With countries and tech stakeholders rapidly moving to claim AI in their interest, such as the US government’s move to restrict non-citizens’ access to platforms like Claude, what does it mean for Australians to have control over their technological future? AI is becoming closely linked to Australia’s economy, security, and everyday life. It’s time to discuss what having AI Agency looks like now and into the future.
In this episode of Tech Mirror, host Johanna Weaver is joined by Tech Policy Design Institute co-founder and lead researcher Zoe Hawkins to unpack their latest report Expanding AI Sovereignty to AI Agency. Together, they explore why Australia needs to move beyond simplistic debates to uncover how Australia’s tech movement holds strategic advantages that build both leverage and resilience when global events expose geopolitical uncertainty.
Links
Anthropic to disable its most advanced AI models after US order limiting foreign access | The Guardian
Tech Policy Design Institute – AI Agency Report
https://techpolicy.au/aiagency
Tech Policy Design Institute
https://techpolicy.au
Australia’s National AI Plan
https://www.industry.gov.au/publications/national-ai-capability-plan
Organisation for Economic Co-operation and Development (OECD) – AI
https://oecd.ai
Anthropic
https://www.anthropic.com
For transcript and full show notes visit techpolicy.au/podcast
Transcript
Johanna Weaver: [00:00:00] The Tech Policy Design Institute acknowledges and pays our respects to all First Nations people. We recognize and celebrate that among many things, Indigenous people were Australia’s first tech innovators.
Zoe Hawkins: It’s a fascinating time to be, um, having this conversation. The AI sovereignty debate has really gotten trapped in this false binary of total dependence or total self-reliance, and it’s just that Australia is actually in a way stronger position than, than most people might think.
Um, but that we really need to engage, understand with the evidence so that we can work out where we have leverage, where we need to build sovereign capability, and where we need to fill critical gaps
Johanna Weaver: Welcome to another episode of Tech Mirror. My name’s Johanna Weaver, and I’m your host. And this is the podcast where we talk about how technology is shaping our world, and how we, the humans, [00:01:00] can shape technology back. And this week we’re going to be talking all about artificial intelligence, because last week TPDi released a report called Expanding AI Sovereignty to AI Agency.
And coincidentally, in a bit of a gift from the US government, at the same time, the US placed export restrictions on some of the most powerful AI models built by US company, Anthropic. So they placed restrictions on Mythos-5 and Fable-5 and said, “You can’t export these to anyone other than US citizens.” We could not have scripted better a geopolitical move that would highlight just the importance of having a nuanced conversation around AI sovereignty and AI agency.
And, you know, TPDi spent the last 10 months having, um, discussions about this, really digging into it, developing a new [00:02:00] methodology, which we call the AI Agency Tool, which helps us to be able to actually assess how much sovereignty does Australia have, how much agency to control the future of artificial intelligence in our c- in our country do we really have.
And I am really excited to have, uh, Zoe Hawkins, my co-founder here at the Tech Policy Design Institute, to unpack this report. Zoe was the lead author of this and has led the team building out what is really a world-leading methodology. So Zoe, thank you so much for joining us today.
Zoe Hawkins: Thanks, Johanna. Good to be back.
Johanna Weaver: I’m gonna start with a really softball easy question for you. Hmm. Uh, i- what is AI agency?
Zoe Hawkins: Well, Johanna, AI agency is a country’s ability to shape its AI future. What does that really mean? It’s talking about the capacity to steer outcomes. It’s talking about, can we promote and protect our national interests, and can we capture value in what is a [00:03:00] globally connected technology ecosystem.
Johanna Weaver: Zoe, why… What was wrong with sovereignty? Why did we, why do we need this concept of agency, and how do the two interrelate?
Zoe Hawkins: Um, AI sovereignty and the concept of sovereignty itself remain super important, uh- Mm … generally, but also in this, uh, AI agency tool. It’s actually a core component. And what we’ve done is take what can sometimes be reduced to an unhelpful binary of, you know, you have sovereignty or you don’t, you know, you have complete dependence, um, or you have complete self-reliance.
We’ve broken that down into what we think is a more useful sovereignty spectrum, which is acknowledges- Mm … the kind of gray spaces in and amongst that. It’s talking about, you know, in a reality where the AI ecosystem is so interconnected, uh, no country can be totally self-reliant, not even the US and China.
Um, and so what does policy look like in that space? It’s actually acknowledging that [00:04:00] sometimes you’re gonna want access to world-leading capabilities, even if they’re not your own. So do you have international access? You definitely need sovereign capabilities, particularly in critical areas, so do you have sovereign control in, in the areas that you need it?
Um, do you actually have enough of both of those, international access and sovereign control, so that you have really resilient choice and supply despite disruptions and shocks like the one we’ll be speaking about today with Mythos? In best case scenario, do you actually have international export leverage?
Do you have so much of a, of, uh, a world-leading capability that other people rely on you? And so really it’s not just about if you control every capability, it’s actually about whether across that whole sovereignty spectrum, do you have enough agency to shape the outcomes in an international context? Mm.
And I know that people find that a little, uh, it can sound a little abstract, but you’ve got a really useful analogy, um, Johanna, that really brings that into, um, a different case study. So do you wanna share that with our listeners?
Johanna Weaver: [00:05:00] Yeah, so I think people, when they first hear that description you’ve given, they kind of balk and think, “Oh, maybe that’s overly complicated.”
And what we use this analogy to help highlight is that this is exactly what leaders are doing globally all the time, every day in a complex, globally connected world. So let’s use the example, totally unrelated to AI, of the current petrol crisis that we have, or the global shortage because of the closure of the Strait of Hormuz, which, you know, may or may not be opening with tolls or without tolls anytime soon.
But in that situation, if we look at where Australia sits on the sovereignty spectrum when it comes to access to, for example, diesel, we have some international access to diesel. Some of that is from countries that are strong allies, and we can rely on that access. Others is a little bit more unreliable because it’s possible that they will divert the access elsewhere.
So we’ve got access to some [00:06:00] diesel. When it comes to sovereign control, we have two refineries in Australia, when one of them isn’t burning. And so we actually have quite limited sovereign control when it comes to diesel, which means when it comes to resilient choice, we don’t actually have a lot of choice when we have domestic supply chain shocks, as we saw with the fire, uh, in the reserve– in the, uh, refinery.
But we also don’t have a lot of resilience when it came to the global supply chain sh-shocks that we saw. And we certainly don’t have any excess capacity or export resilience. But what we do have is a lot of gas. And so what you’ve seen is the prime minister traveling a few months ago to Singapore and to Brunei saying to those countries, “Look, we will guarantee you a supply of gas if you guarantee us a supply of diesel,” or perhaps it was aviation fuel.
And so you’re seeing a capability in a different, a different capability being used as [00:07:00] leverage in order to shore up another capability. And that’s effectively what we’re looking to do with this agency tool, the AI agency tool, is to say, “How do we identify, zoom out, look at the whole chessboard, and understand where is Australia’s competitive advantages?
What are the things that we can use as leverage whilst we’re also shoring up international access, building that domestic capability, and having that resilience?” And so it’s really having that big picture of what actually is the full chessboard when we’re looking at those AI capabilities. And so, Zoe, that kind of leads us to the next obvious question about this is, when we say AI agency, AI obviously means a lot of things to a lot of different people.
How have we defined that? It’s another area that we’ve really sought to bring nuance to in this body of work.
Zoe Hawkins: Oh, absolutely. I think we’ve all observed, um, you know, whether it’s experts, commentators, politicians talking about AI [00:08:00] and, and really referring, uh, to different things, and that, uh, we need to bring that specificity to the conversation.
And this research actually does that by, as you say, defining the full chessboard of 103 different capabilities, and we’ve done that across six layers, and it’s a very broad, comprehensive way of looking at the AI ecosystem. Mm. So it’s everything from infrastructure and resources, so your data centers and your compute, through to the data itself, um, and also the models and applications that use that data to bring AI applications to life.
But also those additional layers that bring this technology into society. So that’s looking at the innovation and adoption ecosystem. That’s the skills that people need to use and live with AI, as well as the governance layers that sit around that to make it all work. Um, and the reason this is valuable is it gives people a shared language, so we can speak about this with more specificity.
And the, what the tool does is it takes each of those 103 capabilities, uh, gives it a maturity rating, places it on [00:09:00] that sovereignty spectrum that we were just talking about before, but also indicates a measure of how globally scarce is that particular capability. And that’s relevant because your influence or power, um, on the world stage is in, is in part determined by how common something is, right?
So I think you don’t have leverage if, uh, in a particular capability if everybody else is also good at it, right? So the more scarce something is, the more leverage you can get from it. And I think the key takeout in terms of this defining AI and the typology that we’ve put in the tool is you do not need, and you shouldn’t try and be good at all 103.
No country can do that. The purpose of defining it out is actually so that we can specialize, so that we can select and be strategic with where we wanna invest and build our agency.
Johanna Weaver: So Zoe, we’ve obviously developed this tool, and I think even just articulating the 103 different capabilities, we’ve got definitions, examples, and metrics.
But then we’ve also [00:10:00] then applied that and done an assessment of where Australia’s agency sits against each one of those 103 capabilities. So can you talk us through what the key findings of the assessment of Australia’s AI agency are? Headline is that we have more agency than many Australians probably think, and I think even you and I were surprised when we saw the results coming out of this.
So where are, where does Australia have the most competitive advantage?
Zoe Hawkins: Absolutely. Australia’s actually world leading in eight of these capabilities, which it has the h- where we have the very high agency. I think we need to own that and, and, um, be bold and bolshy with that capability that we’re sitting on and that competitive advantage.
So it’s interesting because that capability is actually spread across several different layers of the stack, which I guess is also a good thing. So one of them is we have very rich data assets in [00:11:00] Australia. Um, so across our medical, geospatial, our infrastructure, demographic, and environmental and resource data sets are incredibly valuable, um, and compared to many other countries.
So this is something that we’re actually sitting on. This is like a resource in and of itself, not in the traditional natural resources kind of way, but in terms of data assets. And the– another area where we’re really strong in Australia is in model development. Now, not the ones that maybe people most often think of in terms of that sort of generative AI or frontier model development.
Lesser spoken about type, but very important is computer vision, and Australia is exceptionally good at this. Um, and this is really about models that bridge the physical world, so the ability to see and navigate in the s- physical space around them with technology and computer models themselves. And the reason that’s been developed in Australia, um, is a lot to do with our strength in the agriculture and mining sectors.
Another space that Australia’s really strong in that’s come through in the assessment [00:12:00] is actually, well, and you’d know a lot about this in your previous roles, international influence and norm shaping. So Australia has a really strong reputation and, um, leverage on the world stage in terms of writing the rules of technology policy, but particularly around AI governance, and we’re seeing that right now.
And there’s only two capabilities out of the entire 103 where Australia has much lower agency. So there’s the manufacturing and packaging of, um, accelerators, which are the most advanced sort of AI chips. So I think, as you say, Australia’s sitting in a really strong position with a lot of cards to play there.
Johanna Weaver: And of course, knowing our strengths, um, is one part of it, but then it’s also looking at what do we do with those strengths. So can you give us some examples of how we can actually use those strengths to Australia’s competitive advantage? And I guess what we’re talking about here is like how are we leveraging the gas to get the diesel for Australia when it comes to AI?
Zoe Hawkins: [00:13:00] Yeah, I think that’s right. You know, I think some of them lean naturally into bargaining chips, and I think that strength that we have in critical minerals, particularly, like, as the natural resources asset, that’s a really powerful one because nobody can produce these powerful AI accelerators, these AI chips, without those core ingredients.
So how do we actually use the fact that we are the leaders in that, uh, critical mineral space as a bargaining chip to make sure that we have access to those international world-leading chips? I think that’s a really great example of that. I mean, there’s other ways, too, in terms of thinking about how we make sure we’re leveraging our own existing assets, so making sure that our data assets are delivering good public value.
Mm. Making sure that our opportunity to be a data center hub, which we’ll speak more about, that we use that as a lever by which to support our energy transition in Australia. How can… You know, how we use that as a, a mechanism for progress in other areas that works for Australia. And as, as we were just [00:14:00] saying before, looking at the international rule development, how do we do that to make sure that we’re shaping the rules in a way that works for Australia?
Johanna Weaver: Yeah. And I, I think the… Many people, I think, are skeptical of Australia’s ability to influence international rules, and the reality, though, is that Australia’s been leading many of these conversations, whether it’s at the OECD or international standards organizations. It’s Australians who’ve been leading these conversations for years around, well, what are the standards when it comes to ethical AI or, or international standards around, uh, data in AI, for example.
And when you’re… when you have countries like the US becoming less and less stable, people not necessarily trusting China, they are looking for alternatives, and it’s also about how we combine some of these competitive advantages as well. You know, one of the ones that I like to talk about, you mentioned there, Zoe, the medical data, for example.
Do you wanna [00:15:00] talk maybe a little bit about how that can combine with some of the other capabilities to actually deliver an even higher competitive advantage for Australia?
Zoe Hawkins: Oh, yeah, I think that’s spot on, and that’s why seeing the big picture is so powerful because you can see how these capabilities actually relate with each other.
So firstly, when it comes to medical, the medical data, so how does that actually combine with the fact that we have, uh, very high capability, maturity, and agency in computer vision? Actually, that means that Australia has… c- can be a leader in terms of radiography, so looking at, um, cancer diagnosis. The fact that we have that treasure trove of historical data of, um, disease patterns and that we have, um, technology that can convert those images into pattern recognition, that’s a really powerful combination of different AI capabilities across the stack.
And so it just shows how we need to think, we need to think creatively and a bit laterally in terms of what are those ingredients and how can we work with what we’ve got.
Johanna Weaver: And that’s incredibly [00:16:00] valuable to Australians and to our health, but also to others internationally as well. And I think another example for me of where that all comes together is looking at things like, you know, we’ve got really…
Australia’s been collecting information on weather patterns for longer than most countries globally, right? And so how do we use that immense wealth of data, combine that with an area of Australian, um, high maturity in another area of model development called forecasting, and then that’s helping us to better predict fires and floods, for example.
Again, the, the export potential of technology like that is really significant for Australia. And so it’s about zooming out and, uh, you know, I think, uh, you’ve said it so well in terms of seeing that whole chessboard and understanding how we can combine those different pieces together. We’ve spoken there about what the assessment shows us in terms of some of the areas of high agency, high maturity for Australia, but we did [00:17:00] also identify areas where there are what we call critical gaps, and these are areas that we do need to fill, potentially by building sovereign capability.
So can you talk us through some of these, Zoe, um, and why they are important?
Zoe Hawkins: Yeah. I think while I said before, um, and it remains true that, you know, no country should try and do all 103 things really well, that’s, um, nobody has the resources for that. There are particular capabilities that are considered to have higher public interest, and they…
those capabilities maybe are more likely to be the capabilities that should be, uh, under sovereign control potentially. And that’s why, uh, one of the cons- one of the things that’s come through in our research is areas where there’s an intersection of a lower maturity in Australia, but also a high public interest in us having these capabilities.
And some examples of that include public interest compute, so the kind of AI resources needed for public interest. So think, think deep research, think [00:18:00] academia, think, uh, not-for-profits, that ability to do… to use this technology in order to deliver the, some of those positive outcomes that we were just talking about that actually benefit Australians.
Other areas include, um, uh, developing nationally, um, and culturally inclusive models. Um, making sure that we have trusted adoption at home, that Australians have skills to, um, to leverage and build AI. Um, but also regulatory oversight, having enough capacity, um, and regulatory frameworks to make sure that all of this is being, um, developed and adopted in a safe, in a safe way.
And so those are some of the areas that came through in the assessment that, uh, yeah, might be a good target to prioritize next in terms of Australia’s capabilities.
Johanna Weaver: Zoe, we’ve done this assessment. We’ve created the typology. We’ve then mapped, based on a national [00:19:00] stocktake, Australia’s maturity against each one, 103 of those capabilities. We’ve put each of those capabilities on the sovereignty spectrum. Uh- And this is- … we’ve looked at global scarcity and come up with these AI agency scores for each of them.
But then we’ve actually gone one step further, and we have also then mapped our assessment of Australia’s capabilities against the government’s national AI plan, and created, in effect, a scorecard of those capabilities. Now, we can talk about that, um, for a long time, but maybe let’s just touch base on what are the key takeouts of mapping the assessment where Australia has these strengths, where we’ve got weaknesses, where we’ve got critical gaps that we need to fill, against what the government is doing.
How’s the government shaping up, uh, in this analysis?
Zoe Hawkins: I think the reason that this was such a good idea to do is it’s so powerful to put it, this research in context, right? And I think- Mm-hmm … I’m so glad that that [00:20:00] was a, a suggestion from you is to do that. So I think what we’ve seen in mapping the assessment with the government’s National AI Plan is the key first takeout is that the major commitments in the National AI Plan actually align with the evidence of where investment could be well made.
Um, so it really leans in to Australia’s existing strength, a lot of the things that we were just speaking about before. So for example, the fact that Australia’s well-positioned to be a data center hub given our enabling infrastructure, and the idea of leaning into public cloud because that’s a really cross-cutting ecosystem enabling capability that, that lifts across the board.
Um, leaning into general AI applications because that’s actually an area of competitive advantage for Australia. We’re really good in that space. Um, but other things as well, like government and, um, small business adoption, and the international engagement, which you mentioned before. So there’s, there’s sort of commitments around that as well.
So it’s really about the plan does… Where it does lean in, it’s leaning into places where [00:21:00] we’ll get a good ROI for our investment. I think the challenge is, as we said, the stor- the work is never done. Um, and our assessment did reveal places where we could use, uh, those sort of high agency capabilities even more strategically.
So there’s some opportunities for further prioritization in terms of how we leverage those capabilities where we’ve got that high agency, whether that’s our geospatial data or our international influence. You know, I think there’s, there’s some more areas beyond that, whether that’s the other data sets, the computer vision.
How do we use those chips that we’ve got to play really powerfully? And- Mm … as we mentioned before, there’s those critical gaps that we could close, and those ones that came through, which is the public interest compute, um, inclusive and discerning adoption, and those other areas of, um, regulatory oversight and national inclusive models.
Th- so I think it’s, it’s great because it gives a roadmap to say, “Okay, where to next?” Um, you know, we’re leaning in in the right places in the plan, but it also gives a bit of a guidance as to where investment could be [00:22:00] well-placed.
Johanna Weaver: Yeah, and I think, you know, it’s a, it’s a challenging task, right? Government has- Yeah
to prioritize, and I think, you know, it was, again, I think, quite an unexpected outcome for us to see how closely aligned the things that the government had prioritized align with the assessment. And then we hope that this helps to guide where we’re going to be leaning into and prioritizing next. So, Zoe, I wanna, before we wrap up the discussion, I wanna sort of go back full circle to where we started the podcast, and this decision out of the US, uh, you know, a few…
It’ll be a couple of weeks ago by the time this goes live. Where the US government, just to recap, the US government placed an export restriction order on Anthropic and basically said, “You can’t export these two most powerful models that Anthropic builds, Mythos and Fable 5,” and said, “You can’t export these models to anyone [00:23:00] who isn’t a US citizen.”
And the concern about this was not so much about its immediate impact, although there are significant and, and important concerns about the immediate impact of not having access to those models, but it’s also about the precedent that this sets, right? This is the first time that we’ve seen the US government put an export, export order like this out, which doesn’t make a distinguishing exemp- exemption or exception for a country like Australia or the Five Eyes countries.
You know, we’re being treated the same as China, which is, you know, I think, um, quite a stark reality for Australians to have to, to have to confront. But it also has emphasized the danger of depending on AI from, you know, one company or one country, right? So maybe companies are using models from a few different companies.
So maybe they’re using OpenAI and Anthropic and possibly some of the Google models, but, you know, the reality is they could all also [00:24:00] potentially be placed under export restrictions. And so it’s really that precedent that’s kicked off this discussion and having lots of commentators coming forward and saying, “You know, we need to build this.
We need to control it at home.” So how does the AI agency tool and this sovereignty spectrum that we have articulated, how does it help us to understand what levers Australia has in the face of this, you know, quite dramatic shift in the geopolitical environment that we are operating in?
Zoe Hawkins: Yeah, it’s a fascinating time to be, um, having this conversation.
Um, the nerds, the nerds in you and I are, like, enlivened by this, the timeliness of this. I think, uh, as you’ve said, this development really has driven home, and you’ve seen it in the common- in the, um, commentary- a, a kick up in that conversation of, okay, what is our sovereign capability? Do we have this at home?
We need to rely on ourselves. And of course, [00:25:00] that’s incredibly important, and as we were just talking about in terms of those critical gaps we need to close, Australia cannot afford not to have sovereign AI capabilities in core areas. Absolutely, and this has driven that home. But I think that that’s only part of the story, and it would be a big mistake for that to be the end of the conversation because it also shows that we can…
We can’t afford to only rely on- Mm … sovereign AI capabilities. What we were just talking about, there’s 103 different ingredients that go into a thriving AI ecosystem, and Australia is not in a position to do all 103 well. We must get some of that world-leading capability in other places sometimes. So how do we do that in a way that is reliable, resilient?
And that’s really the, where the AI agency tool, and particularly the sovereignty spectrum, comes in and is particularly useful because it, as I said at the beginning, it helps us break down that total dependence, total self-reliance into something more practical, right? And it’s about how do we make smart decisions to prioritize well across those 103, and that’s what the [00:26:00] tool helps us do.
Um, and so it’s about identifying which of those 103 capabilities are the most critical to us. Where do we need to have that sovereign and domestic control and capability in making really, um, ambitious investments in that? And you know, our report points to areas where we need to do that. But also, where do we need to build resilience?
So where we might want domestic control, but we also want to have access to some of these world’s most advanced models. Mm. We don’t necessarily wanna close Australia off from some of these capabilities just because they come from other parts of the world. But it’s how do we make sure that we have that redundancy and resiliency so that when there are shocks like the one in the last few weeks we’re talking about that might be surprising, and won’t be the last, right?
100%. That we can continue to operate. And so, and also to be, as you say, like… You know, Australia is a better place than many people might think. So how do we make sure that where we are world leading, that we’re really leveraging and exporting that capability so that we can use that to plug where we have those gaps, use those as bargaining chips?
So I think there’s a lot in this tool, and specifically in the sovereignty [00:27:00] spectrum, that hopefully will be practical levers for policymakers now looking down the barrel of things like Mythos Five. Yeah.
Johanna Weaver: Yeah. And I just think, um, that point about not being able to afford not to have sovereign capability.
We must build sovereign AI capability, but we also can’t afford to only have sovereign AI capability, and, you know, we were looking at the budget papers ’cause we really are nerds, and looking at, okay, so the amount of money that Anthropic reportedly spent just on compute in nine months is the same amount of money that the federal government is forecast to spend on public order and safety, right?
And so the orders of magnitude of the funding that is being put into building some of these technologies, it’s not saying Australia shouldn’t be leaning into that, but it’s about being strategic about which of the technologies that make the most sense for Australia to lean into so that we can shore up and de-risk our [00:28:00] dependency on foreign technologies, whilst at the same time leveraging our ability to build world-leading technologies that others come to depend on, and that gives us a lot of power.
Zoe, it’s obviously easy for us to sort of sit here and, and say… Okay, well, not easy. This has not been an easy project. I think it’s fair What are you talking
Zoe Hawkins: about?
Johanna Weaver: But I think, you know, it is… Looking at it in its full complexity like this, you know, and it is a huge double A3 spreadsheet when you see it all spread out.
It does bring home for me just how challenging it is for decision-makers right now to be making these decisions and having to make these prioritized decisions. One of the reasons we’ve built the tool is to help people to be able to be making those decisions based on evidence and data, not on, you know, spin or sales pitches.
But if you were in their shoes, what is the one thing that [00:29:00] you think, maybe po- possibly something that we haven’t spoken about, that you think we really need to build urgently in Australia?
Zoe Hawkins: Totally agree. Don’t envy the job of people trying to, um, in those hard, hard positions of, um, finite resources and lots of different important areas to invest in.
I think one area that is really a cross-cutting issue and that is actually brings our conversation, we’ve t- spoken a lot about the technology itself, but I think humans, you know, we’re… At the Tech Policy Design Institute, we really believe in that relationship between humans shaping soci- uh, technology, and technology shaping society.
So Bri, yeah, exactly right behind you.
Johanna Weaver: I’m just pointing, for those who are listening, to a sign behind me that says, “Tech Policy Design Institute: Shaping technology for the benefit of humanity.”
Zoe Hawkins: So I think that something that came through really strongly in the assessment that we haven’t spoken about as much today is the role of inclusive and discerning adoption as a capability- Mm.
Mm … for Australians, and in, in the human sense and the people. Because ultimately, this [00:30:00] technology is to be used i- i- the purpose of it, obviously, is for benefit of, um, Australians ideally. So what that means is, do Australians have the skills, the information, and meaningful choices to be able to access this technology in an equitable but also empowering way, right?
It’s, it’s looking both at the skills, um, information, and also the access, right? Because we already have a digital divide issue in Australia. It’s a, it’s a massive concern in terms of, um, the equitable outcomes that technology can bring. And AI has a real risk of entrenching those divides, I think if we don’t, if we’re not really mindful to make sure that the investment is there, to make sure that the access avail- this is available to people in a material sense, but that they’re equipped with the skills and literacy to be able to take that up.
But the last part about w- that, which is the discerning part, which I think is the bit that’s often lost, is that this is not a predetermined value judgment that every single person in Australia in every [00:31:00] single situation should and has to adopt AI. That’s… It’s, uh, really about empowering people to make, and this is agency again, right, but at an individual level.
It’s like, can you… Are people put in a position to make that choice? And you can only make that choice if you have actual, the option of real access to it, and you have the option to opt out if you don’t want to, and you have the skills and the information to make those. So for us, I think that’s a really important element to this, not to get lost in all the tech conversations.
Um, because otherwise, we’re gonna end up with these advantages entrenched in the hands of a few, with the cost and the sort of access kept from others. So I think that’s… What about you? I think there’s, um, other things to be prioritized here in this long list of, of important capabilities. What would you pick?
Johanna Weaver: Look, I think if I was sitting around the cabinet table in the next week or so, the things that I think we just have to prioritize, and they’re, they’re things that you have mentioned already, is the public interest compute. We need Australian researchers to have access to public interest compute. We only [00:32:00] have two supercomputers in Australia.
They’re both getting a bit old and clunky to the extent that they’re one thing, but, you know, they, they need… We need to have a critical investment into access to public interest compute because the researchers will then go out and do all of this incredible stuff that will drive benefits for Australians.
So it’s– I think that to me is really important. And then obviously the stalemate on copyright needs to be resolved, and I do think we’ll start to see… I know the conversation around data centers has been, you know, a really big one, and it’s important that Australia gets those settings right. I do think part of the answer to that is evolving the conversation from just looking at the data centers that are inference data centers versus the training data centers, and that’s, you know, inference being when you’re using ChatGPT and it needs some compute power, it goes to an inference data center, and the training is what goes into actually making ChatGPT work as a, as a basic example.
And I just think the value add for [00:33:00] Australia really sits in, in the compute training rather than inferencing and… But we need to solve copyright to get around that, and we have to do that in a way that protects and that promotes, um, you know, the work of Australian creatives, the work of First Nations people, the culture and the history.
So, you know, I think that’d be where I would be focusing, and an absolute big plus one to what you’ve said around inclusive and discerning adoption, ’cause I just think that is, that is absolutely key if we want to be shaping technology rather than having technology shaping us. And, you know, obviously that’s a big part also of what we’re focused on here at TPDi.
So Zoe, I mean, we’ve covered a lot of ground. I really do encourage people to go and, you know, have a look at the report. The report itself explains the methodology, then we have the Australian assessment, and then there’s a separate document which is, you know, mapping the assessment against the National AI Plan.
If people take one thing [00:34:00] out of this conversation, what do you think the most important thing for people to reflect on and discuss with, you know, colleagues or around the dinner table?
Zoe Hawkins: I think that it’s just to say the AI sovereignty debate has really gotten trapped in this false binary of total dependence or total self-reliance.
And in reality, in a world of global AI supply chains and finite resources, um, actually talking about whether we have the agency to shape our AI future is maybe a more useful frame that throws up more options for policymakers. And just to be cheeky and add a second takeaway, it’s just that Australia is actually in a way stronger position than, than most people might think.
Um, but that we really need to engage, understand with the evidence so that we can work out where we have leverage, where we need to build sovereign capability, and where we need to fill critical gaps. So I think there’s a lot to dig into there, as you say.
Johanna Weaver: Yeah. Absolutely. And I, I mean, my one plea to any Australians listening to this is we just have to stop being humble.
We have to- Mm … [00:35:00] accept that we actually do have some really strong cards to play, and the assessment really sets these out in a lot of detail. So I encourage people to go and, to go and check it out for yourself. You can view all of these different documents at techpolicy.au/aiagency. Um, and so that brings us to the end of today’s episode.
Um, thank you so much, uh, Zoe, for this conversation, but also for, uh, leading such an incredible body of work, um, at TPDi.
Zoe Hawkins: Pleasure to do it with you. Thanks, Johanna.
Johanna Weaver: That’s it for today. So, um, we’ll be back in a fortnight with the next episode. In the meantime, please do get in touch and get involved.
Well, that’s it for this episode of Tech Mirror, which is brought to you by the Tech Policy Design Institute. We’re based here in Canberra on the lands of the Ngunnawal Ngambri people. If you found today’s conversation useful or thought-provoking, please do [00:36:00] share it with a friend or a colleague, or leave a review and subscribe wherever you get your podcasts.
If you’re watching, please don’t forget to like and follow. For show notes, you can visit techpolicy.au/podcast. And this podcast was made possible with thanks to the generous contributions from government, industry, and philanthropy to the Tech Policy Design Fund, the full details of which are available on our website.
The team at Audiocraft produced this pod on the lands of the Gadigal people of the Eora Nation, and Amy Denmade provided invaluable research support. Music is by Thalia Scopelis. A big thank you also to the team at the Tech Policy Design Institute, without whom this pod would not be possible. Thank you for joining us, and as always, get in touch and get [00:37:00] involved