see it. I see it as a sort of Silicon Valley paternal journalism, that it's not Western, it's not Northern. I mean, the academic community that you're sort of, you know, speaking about, they're not in. They aren't holding the reins with AI. As a matter of fact, that's just a handful of private corporations run by a very small number of people. And like, I can understand that people are a little, little bit skeptical about that. There's no public discourse. And I think that's why people trust public education, is they have a say in public education. They are able to, you know, they. They see themselves represented. There's democratic processes about this. But we're living in a world where people are calling for legislation to prevent legislation about AI. And I think we can all agree that that means that we're just putting a very small group of people, you know, in charge of these technologies. So when you're saying it's this Northern or Western term, paternalism, I'm not sure that's accurate. I think it's often a corporate paternalism. And you're seeing a lot of the people, you know, are now fancy, you know, venture capitalists who now fancy themselves Sort of philosophers of the world and the Peter Thiels of the world and on and on and sort of writing these manifestos and other things that have no. There's no public discourse, there's no agreement. And I finally just want to stop to say that at UNESCO we represent the entire world, the whole world is represented at UNESCO. And UNESCO has put forward the first ever international recommendation on ethics for AI. That's almost 200 different member states that have agreed on a kind of ethical framework for AI. So that is a chance at least to have an international discourse about where we want to go with this technology. I think it's an important and excellent blueprint. But there are venues where we can have these discussions. The UN and UNESCO are one of them.

Yeah, so the two questions there, the first one in education, the second one about the power of Silicon Valley. Valley. So right at the beginning of the talk I said that Europe might be part of the global south, acknowledging this point, in fact. But what should our reaction to this be? So, you know, the printing press got invented here in Germany, but was eventually adopted by the world, as it were. And it tends to be the way with these big paradigm shifts in technology that somebody does it somewhere. Now in the three years that large language models and other models have come about, we had the initial flourish in the US by the big tech companies, but also DeepMind in London, interestingly, who are a real power player in all the of this. But we also have China, but we don't have Europe. And this really saddens me because I've come to this conference for many years, made this point time and time again. If we just diss it all the time and don't do anything about it, we've only ourselves to blame here. If all we do is get involved in the discourse of grievance and complaint, we're in real trouble. And I think I wholly agree with your point about get this out and discuss it. Because there are big problems around, let's say antitrust laws, for example, in the US with regard to these companies. But what do you do about China now? Interestingly, China released loads of open source models, which is very interesting in itself. We need a little open debate around that. But I think with technology it just happens to start somewhere, then somebody else does it and then it goes global, like the Internet. The Internet comes out of the us you have this thing in network that does get bombed by nuclear wars. Then an English guy, Tim Berners Lee, comes along and invents the browser, giving everybody access to it. Then you have this applications layer and we've got a massive increase in GDP globally on the back of it all. And that's similar in a funny sort of way. I think it's a parallel here. You've actually got more tech companies, the platform's already there now you've got more tech companies in competition doing this. Then you have an inference layer, then you have the applications layer that's about to hit us, which will be massive just like web companies. And that's why I think it gets diluted if you look at, I've forgotten his name now, the Nobel prize winning economist. Anyway, but basically, if you take the pie that Google produces in the AI, the benefits of course are massive for Google in terms of the cash they generate and profits. But there is also this massive human benefit for everybody else who's using all those services. So there's a technical name in economics for it, which is that sort of the consumer index plus thing you get now. We all benefit from premium models, but multiplier.

Sorry, they talk about the multiplier.

No, I can't remember. God, I had a whole page of my book on this and I forgot the name of the damn thing. But I think the important thing about the freemium model is so many humanity benefits almost automatically, despite the fact that these companies get very cash rich and at the core. And of course taxation. I've always said this, the problem here is taxation. And as long as you allow Ireland to have half a dozen companies dragging out taxis from the rest of Europe, sorry for my Irish friend at the back, then the EU are in trouble on education. Let me finish on that one because this is really important to me. I spent my whole of my adult life in learning and technology and having kept an eye on, on the studies that are there, so what empirically can we say about this? After only three years of LLMs, I think there are three or four studies that have been absolutely astounding in terms of the sample size, proper academic work that are showing fantastic benefits, especially for teachers. So I think you're sort of right there. Produce all this content if you want, but I think there are two things happening at the same time. AI is bottom up. Andrew Karpathy said this. It's the first technology in the history of our species that's emerged from the bottom. You know, a million people were using it in a few days and then suddenly it was 10 million up to a billion. And that's what's happening in Africa. And everybody's just using it because it's so damn useful that will have a way through education. I use it all the time to learn things or to get solved problems for myself, as many people in this room do. But I think my focus would be on teachers at the moment because that's a pain point, as you know. Well, you know, it's a real pain point in poor countries and giving teachers help to do what good schools do. I'm happy with that. The second thing would be agency amongst learners themselves. I think that will happen anyway because of this bottom up rush. It's so useful for kids. I was in Norway yesterday and all the parents were telling me all the kids actually learn English before they go to school because they're so immersed in the English culture and you church and blah blah, blah, blah. And language learning is an interesting thing because it's an area of catastrophic failure. Hundreds of thousands of kids are in classrooms today and every one of you Europeans will know if they come to your country, they wouldn't be able to order a cup of coffee after five, six, seven years. Have a look. But you've got kids learning through their own personal agency now. And in China they're using these amazing apps that are wholly immersive with chatbots, you know, human being, avatar type stuff. So I think on language learning there's a big rush towards efficacy here and let's go with that. You know, we've spent a long life wishing this would happen faster and suddenly a technology comes along that says, yeah, you know something, this might just be the one. But I think if we kill it, you're right. You were right when you said it's an American power play, but also now a Chinese power play. Almost an unequal plate there. Nevertheless, I don't think it matters. Just like the Internet, there's loads of application stuff that needs to be built that will be of benefit to all.

Going to have to draw it to a close that we're a mere 10 minutes over time, which for must be a record done. I don't usually we go an hour over to rough with great magical questions.

Good questions.

Yeah. I'd really like to thank you not only for coming, but I'd really like to thank you for the quality and thoughtfulness of your questions, which I think is really unusual and thanks a lot for that. The perspective for UNESCO was very good to have as well. If you want to hear more of Donald, he's presenting tomorrow at 12:15. And don't forget to buy the books. I'm sure it's just as magnificent as all the other stuff. Thank you very much for coming today and I'll disprove what Donald just said there and say thanks. Sure.

Thank you very much.

Great Minds on Learning comes from the Learning Hack team and is produced by John Helmer. The podcast is based on a series of blog posts written by Donald Clark and we thank Donald for his kind collaboration with in this project. We'd also like to extend a heartfelt thanks to our sponsors.

Sa.