By Phil Wickins

(Title Image by Gerd Altmann from Pixabay)

These are my own developing ideas about how these complex new technologies work, so forgive me if I have oversimplified or overlooked!

In the KS2 computing curriculum, it states that pupils should be taught to:

• select, use and combine a variety of software (including internet services) on a range of digital devices to design and create a range of programs, systems and content that accomplish given goals, including collecting, analysing, evaluating and presenting data and information

This objective already covers a multitude of software from creative to academic, but must now also be applied to applications that did not exist when it was conceived. Furthermore, the DFE have recently released some guidance on Generative artificial intelligence (AI) in education which states that we should “…teach students how to use emerging technologies, such as generative AI, safely and appropriately”.

A.I. like it or not, is going to be part of our world and even more so, the world of our children and it’s growing fast. It would be irresponsible of us to ignore it, even if we do not yet have to tools or knowledge to teach it. So what can we do?

First of all, let’s talk about it with our classes. What is it? What does it do? Is it actually sentient artificial intelligence, or is machine learning something different? What is a large language model? Are we allowed to use it in our learning? Have a look at this interesting article: Summary of Envisioning AI for K-12: What should every child know about AI?

I had a very interesting discussion recently with a good friend and colleague of mine around the dangers of anthropomorphism when it comes to computers. That is essentially where we personify machines and begin to expect them to behave like humans. Which is easy enough to do, right? We are bombarded with films and media where robots push back after being treated badly; we get emotional when a machine that’s been programmed to defend a human meets its end. It’s a trait we’ve had for years without even realising it; how many of us have stated ‘this photocopier doesn’t like me!’ or how many pupils have come to you saying ‘the computer has lost all my work’?

I conducted a little experiment on social media recently where I asked, ‘should we teach our children to say please and thank you to Alexa?’ The response was not surprising; 8 out of 10 people said something along the lines of: “I always say it myself, because during the robot uprising I will be spared because I was always kind to them!” Joking aside, it can’t hurt to be polite and practise manners? Can it? There is of course the argument (I think in Suleyman’s ‘The Coming Wave’) that we should mind our manners with the AIs because we are literally training them – if we are polite, then they too will be polite in the future.

The danger comes after we’ve attributed these human personalities to machines, we begin to trust the machine’s response, we trust it to be able to think and adapt and reason like a human. Because of how articulate and knowledgeable it is, it fools our subconscious into thinking of it as sentient, as everything else we encounter with that command of language is a person. At the moment, to my understanding, we are not at those levels of AI yet. Where are we then?

Large language model systems, like Chat GTP, are being trained on enormous amounts data at mind blowing speed, to simply predict what word comes next. What is the most likely word to follow the previous one (or the second or third most likely, which is why you get a different response to the same prompt), based on all of these similar documents that have been written by humans. Machine learning systems do not think; the output from these systems is just a prediction, one word at a time.

So there are going to be mistakes, there are going to be inaccuracies, there are going to be facts that sound like facts, but are simply not true (Try this; get Chat GTP to write you a piece on a certain subject, but ask it for citations. The likelihood is some authors won’t exist, others will be misquoted, or there will be no citations at all). That’s what we need to talk to our pupils about. Not to personify, not to trust implicitly, but to check and double check. Because after all, it’s a machine running a program, using material that has been written by humans (which can also be incredibly bias which is an added problem!).

I’d love to develop an unplugged activity for KS2 where pupils emulate a large language model by submitting words into buckets, which are then selected based on the previous word to see if a sentence that not only makes sense but is also relevant to the request…  but this needs some thinking through!

I have heard of a fantastic use of an AI image generator being used by a teacher, where the pupils would feed in their setting description and an image would be generated. If it did not look similar enough to the picture the pupil was imagining, they are invited to then refine their description in an effort to give the reader a better image, and so on. Now I’m not sure of the implications of this in terms of cost or age restrictions of the software, but I can see this being the kind of activity that would suit a primary classroom.

So how about you? What do you think we should be doing in the classroom around AI? Teaching how to use it? Talking about it? Do you have any amazing resources or ideas for the classroom? I’d love to hear them; why don’t you reply to this thread ‘Using AI in the primary classroom’ on the STEM Community and let us know your thoughts…