Skip to main content

27 April 2022

What I learnt about AI on #caschat last night

Rachael Coultart profile image
Written by

Rachael Coultart | Computing Subject Lead

On a Tuesday evening most weeks between 8 and 9pm a group of enthusiastic educators come together to Tweet answers to questions presented by the host on a topic related to Computer Science and Computing. Last night the topic was Artificial Intelligence – something I am both curious and very nervous about. So I readied myself for an interesting (and frantic typing) hour following the discussion with the aid of Tweetdeck - and I was not disappointed.  Hosted by Michael @MikeJonesCSTalk the evening went something like this...

Icebreaker: Why bother with AI in the curriculum?

It’s a big part of our present and will be an even bigger part of our future. @wnfranklin

Feel it’s important to examine how AI will feature in society. Lots of decisions to make for future generations. Vital that students see the benefit as well as potential harm, and so they can hope to influence its direction going forward @TeechGeek

I think we should teach AI in school because we want the next generation to understand ALL it's capabilities and develop it in the best ways possible. @rcoultart

So there was a definite feeling that AI is something we should be teaching in school, but the answers to Question 1: Are you teaching AI at your school? Why/why not? Do you plan on doing so in the future? showed that we aren’t all doing it yet, for various reasons. Where it is being taught it is clearly a topic that students enjoy and engage with well.

Some amazing resources were then shared as part of Question 2: What are your go to places for information on AI/curriculums you use? and this is where the tabs opened for me to explore later multiplied! Too many to add here so I’ll share them as a resource separately.

It was good to hear that the general consensus for Question 3: Does teaching AI require investment in hardware? was ‘No’, again linking to lots of the free online resources as mentioned above.

Question 4: Do you need to be an expert in AI to create an AI/teach about AI? was an interesting one for me because I’ve always maintained that you don’t need to be an expert in anything to teach it, you just need enthusiasm and a willingness to learn. And this was reflected in others’ answers too @LauraKeeney01, @sneekylinux, @clickschool, @MWimpennyS and @TeechGeek. Yet somehow I haven’t quite embraced this ideal when it comes to AI in the curriculum! Possibly because it isn’t a statutory part of what we have to deliver and is therefore easy to avoid? I certainly think it could generate some amazing discussions and ‘big questions’ that would rival anything that comes about in Religious Education lessons!

*I did have to ask one question eventually… ‘Is AI the same as Machine Learning?’ and rather than making me feel like it was a silly question (which it might have been!), some great answers came my way:

Machine Learning (ML) is just one type of approach to building models in AI. Others include Deep Learning and rules-based. @caschat_uk

Arguably, AI is what we get the kit to do. ML is the conclusions the kit comes up with once we've trained. Sorry, a really fuzzy answer. One way is to think - AI is the way I taught you ML is you being able to decide whether the object is a person or a bin bag in the wind. @MikeJonesCSTalk

Finally, we arrive at Question 5: Where have you been able to map AI into your curriculum/where would you map AI into your curriculum? Answers weren’t just limited to the Computing/Computer Science curriculums either, which was fantastic to help me consider some of those wider ramifications I was wondering about.

And to top all of this, today, Ben @BenberryPi tweets about this Future Learn course Introduction to Machine Learning and AI where I can “Discover the fundamentals of machine learning, how it works, and learn to train your own AI using free online tools.” So I’ve no excuses now for not being able to get my head around AI and how to teach it in Primary school.  Thank you #caschat once again.

Discussion

Jo Hodge
30/04/2022 07:50

Rachael sounds very interesting. Wondered if you are going to introduce it and if so where will you begin? Thanks!

Paul Powell
01/05/2022 11:10

The issue I have always seen with AI is that much of it requires a level of coding that is far beyond all but the most talented coders at GCSE (and probably most at A-Level.)

We can use a library that does the heavy lifting for us, but then what is it that we are actually teaching? This isn’t an outright criticism, but an attempt to firm it up.

For simple rule based systems we could look at a diagnosis tree - this could be just a set of nested if statements that try to get to the bottom of an issue.

Going more advanced we could look at free text entry and how chatbots can pick out words from a string, turn them around and complement them with rules. This could extend the diagnosis idea. This would probably involve teaching regular expressions (A-Level link to finite state automata).

An alternative would be rule based AI in games. Kodu is pretty accessible here - making an enemy patrol a line until it sees the player and then chases the player. Alternately coding an AI to play the game nim perfectly.

All of the above are rule based and are entirely mechanistic. At best they may remember previous input and use that in their rules. A whole heap of extra skills are needed for AI that doesn’t just follow the rules set by humans.

I have seen activities where students train an AI and the AI learns, but I’m not sure how empowering that is. It seems to teach that 1) AI learns by example and 2) AI can learn. The hard stuff is often left out / done for us.

Perhaps there are better resources out there - if so then please do share.

Rachael Coultart
04/05/2022 17:02

Hi Jo,
where to begin is certainly a good question! At the moment, I’m not sure… I’m thinking maybe some discussions with my Digital Leaders or Code Clubbers to see what they already know and understand about AI. I certainly haven’t got any grand plans to revamp our Computing curriculum and introduce it across the board! I still feel I want to know more myself, and talk to more people like yourself about what it might look like in a classroom situation…
What do you think? Topic for an online Community Meeting?
Rachael

Rachael Coultart
04/05/2022 17:19

Hi Paul,
thank you for joining the discussion and adding your thoughts on this important topic. I totally agree with where you’re coming from in the ‘what is it that we are actually teaching?’ and the need to be very clear about what this looks like in actual practice.

As I teach mainly in Early Years and Key Stage 1 I’m most definitely not looking at developing coding skills to this level! I think what I’m most interested in is how we raise awareness and understanding of the implications and applications of AI. I’m keen to ensure the next generation develop their coding skills to this level with a firm grasp of the ethical and moral issues surrounding AI and I believe this starts in the Primary classroom.

But quite what this looks like in the classroom, I’m still not sure, I’m just enjoying the process of thinking and learning about it myself.

Thanks again for your contribution to the discussion,
Rachael

Paul Powell
04/05/2022 19:54

Sorry, I entirely missed the mark there! Perhaps one place to start might be looking at the Turing test. Maybe you could find some child friendly chat bots and see what they think. I have done this in secondary with ELIZA. Also, you could look at doing some pattern recognition work - perhaps something like accepting or rejecting a sequence of symbols with hidden rules - getting them to build up associations. Even better if the symbols actually mean something and the accepting or rejecting does useful work - e.g. identifying if it represents a CVC word? Could code something in scratch for this?

In terms of implications, I wonder if story and imagination might help rather than technology?

Simon Morris
05/05/2022 16:55

On a more practical note, this video is a pretty fun introduction to neural networks using racing cars trying to work out the best line to lap a track at speed.

If the students had access to a software like this, where the students get to choose the winners that seed the next generation and see (in a visually engaging way) how their choices shape the AI, perhaps it might provide a non-technical and fun way of introducing the topic. It would also go some way, I suspect, to demonstrating how AI can go astray if its ‘winning’ conditions are not clear enough.

Please login to post a comment