Download transcript PDF

00:00:04
and I want to show you this uh image uh first oh I need to somehow uh see the chat bot I hope I can get it out yes my question to you is what do you think this is okay so what does this image remind you of and feel free to put your ideas in the chat uh a snake I I love it I love it a snake a worm a dove th now that is very very creative okay uhuh string stream okay a noodle I love noodles okay someone said oh my God like you're getting more and more creative like we had Lions as well oh learning

00:00:52
path you're getting close to what I or originally wanted spaghetti hat uhuh very good very good um those who are um thinking about learning Journeys um then then they are they are getting close to the right solution I'm going to show you thing uh I'll I added a couple of emojis here um what does it remind you of now a way to a goal uhhuh okay process of learning very good Trends ups and downs in life uhhuh o someone's actually got it our attitudes to chat GPT okay very good um I keep going but I'll I'll give

00:01:45
you the solution um someone in the morning said it's an emotional roller coaster which I think is also a very very good response it's actually it it can be a lot of things because this curve can be applied to many things it's actually the curve of um stages of culture shock when you're when you move to a new country which stages you go through it's also it also can illustrate the stages of a romantic relationship but it's also um and that's really relevant to us it's also the Gartner

00:02:23
hype cycle um that shows our relationship to new technologies and basically as you can see and it seems like we all go through the same steps whenever something new appears in our lives whe be it uh a romantic partner a new country or new technology we've got something at the beginning uh the trigger or this new technology or new thing that happens or appears in our lives and then at the beginning we are super excited uh typically I know that even with C GPT not everyone was super excited at the beginning some people

00:03:05
were super afraid but um most people most teachers got really excited they were all over about raving about chpt and that is typically followed by what you can see there the bottom which is called the trough of disillusionment basically at the beginning you're all about um which we can call the peak of inflated expectations that you just see the good side of it then suddenly from this very extreme High you fall to an extreme low of disappointment and you see that oh come on this is not what I expected it

00:03:49
to be it's not as good I as I thought it would be and I hate it and suddenly you you end up in this other extreme you hate it you don't want to use it and um if you remember at the very beginning when Chen C GPT appeared everyone was saying like how it's going to revolutionize education how it's going to change education um people were testing it how it can pass certain exams and then suddenly people started saying like oh come on this cannot even solve a math problem it's not good for this it's

00:04:23
not good for that it's making mistakes and lots of uh hallucinations for example factual mistakes so we hate it but then as you uh gain even more experience um and experiment with the tool more you come to this next stage with the thinking face and that is this slope of Enlightenment so you suddenly not suddenly but gradually discover that okay maybe I I was too harsh and it's not that bad as I thought um I can I can see how I can use it and um actually the Gardner hype cycle finishes with a

00:05:07
so-called plateau of productivity while the plateau is typically this um straight line I wanted to draw something else because I think that it's not we don't want to arrive at a plateau we would like to go further so what I or how I imagine it is that after this um rational look at um what AI can do for us we can actually move up higher and reach this point where we really see how it can Aid us how it can benefit us and that's where I think we should all end up in this year okay so that's that's

00:05:50
what I think uh 2025 could be about but let's see uh what I like things in detail um um what you're going to see in the coming slides are um let's say let's imagine a generic language teacher who um says these things or actually said these things at the very beginning it could be the very beginning in 2022 when CED jpt appeared or the moment they meet AI or CAD GPT or any other chatbot Gemini Claude and so on they are very enthusiastic like come on I love these I can generate entire lesson plans tasks

00:06:36
images videos I can do this all day long I'm just going to prompt away but um there's always a butt and this is why we are here to think about the other side of this um so let's take on or put on our thinking or more critical hats and think about what wrong with this idea or with this um approach I'm going to invite you to put your ideas in the chat I'm going to be monitoring the chat so what do you think can be wrong if someone says this or someone so excited about prompting all day long

00:07:20
feel free to add your ideas in the chat I'm always going to give you further um hints and and input uhhuh someone said AI can make mistakes every St oh someone mentioned energy and water usage very good very good no human interaction exactly uhhuh losing creativity wasting time and creativity exactly now everyone's talking about creativity exactly so there are several things here to to take into account it's it's okay we love it yes but first of all AI models um so their environmental impact

00:08:03
is huge the the environmental um requirement or sorry the energy requirement that's needed for the training of models is is huge but even if you just think about your own personal level the amount of energy that's needed to to just create one prompt according to Nikki hawkley uh who wrote this amazing book uh 30 Essentials for using AI which just came out this year um she says that one prompt uses half a l liter of water or we can look at it from another perspective uh other estimates say that one prompt uses as

00:08:46
much energy as 10 Google searches so it definitely uses energy um the the op the maintenance of these um servers also need water as someone said it to cool them so their their environmental impact or footprint is huge but apart from that you mentioned very well that AI can make mistakes and if uh we rely solely on AI and and its Genera responses we might fall into this trap that we accept hallucinations incorrect information as truth or uh we are going to lose our creative anger critical thinking because

00:09:31
we're just not exercising our brains that much as before and you also uh mention no human interaction so yeah if you just imagine this that I'm sitting in front of my computer like right now and I keep prompting away it's fun but actually I'm not communicating with anyone I'm communicating with a machine and it's it could be a great feeling to have teacher colleagues real Flash humans who who you can discuss your ideas with and real human uh humans you can teach so human

00:10:06
interaction is also very important and we're going to continue in the same fashion so here's another quote um for example a very very happy Teacher as many teachers were um at the beginning um can say oh look I can upload this entire book and ask Chad GPT to modernize it change the level of the tasks add more Tas adapted to different levels um I can do so much or look I can upload this activity and do this or that with that um but what's what's the problem with this um approach um

00:10:50
anduh I'm going to help you with this it's actually that uhhuh okay uh more ideas are coming in it might not be appropriate very good um we oh someone mentioned copyright exactly I was thinking of copyright but those who mentioned that the level might not be correct they are also correct um we were and all teachers um have been experimenting with this level adaptation and while AI chatbots are sort of getting better at that they're still not there so there level even though they mention and there might be lots of

00:11:33
websites that say that um look I can adapt this one task uh to all the CFR levels it's not 100% uh what we would imagine or what we would um expect but the real point is copyright indeed if you upload a book uh an activity from a course book or an entire course book you are uh violating copyright and unfortunately at the very beginning people were like this was the best use of AI they thought like this is what a these text based chatbots were made for why can't I upload a book and just change everything well because

00:12:20
you're you're violating copyright so it's very very important even though it's it's um tempting um and a rule that you can keep in mind is that the the fact that you can do it doesn't mean that you should do it um so don't upload books s um contracts tasks and other copyrighted materials into these chat Bots because that's not legal okay what about this other one I don't have to spend all my time grading essays and writing reports I just upload everything into a chatbot and have them

00:13:02
analyzed and written up for me um but what's what's up with this one so what do you think because again a lot of teachers um think and are advis to just upload student homework into AI chatbots or AI powered websites um or why can't I just upload their grades have an AI website um analyze it for me oh nice Katie got it exactly and someone um someone else as well has said it privacy exactly so private data data protection or gdpr that's the issue without your students consent you shouldn't upload

00:13:50
any of their essays homework tasks grades you might want to give them fake names and then upload an Excel sheet and analyze their grades with fake names but with their real names you cannot just upload um their homework their essays and so on so you should always ask for their consent and this is also important when you're finding these AI powered websites that promise you that you can cut your grading time in half if you use their website and upload stuff be careful what you upload okay let's see another

00:14:33
one I'm not sure maybe it's new information but uh CH GPT has this amazing function and I I I truly uh agree with this quote have you tried voice Mo voice mode yet it's insane it is uh you can have these amazingly realistic and authentic conversations with CAD GPT you can ALS like get it on your phone in the cad GPT app or you can have it on the website as well and more and more chatbots um offer this speech to speech technique so you speak to your phone or um laptop and you get authenticly sounding

00:15:16
realistic conversation um or response but what's up with this one um and the issue with voice mode uh there can be several issues actually here um the main thing is that we're teaching English right and English is now a global language and there are many englishes many accents many dialects native non native alike so the problem with C gpt's voice mode is that currently they have nine voices out of those nine voices uh two are UK accent and the rest are us ACC accents but where are all the other

00:16:02
accents um Australian South African uh New Zealander but these are just the native accents where are the non-native accents if you think about it often you've got a non-native speaker of English who are who's the teacher teaching English to non-native speakers who are going to communicate maybe their entire life with other non native speakers so we should need a model that's going to help those non-native speakers if only if they only listen to native speaker accents they cannot really get that far they are going to

00:16:44
try and try to obtain or an accent which is which is difficult to get so we are not really providing them with the right model um and Jennifer Jenkins said um in one post uh that we could even call AI artificial imposition because it's imposing a native uh accent or native view on our students so to make this work we could teach or train these models with other accents so that they can provide a more realistic model to our students also I heard from my um the participants of my teacher training

00:17:31
courses um Spanish speakers said that CAD gpt's voice mode uses a Latin American Spanish accent and Portuguese speakers said that it uses a Brazilian Portuguese accent so for them they could say that come on my Portuguese Portuguese is not represented so we need a bigger training pool okay and here's another one um I'm I'm all about new new functions and new tech so um for example imagine me saying have you tried deep seek uh which which is the new chat bot what about this new tool and that new tool and image

00:18:13
generation and video generation and open AI deep research which also just came out what could be the issue with this oh someone said too many absolutely so that's what um we can also call AI fatigue that you just get burnt out there are just too many new tools too many uh new functions added it's very very difficult to keep up cognitive overload very good someone said exactly that's actually one part of the story um it's definitely cognitive overload it's just too many tools but if you think

00:18:56
about it not everyone has access to these tools so um we oh I hope you can still hear me because there might be a issue with my earphones um the thing is that oh good I got a reaction thank you so the problem is that these tools are not available everywhere and this is what we can call the digital divide the digital divide can be uh understood in many ways it can mean that in some countries there's not there's no reliable internet connection um so they cannot even get to this far to access AI

00:19:40
tools um and in some countries they only have mobile phones but they don't have laptops so that's another issue they are app based and not website based but there's another level to the digital divide and it's like where these tools are introduced for example as far as I know open AI deep research has not uh or is not um available yet in Europe it's only available in the US so if unless I use a VPN I I can't really experiment with it I can't really try it so that's

00:20:18
one thing that they are introduced in various uh locations you can't really access them but another um point is that some countries they do it well they respect gdpr and they are not going to let students or teachers experiment with tools that don't really have a solid data protection um background or that data protection is not uh well expressed in their terms and conditions but some countries don't really care and I'm not going to say which but those countries that don't really care they are going to

00:20:56
try more tools this actually came came up in one of my teacher training courses that um there were participants from various countries and some countries don't really have that much experience with so many tools because they really respect uh data protection and gdpr while other countries just said we we don't care and so they were more advanced technologically however they didn't really care about data protection so it's a double-edged sword definitely and this is the final one um I'm going

00:21:33
to be very quick from now uh because um I've been talking a bit too much but let's look at this one someone who's very very excited about generating um art or music in someone's style what about this one um the problem is that all of these styles are taken from from the internet and and we might think that well whatever is on the Internet is is up for grabs it's for free but not really um and all these models scay the internet for text audio um images and they are usually not

00:22:15
asking for anyone's consent they need data to become better and to avoid hallucination and so on but in order to get that data they might not use like totally legal um tactics and that's a very very important thing who owns the data also who owns what you generate for example you generate a song or an image can you sell it can you make money off of it and it's a very very gray area it's very hard and there are still no hard rules um like who owes what and what can be done but let's look at the uh less

00:22:59
gloomy side so the positive side and I still want to show you how we can work together with AI in the future and and build a positive relationship um so what I um mention most of the time is that I truly believe in a balanced human AI collaborative relationship that can really boost creativity and productivity I'm going to be giving you a couple of examples so I think that AI is amazing we can def use it for idea generation brainstorming creating worksheets um additional activities uh creative twists um that

00:23:38
you want just want to apply to an existing activity that you generated not taken from a book um and you can play with it in a in a totally different way uh not just by generating an activity but by generating something by accident so this image was sort of generated by accident um some of or most of the chatbots are now capable of image generation but not all of them Claude uh Claude is a very good chatbot but it cannot generate images it's based on code generation so it can generate an image that is that comes from

00:24:21
code that's how you end up with these abstract and minimalistic images so my question to you is what do you think this is and I'm going to give you a couple of seconds so this was generated uh with claw and it's a very minimalistic image of something someone says confetti yeah glass of juice a sad smiley a oh it's interesting that you see sad faces be more positive um oh Pizza yes in the mornings a lot of people said Pizza it's actually uh it was generated by one of my teacher traines who generated a

00:25:04
p or P art um and it was supposed to be paa but um as you can see it can start this creative process of trying to see something into this very bad image uh but at least we can use it for a creative purpose this is another image generated with the help of Claude uh a very minimalistic and Abstract image but what do you think it could be so again I'm curious about your thoughts students very good a classroom right nice exactly it is a classroom it was generated by a Portuguese participant and um if I'm right it means in

00:25:50
Portuguese respect and tolerance and it would be a very interesting activity to see how respect and tolerance is represent presented in this very minimalistic image and what Claude was thinking when it generated this image another example I want to give you is like this this one um just to tell you this entire thing wasn't generated with AI as you can see in the bottom of the or in the corner of the image this is a genially template however the activity was created with the help of AI and I just want to show you very quickly

00:26:28
like how I created an escape room for my students with AI and this is something I think could be a a nice AI human collaborative relationship I came up with the idea but AI created the background Story the names of the planets and and the stage descriptions I came up with the idea of the tasks that I co-b brainstormed with ai ai I I asked AI to rewrite some of the activities that I already had and then I put everything together so this was a very nice AI human balanced uh relationship and another uh point that I

00:27:09
really want to emphasize uh where ai's potential lies is being or creating highly personalized and adaptive materials um so with the help of AI with chat Bots we can create entire sets of customized materials and we know that now we have lots of students with uh various needs in order to meet their expectations and needs we can use these chat Bots to generate different versions of the same activity but if we don't want to uh use chatbots and do this work all by ourselves we can use McMillan

00:27:48
englishes or sorry McMillan education's pivot English and pivot English is a is a platform where students can learn by them elves it gener it works with AI technology generates tailor made activities based on their individual performance it uh constantly tracks their performance um and adapts materials uh to meet their individual needs but it's very important just as in the previous case the teacher still in control so the teacher can still decide if they want to um guide because they can oversee the entire process and they

00:28:32
can decide what individual students should see whether they are ready for this activity or not so they can still be in control however they individually get support from pivot English using this rather complicated system but you can see how the AI is behind all of this because AI is not just the chatbot AI can be this technology helps personalization and adaptation and another thing where AI can really help us is automation of robotic or mundane activities such as writing reports and filling in tables I

00:29:12
hold like I definitely you can see this mentioned many times but I truly agree that this is where AI can help us gain some time to automate a couple of uh workflow items that we have it can also help with the objective part of assessment so grade scores points however with the subjective parts of assessment I believe that it won't actually revolutionize education but it pushes us toward more authentic integrated and performance-based assessment um so for example in connection with essay writing uh

00:29:52
composition writing of course you might be thinking well how will I know whether AI wrote this or my student well maybe it just opens our eyes to other approaches to uh to assessment instead of essays and compositions we might want to go in the direction of more authentic and performance-based assessments um and what about our future I believe that may like maybe perhaps our roles are going to change from teacher to facilitator because if think about it AI is going to be available it's already available on our phones on

00:30:33
our laptops so students might think that I can do this all by myself I don't need a teacher so what is the teacher or what can the teacher really give their Human Side so instead of teaching they might be facilitating instead so emphasizing assessment for learning emphasizing their help they can give in AI literacy because our students will need uh AI literacy to use these tools well and we can also help them give live skills that we can give them and within AI literacy we can also help in these two cases help

00:31:14
them with global englishes showing proper models of non-native um speaker accents world englishes and also teach them about sustainable AI use that not everything needs to be prompted but we can help them guide uh or we can help them show the way um and this is where I'm going to finish my talk so thank you so much for being here thank you for listening