I find AI fascinating – and I know extremely little about it, but still use it probably every day. If you are someone who is interested in AI in education, and want to hear from better experts than me, as well as y’know, still hearing from me, you can join me and others at CredimusAI on the 16th November. Find FREE tickets here.
Picture the scene. It’s 7:32am. You’ve had your first of your two morning coffees, and you open your e-mail inbox. The third request for some really, not integral, data is there, as well as some key requests for meaningful student intervention – support plans, academic intervention requests and safeguarding concerns. Parent e-mails asking why decisions have been, or haven’t been made. Some curriculum documentation that needs a once going over. Wouldn’t it be great if you could just slap all these requests into one all encompassing LLM (large language model for the uninitiated) and it just did it all for you, and all you had to do was double check it was right? How great would that be?
Now at this point, you’re either at one of two schools of thought. You’ve answered “Yes!” and are looking forward to the link that I share at the end of this where I charge you a lovely monthly fee for access to SuperEduGPT…or, hopefully some of you have answered “actually, that doesn’t sound like a good idea at all.”
Hear me out. All these fantastic packages offering to plan my lessons for me, mark for me, give students’ feedback for me, write my e-mails, organise my calendars, take minutes of meetings, design curriculum elements – whatever it may be that you need doing, they’ll be an AI-driven EduTech product waiting for you. Some of them might be very good! Some of them might not be. But this isn’t an evaluation of said products – it’s a critical thought for this watershed moment in education.
Why AI probably can’t do it on a cold rainy night in Stoke
Back in early 2024, there was a post published by the University of Oxford, where they asked Claude, a key player in the chat bot AI world, from Anthropic, to draft an article on the opportunities and challenges for AI in education.
“One of the biggest opportunities with AI in education is enabling more personalized and adaptive learning experiences. Through machine learning and neural networks, AI systems can analyze large amounts of student data to identify learning patterns and preferences on an individual level. AI tutors and virtual assistants can then tailor instruction, feedback, and recommendations to meet each student’s unique needs and strengths. This level of personalization can increase engagement, motivation, and academic achievement, especially for struggling students.”
University of Oxford, 2024
That’s a great quote from Claude. Isn’t that a goal of all educators, really? To make sure that every member of their classroom has such a personalised experience that everyone makes the most incredible progress that they can? It’s quite the aspirational goal from Claude. Can a piece of software, no matter how capable or advanced, really deliver such an aspirational goal?
In February 2024, Google DeepMind created a transformer-based AI that could score a grandmaster ELO in Chess.
It did this by training the AI model with a dataset of 10 million chess games, leading to roughly 15 billion data points, and led to the successful solving of several challenging chess puzzles, without any tweaking or ‘search’ – just by figuring out what the next best move could be. Sounds amazing – and it is amazing. I am terrible at chess and it astounds me that someone has created a programme that would get to heights that I could only dream of.
But here’s the catch – there are rules to Chess. Only certain things can happen. The variables are dictated early and are clear. The parameters, and hyperparameters, are in place. This isn’t based in the reality of human nature and culture – often something that teachers are tasked with improving or changing. Whereas the rules of games and codes are set, an AI will only ever be able to deliver what it thinks the best outcome is – and that’s what we do as educators already. Now, there is an argument to be made that an AI might be able to do it with a greater degree of accuracy – but as the complexity of the situation increases, the less chance of success occurs.
AI won’t, at least at the moment, be able to appreciate the silent victory of a student who finally cracks electrolysis – or that Friday Period 5 lesson where you’ve spent the first ten minutes just telling everyone that they are all okay, and functional human beings who can lift a pen and write a few answers to a few questions – or that difficult student who, no matter what you try, doesn’t respond in the way you want them to. And to be fair, there might be a future where it can – but it is very far away.
Is there any point then?
Of course there is. Where I believe generative AI will change education in the next few (<5) years will be focussing on areas of administrative tasks, freeing time up for teachers to get back into the classroom and delivering more meaningful academic experiences, learning and interventions. Arbor’s AskArbor AI-driven tool for communications and data is something that will revolutionise the workload of both teachers and leaders – and as other MIS systems catch up, I can imagine we will see more inroads in using AI in reducing teacher workload. But even more critically – one of the areas I think we will see the biggest change to education is the evaluation of this after it starts being used.
If we begin outsourcing tasks that were once integral to our roles, it prompts us to reevaluate the true value of those activities. Were they ever essential, or did they distract us from more impactful work? By allowing AI to handle routine or administrative duties, we may uncover that our real contribution lies elsewhere—in the human connections we build, the critical thinking we inspire, and the creativity we foster in our students. Teachers are irreplaceable when it comes to nurturing and mentoring students; no AI can replicate the empathy and insight a teacher brings to the classroom. This transition isn’t about reducing our focus on students but enhancing it. By ensuring that administrative tasks are efficiently managed, AI empowers teachers to spend more quality time with their students, making sure that everything is right for them. Ultimately, it’s about using technology to support the irreplaceable human elements of education, allowing teachers to focus on what truly matters.
This final paragraph was written by ChatGPT rather than by me.
Tickets for CredimusAI available here.
Leave a Reply