Why Generative AI Feels Unsettling

Thank you very much for having me. Uh great to to meet you all. Um yeah, so I I guess I I really enjoyed Fiona's talk earlier and she was positioning her her sort of worldview very much from the center of pedagogy and the sort of process of learning in a classroom. I guess I guess my position is probably in the middle of the tech bubble. So my background is a techie. I've spent most of my career building ed tech tools. So I'm I'm sort of sitting in the middle of the tech bubble. Um but with a remarkable amount of overlap. Um just to sort of introduce myself a little bit. I collected a fairly random collection of of sort of postcards of my past just to give you a bit of a flavor who I am. I'm from South Africa. I'm a zealot for how education can change the world and how it really means so much. Um I'm a technologist so I build tools and I'm particularly fascinated in how these tools can enable a kind of learning that doesn't happen at the moment. So what is the thing that it could do that's a valid type of learning that actually isn't happening in the non- tech world. That can be about reach impact scale just different aspects of that. Um, I've I've spent a while. This was a Guinness World Record that I built with my team, which was an augmented reality artwork in California. I was working in California for a while. It was the biggest augmented reality artwork that you could scan your phone up and down. It was a kind of history of a particular of of a particular engineering company. Um, at one stage I was interviewed a lot on the BBC particularly about phone use in schools, which is a big hot topic at the moment. This was this this was probably 8 10 years ago when I thought it was a really good idea to have phones in schools. Now I'm not sure that's I'm not sure that's still still true. Um I I've had the privilege of having millions of users, learners, language learners using tools that I've built and apps that my teams have built at Babel. We have two and a half million people. So it's just a sort of crazy scale and trying to think what can I do that adds some sort of value when there's no way that you ever have individual relationships with all the users. So this is part of part of my uh my background. Um yes that I did write a book a year ago about AI and it's kind of a philosophical approach on on how you use how you can think about AI as part of as part of learning. We took the slightly odd decision to sell it and give it away for free at the same time. So, it's a free download and and it can be bought. Um, yeah. So, I'd like to um I'm I'm borrowing some of the pictures from the book, but actually the the the talk itself is more structured around what we're doing here. And I'd like to kind of break it into four chapters. I'd like to talk about this sort of crazy AI mechanisms and mechanics that exist at the moment. how that makes us think about our own intelligence as humans and what intelligence really means when we start talking about artificial intelligence, how I think that that's threatening and evolving our educational institutions and then some practical anchoring about what we could do with learning and what what these tools can do that are really helping learning. I hope this is helpful. This is the sort of chapters that I wanted to talk you through. I guess to start with this is a slightly weird time. I started off with a picture of times of flux and water slushing and it feels like that. It feels like that politically. It feels like that environmentally, the weather, it feels like that sort of techwise. It is a slightly strange time and it is slightly un unsettling. And the the power that's within some of the AI tools really is extraordinary. It's the kind of things that was science fiction 15, 20 years ago and it's kind of brought to life. the it's quite yeah it is extraordinary and I think nobody really fully understands what that application is like. Um you know we've we've created these engines that don't serve us. They outperform us. They they free us and potentially indenture us. They feed us and they feed off us. They may take your job or they make might make you better at it. And this is this kind of uncomfortable reality. Right. So I I I do inhabit this tech world. I'm enthusiastic about some of the power. I'm also a bit terrified. I also don't know quite how it's going to be applied. Similar to Fiona, we don't always have all the answers. Um, so I'm sure you've heard this countless times before. Um, and I'm sure you've also heard sort of GPT is a term that's used describing the generative AI. So I'm mostly talking about generative AI. The the world of AI is a lot bigger than that. But generative AI is what's in all the news and it's what's in open AI and it's what you stumble over all the time. And I just want you to pause a little bit to reflect on GPT. So you hear this again used all the time. What what is it actually telling us about what this type of generative AI is doing? So it's generative, which means it makes stuff up. It creates stuff. It originates things, but it it's sort of creating. It's making up. It's pre-trained. So it we didn't teach it. We just gave it lots of stuff to look at, to read, to listen to, and it's using those for what it makes. So, it it has to ingest a vast amount of data. And it's a transformer. And this is the bit that nobody fully understands. Even the people who invent the transformer, somehow based on everything it's everything it's read and the the the feedback you've given it, it decides the context and what you really meant and what this moment is. and it goes through this sort of very clever algorithmic transformation process. But but truly almost nobody really understands what's happening at that moment because it's sort of learned itself. And it's useful just to I I find that really useful to regground myself when I'm thinking about all the magic of AI that that's actually what's happening. It's making stuff up based on vast amounts of our own materials that we've given to it in some sort of mapping that we don't always fully understand. Which is why you get all these weird things that happen. You get if you say thank you, it writes slightly different text than if you didn't say thank you. Or if you say please, like weird, it doesn't really make sense, right? But but there's all sorts of weird things that happen because of this magic. So I guess the one thing as a techie person that's uh amazing is that if OpenAI releases some startling new release of GPT and power and transformation actually app developers all across the world suddenly have all of that power at their hands. We we all do but but but also individual apps do. So there's kind of a a rising tide that comes up and so uh Gemini releases something, Anthropic releases something, they're competing with each other. The sort of tide is rising up and it's available to everybody. So I I could make a quite nice chatbot, take me a week, maybe it can be quite contextualized. The chatbot could be based on all of the work that we've that we've collectively done. Would be pretty good. Um two months later, a new a new openi release comes out. I can map onto that and suddenly it's got all of this new power. Suddenly it can do speech and it can do images this based on our text. So it's sort of like a new global operating system which is quite like fascinating in terms of how rapidly it evolves. You've you've probably heard of most of these but I'll bounce through some of them quickly because these are quite size seismic moments. I mean chat is popping up all the time. Open AAI chatbot talk to it. That's the main interface. the way we we sort of interact with a lot of the GPTs. Um and then Agentic was a huge buzz about six months ago maybe. And this is where individual chat bots become expert in particular things and you can chain them together. So there's one that's good at doing your online shopping and there's one that knows about your holiday choices and there's one that's been trained on all the academic papers that you're reading for your thesis and and and these these can talk to each other. So you could ask a problem and it can chain chain things together. So it's it really like becomes like a little little army of of robots that each is good at one thing. And then the reasoning or thinking uh AI is fascinating. So this is this has mostly been out in the last 3 months, three or four months. So if you if you're using GPT5 and you go in and ask it a question, it's kind of got that baked in. But what we understood was if you let that if you let the AI think a bit longer before answering you, it gives you a much better answer. So they just slowed it down a little bit and they said, "All right, ask the question and take longer to think about it." And so if you're using one of the um if you're using paid versions of of any of the main Open AIs at the moment, you get that as an option. Sometimes it's called reflexive or thinking, and it literally pauses and waits for 30 seconds, 40 seconds. And the more the more more recent ones are clever enough to understand this feels like a complicated question or a deeper question. I'm going to allow more time or oh this is a trivial one. I'll just give it to one of the agents that does a search and finds things. So it's a kind of a meta level of figuring out what to do next. Multimodal. Um that's fascinating. So I I guess the the funnest one of the moment is nano banana. If you have if you ever heard of that, it's it's um Gemini's latest version 2.5 Pro. 2.5 Pro Flash is the proper word, but it's the image generating. And it's better than any previous ones at you writing something and it creates a picture. But you can give it a picture and write something and it creates a picture. So you upload a picture of you and your family and you say, "Please, can you make my sister-in-law look a bit uglier?" and and it's it's smart enough to understand the picture and make changes in the picture. Um but so this is multimodal where you use words and you get back audio or you use audio and you get back an image and that that transform I mentioned before doesn't actually mind. It doesn't care. It can somehow map from all of these words into a picture again in ways that we don't fully understand. drifting across this. These are these are mostly the big the big players in the open AI in the AI space. There are others. They're not the only ones. Um and sort of copilot is here, but actually copilot uses uses from draws from the others. Deepseek is particular. This end are quite special because they're doing more open source and they they're giving a you can see what's under the under the the lid of how they work whereas these ones tend to be all more private. Um, increasingly people are figuring out that sometimes small is better. Use much less power, much faster, can run on your phone. That's they shouldn't just be bigger, bigger, bigger. That was an arms race until again about 6 months ago. And now people are realizing sometimes smaller is better. Sometimes small agents that do very specific things is better than one mega mega thing. So this is the rising tide. So, I just really wanted to help you see that these are all actually there's a sort of crazy arms race, but really it's boosting it for everybody else who's building digital tools on top. Um, and and if you look at some I just wanted to pick out a couple of the threads of these stories. So, so we're here thinking about language teaching, language assessment. Um and so you know I guess traditional big exam IELTS type type thing. Um has no real AI in it except for the last 5 years or so the IELTS team have been trying to understand can we use AI to automark things? Could the speaking test have AI in it? What what might that look like? So there's a lot of thinking going behind the scenes about how to map a a non non technology type exam into nuanced ways of working. Duolingo obviously shocked the shocked the test the the sort of assessment scene of coming in with something that really wasn't that good but was very fast and very cheap and hit the spot that people needed and that shocked the whole system a little bit. Oh, maybe maybe we need to be faster and quicker and maybe maybe being good isn't enough to actually be useful. Um, Jeep is a project that the British Council's doing. The actual product is called something different. And I'm sorry, I've forgotten the name, but that's trying to unpack assessment to say, "All right, what if it was a series of modules that were that were skills that you demonstrated and you actually demonstrate the skill?" You have a chat with a virtual shopkeeper and you debate something and you and and that whole chat is analyzed and turned into building blocks of an assessment which is pretty cool. It's a sort of fairly they're doing quite a lot of research on it at the moment. So you'll find more information on the research side but it is wrapped in some of their products. But so this is a kind of journey up this rising tide. Um and the tide being pulled up I guess mostly by GPT5 Gemini the Google one claude anthropic one. They they're they're the biggest biggest ones of pulling the tide up. Grammarly have been around for a while, but they're an example of AI just just in your browser helping you write better, understanding text. Then there was a awesome piece of work that I think Ben is somewhere here. Ben and I would were involved if we were at Cambridge. Um so this if you haven't tried it, do try it. It's it's free. It's a sort of writing support. So you you're writing paragraphs and it's analyzing your text and roughly assessing your level and making suggestions of how you could change it. But this took years. We worked with super smart academics who were state-of-the-art neural linguistic. They they had done years and years of research. We worked with them for years to build this product on top. And probably today I could build you something almost similar in a week just using these tools up on the top. So it really is this tide is is is rising up. Um the speaking and voice is an amazing area as well, right? So there's a there is also a speak and improve um coming out of Cambridge research which is again using your voice and pract practicing for sort of English English exam type practicing and leveling again years and years of research and practice. But today I can go into Duolingo and Babel and there's a pretty good chat interface that I can have a conversation with and it gives me pretty good feedback on how well I'm speaking and what my levels are and gives you suggestion of what I could do next. And there's just come from GPT5. It's it's taking Open AI and it's wrapped it's wrapped a bunch of stuff around it that makes it useful for students and sets the levels, but it's mostly coming from that. And then there's a couple of other language learning apps which are really leaning into all AI, all about the AI, but again they're also using some of these tools on the top. So I guess my main message here is that there really is a kind of tide rising and if you're using one of these tools that's been around for a few years, probably you should be looking at one of the ones that's further up. If you're using one tool, you should probably be comfortable to hop to another one. If you're not using any of these ones at the top, it's worth your while to have somebody in your organization trying to understand these because you can probably do almost as much with these for 15 pounds a month than you can with the ones down there. So yeah, I guess that was my my message. Um this this sort of rising tide and it's really helpful to learn how to ride on that tide. Uh this picture is from the seven boar between Wales and England. It's this insane wave that happens a couple of times a year where the tide rushes up the river. I have surfed it a few times. Um, uh, it's it's a bit mad, but I just thought that would be a fun metaphor because there really is this sort of riding tide and what's important is that we keep moving forward and trying the different tools out and understand the evolution rather than that we stick on one approach and one tool. So that was about the kind of machine the machines right. So this one is about intelligence. So we talk about artificial intelligence, intelligence, are the computers intelligence as humans? What what does this intelligence mean? And kind of why are we so freaked out about the idea that computers might be cleverer than we are? Um so we we've built tools that in some in quite a few things perform better than we do and and that's actually quite hard to weave that into our regular life. Um why why is that why is that such a problem? I I think it's because we as humans value words very highly. If I if I if I want to understand the how how how meaningful the document is I weight it by how easy it is to read, how clear the argument is, how well it's written. Words matter. When I listen to you're listening to me speaking now, you're probably partly judging me by what you see there and you're partly judging me by by what I look like. Do I look like a techy guy? Am I speaking clearly? And and these things matter to us and we we've over over generations we've built up the sort of filter to understand is this real, is it a scam, is this person trustworthy? And and certainly in the written form, we really depend on how well it's written, which journal, which journal has this publication in it, which authors are behind it. Um, and the problem is that that's what GPTs are very good at, the generative stuff. They can create very authentic sounding, legitimate pieces of text, audio, videos, and we're not very good at filtering that out. It sort of freaks us out a bit. So, you know, we we we we think um we value knowledge, educated, trustworthy, it's coming from peer-reviewed sources. That's also why when our students use it in a classroom, we're not quite sure how to deal with that. Is it okay to use it for homework or is that kind of cheating? Is it okay to use it in exam or is that is that plagiarism or actually is that fine because they probably will do that in their career in the futures? What about the bias that comes in that's is the systemic bias that's inside some of these documents? Is that the students fault? Is it the system's fault? So, so and and it's hard for us to really navigate spam emails, fishing, fake news, radicalization. A lot of the stuff is happening, right? But I think one of the reasons we're so freaked out by it is that we're not very good at filtering it out. We we we sort of it's catching us by surprise. And it's partly our own fault. So this is just a the latest release from Anthropic when they were telling telling the world about Opus 4.1 which is the which the latest and what they do is they tell you how it scores against a whole lot of different mostly human exams. Um and it compares it with previous ones. Um but but these are mostly human exams. And the thing is we humans got a bit lazy. We started measuring intelligence by exams that had scorecards that you could grade. And we've produced thousands of these. Students have produced thousands of exams with scores. We've produced hundreds of exams that we've sort of forced students to sit through. That's what AIs love. They read the exam, they read the answers, they learn from that. That's kind of that that's how they learn. So we've we we narrowed intelligence down to these things that could be examined on paper. and then be surprised that the AI could read all those papers and learn from it and actually do it better than us. So I guess it's partly our own fault for thinking that that's how you measure intelligence and it's forces us to think a little bit more openly about what what is intelligence really. But it's it's also just a a useful reality check. So I I yeah I guess it leaves you thinking about intelligence in slightly more open ways. I have you come across this fellow before? His name's Ethan Mollock? Put your hand up if you have. Okay, cool. So, I would if I have to recommend one, if you're just a little bit interested in AI and you have to recommend one thing to do or follow after this talk, look him up on LinkedIn. He writes a blog called one useful thing. He has a book called Co- Intelligence, which is good. But actually, if you read the blog, it's the same it's the same stuff. He's he's a um professor in in Wharton, I think. Um, but he he's he comes from, I think, business or or psychology or something. He's not coming from AI, but he's very deeply reflects on AI and how to use it. He's continually testing the latest ones, and he's he's written a series of really thoughtful essays about how you might weave it into your life. And the one the one thing um that it's a really sticky idea is that it's very AI is very intelligent, but it's intelligent on a jagged frontier. So if you imagine a a wild landscape of of hills and forests and rivers and there there's a wall around that around that landscape and and that inside bit is what AI can do very well and the outside bit it really can't do but the wall's not even it's not not obvious in one section the wall is way over there and the other one it's quite close in and was his idea is that it's a jagged frontier of what it can do and what humans can do quite capably and it's not obvious to understand where the wall is. It's not a clean easy map, but each new release of AI that that that frontier changes a bit. So, so it's a really nice metaphor and I think that's how I that's how I approach it. I think there is this landscape of human intelligence and there's a growing section which AI can do better than us, but not all of it. And it's as as professionals, it's really helpful to start to build your own idea of what that edge looks like. His other his other sticky idea is it is just weird. AI is weird. Literally, you say please and thank you and it gives you better answers than if you don't say please and thank you. Why is that? You're talking to a computer, right? There's there's countless examples of it just doing slightly odd things. You can ask it the same question four times in a row and get different answers each time. Again, it's just slightly odd. And you you need to appreciate that. You can't take every interaction with it totally seriously. In the in the early stages, he referred to it as your drunk intern. So he he said it's this helpful machine, this helpful robot that you can ask for advice and it's very quick and very smart and sometimes totally wrong. And that's again you need to kind of understand that. If you understand that then it's easier to work with it. Um he's just recently reframed that a little bit and he thinks right now it's getting clever and clever. So it's it's actually harder to even understand when it's drunk and when it's wrong. So he now is describing it as a wizard. It's the sort of magician that can produce marvelous tricks. But the question to ask ourselves is are you next to the magician? Are you kind of his are you are you his peer? Are you maybe his assistant? So actually the magician is doing things or the scariest are you sitting in the audience and just watching in which case you lose your critical review of what's going on. And I think that is a a serious worry that as it's getting cleverer it's harder to be critical of what's going on and more easy just to assume. Um so um his other sticky idea is that if you haven't spent 10 hours actively using AI to solve meaningful problems, you won't have any hope of understanding the edge of that jagged frontier. If you really want to get it, I I can't help you do that. All I can advise you is to try spend 10 hours. And I will I will advise that in a moment. I'll remind you in a sec. Um, so, so I guess my my main challenge in this section is that to really understand this funny world that we've that that we're in and we can't get out of, you need to become a bit of an AI explorer. If you haven't spent those 10 hours, you need to spend those 10 hours. Start small. Just 10 hours. Um, do do but do real meaningful things. Ask for feedback on a curriculum. Upload a complex email trail and ask for a summary of what's going on in that email trail. Create a lesson plan. Um, create a whole bunch of learning resources for a for a group of students, but target them differently for their different special special sort of enthusiasms. Um, analyze data, put a spreadsheet in with a whole bunch of learner data and ask it to look for outliers or anything that you worry about. Just try, but try with meaningful tasks that you do you do in your day job day-to-day. It doesn't really matter which tool you start with. Start with Gemini, Claude, Chat GPT. It really doesn't matter. If if you can, do pay for it because if you pay, you get a slightly better version. Um, but just try and and and stick with it for for at least 10 hours. And I at the end of it, I hope you have a slightly better read of what it's good at, what it's bad at, ways ways it could be useful. So that's about you building your own kind of explorer mindsets. I'd love a show of hands who feels they're already a bit of an AI explorer. >> Nice. Very cool. So that's half of you. So the other half that's your mission for the so this this section um I sort of wanted to think a bit about the educational institutions and it feels a bit like there's a a sort of storm waves bashing at the foundations of of institutions particularly where where um institutions for various reasons can get quite tangled up into tradition and what how they used to do things four or being compared with other institutions on grade scores and grades that were scored in the way that they've been scored for 50 years and it's sort of quite difficult to move beyond that. Um and and so that's the kind of institutional view. I I work in the tech scene. So I I work in with startups. I'm advising startups. I work in startups myself. That's the absolute opposite. It's about speed above all else. It's this crazy race, this rising tide. Everybody's trying to figure out what the new tools are and how they can be faster and how they can reach more people and there's just such a more energetic exciting pace of velocity. It's not always right. And uh to to some of Fiona's points earlier, some of the tech is entirely misguided and it's not really solving a real learning problem, but there's a kind of movement like the Dualingo exam is a good example of that. There's this rapid movement and adoption kind of regardless of whether the institutions are on top of that or not. So it sets up a really interesting tension between those two those two the sort of speed of evolution happening with the tools and the very cautious gentle move sort of approach that's that's within institutions and I I I guess that's the the sort of the dichotomy that we're all stuck in a little bit. So random Google search AI in education is full of this sort of stuff. Some kind of robot magically doing things or some kind of magic information flying around and it's it's a bit cheesy and it's not really like that, right? But it's it's somehow this very seductive idea that there's that there's a robot doing everything for us and that information sort of magically flying. I'm always puzzled by the fact you still have rows of desks in rows and doesn't actually evolve too much. Um, but it's it's easy it's easy with that to just think that there's this mad rush to te to tech adopt and I'm actually I'm totally with some of the way Fiona was positioning the use of tech and trying to understand the human part of that and what's the value of the human piece. Um, but you don't need to look very far to find exactly that discussion going on. There's a whole group of sort of tech ccentric organizations Microsoft and then the gov UK AI for education teach AI Khan Academy who releasing really thoughtful information about actually how what are the ways we should be thinking about tech and teaching and how do we connect the dots there. This particular visual comes from teachai who I find quite good. It's a consortium of some of the tech companies and some educators and some uh educational charities. And I mostly I I kind of like that they're framing it as a yin and a yan. The benefits of AI, some of the things that are really useful, differentiated content, assessment design, personalized assistance, creativity, um efficiency, and then some of the risks, diminished teacher and learner agency and accountability, over reliance, loss of critical thinking. These are these are real. They're real issues as well. And so I like that it's slightly more holistically framed. And if you haven't come across Teach AI before, I would I would give them a look. Um I also really like um this this is a it's in a bunch of German schools have put this together and there's not a huge amount of English variants of this. So so do take a picture of it, but you might struggle to find it. It's there's no magic to this particular competency model, but they were trying to think about on a more holistic level what what are the levels of using tech in teaching and what are the different levels that I would need as a teacher to understand. So that I I I like the fact that they think about understanding then applying then reflecting and then co-creating. I I like the the fact they sort of think up three different levels. Um which some of those levels also apply a little bit to to Fiona's bar charts of the different sort of behavioralism at the top um and the social constructivism and just that that sort of stack some of them map a little bit a little bit to that as well. So I I found this really helpful and I guess my main point was there is a lot of interesting dialogue going on about healthy holistic connections of teaching and learning and AI despite the the crazed noise and the bonfire of kind of of hyper accompanying it. So, so my thought here is if you're looking at if you're responsible for a whole school or for for a large group of teachers, one of the best things you can do is make it easier for those teachers to become explorers, too. It's not just about your personal exploration. It's it's sort of helping enable that kind of curious mindset and bravery and risktaking that it does take to try out some new tools with your students and try to build a new narrative with your students. So this is now the last section um and I was uh this is about the actual learning bit um and how how Gen AI does shift how we think about knowledge and how we think about our role as students and as teachers and I just wanted to sort of share a few ideas. I guess the one um that there is a real difference. I'm not sure if you find this. I I find this regularly that there's a such a hype about AI in education, but actually it's it's often not very new nuanced. It's just this very broad generalist idea of AI will be teaching or um teachers won't be needed or some sort of huge generalization. And I find it quite helpful to unpick some of the threads of that. So the thread at the top is the sort of hype that there's going to be this robot or general intelligence taking over. The reality is probably somewhere slapbang in the middle. There's some very cool things that it can do and very helpful things it can do. But actually the human plays quite a critical role. It's just a shifting role. Um there's a sort of a a line about personalization and and logistics. And I think I think there's some quite amazing things that you can do with some of these tooling just to make the organization parts of your classroom better. Classroom, you know, the logistical management things. Actually, it's great at that. Um, we we I've had massive success working with customer support teams and managing to process 10 times as many requests with the same number of people just by being a bit adding some smart AI. Um, I think there's a lot of hype about sort of personalized importance of very very hyperpersonalized learning. I'm not I think sometimes that's overblown a bit. It doesn't need to be quite as personalized as the AI enthusiasts will tell you, but it is quite powerful at generating things. Um, and then this bottom one and I I I I advise quite a lot of AI startups and in education and a lot of them start with this really weird idea that their job is to productionize the passing on of information. I teach you this thing and then when you've answered that I'll teach you the next one. I'll teach you the next one. but you want different things. So I'll get it's the sort of flow of nuggets of information that need to be shoveled down a sort of pipe delivery of knowledge and it's a it's a weird idea and no teacher I've heard has ever thought that's a good idea but a lot of the AI startups are sort of starting on that end. There's a lot of hype about that. Actually I don't think I don't think any of those will really last too far but what it is quite cool at is thinking collaboratively with teachers and students about new skills that you need for your future. what that might look like, how you build, how you how you build with AI rather than without. So, I guess I my main point here is that there's lots of dimensions to thinking about AI and education and don't don't get bombarded into thinking it's all one thing. Um, this came out a week ago. So, so OpenAI looked at all the chats that had happened uh sampled a million chats over the last year and just to say what what are people talking about? What what are they asking AI about? What what is it? And it's kind of fascinating because almost 30% is asking for help with stuff. The green bar over there almost 10% is literally tutoring or teaching. It's asking AI to be a teach me something which is kind. So that's general public people going to chat GPT and saying how do I dot dot dot. Um the sort of red one is a bit more like search things but this purple one is quite good. The whole purple one is again it's about 28% is about writing um translation, writing fiction, personal writing, communication, critiquing a text, helping structure an argument or or building a summary. That's kind of interesting, right? >> Because these are your students busy doing that or maybe >> or maybe it's the teachers doing that, right? Which is good. But that that's just the reality. It's not no no hype about it. is just what it's used for right now. And so that tells me if you're doing any of those things in your classrooms, teaching, practical guidance, writing, practical guidance, kind of makes sense to to figure out how to weave this in even though it's bit awkward. So we're we're here all enthusiastic about teaching English. um what that looks like and I I I just thought it might be nice to think about what are some of the uses of AI in English teaching and learning that you can think of. Can you have any suggestions of of the type of area that AI can do good things? Just >> reduce uh reduce teacher time in preparation and free up their time potentially for >> freeing time on students. >> Nice. Anything else? >> Engaging activities for the young generation. >> Engaging activities. Yeah. Nice. >> Material design. >> Sorry. >> Material design. >> Yeah. Material design. Nice. >> Yes. >> The language that they're using is right is correct. >> Okay. So, correcting language. >> Nice. >> Yeah. Personalized feedback and personalized curriculum. speaking time uh outside of the classroom to practice >> practicing speaking very nice yeah I was waiting for somebody to say that that's the one of the bits that infused me the most when I was at Babel because all of a sudden previously it was a sort of it was a screenbased and you couldn't talk and suddenly you could talk to it yeah >> lesson planning >> lesson planning yeah sorry looked at the wrong personal approaches like trans languaging you know there's there's a lot of opportunities there working cross languages. >> Nice. Very nice. >> Creating nonjudgment environment actually. >> Mhm. >> With AI normally in for example in Turkey in Turkey we have a judgmental uh language learning environment. Mhm. >> With AI support, students can find a nonjudgmental environment where they use AI. >> Very nice. I I used to work in a tutoring business with human tutors online online tutors and um we were trying to understand what works with tutoring. What why is why is it that when you tutor a student that their grades go up? Is it because of the quality of the teaching? Is it because of the personalization? And we came to learn that one of the most strongest reasons for success was confidence that that we were catching people who were unconfident in certain areas of maths or certain areas of English and we were just helping with those bits which enabled them to then speak up in the class and so that same knowledge judge. Yeah. Yeah. So, I thought of the sort of pra p practicing speaking, the simulating social skills, personalized materials, um, which you suggested, um, feedback and sort of self-evaluation, being able to get kind of more meaningful feedback. Um, and don't forget the admin automation. It really, it really is quite powerful at taking away some of the tedium for teachers. I think one of the first comments was about freeing up time um, and helping the students build new skills and the teachers build new skills. It is a it is a weird moment. It is a dangerous moment, but it's actually kind of exciting, too. Yeah. So, back to I've been using this water metaphor all the way through. Um, I'm slightly obsessed by waves and tides and and so I guess I I guess my water metaphor here is to try and create a safe space to be a bit playful because the best way to understand these emerging tools is to play a little bit and is to have a sort of lower risk way of experimenting and trying out new things and um and so this is this is sort of to allow students to be student scientists and teachers to be teacher scientists to try and experiment and treat treat these tools as a interesting thing to figure out how to use which is very different from the classic teacher training where we tell you exactly how to use it and it's it's not that it's kind of the opposite it's giving you enough space to navigate and share with peers so I mentioned the book before it is free if you'd love to if you'd like to have a look at it um that's the QR code I'll pause for just a second um um but but it's it it's slightly philosophical and meandering. It's not just about language. It's more about how we think about it on the whole in the workplace. Um um we it's very kindly been picked up by a bunch of business schools, but so we we're quite enjoying that. But really I my my kind of summary that I just wanted to share with you was these machines are there and we're not escaping them. the tides keeping rising, but it's up to us to figure out how to use them wisely and how to make how to make use of this power because it is kind of there. Um, I'd really encourage you to spend the time to understand that jagged frontier and understand what it can and can't do because it's something I can't tell you. I can I can point push you in the right direction. um ask you to think about how you might build a team of explorers in your own institutions um to create the freedom where they can be explorers and we just discussed some fairly sort of loose particular ways of of taking this further in language. And that's me. I hope that was helpful. Um, I I hope it sort of dovetailed a little bit what Fiona shared because we're both very enthusiastically human about about how what AI could do, but I'm coming from the tech side about where we could point it and what's what's useful and she's been coming from the more school side. Um, the three of us are on a panel later discussing AI, so you could together with Erica. So, you'll be able to um ask us ask us questions then as well.