Tuesday's Plenary - The AI factor have we figured it out?

good morning everyone I'm Vicki s um I'm afraid there is no AI artifact that can deliver this presentation for me so here I am in the flesh all the way from buenosaires Argentina I would like to thank a Teo and the Board of Trustees for having me here today and allowing me to share my views on what on what seems to be one of the topics of the moment so I would like to start with some questions that I've heard around the use of AI in our field I've been reading a lot and I've been listening to people and I have put together 10 questions that I think kind of summarize what people are wondering and asking about AI these days now let's have a look at them the first is what strategies are there for AI to empower teachers rather than replace them how can AI be seamlessly integrated into existing elt curricula without overwhelming or alienating teachers but there are questions what are the big commercial players in elt doing how can we exploit the potential of AI regarding personalization how can AI support language assessment and evaluation in elt what are the current limitations of of AI in elt and how might they be addressed in the future what are the potential future Trends and advancements in AI that could further revolutionize elt but I think more important than these questions are the last three what if any work is being done to counter the environmental damage that AI is causing and how bad is it what are the ethical considerations and challenges associated with using AI in language education and is AI going to democratize education or further the digital divide I would need much longer to be able to answer all these questions now and today so some of these questions will be addressed the others will remain questions for us to ponder discuss and hopefully try to find an answer to during these four days so before we can even look at language learning itself I think we need to look at some of the basic six so I'll be looking at some definitions I'll be looking at how AI works and I'll be showing you some relevant facts and figures now in order to understand generative AI we need to look at the broader terms so when we talk about about artificial intelligence there are other terms that come into play machine learning is a subset of artificial intelligence deep learning is a subset of machine learning and generative AI is a subset of deep learning but what do these mean machine learning is a datadriven approach that involves training algorithms using large data sets that are labeled manually labeled so that they learn and adapt some examples of machine learning are product recommendations on social media or on streaming services um spam email filtering and algorithmic trading now deep learning can process unstructured data which means that the data sets don't need predefined formats or manually created labels in health care deep learning is used to create predictive models for disease diagnosis prognosis and treatment but other examples of deep learning include facial recognition speech recognition virtual assistance self driving driving cars and automatic machine translation so what is generative AI this is one of many definitions available and I'd like to focus on the word new generative AI creates something new but how does it do it it analyes data and uncovers the underlying grammar within the data sets by recognizing objective patterns and working out frequencies within them it then mimics the language to create something new though necessarily derived from the data when we talk about generative AI chat Bots like open AI chat gbt or Microsoft co-pilot or Google Gemini for example we are talking about tools that are based on large language models or llms these llms have been fed huge amount of data they have analyzed it and they have come up with statistical correlations to articulate these interactions so generative AI does not really know or understand things they are doing Advanced mathematical calculations and when they don't know when it does not know it hallucinates it invents it fills in the gaps now in order to recap how AI works these are four features of AI it is Data driven as we said before but this means it's only as good as the data it's been fed with and we I will come back to this because it's an essential aspect that needs to be addressed in more detail it is creative which means it creates something new each time as it was mentioned in the definition I know this might be controversial because some people argue that it merely regurgitates what it's learned but the fact that most results are different based on the prompt wording the context and the data available might help to back up this creativity claim it can improve over time as it is fed information in two distinctive ways first by its creators from the training data sets second by the constant use by anyone so everything you input into a publicly available open AI tool becomes part of its data set and when we are asked for positive or negative feedback of the results that is used to retrain the models continually so it learns but this also raises questions about data privacy it is versatile which means it can carry out a wide variety of tasks not just searching and of course I think we will see we will see new applications being disclosed as we speak right now I think we need to look at some facts and figures that will help us see the magnitude of what we're talking about the global artificial intelligence in education market value is rising with very steep projections so we are only seeing in the beginning of the AI Revolution now this revolution is evident as we look at this other data the annual number of AI research papers has increased 40% the number one researcher in the world is the US with 30% of the total search number two is China with 18% Africa South America and most Asian Nations have only contributed less than 5% of the total research so this might be an indication of the global divide now the number of generative AI related jobs has also increased 36 times compared to last year prompt engineering courses abound in both formal and informal settings but people hired for these positions are being selected on a skills first mindset because there didn't used to be certifications before there is a website called there is an AI for that it's a repository of AI tools where you can search for an available tool to carry out the specific task it currently lists 12, 261 AI tools for 15,34 tasks in 4,846 jobs how can we keep up with that can we now the facts and figures related to the environmental impact of AI cannot be ignored the carbon footprint of creating training and prompting generative AI is considerable both in terms of electricity and water resources needed to run and cool down the servers used these are other figures that might help us to understand this these figures uh make it seem that we are at odds with a much agreed sustainability stance now uh quantifying comparing um figures with more visual representations as countries in the other slide and cars and swimming pools and bottles of water later have helped me to make better sense of the data image generation is particularly energy intensive data centers server rooms must be kept cool at around 10 to 27 degrees celsus to avoid equipment malfunctioning they must use clean fresh water to avoid corrosion and bacteria growth associated with seawater fresh water is also necessary for humidity control in these rooms this is one of Google's hyperdata centers one in Oregon and what you can see is the steam coming out from the four cooling towers used so whenever you he don't worry if you don't get your result right or a precise result you can refine The Prompt and try again you may want to think about water consumption each time you prompt a generative AI tool now the next question I want to address is what can AI do now these days is rather what can't it do but what can AI do and I'm sure you've seen loads of examples of the kinds of things that can be done using AI impressive videos of people who cannot speak another language speaking impeccable English Japanese Swedish German whole videos created from text prompts voice cloning and much much more so you may be disappointed but rather than showing you examples I have created two visuals to summarize this these are some things AI can create some things that teachers or students might need or want to use AI can also carry out all of these tasks and we expect this list to grow considerably now these are just possibilities not an endorsement of what I would particularly use AI for so whenever we talk about AI people tend to adopt a particular stance in this debate so I want you to invite you to think where you stand and if we look at this as a Continuum of opinions the first group is their deniers or resists oh I don't want to go anywhere near it or this will be our downfall the second group is the indifference oh yeah I've heard about it but haven't tried it yet or I don't think there anything in it for me really the third group is the cautious optimists those who prefer to have an open dialogue with it in order to assess its value the last group are the enthusiasts or preachers those who swear it's the best thing ever those who want to do everything with AI so where do you stand and I think it depends on how AI affects you how does it affect us and who is us although AEL is a Teachers Association its members are not only teachers we are part of an industry that includes many more roles so for the past few months I've reached out to colleagues in these groups because although I wear some oh let's sorry I'll go back to this so this is the groups I thought we could have this is how I grouped the roles we have teachers and teacher trainers we have directors of studies and administrators we have materials writers and illustrators and we have Publishers and editors and although I do wear some of these hats myself I could not possibly pretend to wear them all or know about them in detail so I have reached out to colleagues in these different roles and asked their opinions in order to inform this talk some of them will be named with their permission some preferred to remain anonymous they all selflessly shared their views with me and I am really grateful to them so let's start this is what I did I asked them about their role in a closed question so that I could then classify the information they gave me and then I asked them three very open ended questions so that there would be enough room for personal expression I asked about their positive and negative views about Ai and I asked them what the impact on the role was how has AI impacted Your Role already um so it is very difficult to quantify impact uh but I found that there is a an AI job impact score that someone created and uh this impact this is shown as a as a percentage and it is calculated based on the tasks that can be done using AI the the tasks that are relevant for a job and the AI current capabilities for you know this tasks and so this impact score is expressed as a percentage where zero% indicates that the job has no impact that AI has no impact on the job and 100% indicates that this job could theoretically be fully automated using AI based on current capabilities there is no certainty that this is accurate enough but I have included it as a curiosity okay so let's have a look at our first group our first group was the teachers and teacher trainers the impact score the teachers is 20% so it's not that high but what did they have to say on a positive note many mention they are using it as a personal and research assistant that was the main reply but they also mentioned that they use it to generate ideas to summarize to generate texts and questions and for writing support a common view a common negative view related to students using AI was uncritical use misuse and academic Integrity as a language lover devia maman's concern was that students may be falling out of FL love with language really resonated with me I fear there are students will be trying to find shortcuts that will undermine their learning now when asked about the impact on their roles they mentioned that AI mediated exchanges seem to be more efficient but less authentic they recognize that AI literacy has become essential as and they feel the need to redesign take-home writing assignments in light of plagiarism and I will come back to plagiarism later so it seems that teachers and teacher trainers will use Ai and might use AI for things that will help to reduce their workload and that may free time for them to pay attention to other tasks where AI is not so good at I I guess the key will be finding out and figuring out which are the things that AI is good at and which are not now let's look at our second group directors of studies and administrators this group was not represented in the job impact core might this mean they are safe their positive and negative views about AI were very similar to the previous group so I'm going to jump straight into the impact on their roles they all agreed that they the there's a need to Monitor and ensure academic Integrity on all qualifications and assessment and this leads to AI plagiarism detection that I will address later they also mention that some are using AI for marketing and one very prestigious institution is working on including AI practices in all of the courses I guess that for directors of studies and administrators we have something similar than for teachers they will need to find out which things AI is good at like process Automation and routine tasks and that will free time for them to focus on more human tasks our third group is material writers and illustrators for Content writers the score is 35% and for illustrators it's 40% so this is a bit higher but what did they have to say they agreed that AI is saving them time and helping them to produce certain content more easily especially images on a negative note they were worried again about ethical aspects like copyright and also about AI misuse although creating images is very quick in the words of Katy BSB and illustrator it is hard to replicate a personal style in illustration in illustrations so there we go as regards the impact on their roles most of them are using them as a research assistant some have been getting uh explicit guidelines from publishers about using Ai and some have seen air related Clauses included in the contracts now in some institutions multiple roles have been replaced by one prompt engineer so I guess for material writers and illustrators it's seems AI will be very useful in assisant research but I believe that their expertise and their personal style is really hard to replicate and might should be their trademark and maybe path to survival now the next group and the last group is Publishers and editors it is in this group where I saw the most changes already happening the impact score for Publishers is 25% it's not that high so let's have a look at what they told me positive aspects of AI were reported extensively like there were lots personalization opportunities like personalized feedback and support was one of the main Trends especially in regions where there is a teacher shortage they also mentioned the east2 prompt illustrations and create more authentic materials now creating more accessible and affordable content for certain regions which would benefit from this is a welcome option AI mediated technologies that do not require changes in infrastructure or digital skills will benefit students in regions with limited resources so we could be taking a step towards the democratization of Education here Ben Knight's belief that AI can cover more mechanical aspects and teachers can focus more on engagement and more nuanced aspects of teaching is a view I share as well there are quite a few negative opinions here materials shifting to areas which AI can handle better and neglecting others and thus leading to the dumbing down of content is a concern assessment is also in the spotlight what if AI assessment is given more value than human judgment personal data collection is another concern and it should be noted here that there is a difference in how data is collected and used between publicly available open AI tools and closed AI platforms further concerns are D Skilling on the loss of human connection Publishers becoming silos of wealth and power and therefore furthering the digital divide is a scary but real concern and what about producing content without editorial scrutiny to prioritize a fast turnaround some of these seem exaggerated but are they when asked about the impact on the road they mention they are researching the potential impact of AI on teaching and learning materials they are looking for ways to improve learning materials and tools some Publishers are creating prototypes for experimenting with prompting within a designed interface for example to create lessons one publisher is already using AI in in these areas ethics and legality internal efficiency through process Automation and lowcost content creation so it seems that Publishers and editors are quickly finding ways to make AI work for them but I still think that the vast experience and the Myriad of decisions made when creating and Publishing materials is far from replaceable so in order to recap and reflect I have put together some ideas that for me are the core issues around AI looking ahead at the beginning of the talk I mentioned that AI is datadriven now the question is which data unfortunately most of it tends to represent a subset of our world's geographical and cultural perspectives St shats instead mention the acronym weird in their book engines of Engagement I must read weird Western educated industrialized rich and Democratic cultures are over represented among the creators and curators of AI this means and this has a direct result in output that is ridden with inherent biases and assumptions like discrimination gender assumptions and sexist views now most of our most of our social attitudes are slowly changing but the vast majority of our historical and visual records represent models that have become outdated we cannot escape the fact that we are biased by Nature there are lots of systematic inherent biases interwoven into our reality and this is obviously transferred to AI let me tell you a very simple example of gender and race bias if you are on social media you may already have seen it but if you prompt a generative AI tool chat Bo or image creator with a prompt that includes the words a doctor and a nurse guess what most results will show the doctor as male and the nurse as female and probably the doctor will be Caucasian and the nurse won't now another problem associated with data sets is that AI is poisoning AI so AI generated content is filling the internet and at an unprecedented rate so there's a fear of the unintentional introduction of Errors into the data sets that are used to retrain the models or train new models this could lead to model collapse where compounded error lead to significantly compromised output this could have drastic consequences for jna models and their training now another issue that is related to the data is multilingual representation the British Council report states that generative AI is standardizing languages and ideologies so we have two questions around this the first is uh data most data is in English most models are trained with the majority of data in English most data sets in English so a lower percentage of data sets in other language is reflected in the quality of the results in those minority languages and also which English there is a lack of recognition of nuances and exclusion of certain groups and varieties of English all together this has a direct effect on plagiarism detection we cannot Overlook the issue of copyright and this again has two sides the first is who owns the copyright of AI generated content and this is being discussed at all levels but it is not clear at all yet is it some say it's public domain others that it's the person or legal entity that generated it the second is the claim that large language models have been trained using copyrighted data and this is also under scrutiny in various International courts and is undergoing ethical examination one such example is the case of the New York Times suing open Ai and Microsoft over the use of copyrighted work it's still too early to know but these too could have drastic consequences on gen AI models and their training as mentioned before by several colleagues there is a real concern related to academic integrity institutions are trying to figure out how to address plagiarism and cheating most or you can see this in the development of AI use policies and guidelines but institutions are also using AI detectors and they are these detectors are being used to punish offenders now the problem with AI detectors is that they still prove quite problematic because they turn a high frequency of false positives and they have been shown to discriminate against non-native speakers of English I have already shown you several figures related to the environmental impact of AI and this only goes to show that technological progress is not always in line with other Global issues most or many organizations are working on ways to reduce the environmental impact of their operations such as solar Farms to power their plants some experiments repurposing the fresh water used or using seawat but it's still too early to know whether they will succeed one last concern is that of unequal access leading to furthering the digital divide there is a huge a huge concern about the research being done that can widen the global divide especially between the global North and the South there is it's no surprise that most companies have done more research about AI than governments since 20121 and the analysis of the data shows the Stark divide in AI research and development so countries like India are gaining ground in AI research but the leaders in in traditional the traditional leaders in Europe and North America are still producing High volumes the rest of the countries developing countries are trying to get more seats at the table still so where do you stand now has your opinion changed where do you think I stand so depending on where you stand now you may already have a closing argument in your mind so let me share mine we will continue to see AI develop but I think that What Makes Us human is very hard to replicate our subjective patterns humor wi empathy our social emotional intelligence I think we will be okay as long as we play to what makes us different and sets us apart as long as we understand that our critical thinking our human touch and connection will allow us to navigate these times and Empower ourselves thank [Applause] you if you have questions for Vicki you can uh get ask them this afternoon she will be answering questions at 1535 in meeting room 3 okay if you missed the beginning or you want to watch this again the recordings of the plenaries will go live each night at 800 pm on our social media channels so you'll be able to see Vicki do it all again from 800 p.m. tonight thank you very much enjoy conference and thank you Vicki thank you

Channel: IATEFL

Views: 1,941

Keywords:

Id: 2IkCub2jFXs

Length: 47min 48sec (2868 seconds)

Published: Wed Apr 17 2024