SPEAKER_00 0:00

Hello.

SPEAKER_01 0:01

Hello.

SPEAKER_00 0:02

Thank you for participating in this episode. So would you happy to share your current role and the area you work or research in?

SPEAKER_01 0:09

Yeah, I'm a professor of educational technologies and media. And so mainly I'm working around common technologies, but also obviously the brand new ones, artificial intelligence and immersive media, so virtual reality, mixed realities.

SPEAKER_00 0:23

Yeah. When you hear the word AI, what's the first image or feeling that comes to your mind?

SPEAKER_01 0:31

A giant breadth of things. AI is used for just about everything at the moment. So where things might have been like normal machine learning or algorithms, AI is kind of being used in that terminology as well. So I guess artificial intelligence is for me is something which produces a output that is human-like in its um quality and um the way that it's the the tone is there as well. So it sounds like a human, behaves like a human, does the work that a human would do, um, but it's not got any humans involved at all. At least not at the tail end.

SPEAKER_00 1:11

Yeah. Do you use AI for like for your everyday tasks?

SPEAKER_01 1:15

Uh I use it for um a lot of things. I probably would say I use it every day, um, but I don't use it for common things like emails and stuff like that. I when I get an email from someone who's clearly used AI, it just bugs me. And I don't want to pass that on to anyone else. And I don't send out that many of them that it's really worth it. Um but for something like uh drafting a report or something like that, um then I'd definitely use it for those sorts of things.

SPEAKER_00 1:41

What would you say if I ask um the the everyday task, you would never give up using AI?

SPEAKER_01 1:47

I wouldn't say I'd never give it up. I just feel that um where you want to have personal contact with another human, that I'd rather not be re using AI to write an email. Um because someone can just respond with an AI, and then essentially you're just having a communication between two um, you know, artificial entities. Yeah. And the humans actually aren't involved in all that at all. And if that's the case, that's fine. But don't pretend you're trying to do something which requires humans involved in it.

SPEAKER_00 2:16

Mm-hmm. Yeah. So this question will be a little bit general, apart from AI, but uh what's the sustainability habit you've picked up recently?

SPEAKER_01 2:26

Sustainability habit that I've picked up recently.

SPEAKER_00 2:28

Being small, like anything.

SPEAKER_01 2:30

In general? Um or related specifically to AI?

SPEAKER_00 2:34

Like in general.

SPEAKER_01 2:35

Okay. In general, I think I've always been um working in that sort of sustainable mindset. So even as a kid, um we would pick up bottles and cans and get them recycled. Uh I guess one thing that's been that's that's become a bit more common is that if I go out for a walk and I see cans, I've actually got a bag now to pick up that stuff and then I can just take it back and put it into the recycling. So there's a little thing that I do there.

SPEAKER_00 3:02

Yeah. What about uh if what would you just say um for AI specific?

SPEAKER_01 3:08

AI specific, use it when you need to, um, and not any other time. I think one of the things that really bugs me about AI is when you're doing a Google search and AI just spits its garbage out at the front. Um and it's and it's I didn't ask right and I want to turn it off and I can't. So that's that's one of the things that irritates me around that.

SPEAKER_00 3:26

Oh yeah. So what excites you about AI or worry is most?

SPEAKER_01 3:32

Well, they're two very different questions. Yeah. Uh I guess the things that AI is genuinely going to improve people's efficiency, uh, where they have the expertise to start with. So from my perspective, if I'm going to uh write a a report about or an introduction to maybe a a session in in class, I could use AI to just quickly whip up something like that. And then I could fine-tune it in harmony with it. Yeah. Um, and so that speeds things up. I use it for generating role plays. They take hours and hours to build, but AI can build them in a few minutes. Uh and once again, it's relatively quick to fine-tune those sorts of things. Um the worrying part, of course, is that um I've got the skill set and the expertise to recognise what's good and what's bad. Uh, and I can make very quick changes to things. So the efficiency comes from my interaction with something which is providing me with cues, essentially. Um, I would never really just cut and paste stuff and just put it in there because it's by and large, more than half of it is just not appropriate for what I want to use. But it's a great starting point for me. And it's uh it's a good thing for just fixing up grammar, um, which you know it does a better job than I will ever do. Yeah. It knows the rules. But if I don't have the critical analysis skills, if I don't have the foundational skills to actually assess whether this output's worth anything, then I think that's hugely problematic. And that's what worries me probably most of all. Um so it's that it's that learning that comes from a and the use of something from a tool which is renowned for making crazy mistakes.

SPEAKER_00 5:05

Yeah. Is this something also our people in your field talk about whilst sleep?

SPEAKER_01 5:11

All the time. Yeah. Um when AI became big on the scene, what, five years ago, maybe a bit less, um, everyone was talking about, oh, my assessments will have to change. Um, how do we make sure that students don't cheat on their assessments? Um there was quotation marks there around the cheat. Uh is not a great terminology because I think students don't intend to cheat. Students tend to get pressured into situations where they make bad decisions. And one of those bad decisions is now really easy to to actually activate, which is AI.

SPEAKER_00 5:44

Yeah.

SPEAKER_01 5:45

Uh and unfortunately a lot of the students are very poor at using AI, and so they produce work which is clearly identifiable and clearly just not up to scratch in terms of quality. So it's problematic.

SPEAKER_00 5:55

Yeah.

SPEAKER_01 5:56

Um, what was the question again?

SPEAKER_00 5:58

Um is it is it something that people in your field want to do? Yes.

SPEAKER_01 6:02

So the the the discussion progressed relatively quickly from assessment to learning, and it's about assurance of learning. So, from my perspective, I don't care what tools people use to learn, as long as they actually do learn. If they're using tools like AI to just produce work and then they submit it and they get a pass mark, they probably haven't learned anything. And from a university perspective, that's really problematic because we basically assure um our accrediting bodies and the community that when we pass a student, they've actually got some skill sets to demonstrate that. Yes. So that's the big problem for us, is assurance of learning. So that's going to require modifications to assessment. It's going to require modifications to institutional approaches to things. It will probably require modifications to the way we teach. Um that's where the conversation is at the moment.

SPEAKER_00 6:59

Okay, yeah.

SPEAKER_01 7:00

And it's a much better place because honestly, that conversation's well overdue. I think our assessment regime has been a little bit ordinary for the last 50 or so years. And we've taken a pro and we've been able to get away with it because the inferences we make from assessment tasks have been relatively robust. We know that some students have gone to you know paper mills and things like that, and they produce work that's not their own. They're often much easier to catch, and they certainly are much easier to catch than AI will be in the next couple of years.

SPEAKER_00 7:29

So how do they explain AI to non-experts and what do they usually get wrong?

SPEAKER_01 7:36

Oh, I don't usually explain it much. I think people have an innate view of what AI is, right? Um they've either seen it in science fiction movies and and they're they're you know they're quite quite um diverse in what they cover. So, you know, if you want to talk about AI as a dystopian endpoint, essentially, Terminator, all those Terminator movies where the world is basically ended by machines taking over, everyone kind of knows that sort of stuff. So they've got that side of things. They also know that they get playlists from Spotify and Netflix and stuff like that. They know that Facebook is watching everything they do and recording stuff and feeding them marketing and stuff like that. And it's not just Facebook, of course, it's all companies that have any sort of opportunity to um to leap into our spaces and see what we're doing, they will do so and then they'll push out the appropriate stuff to fine-tune their marketing. Nothing intrinsically wrong with that. Um so they're familiar with all of that stuff as well. So underneath it all, it's like um people will have a bit of a play with it and they'll make some images up. And a lot of people outside of academia and corporate spaces, they'll use it to make pictures of things that are amusing, um, things where they've got family members in there that weren't actually present at certain events. Um, they use it to generate some texts, and and it's pretty straightforward. Like the interface for you know ChatGPT, for example, is really straightforward to use. Once you've actually downloaded it and put in one question, you know exactly what it's going to do. So where people haven't seen that sort of stuff, I try and think of it as a um a personal assistant for everyday life. You can you can try and describe it in that way. Um for students, professor in your pocket is kind of the sort of thing that I talk about. Yeah. Uh and it's it's more than a professor in your pocket, it's literally the entire faculty is sort of sitting in there as well. Um, because you've got access to chunk just about everything. Uh just depends on how well it's been trained in the AI interest that you're using.

SPEAKER_00 9:30

Yeah. Um, have you come across any misunderstanding about AI?

SPEAKER_01 9:35

Misunderstanding. Um I think so. I think I think there's a lot of misunderstanding because what I've just described to you there is quite superficial in terms of what it does.

SPEAKER_00 9:44

Yeah.

SPEAKER_01 9:44

And honestly, that's the general public. That's all they really want to know is what does this thing do? Can I use it? If I can, I will. If I can't, I won't. End of story. Um so they don't really understand how powerful it is. They don't understand the impact that it's probably going to be making on the workforce very soon. They don't understand the environmental issues associated with it. Um and no one's really communicating that to them. Like the big AI companies are not really concerned about any of these issues. They're concerned about people getting their product and make and paying for the use of their product. They're a business. Uh, and of course, they people probably don't also understand that AI is probably in most of the apps that they use now on their phones, uh, and certainly on the apps that they would use on their laptops and desktop machines. So there's a there's a lack of understanding about the breadth and probably how far this is going to go ahead. At the moment, it's a bit of a novelty. Um, it's a faster way to do a search, perhaps. Um, but um when it starts impacting people's jobs, I think there will definitely be a a greater understanding required. And I think there's a there's a general obligation, I think, from the government, from universities to try and help the community understand this a little bit better. We're not very we're not doing a very good job of it at the moment.

SPEAKER_00 10:56

Mm-hmm. Okay. Um so what skills do you think people, like especially students, will need in an AI-shaped future?

SPEAKER_01 11:05

Uh number one, when to use it and when not to use it. Um we can guide them there. Like and we can give them the re the reasons behind it. It's not a case of like we don't want you to cheat. It's a case of like, if we do it this way, your learning is going to be better. If you do it with AI, your learning is going to be compromised, and we don't want that because it's going to be a cascading effect. So we can communicate that better to students. Um, we need to teach them basic skills on how to prompt. It's prompting's still a thing, it won't be for you know forever, but for the moment, prompting skills are kind of useful. Uh and I don't know if you've you've got kids or or something like that. Have you ever seen a kid or a student doing a Google search even before AI was there, they really don't know how to use those tools very well. They search weirdly. Um, and so they can't find anything. And I'll go in there and just five seconds later I found a whole bunch of stuff because I know how to use the tools, I understand the subtleties that are required to try and get the information out the most efficiently as possible. So students don't have those skills. They don't, um, they also won't understand the breadth, they won't understand how it's got to impact on their future careers. So we need to get career people coming into the university and saying, this is what it's going to look like, this is what it currently looks like, but this is where we're heading. And we need you to have these skills. Industry's clamoring for our students, our graduates to have some sort of AI skills. They don't want to see them come out there with luffy uh in that space. So we've got to push that um in our courses. And it can't be a case of like we don't want you using it all the time. It's got to be it's got to be a metered sort of approach to things.

SPEAKER_00 12:37

Yeah. Do do you think our education systems now are preparing students for done?

SPEAKER_01 12:44

Everything's there. Like the foundations for all of this stuff is there, the curriculum design skills are there. Um AI's just been so quick. Uh and and universities are slow to respond, they're low-risk entities. So they don't often take punts at these sorts of things. Um they're stepping up now. That's that's very clear. Um whether they can do it quickly enough, I guess it's just wait and see. Uh, I think what we're seeing at the moment, um, and there's a there's a thing called the Castlereagh Declaration, um, which is worth having a look up on the on the internet, castlerey.ai, I think is the U URL. And it's all about getting us ready for what's what's happening. It's it's basically a call to action. Like we've been sitting here for the last few years, not doing very much, not being proactive about these things. Now we have to be. Otherwise, we are definitely going to get left behind. We're already being a little bit left behind, but uh it's hard to catch up, actually.

SPEAKER_00 13:40

Yeah.

SPEAKER_01 13:40

It's incredibly hard to catch up, and especially when it's moving so quickly. What it what it could do a year ago and what it can do now, and vastly different things. Yeah, exactly. Yeah. And it's beginning to govern a lot of the things in our lives. Um, if you go to the doctors, the GP is probably taking notes with an AI assisted tool.

SPEAKER_00 13:55

Yeah.

SPEAKER_01 13:56

When you're getting imagery done, um x-rays and things like that, there may well be an AI in the background that's actually providing another opinion to the human that's actually involved in that. So there's already these spaces. Um businesses are using it for predicting um stock um rollouts. Um, you know, that whole supply and demand chain is is being um seriously attacked with with AI because it can actually provide really meaningful efficiencies.

SPEAKER_00 14:22

Yeah, I agree. Yeah. Um one thing that's coming up more and more in AI conversations is sustainability. So uh people are starting to talk about more. Um about more about AI's energy use and environmental impact. So what's your perspective on that?

SPEAKER_01 14:40

It's a problem. Um without a doubt, it's a problem. It's I don't believe it's an unresolvable problem, but it's it's definitely something that we the general public and governments need to start paying a bit more attention to because industry is going to be requiring a lot of resources to keep these these AI tools going. So if any individual query with a with an AI tool doesn't really use that much power, like running the mic away for a second or something like that.

SPEAKER_00 15:08

Yeah.

SPEAKER_01 15:08

Even even creating an image, which you might think is much harder and might consume more energy, actually consumes less than a lot of text queries. Video is much bigger, so that's harder. Training these tools consumes massive amounts of power, like um in the in the high gigawatts to terawatts of power. And we're looking now at some of these AI engines on an annual basis, using as much as cities and and even small countries. So we're getting to the stage where in the next few years that we're looking at hundreds of terawatts of power being required to run data centers which are predominantly being used for AI as as time progresses. So that that that change is happening. So these data centers, they're just sitting there, they consume massive amounts of energy on their own, but they also require massive amounts of energy to keep them cool.

SPEAKER_00 16:03

Yeah.

SPEAKER_01 16:04

And they require resources such as water or other cooling agents to keep that cool. Which is fine, except a lot of these data centers are built near cities and or they're on waterways and things like that where the loss of water in a local area can be hugely problematic for people, anyone downstream. So you've got a water loss issue, you've got an energy issue, and it's it's so acute. Um Meta has just secured, well, beginning of the year, secured a couple of nuclear power plants to provide them with gigawatts worth of energy. Um Microsoft has recommissioned Three Mile Island, which is a famous nuclear power plant in uh in the US for that sort of purpose. And that's good because nuclear power is essentially clean. Um it's um it obviously has issues, but it is better than using coal-fired areas um and and gas and and other things which are, we know, are major issues for us in terms of the environment. And of course, the US is predominantly runny on coal and gas and fuel. Uh and data centers are often in places where those resources are actually expensive. Um, and they so you might actually end up consuming more of those fossil fuels in those spaces where data centers actually are than you might in other spaces. But on the positive side, um, you can improve the algorithms so they use less energy. You can improve the training processes so that it uses less energy, or you're much more efficient with your queries when they come from that. We can definitely improve how we create our energy and where we put it. So my real concerns around this is probably the water and the local effects that it has on communities. Because a lot of these data centers will consume water that can't be replaced. Tell me. Uh and yeah, I mean, it doesn't get destroyed, it evaporates and goes into the atmosphere, but it won't land where it has been landing in the past. So there's problems for farms, just for how housing water supplies and things like that, potentially on the horizon and maybe not that far away. Yeah, they're building a lot of these things.

SPEAKER_00 18:08

Yeah. We're we're talking about huge impacts actually here, but uh what should be the most responsible? Who should be most responsible for AI use, for reducing the AI use? Like governments, users, or like um companies? What do you think?

SPEAKER_01 18:24

The companies will push as much as they can back onto either governments or individual users. So any algorithms they can build where the AI power is being used by the consumer and they're paying for it, something that they would target. Um from my perspective, it's the companies. Um I know, yeah, we're talking lots and lots of money that's that's going to come out of these things in the future. It's it's actually is a lot of RD happening at the moment, which is quite costing companies quite a bit of money. But at some point, they're going to monetize all of this stuff and it's going to produce massive amounts of money for them. And so therefore, the responsibility lies with governments to basically make sure that they're accountable for it. And governments are pretty bad at this because they would rather have, I wouldn't say they would rather have, I'll reword that. Governments see opportunities when someone says, I'd love to build this gigantic data center in your area, it will produce a lot of jobs, we can give you this much revenue, um, there's probably some lobbying that happens around that and and other benefits that come to the country or the government. And the government has just got to be able to resist that push and consider all of the things that are actually necessary to make sure that it runs efficiently. So if part of it is like, well, yes, we could we're happy to have this data center here, but given it's Australia example, uh then potentially we want you to build a solar farm that's also going to go next to it. It's going to be put out um in other spaces, and you kind of have to think about how you're going to call this because you're not going to compromise our water supply. And we've already seen that agriculture consumes a lot of the water and it sometimes hasn't been negotiated as well as it could have been. And I think we've just got to get up front with this stuff. Uh otherwise it's going to um it'll take us by surprise when towns start running out of water and complaining.

SPEAKER_00 20:12

Yeah, yeah. So um would you pay more for uh for a greener AI service?

SPEAKER_01 20:20

It's a million-dollar question, isn't it? Uh and people have that option with flights, with their with their energy costs.

SPEAKER_00 20:27

Yeah.

SPEAKER_01 20:27

The people that have the money will do that. Um more often than not, I suspect. People that don't have the money can't afford to consider the luxuries of that. So the government's got to force it. Has to be forced. I mean, I don't know why in some countries where there is an abundance of wind energy and solar energy, why it's even an option that you choose a green energy source. Um in other countries it's much harder, of course.

SPEAKER_00 20:50

Yeah, that's right.

SPEAKER_01 20:52

So once again, I think people and companies need a little bit of support from the governments, and it's it's not people, governments telling people what they have to do, but it is actually governments being responsible about looking after the future.

SPEAKER_00 21:06

Yeah, I agree. Yeah. Um one last question. Okay about educational institutions, back to that, back to education. So, what role should educational institutions play in modelling or teaching sustainable AI use to students?

SPEAKER_01 21:21

The same. Yeah, absolutely the same. So it's unreasonable to think that students are going to pay for the costs of um the courses they're doing. And the more AI we put into our courses, the more we have got to consider the cost of actually running these tools. Yeah. So um my university has been particularly good in that space. Um, they've been supportive of um essentially growing trees to offset the costs of the um of the AI use anticipated in courses. So this is predictive maintenance, if you will, of the of the environment. Um and I would suggest that all. All universities should basically say we're going to work on a carbon neutral approach to this and make the best assessment they possibly can of how much energy they can chewy. That's not always easy because the the businesses don't release enough information to tell you exactly how much energy is being used, so you've got to make your best guess. Universities are full of smart people, they can do that.

SPEAKER_00 22:21

Yeah, agree. Yeah, it it's such a great conversation with you. Thank you for joining us today. You're very welcome. I think the listeners are going to take a lot from this conversation. So thanks again.

SPEAKER_01 22:34

Okay. Pleasure.