March 9, 2023

What ChatGPT means for schools

Mike Broadstock talks with education consultant and author, Leon Furze, about the latest tech already disrupting classrooms – AI program ChatGPT – and how educators can make the most of its potential.

Transcript

Note: isPodcast is produced for listening and is designed to be heard. We encourage you to listen to the audio, as it includes emotion and emphasis that’s not on the page. While every care is taken, our transcripts may contain errors.     

Shane Green: 

Hi, everyone and welcome to isPodcast, ISV's show for schools and the wider community. I'm Shane Green. 

One of the biggest challenges schools face is integrating ever-changing technology into the classroom and preparing students for an uncertain future.  

On today's episode, Mike Broadstock talks with education consultant, Leon Furze, about the latest tech already disrupting some of our lives – AI program ChatGPT – and how educators can make the most of its potential. 

 

Michael Broadstock: 

The artificial intelligence program ChatGPT has become a hot topic, especially in education circles. It can give a detailed textual response, one that looks like it was written by a person, to almost any question you might ask. It can even write you an essay.  

Some, in schools and even universities, are concerned that students might use the program to complete written assignments. Recently, the Victorian Education Department banned the use of ChatGPT in government schools while they undertake an assessment of the implications of the software. 

Independent schools are also working through the implications. So what are the risks and what are the opportunities? 

I'm joined by Leon Furze, an education consultant, who's currently undertaking his PhD on the implications of artificial intelligence on writing, instruction and education. Leon, thanks for coming on isPodcast. 

Leon Furze: 

Thanks for inviting me, Michael. 

Michael Broadstock: 

That's usually the point in an article about ChatGPT where the author reveals that the intro you just listened to was actually written by an AI, but not in this case. 

Leon Furze: 

Yeah, this is becoming a bit of a classic trope now, isn't it? 

Michael Broadstock: 

It is. So for listeners who are out of the loop, what is ChatGPT? 

Leon Furze: 

ChatGPT is an app really developed by OpenAI that sits on top of a large language model called GPT-3 or GPT-3.5. So the language model is actually the AI part, and a language model is a type of artificial intelligence, which is essentially an algorithm sitting on top of a lot of text-based data.  

OpenAI's model contains a large chunk of the internet, things like Wikipedia and other publicly available texts, and they train this algorithm to read and respond to written prompts. ChatGPT then sits on top of that model and puts the ‘chat’ interface on there, so that's been further trained to respond like a chatbot. 

Michael Broadstock: 

So, what's the writing like? How good is it? 

Leon Furze: 

Look, the quality of the output really depends on the quality of the input. So, there's an expression pretty common in computer science, garbage in, garbage out, and that certainly holds true for ChatGPT. If you craft a really well written, really succinct, really precise prompt, you can get some very high quality output. 

Michael Broadstock: 

One teacher told me that her colleagues are terrified about the potential impact of ChatGPT. Should they be? 

Leon Furze: 

Probably terrified, no, but cautious, yes. So there's no need to go charging into this. The technology is going to be around now for a long time, and being a little bit cautious and a bit wary of the technology and its implications can be a good thing.  

The phrase that I've been using with educators is being creative but critical. A lot of the immediate concerns have been around cheating and plagiarism, and that's understandable. I think more pertinent to education though is the risk that these technologies come into our classrooms, unbidden, and we lose control of them, so I think that really the risk is that educators don't get involved enough in the process of bringing these technologies into the classroom and into education. 

Michael Broadstock: 

In the era of Google, we've seen a move away from facts-based learning, I suppose, towards an approach that's encouraging critical thinking. But with ChatGPT, we have a tool that not only gathers the information for you but then makes your argument. It seems to me to be a move away from critical thinking. How can ChatGPT add to a child's education? 

Leon Furze: 

Obviously, it's important to note that ChatGPT is not the only AI and that there will be other things very soon like integration into Microsoft Bing search, integration directly into its browser, which is called Edge, integration into all the Office products.  

But I think that tools like ChatGPT and ChatGPT itself need to be seen as creativity tools rather than critical thinking tools at this stage. I think we can apply critical thinking to the output of the tools so we can question the veracity and the accuracy and the bias in the output, but we can also use these tools for very creative brainstorming processes and future integrations with search engines like Bing and Google are going to add maybe a layer of truth or a layer of accountability to be able to bring those critical skills in. 

For now, I think that ChatGPT is a great tool for creativity, ideas, brainstorming, but I wouldn't be using it as a critical tool. I would be critical of the tool but not using it as a tool for encouraging critical thinking in itself. 

Michael Broadstock: 

What sources do they draw their response from? You mentioned Wikipedia, but how do I know when I ask ChatGPT a question that it's not going to some alt-right website to get an answer? 

Leon Furze: 

ChatGPT, I think, I would avoid asking it questions about anything that you want any truth out of because it's particularly prone to doing this thing, which it's called hallucinating in AI terms, so it will just hallucinate. It will just fabricate information and if it doesn't know the answer, it won't say, ‘I don't know the answer’. It will just give you an answer regardless. And there's also no way of checking where that answer came from. 

With these new technologies, so Bard from Google, Bing, or at the moment, they're calling it New Bing  and also new platforms like You.com and Perplexity.ai, they actually provide a reference for where they got the information from. So they will give you that ChatGPT style answer, but they include footnotes often with a little link to where that answer came from so you can manually go down the rabbit hole and check that out. 

I think that aspect of large language models is very similar to the kind of instruction that we've given students for years now around search. Always be critical of the output. Always be questioning the output. It's not flawless. 

And the other thing to remember is that they're trained on human produced data, and humans aren't flawless either. So, it's the same old garbage in, garbage out that we talked about earlier where these models pull their answers from, even when they can provide references, even when they can say, ‘Oh look, this is the Wikipedia article that this came from’." There's still an element of trust or of truth there that we need to be questioning. So we can never take anything that these models produce at face value.  

Michael Broadstock: 

So what are the ethical considerations with regards to ChatGPT? 

Leon Furze: 

There are many of them. One thing I would encourage listeners to check out is Kate Crawford's book from 2021, which is Atlas of AI. It's a fantastic exploration of the ethical considerations in the artificial intelligence industry as a whole.  

There's obvious ones that get discussed in the media such as copyright and privacy, particularly around ChatGPT but also image generation.  

Bias in the data. So, because of the corpus of texts that these models are trained on, there's a lot of bias in there. There's a lot of bad stuff out there on the internet. Part of the training data includes information from Reddit, from Twitter. We've seen in recent years a Twitter bot created by Microsoft, which was responding incredibly racist and sexist remarks because of that training data. So that's a primary ethical concern. 

But then on top of that, we have environmental concerns, so training a model like GPT uses about a thousand times the electricity consumption of one US household in a month.  

Kate Crawford refers to them as extractive technologies in many ways. So, there's lithium mining and an impact on the environment, but they also extract a lot of data from humans. They rely on a massive amount of human labour. We've seen articles recently about ChatGPT outsourcing the labelling of bad data to workers in Africa for $5 a day.  

Huge number of ethical considerations that really are wrapped up in the whole ecosystem of artificial intelligence. And then ChatGPT as an app just sits on top of all of that, and it's a really complex issue for education because fundamentally, these technologies are going to find their way into our classrooms in one form or another. 

So I think educators, we need to grapple with all of those complexities of the ethical considerations. We need to be aware of them and ultimately, our students will probably be the next generation of people dealing with those ethical concerns. 

Michael Broadstock: 

Someone on the radio recently said that ChatGPT was a flash in the pan. Do you think that's so or is it here to stay? 

Leon Furze: 

I think that ChatGPT is a watershed moment for artificial intelligence. It's not particularly anything new. So GPT-2 and GPT-3 previous iterations have been around for a while now. I recall my PhD supervisor, Lucinda McKnight, wrote an article in The Conversation in 2021, talking about AI written essays. ChatGPT is really the thing that's brought it into the mainstream consciousness.  

It's not going to be around forever. It's difficult to see how ChatGPT will remain relevant when that function is built into Microsoft's Edge browser, built into Word, built into PowerPoint, and you've got a little sidebar where you can essentially use ChatGPT style functions in the document that you're working on. 

I think the next iteration of this technology, whether that's GPT-4, the next model or whether it's Google Bard, I think that's really what's going to be interesting to watch. So these are definitely going to come into the kinds of products that we use in the classroom every day. 

Michael Broadstock: 

How can schools prepare students for a career in a world with technology like ChatGPT? 

Leon Furze: 

Flexibility I think is key. So these tools, they don't replace human creativity. They don't replace that human ability to improvise, to make connections, to make social connections, to connect and build on knowledge. All of those things that we're really good at, artificial intelligence at the moment doesn't really do a very good job of any of that. I think they can augment human creativity and they can replace tasks that maybe shouldn't be done by humans anyway. 

So in terms of preparing students for careers in the future, there will be a certain number of careers that technologies like ChatGPT ultimately end up replacing or significantly changing. I think that these students, they're going to go out, they're going to leave school and they'll be the ones driving these technologies, so maybe we need to prepare them for carving their own path there through these new careers. 

Michael Broadstock: 

We've been talking about ChatGPT, which is a text-based AI, but there are other programs like Midjourney and DALL-E 2  and so on, which give image-based responses to your prompts. I've seen some of those and they can be remarkable, but there have been concerns about them drawing on real artists' work, using their intellectual property without attribution. What would you say to a student who's worried about their prospective career being replaced by AI? 

Leon Furze: 

Another one of those tropes says you're not going to be replaced by AI. You'll be replaced by a human using AI. And I think that's an important thing to bear in mind for educators, that teaching students how to not just work with these tools but also how to develop these tools, how to grapple with these tools. 

I've been talking to a guy from Stanford University who's working with students on using really low code tools and no code tools to develop their own AI programs and create their own AI chatbots and assistance. I think we're going to need to equip students with the skills to work with these tools at a variety of levels. 

Coming back to that concern about image generation and the copyright and the stealing essentially of artists' work, I think that this is one of those really big ticket items that we need to be discussing in schools. We need to get young people's opinions on this. We need to co-design a way forward through these technologies.  

The big developers behind these technologies – so OpenAI, Google, Stable Diffusion with image generation, Midjourney – they often go crashing right into these technologies. OpenAI in particular with its launch of ChatGPT really exploded onto the scene, and they've been criticised for maybe the lack of ethical consideration in releasing these into the wild without properly addressing those concerns of the data and the copyright and all of that. 

So I think that one thing that we could really proactively do in education is work with our students in doing it better the next time around. So how do our students become the people who do a better job of these technologies in the next iteration and in future iterations? 

Michael Broadstock: 

Government legislation often trails technological advances, especially in modern times, by several years so it could be a long way down the track before someone suddenly says, ‘Look, it's gone too far. We need to fix it’. 

Leon Furze: 

There's two terms that I keep coming across in my research and then I keep coming back to, and they're a bit of a mouthful, but one is techno-determinism and the other is techno-solutionism. The first, techno-determinism, is this idea that the technology is inevitable and that it's just going to roll over us whether we want it or not, and I think we need to push back against that in education. We need to say like, 'Okay, yes, these technologies are here, and yes, they have the capacity to be very helpful in personalisation, in differentiation, in all of these things, but they are also ethically problematic and we don't just sit back and take that. We actually do something about it’. 

And then the second, techno-solutionism, is that real Silicon Valley approach to technology is going to solve all of the problems of the world. It will solve climate change. It will solve the food problems, food chain supply problems. It will solve everything. We have to acknowledge that AI causes as many problems as it might potentially solve at the moment and that we have to do something about that. So I think that's really the role of education in this piece. It's to push back against that techno-determinism and the techno-solutionism. 

Michael Broadstock: 

I think it was Bill Gates who said, with regards to technology, we've got a better chance of protecting what will happen in 10 years than what will happen in the next couple of years. 

Leon Furze: 

Yes, absolutely. And look, this is why there's a groundswell in education, in Australia in particular but broadly across the world. I think, towards those human skills, those skills of creativity, of ethical consideration, of collaboration, teamwork, all of those. I guess in Australia, we have the general capabilities and there's many, many different frameworks around these types of skills, but they're fundamental skills which can help to prepare students for this future. 

Michael Broadstock: 

You're running a seminar for schools at the end of March for ISV on how teachers can use ChatGPT and AI to enhance their teaching and learning. Can you give us some examples? 

Leon Furze: 

Yeah, absolutely. So my sessions will be aimed at how educators can use the technology for themselves. 

So I'm not talking about using it in the classroom with students. I know that's still problematic because of the terms and conditions with ChatGPT and because of some schools and sectors blocking the technology.  

So, I'm talking about teachers using it to reduce workload, automate certain administrative tasks but also augment their creativity, their planning, their resource creation, and the ability to improvise during lessons. So one thing I'm really interested in is how we can use these technologies on the fly to make those processes a little bit more seamless and a little bit more fluent. 

Michael Broadstock: 

We will link to that in the show description. Thanks very much for joining us on isPodcast, Leon. 

Leon Furze: 

Thanks again for inviting me. It's been a lot of fun talking about this. I could talk for days on this topic and I'm really looking forward to the ISV sessions. 

Shane Green: 

ISV loves celebrating student talent, and one of our favourite ways of doing that is our annual statewide poetry competition. We'll be launching this year's competition on Tuesday, the 21st of March, World Poetry Day. If there's a young poet in your life, please encourage them to take part. It's open to Victorian students of all ages in all schools. We'll put a link in the description. 

We're going to leave you with one of the winning entries from our first poetry competition in 2020. A personal favourite of mine, Olivet Christian College student, Haelie Roberts, and her poem, Hope in a Whistle. 

 

Haelie Roberts: 

The sky was fresh, the landscape too, the dew was on the ground 

The sun was slowly creeping from the east. 

A world of beauty wakening up from slumbering all night round 

Sounds of life heard never to be ceased. 

 

Then high up in the crimson sky, a bird of prey appeared 

Something like an eagle to be seen. 

A whistling kite flew swiftly, though as one meant to be feared 

Soaring from the ranges to ravine. 

 

He circled low above the brush, keenly seeking for 

Dormant prey just waiting to be spied. 

Circling still, he whistled out, a single piercing lure 

‘Teeee-ti-tiiii’ he cried. 

 

Then his beady eye caught sight of movement down below 

He caught the passing breeze to make descent. 

Plunging in a headlong dive he gained the bush plateau 

A spiralling ball of menace on the scent. 

 

Talons wide, his wings now spread he hovered just above 

Then plunged and took his prey in swift surprise. 

His talons closed on furry frame, he caught it by the scruff 

It had no time to utter any cries.  

 

Bird tensed its grip on squirming quest and caught the wind’s updraft 

Triumphantly, he mounted for the crest. 

His cunning eye rove peak to pile, seeking out the craft 

Tucked away inside a homely cleft. 

 

From a distance watchful eyes regarded all ago 

A boy, intent on catching every act. 

Often up before the sun appeared and out to see each show 

He saw the kites’ performance most exact. 

 

His interest was in birds of prey; he studied them with zeal 

His entire life depended on this form. 

It was he alone who gazed on them; observing to the meal 

And wrote and marked down routine to un-norm. 

 

Since he was just a toddling child, he loved to hear them call 

His family shared the interest of his youth. 

But then the day of fate did come and tragedy did fall 

Leaving him alone to face the truth. 

 

His mother gone, his father too, he fought to stay alive 

While uncles, aunts and cousins fussed around. 

They shipped him off to Dad’s aged bro to learn to tend to hives 

Though never giving him a little ground. 

 

He’d found a chance to be himself when Uncle Jesse said 

‘We’ll go a–watching for them birds you like.’ 

They waited still until they saw kites soaring from the head 

Of the rugged cliff tops so alike. 

 

Then he’d heard a high-pitched sound, a whistle so unique 

The fire started in his heart anew. 

‘Teeeee-ti-tiiiiii!’ he had that day heard shrieked 

From the creature fully in his view. 

 

And still he loved to hear them call; they gave him hope afresh 

Telling him the past was now behind. 

He now had thought to look ahead; and one that would refresh 

His sorrow stricken body and his mind. 

 

There was one thing that was hope to him; the call of freedom fire 

Still burned strong and never would go out. 

‘There’s hope in a whistle, a whistle I say,  

a hope that’s still rising higher.’ 

‘You can still live your life, though in sorrow or strife,  

just keep hope in your heart today.’ 

 

Thank you, everyone. 

Natalie Moutafis: 

IsPodcast is brought to you by Independent Schools Victoria. It's produced by Duncan McLean and presented by Shane Green, Michael Broadstock, and me, Natalie Moutafis. Our podcast theme was composed by Duncan. There are transcripts of our show with links to what we've discussed at podcast.iseducation.com.au. Please follow us wherever you get your favourite podcasts. And while you're there, we'd love it if you could rate and review the show so more people, just like you, can find us.