Generative AI P3: Collaborative Inquiry Into AI and Higher Education - The Conclusion

This study was shared on May 4 at Middlesex University Dubai at the 2nd International Conference on Technology, Innovation and Sustainability in Business Management (ICTIS 2023). 

Find more information: https://www.mdx.ac.ae/ictis2023.

AIMS OF STUDY
The aim of this auto-ethnographic study is to provide insight into the lived experiences of higher education faculty in their attempt to design a new educational module with an advertising/marketing specialism making the best use of  Artificial Intelligence (A.I.). This is in light of increasingly prevalent social robots, such as ChatGPT4, Dall-E, Designer etc., which are particularly disruptive within the field of advertising. The acceleration of A.I. applications across various fields of content creation has elevated the importance of a graduate’s self-efficacy in working with A.I. in terms of employability. 

The auto-ethnography will be presented in the form of an evocative narrative and will share feelings, hopes, aesthetic reactions, and moral dispositions.  

As befitting a rigorous institutional auto-ethnography, the study will aim to explore the following research questions: 

• Can AI be used to help develop teaching and learning materials?  
• How should we teach AI to advertising students?  
• What are the most effective and useful programs that infuse AI concepts within an advertising curriculum?  

The authors are both affiliated with Middlesex University Dubai:

• Stephen King, Senior Fellow HEA, Senior Lecturer in Media
• Judhi Prasetyo, PhD Candidate, Senior Lecturer in Computing, Informatics and Engineering

Correspondence via s.king@mdx.ac.ae.

Chapters

• 1:12 Podcast Introduction
• 2:18 Start of podcast

Transcript

Welcome to the third and concluding episode of our podcast on AI in higher education. In this episode, your host - that's me Judhi Prasetyo and Stephen King will share the conclusion of our journey into the use of AI in higher education for course, design. 

Over the past two episodes, we explored the ethics and methodology of autoethnography and shared our ambitions for the study. We also covered the first two weeks of our research journey, where we familiarized ourselves with the literary universe and conducted some initial familiarization experiments. Our efforts culminated in the successful generation of a module narrative. 

Now, in the final episode, we'll be sharing our conclusions, including our perspective on how valuable AI is for this purpose and potential for misuse. 

Overall, this podcast series has been an exciting and thought-provoking journey into the intersection of AI and higher education. We hope that it has inspired you to think critically about the potential of these technologies and the ways in which they can be used to support learning and teaching. Thank you for joining us. And we look forward to sharing our final thought with you . 

Hurrah! We are here. We made it. And so Judhi, so thank you very much again for being with us over the past couple of weeks, we've done this adventure. I feel like you're like Indiana Jones and I'm the young boy as we go through the Temple of Doom. We've dodged boulders we've we've encountered strange new worlds of alien AI.

I think you are Indiana Jones. I am more like watching you, or maybe like those agents on the field. And I'm the one that is behind the desk and keep calling you like: "hey, don't go there. Go here!". 

So let's see, where do we stand? Okay. So there's two things before we start today, because this is the conclusion. I think we should revise the weaknesses of the auto-ethnography for everyone who's listening at home. And maybe the first time they listen. 

Yeah, I think this method is quite controversial in some ways. 

Yeah. Although it is academically rigorous and there are a number of studies using this method, especially over the COVID-19 period. There are weaknesses. The obvious one is that there are very few people involved. The two of us. 

Yeah. 

And sample of two or a sample of one is obviously got to have its own weaknesses. 

Yeah. 

 And on top of that, because we are presenting our own life experiences. There is obviously great risk of bias. And of us trying to make the data look as, as positive as we might want to just for the sake of the fact that we're friends. And we want to make the experiences that we've had, present ourselves in the best light. So there is a certain level of positivity that you might expect from this conversation. And finally the most important thing is that. Whatever we say, this is our experience. 

Yep. 

It is not the experience that everyone will have.

Yeah. 

 It might not be the experience that anybody has apart from us because we are in a very privileged. 

It's very subjective, yeah. 

And we're in a very privileged position to be working in a British branch campus of a large university which has significant equipment and investment in faculty, such as having specialist robotics, faculty like yourselves and having an advertising program like I have. 

Right. 

So with all that online. Should we go on. 

All right. Go ahead.

What most people are wanting to know is how successful were we? Because at the end of the last session I had just completed the module narrative. I was very excited. 

I was banging down everyone's doors.. You remember? I was literally knocking down everyone's doors at this point. Saying, look, I have this transformative piece of software. And it's going to change everything. 

Yeah. 

And I was very lucky that I had to take a breather. We had our discussion, which was very important. So again, the methodology of auto ethnography and the collaborative side of it is the fact that we have someone to hold me back. 

Yeah. It could go in the wrong direction. 

I could've gone too fast. I think I would have gone too fast. I think I'd put an, I think I'd put a number down in an email of how much I thought we could make from selling this course. 

Yeah. It's not wrong, but it's, that's not the time yet. 

I had overestimated what was possible. 

Yeah. 

So my first experiment had been phenomenally successful in my own mind. It had created a module narrative. And I was therefore thinking anything was possible. If it could do this. Then we can start this process of getting validation. And then at the same time, we could be building the content using the AI and we can start marketing it. We can start selling it. And by the time we've got enough. People booked in. 

Yeah. 

 We would be ready. 

Yeah. 

That was the dream. 

Yes. 

So the next week having come back after a breather and having a chill out period. I started to develop the content and I went through some of the slide development, AI that is recommended. There's a number of these images that you get. If you follow the social media and the popular media, they have these little charts, which show which apps are good for different things. 

For example, for audio podcasts, they have Descript, which is what we're using right now. 

Yep. 

So I used the first thing I used was Tome.App. That's T O M E Dot App. And this is promoted by a number of startup entrepreneurs as a way of flexing the fact that they were able to fly all the way across to the U S from Europe and make a presentation. And the presentation they will say, oh, this was all done by AI. Oh, wow. This is wonderful. So I'm going to try it. And the thing with TOME is you just type in the subject of the presentation, and it will come up with supposedly an entire slide deck for you. 

Yeah. You can specify how many slides you want? 

It gives you eight. And one of those is a thank you slide. And one of those is a, is an agenda slide. So it gives you between I, because of the topic I was asking, it was giving me about three or four slides. 

Okay. 

And I try to over eight weeks because the Chat GPT module had given me eight weeks of a syllabus. So I tried to create lectures for each week. And it didn't give me many slides at all. And if you remember, but we went through some of those slides. And one of that, you spotted that it was a mistake. 

Yep. 

And the mistake could only have been spotted by yourself. 

Yeah.

Because it was a very specialist piece of knowledge that you were currently researching yourself. 

Yeah. 

And with the same weaknesses that normal chat GPT has, and the fact that it has only two years or something, it has. It's it finishes at 2021 is database or something as. 

Yeah. And even with that, sometimes they correlate things that don't correlate at all. It doesn't make sense sometimes. 

So this particular case study, I had absolutely no idea it wasn't real. You pointed it out to me. And at which point I was, how would I have even known? 

Yeah. 

The other problem that TOME has was that the slides weren't well-designed. So I was still going to have to play around with the design. There is an option to do branding, if I recall rightly. And I was planning to put the brand guidelines in. But the other problem was. It was just too text heavy. 

Yeah. 

And when you're an educator, you don't want to have all your text on the slides because the students will end up reading them, they are supposed to be engaged with you. 

Yeah. 

 So it wasn't very useful. 

What about the pictures though? 

The pictures were terrible. 

Okay. No that was subjective. But the pictures don't say anything. So again, with the, when you are presenting or when I'm teaching, I like to have images which actually support. 

Yeah. 

And are a lot more. A little bit more literal than what they were presenting. 

Yeah. I think we are not reviewing the software here here, but I think It has limited number of images in its library. So it keeps on using it again and again, 

, I think it's based on Dall-E. So it uses the Dall-E Image generator. And so the images like, were cool. They were aesthetically pleasing. They just weren't educational. 

Yeah. 

 And that's the whole point. You have this huge real estate of a slide, which is full of words. And an image, which is nice, but it's just decorative. It's not informative. 

And which brings you the next one. 

So there was another piece of software that was promoted. This is called SlidesAI.io. And this is a, an extension for Google Slides. 

Oh, okay. 

So this falls into the Androids. AI system. Now the weakness of this particular software or app is that it doesn't work like Chat GPT or it doesn't work like TOME. So you can't put in a prompt. And ask it to just come up with a whole deck for you. 

Okay. 

You have to put in the content and then it will summarize it for you. 

And then it will lay it out across however many slides you want. 

But then you can put some of the content generated by Chat GPT10 or TOME?.

So that's what I did. 

Yeah.

So I took from TOME. And I took the text from TOME, which was a lot. 

Yeah. 

And I put it into SlidesAI.IO now. Then I come up another problem, the SlidesAI.IO only allows 500 characters. On the basic model. 

The free one, you mean?

On the free one. And when you have so much text from TOME, that was like one or two slides. So I'd have to keep going backwards and forwards. 

Yeah. 

Plus. For the SlidesAI.io it was taking two minutes to process everything every single time. 

Yeah. 

And so I was taking the texts across two minutes to come back to copy paste. Copy. It was taking a long time. So eventually I paid for the premium version. 

There you go. 

And it allowed up to two and a half thousand characters. 

Okay. 

Which is a lot more useful. And I was able to dump the whole TOME deck across. And it would come up with a slide deck. The problem here. Was that the text on the slide was too little. 

Okay. 

So it's the Goldilocks scenario. 

Yeah, 

TOME gave us too much.

Yeah. 

Then slides.ai gives us too little. And it left too much up to you to present. 

Yeah. 

And because I don't know too much about the topic. The AI were not helping me

Is there an option to fine tune that? 

Yeah. So, then the option to find you would be to do it as a human being. And to actually research it yourself because at the end of the day 

 The slides needed to be laid out properly still. You had to verify all the data still. And you had to be able to know what it is you're talking about. So then it defeats the whole purpose. 

Yeah. 

Of asking the software. To develop the slide deck. If you're having to go back and go through it all. Be cause that's the time consuming piece. 

So the idea of just pressing a single button and computer will do everything for you. 

No,

Does not apply here?. 

No, not at all. No. 

Do you regret paying that much?. 

Yes, I actually canceled that subscription. But it w it was a, it had to be done. I had to just test it for the sake of this would have been really helpful. 

Okay. 

And I was willing to take a chance for it. So the slides didn't work. I had one last chance. One last thing. We discovered. Agent GPT. Now these agents, I don't know whether you might be able to explain it better than me, but these appear to be AI programs which develop and generate their own prompts? 

It's an intermediary AI program to bridge from human, who have challenge in writing a prompt. Because for example with Chat GPT you have to be. Specific in order to get a specific result. Just like Google search you need to be specific in the keywords that you entered. Some people are not really capable or probably because of the language barrier. They have this challenge. The idea of having this Agent GPT is just for specifying keywords, and then it will generate the prompt, which then can be fed into Chat GPT for a more elaborate result. 

Yeah. And so that's effective. That's exactly what I experienced. But it does, it goes beyond that. So you give it a task and it generates a series of prompts, which it believes are related to that tasks. And then it processes through them all one after one generating new prompts as it goes. 

Yeah. 

And so I use this to develop workshop activities. And that was actually quite helpful. I would just say, develop me a workshop related to chatbots. 

Yeah. 

And it would come up with two activities, which would be awesome. Again, the content is not there. Top line. This is what you need to do. Introduce it this level. 

Let them build. This step one, step two, step three, step four. There's no real content. Although it does give a small introduction. script, which would last about 20 seconds. But it was, it's helpful to create a top-line lesson plan for a workshop. 

Okay. 

I was able to produce five of those. And. Out of those five, three. See the Agent GPT only wi ll process four or five prompts and then it stops. Because they don't want it to keep running indefinitely. And I think that might cause it. It was, that would be a problem. 

No, hopefully with five example. Five prompts. We can use our own intelligence. 

I was able to. So it stops after four or five. And so at the end, it will say, look, sorry, we're not going to give you any more answers at this point. And you have the final the final task of the final activity. And for the final activity I put as homework for the following week. Which meant I actually had eight weeks worth of workshop activities, which is great. 

Great. 

So at this point I had got a module narrative or had a couple of module narratives, which were recommended to us from traditional literature. And we had tested a variety of tools, which had been recommended by social media influencers, and experts that a re out there on, on the social media channels, whether it's LinkedIn or on podcasts. I have developed eight, very rudimentary sets of slides. Which provides some kind of structure, but ultimately I think when you try to rewrite someone else's essay or someone else's work. 

Yeah. 

It's more effort than if you did it yourself? 

Yes. 

That's what I see there. And I have these workshop activities, which I think will be quite fun. They're require development. But the structure is there. And I think that's really helpful. It gives me direction. And that's what, that's how I've, that's how I've we've concluded. So it's not phenomenally successful. 

Yeah. 

In the sense that it doesn't give everything like what the TikTok people are saying. 

Yeah. 

But it is. A little bit helpful. Yeah. Which is, I think is probably what you expected. 

In case someone listen to this podcast way after we recorded it. This is recorded in April and May 2023. 

Correct. 

It's not because of just us, but also the limitation of what is avail able right now. 

It could be. Or it could be me. Cause at the end of the day, this is my first effort. I'm pursuing this in the same way. As the social media influencers are recommending it. I'm the target audience for these people. Once I've gone through more discussion with yourself and exploration. I'm so much more sophisticated. My usage now. 

It's always user-centric. So imagine when the first phone came up, probably the first mobile phone came up. Smartphone probably only a geek can use that. The first computer came up and then evolves even now. Not just the little kid - a chimpanzee can use an iPad for example. There's a lot of development and improvement there. So similarly with these aI tools.

 What do you think is, have you found interesting from this journey? 

 It is interesting to see a technology used by a person from non-technology domain to produce something that supposedly can help them to do their job. So for me, my job is in technology. So using technology to produce something related technology is related. For me personally is interesting. 

Did I do anything which you were not expecting, is there anything which you think? 

Nope. Actually, because most of the activities that you have done I have done it myself and also the state of mind. And then the feeling that you had also I have experienced that before, so kind of expect it. 

Yeah.

Probably I give you some tips, for example, how to write the prompt, how to do a follow-up prompt. 

Yeah. 

Those things. Probably you will discover it, even if I don't tell you, but it probably takes a longer time. 

Yeah. Okay. Good. We talked about the emotional state. And I do think that the Gartner's hype cycle. Which I think is a famous image. I think anyone who's covered technology or innovation will know what this image looks like. It's like a. It's what is it like a up and down curve. And it starts off with the innovation trigger, which is all of the, I suppose the TikTok people and the Chat GPT launch and everyone's conversation about it. 

Yeah. 

And then I drove up that through all the experiments I had with the generative text Chat GPT experiments or getting it to create my social media. 

All right.

And then trying to twin myself. 

Yeah. 

With the investment in descript. And then ultimately trying to use it to do my work or take on a big chunk of my work. 

Yeah. 

I think that was when we reached right at the peak of inflated expectations. 

Yeah. 

And at that point, I crashed. All right. I remember we talked, I tried to use it with a football. I just perhaps tried to have a I tried to have a proper relationship with the software. 

Yeah.

Because it was talking to me normally. And I'd start earlier. I had started using please. And I tried to think of it as a, in a human kind of a way. 

Yes. 

And then it just rejected me. And so that sent me on the trough of disillusionment. And then as I reached the bottom. I was using the slides. 

Yeah.

And the slides were not working. And I was so depressed that I spent money to try to rectify the situation. 

Yeah. 

I threw money at it saying please, where's my AI friend gone to do all my work for me. 

Yeah. 

And then the slope of enlightenment is through our discussions that we've had. 

Yeah. 

We're our links to the Hitchhiker's Guide to the Galaxy. 

Where are you adjust your expectations here? 

Yeah, the use of agent GPT to assist I've become a lot less fearful. The anxiety is no longer there. Although when I'm in a restaurant and a robot does come up to me, I have gone up to the waiter and said: "Hey why are you even bothering? Is it saving you that much time?" I think it was a bit of a shock. I, now I, you know how much time I use it? 

How much.

I think I've used it once. Since the conference. That was just last Thursday. So once in five days, and that was this morning. 

Yeah. 

And that was just purely to to make a piece of my writing less interesting. 

All right. 

So more formal. So I had my style, my personal style is, tends to be a little bit more salesy, maybe a bit more promotion. I've had more promotionally and I needed something which was a lot more conservative. 

Yeah. 

So you use the chat GPT to give me an option, which is more formal, but that is it. I have not even opened any of these Mid Journeys or similar. I haven't even thought about it. 

What about other tools? 

I haven't, I just I've gone back to my own writing. 

Okay. 

Although there is an element that it can save your productivity. 

Yeah. 

I just flipping between screens. 

Yeah. 

Copying and pasting. And my typing speed is already quite good. And my writing style I've gone back to being like, okay, find this is the same savings is a. 

I'm even appreciating spelling mistakes in my work now. 

That's the work of AI as well. It is a spell-checker and grammar checker..

But I'm appreciating that I have spelling mistakes in it. Yeah. And people saying, this is Hume. I can. No, this is proper human content. Yeah. Because it has an error. 

I know some people who purposely made that the typos itself. To say that, Hey, it's me. 

Yeah. That's my new excuse here, if there's any mistakes in there. So this is deliberate. It's human generated it's my real voice. 

Okay. Everyone's experience will be different. 

True. 

I think the value is in the journey itself. When we discover something we would be able to utilize things better if we discovered ourselves. Yes. Take it for granted. We do whatever it is taught to us. But during the implementation or during practicing something we may discover something that is not taught before. And that's valuable stuff and this is exactly what you have I could share with you my experience like what we share with the audience right now our experience But you have to experience it yourself in order to discover something that we don't discover or maybe it's only applicable for you.

Very wise! Go forth And Experiment with the ai Unafraid. Thank you all very much for following us on this limited series And I hope we will be back with more in the future as we explore more of web three and its intersection with higher education I'm Stephen King 

And I'm Judhi. 

Bye for now.