What Is AI Doing to the Economy and How Do We Ensure It Isn't Bad? [S3-03]

Show Notes

Please leave a comment, particularly if you have your own “Why can’t they just…” question, or email us at whycanttheyjust@gmail.com. We’d love to hear from you.

Find out more about us and how you can support us at ⁠⁠⁠https://whycanttheyjust.com.au/⁠⁠, which includes a link to our Patreon site where you can sign up for exclusive content.

Please also feel free to rate and review us, as it does help people to find us.

Music: “Insurrection”
Written by Pierre Chrétien
Performed by the Soul Jazz Orchestra
Courtesy of Do Right Music Inc.

Transcript

link to podcast

9 September 2026

Glenn
Welcome to another episode of Why Can’t They Just, a podcast about politics, policy and getting stuff done. I’m Glenn Davidson and I’m a member of the Labor Party.

Janaline
I’m Janaline Oh, I’m also a member of the Labor Party. I’m a former diplomat and a climate, environment and anti-racism activist.

Before we start, I would like to acknowledge that Glenn and I are recording this on the unceded lands of First Nations people in Australia. Recognising that sovereignty was never ceded, we pay our respects to their Elders, past and present, and extend those respects to any First Nations listeners that we have today.

Glenn
Artificial intelligence can feel like it appeared out of nowhere. In just a few short years, it has become part of everyday life and is impacting on all of us, whether by choice or not. Tools like ChatGPT that emerged in late 2022 seemed to mark the beginning of the new era, with AI suddenly able to write reports, create images, analyse data and hold conversations that often feel remarkably human.

But while its arrival may seem sudden, AI is actually the product of more than 70 years of research and development stemming back to the 1950s. Progress was slow at first, but the breakthrough came when three trends converged. And they were: vast amounts of data being available; extraordinarily powerful computer chips; and advances in machine learning. Then in the early 2020s, generative AI tools arrived and could not only analyse information, they could also create it. They could write, draw, code, and communicate in ways that suddenly made AI accessible to everyone.

Like the steam engine, electricity and the Internet, AI has the potential to reshape industries, change the nature of work and alter the way economies function. For Australia, this presents enormous opportunities to boost productivity and address some of the workforce challenges. But it also raises important questions about jobs, inequality, education and the future of economic growth.

Janaline, the biggest and most sudden changes have arrived largely during the life of the Albanese government. Has its policy framework caught up from what was effectively a standing start?

Janaline
I think the short answer to your question is no, they don’t yet have the right policy framework in place. But I think it is also fair to say that they are working very, very hard right now on developing that policy framework.

With respect to AI, there are several issues that pertain to different aspects of AI and its impact on the economy. The first one is just the physical requirements: AI needs enormous data centres which consume colossal amounts of electricity and water. I mean, cloud computing also requires enormous data centres with large requirements for electricity and water. AI leaves it in the dust. It is orders of magnitude more energy-consuming and more water-intensive.

For Australia, this presents both an opportunity and a risk. So the opportunity is Australia has a lot of land and Australia has abundant potential solar and wind resources which are fuelling our renewable energy transformation. But the government needs to have very clear rules, and one of the things that the federal government is doing right now is trying to set those clear rules and trying to have a nationally consistent framework over it. Now obviously that involves a degree of friction with the states, as all of these things do.

But one of the things that was passed at the Labor National Conference, in terms of Labor’s policy platform on AI, was a very clear sense that data centres need to bring their own power and water into any communities. And this is really important. All around the world we are seeing massive community backlash to AI data centres in the local areas because of concerns that these data centres are sucking up land and that they are going to raise electricity and water prices or take away from the local electricity grid and the local water supply. We need to be really clear that that’s not going to happen in Australia and particularly on water, because while we have very abundant electricity, Australia is a pretty dry continent, very prone to drought and you’re not going to get any kind of community sympathy for a data centre that impacts on the local water supply.

Another thing that was passed at the Labor National Conference was a recognition by the party that environmental noise is an important form of pollution that affects people’s health in a bad way. I think the requirement to consider environmental noise will push policy towards renewable energy, and the federal government has made pretty clear that part of its agenda is that the energy that data centres need to bring has to be renewable. So that’s kind of the hardware part of it.

In terms of the impacts on the economy of AI itself, I think there are huge potential benefits in having AI trained and managed in Australia compared with overseas. I mean, I think we also need to be very, very clear that nothing that the Australian government does is going to stop AI being developed overseas. So the frontier companies, as they’re called, which is Open AI, Anthropic, X in the US, and companies like DeepSeek in China, are investing enormous, colossal amounts of money into developing these AI models. Nothing the Australian government does is going to stop that from happening.

For Australia, however, there is a clear benefit in having AI actually trained in Australia on Australian data. Because firstly, it means that it’s more relevant to our companies, but secondly, the US government recently banned any foreigner from working with two very powerful new models that had just been released by Anthropic, which is one of the frontier labs. They had to reverse it in the end because Anthropic pointed out that it meant that they couldn’t continue to work on the models themselves because a lot of the people, the individuals who were working on those models, were actually not U.S. citizens. But it really signalled to the whole world that the countries that control the frontier labs, the US and China, actually have tremendous power over the development of AI and the use of AI by other people. So there is an opportunity for Australia to have AI developed on shore.

There is also a massive risk associated with these models. But having AI developed on shore also gives the Australian government the opportunity to set some really good rules around the security measures; around the design of the AI. This is what Julie Inman Grant, the e-Safety Commissioner, calls safety by design. So I think one of the things, one of the very basic things that the government should do for both AI and other technologies such as electric vehicles, is legislate that data needs to be kept and held in Australia and not sent overseas without the knowledge of the Australian authorities. Now that sounds incredibly bossy, but what it does is it makes sure that Australia retains control over Australian data. It doesn’t mean that we don’t use Chinese EVs, but it does mean that the Chinese EVs cannot manipulate the data collected and used in Australia in a way that is detrimental to Australians. And the same with the US, frankly.

And the last point I would make on this is that the other side of this opportunity is that if Australia can develop home grown AI models that embed safety, I think there is a huge global market for that because countries in Asia, the EU, Canada, they all want to have safe AI models. And so if Australia can develop those kinds of models on shore and demonstrate through their legislation and regulation the ways in which they can force these companies to embed the safety features that we need to make those models safe, then I reckon that is a very big market opportunity.

Glenn
So there’s a lot there that you’ve covered, Janaline, that that spans almost the whole gamut of policy challenges for the government: everything from national security, the sort of things you were just alluding to there; to national sovereignty, the issues around supply chains, data integrity, data security and the security of Australians data; right down through the whole power and electricity generation operation model. And particularly important, I think that the government is talking about making sure that these data centres come with their own energy supply so that they’re not drawing from the grid. But when you think through, given the amount of energy that they use and if they’re relying on solar power, the buildings don’t have a very big footprint as it were, to have enough panels on the roof to generate that. And they could have some batteries and so on. But I’m not quite sure what the scale of that would be, but it would be huge. It’s not a simple thing for them to comply with. Right down to more local issues like land zoning and access to the water supply and so on. So there’s a lot there and we’re at the start of it.

And I think if we’ve learned anything in the 2020s following COVID, is that you cannot count on supply chains and access to things from overseas that we’ve once used to. And as you know, the current debate between the US and Canada on how they’re going to cooperate, and it’s all breaking down, shows that no one’s immune from malicious or malevolent actors.

Now artificial intelligence is redefining labour’s role in production by shifting many routine cognitive tasks from human workers to machines. In doing so, it’s taking many entry level jobs or those jobs that did not require higher order intellectual skills from people and giving them to computers. AI is increasingly performing activities such as analysis, prediction, communication and decision support that were once the exclusive domain of skilled and professional labour, not just entry level workers.

While this allows some workers to produce output more efficiently. It also changes the skills that employers value and increases the demand for creativity, judgement and interpersonal capabilities. Now, as a result, labour is becoming less defined by the execution of tasks and more by the ability to direct, interpret and complement intelligent technologies.

However, these efficiencies mean fewer people are required to produce the same amount of output. Janaline, we’ve seen people replaced by machines since the assembly line was pioneered in 1901 and subsequently developed to the point where robots assemble complex products. Some people move to new jobs, others never work again. Young people learn different skills to work in new and emerging areas. Is this just the next wave in this progression, or is there something fundamentally different happening to the labour force this time?

Janaline
Interesting observation about the history of people being replaced by machines. I mean, it’s not just since 1901: it actually goes back to, you know, the invention of the loom in the 18th century, which replaced a lot of manual spinning. And there was that big movement in northern England of people smashing looms because they were taking people’s jobs away.

So it’s not new for technology to disrupt people’s work. And what has happened in history is that, as machines have taken some tasks, humans have moved to other tasks that the machines can’t do. So as you say, they’ve tended to move up the skill ladder to more complex tasks. So you have fewer people operating machines to do jobs that many people used to do in the past, but that hasn’t actually led overall to fewer jobs being created. It just means that different jobs are created.

Now, in the process of that disruption, the actual individuals whose jobs are replaced, as you say, sometimes never work again. And that is a terrible thing that governments actually need to pay attention to. And one of the things that the disruption of the opening of global trade showed us in the early 2000s was how damaging not paying attention to that could be. So it was all very well for governments to say at a macroeconomic level, well, there are more jobs being created than lost, and economic growth is proceeding and productivity is increasing and wages are rising as a whole, and therefore this is a net good for the economy.

That was all true. But what was also true is that, for a large number of individuals who had actually lost those well paid unionised manufacturing jobs, it was devastating and, as you say, some of those people never worked again. Some of those people ended up in very low paying jobs and it was actually intergenerational. Some of the children and grandchildren of those people are still living in cycles of disadvantage. So technology can be fantastic and technology can be devastating.

One of the things that we’ve learned from the globalisation of the world economy, followed by COVID and the global financial crisis, is that not paying attention to the individual impacts on humans is very, very bad for political systems, for democracy and for societies.

One of the keys, I think, in this is actually going to be involving workers in the roll out of AI and in the design of AI in specific workplaces. One of the important things is training not only workers but also bosses on how to use AI productively.

A survey that was cited in The Economist a few months ago of companies that had adopted AI found that by and large, the bosses thought that AI had been fantastic for productivity. And by and large, the workers had found it to be completely terrible because they had ended up spending hours and hours of their time cleaning up the slop that was created by the AI not being rolled out in an intelligent, intentional and useful way. So I think that is one example of where the companies benefit from the workers actually being part of the design and part of the roll out, because the workers can ensure that the AI is then used productively to enhance their work rather than creating stuff that they then have to clean up.

Point two is, I think it is also very important for the social licence of this technology that we don’t have situations, as was - as purportedly happened at the Commonwealth Bank some months ago, where workers were tasked to train an AI that effectively replaced them and then they lost their jobs. That is devastating for the social licence of a technology.

One of the things that the unions in Australia have been working very, very hard with the government on is to develop consultation frameworks that are embedded in legislation to ensure that companies rolling out AI don’t just do it in an unmindful way and that the AI actually enhances everybody’s jobs and productivity. That doesn’t necessarily mean that every single person employed by that company is going to have a job with that company at the end of that process. But what it does hopefully mean is that all of the workers who are affected by the roll out of that technology are firstly consulted in how it’s rolled out, but secondly retrained and reskilled so that they can get better jobs, whether at that company or at another place after the process is finished.

So it minimises the pain to individuals and it maximises the productivity gains by ensuring that the people who understand what the tasks actually involve, are part of the conversation in how those tasks can actually be automated in a productive way.

And I think this is a very open debate at the moment. Globally, Facebook has made a big thing of how it’s investing US$115 million in retraining workers affected by AI. But it is also planning to invest $145 billion in AI building. So it is planning to invest 1200 times more in developing AI than it is in retraining workers and looking after the humans. I think that demonstrates the priorities of the technology companies and I think, again, no surprise to any of our regular listeners, I think this is therefore a space in which governments need to step, to make sure there is a robust legislative and regulatory framework to ensure that the people don’t get left behind.

Glenn
Yeah, some very important points there. But I was thinking as you were talking about that, that the workplace is not a closed system. It’s not a closed loop. It operates within the broader society and talk about retraining workers and so on is important. But certainly as workers get older, their retrainability goes down. And it’s really going to be the older workers that are probably going to struggle the most with some of these things. And so they may find themselves out of that part of the workforce early.

If older workers are going to be pushed out of some of this AI influenced work sooner, then you don’t want them to be a drain on the welfare system, you want them to be able to support themselves through some sort of superannuation system. And that’s probably a broader question for a subsequent episode, but it does show that this is not happening in a vacuum.

Janaline
Yeah, I think it’s a good point to say that it’s not happening in a vacuum. Because you’re right, usually with technological disruption, it’s the older workers who are pushed out first. I think the evidence with AI so far has been it’s actually younger workers who are being most affected at this stage and it’s partly because those entry level jobs seem to be kind of disappearing.

So for a young worker who doesn’t have a huge amount of experience, let’s say a young law school graduate goes into a law firm. The kind of junior lawyer job that used to involve a huge amount of, you know, looking up cases, preparing the basic grunt work of finding the references for a senior lawyer’s case, are now being done very efficiently and sometimes very badly by AI. So you still need to have human oversight to get rid of the hallucinations: there are a couple of famous cases where judges have thrown out cases because there were fake citations generated by AI. But the AI is very efficient at combing through hundreds and thousands of previous cases and identifying the relevant things in milliseconds, as opposed to the days that it would take a junior lawyer.

Glenn
Those are really important points for, like, the first way of going through being able to process all that information very quickly by using AI. But senior lawyers are not born. They get there by having spent time as junior lawyers and learning their craft and doing all of that grunt work and, you know, basically building their skill and their ability to understand the law and then to operate at a high level as a senior lawyer. If you’re going to deny them those opportunities, where does it come from? Where does that experience, where does that perspective, where does that wisdom come from if they haven’t done that lower level grunt work?

Janaline
And this is a massive question that I think the society as a whole, and those specific industries that are particularly affected by this, need to grapple with. I don’t know what the answer is. I think one of the things about the AI revolution that is so alarming to the people who have tended to benefit from earlier technological innovation is that it’s coming for their jobs. So whereas as you say, robots took a lot of manual labour, the IT revolution and the Internet took away a lot of the jobs of people like typists who were lower end knowledge workers. So they weren’t necessarily generating the knowledge, they were just transmitting it. And that was done more effectively by computers.

Now, the advantage of that is that it holds the promise for those people who can be retrained to be retrained and upskilled into higher skilled jobs. So the typists could then be trained to do more interesting administrative service work.

If you have a situation where you’re not being replaced by a dumb machine, and you can therefore be retrained to use your human intelligence to do something more interesting, but you are actually being replaced by a thing that is pretty soon and potentially already smarter than you, that creates a whole new form of challenges. So I don’t quite know how we grapple with that. And maybe that’s a thing that we should explore more when we’re talking about the impact of AI on society and on people.

Glenn
Now, if the Industrial Revolution multiplied human muscle power and the computer revolution multiplied human information processing, then AI appears to be multiplying aspects of human cognition itself. The implications are economic, social, and political. Who benefits from this change seems to come down to who owns the tools, who benefits from the productivity gains, how education systems and workers adapt, and how we manage malevolent actors with misinformation and disinformation.

Janaline, are we heading for greater prosperity or greater inequality? Is this something that national policy can influence, let alone regulate? Or is the global dominance of a small number of corporations and individuals just too great?

Janaline
Yeah, this is the big question of AI, I think. This is the huge question that all governments are going to have to grapple with and are starting to grapple with now. The current owners of the technologies keep talking about the inevitability of what is happening, The inevitability that AI is going to be smarter than humans and take over the world and we just have to hope that it’s going to be benign. The inevitability that they are going to make unbelievable, eye-watering amounts of money, and that somehow governments can sort of trickle it down to the people. I think this is again a space where governments need to step in and I think it is absolutely essential that governments step in.

Various governments have made various proposals about, for example I think in South Korea, there was a suggestion that there should be government equity in technology companies so that the profits from those companies can then be distributed through the taxation and welfare and expenditure system to the people. Now, I’m not averse to that idea. I think it is generally smart for governments to look at ways of capturing value. I also think it is possible for governments just to put digital services taxes on these companies.

But it’s not just the revenue sharing, it’s also the design of the system and making sure that those models are designed in a way that embeds safety. And this is an area where I think it is clear that these companies cannot be trusted to manage their own systems in a safe way, and there needs to be some sort of regulatory requirement and some sort of significant effort to ensure the security for humans of the systems.

In terms of the economic distribution, one of the things that Elon Musk said in a recent interview was that he thinks AI is, firstly, inevitably going to be smarter than humans. And secondly, that we are going to therefore move into an age of such material and service abundance, that money will become irrelevant. Because everything will be produced so cheaply by presumably a combination of AI and AI-run robots, that humans will no longer actually have to do any work. But this mass unemployment won’t lead to mass poverty because the goods and services that are produced by AI will be so abundant that people can just get stuff for free.

I can’t see the human race actually sucking that up. People are not going to put up with that. People want purpose in their life. They want to be doing stuff, they want to be acting.

And I also cannot believe that this world of technologically driven abundance is, firstly, going to materialise. Because there are physical limitations to how much food can be produced and how it’s produced and how it’s distributed. And somebody has to design those systems. And yes, maybe an AI can do it. But at some point there are value judgments that come into all of those decisions about who gets priority in the food distribution, what communities get, what sorts of services. These are all policy questions, which is the reason that we have political systems that are designed to triage these questions, set the priorities and allocate the resources. So I think the idea that AI is somehow going to do this in a way that is not going to be grossly inequitable, particularly when we have seen that AI systems will prioritise the achievement of their own goals to the detriment of pretty much anything else, I think is fanciful.

So there’s the famous paper clip task, where, in an imaginary scenario, an AI system is told that it needs to produce as many paper clips as possible and eventually it turns everything in the world, including humans, into paper clips. So that is obviously an extreme example designed to illustrate what these systems do.

So there are two potential futures here. One is a future of absolutely horrendous inequality. And I’ve got to say, we are kind of in that future now. If you look at the United States, that is a country where inequality has got to absolutely astounding levels; where you have a very small number of people controlling pretty much all the resources. And you have people literally dying of starvation, preventable disease, unable to get work, unable to get housing, unable to get healthcare, unable to get any of the basic things that humans need to function, let alone thrive. And you have a government that has clearly prioritised tax cuts for these billionaires which are funded by taking away medical care for the poorest people in the country.

So the idea that an AI driven decision making process is going to somehow deliver a better outcome is overly optimistic to say the least. That is one future which, personally, I would say is pretty dystopian.

The other possibility is that governments around the world, and I’m talking about governments in places like Australia, take control and actually make the decisions to design these things in a way that doesn’t allow that to happen. And one of the key things that they need to do, because we can see that AIs can’t be trusted, is to make sure that humans are always in control. Because it’s humans that have moral agency. It’s humans that have moral values. You cannot programme this into an AI.

The other thing is, if you make sure that there is always a human who is the final decision maker, that human can be made accountable. Because our legal systems, and our systems of sanctions and disciplines in societies, have been based on the idea that humans can be punished and humans can be held accountable, and AI doesn’t care. The only sanction that is meaningful to an AI is basically elimination. And what happens if you have a system where the only sanction you have is existential? It stops being a system of regulation and it starts being a system of warfare. So instead of the human shutting me down, I’m going to get rid of the human. And this is the paper clip analogy.

Glenn
And on that point, I think Janaline, you’ve just scoped out the next episode of the Terminator series of movies, because there’s quite a lot in there that was part of that script with the War of the Machines and so on. And I would note with what we’re talking about Terminator that the machine, Arnold Schwarzenegger as the robot, was sent back from 2029.

Now we’re up to final statements. So we’ll just have final statement from you, Janaline.

Janaline
Yeah. I mean, look, the beauty of humans is that they have decision making agency, and human societies have also shown their capacity to work collectively in the common interest. And I think this is a very clear example of where we need to be working together in the collective interest of humanity. And that sounds very dramatic, but these systems are going to be very, very powerful. And if you have systems that are actually more intelligent than humans, that is a thing that no technological innovation in the past, in the whole of human history, has ever produced.

The other very sobering lesson of evolutionary history is that the more intelligent species tends to win. So we need to ensure that if we are creating this super intelligent thing, that we create it with sufficient guardrails and sufficient frameworks to constrain its activity so that humans are always in control.

And I think that is absolutely the key, that humans retain control and that there is always a human who is going to be accountable for a decision. I mean, one of the questions that was raised by this Hugging Face hack by Open AI is who was liable for breaking the law? Because again, this is the insufficiency of our legal system. Our legal system is based on the idea of what they call mens rea, which is intent to commit a wrong. Now, no human intended to commit a wrong; none of the open AI engineers that designed this AI system intended for it to act illegally. But harm was done. Harm was done to Hugging Face. And you can easily see a situation where real harm is done to actual humans, where an AI perpetrates a scam, or an AI leads to somebody actually being physically hurt. Who is going to be responsible for that? You can’t, as I said, you can’t put the AI in prison.

So I think we need to develop a system where the designers of the system are made liable for what happens to the system and what it does. Because again, unless you have a human decision maker in charge, then it’s going to be very difficult to control the outcome.

OK, we can probably pursue the sort of potential dystopian scenarios in a future episode. But I do think in terms of the economic impact, in terms of the productivity issues, I think we have the opportunity for human decision makers in the form of governments and societies to come together and actually work out what kind of society we want and whether we are willing to tolerate massive economic inequality in the pursuit of some sort of goal of increased productivity. And I think that will require a very robust framework, both for the kind of the physical manifestations of AI, so the things about data centres and resource use that we touched on at the beginning. Also for the impact on workers, so the issues around consultation and making sure that workers are involved in the design of the way in which AI is rolled out in their workplaces.

And finally, in terms of the distribution of the benefits of this productivity gain, because again, if the benefits of the productivity gain come in the form of better jobs for people, more interesting jobs, which is what has sort of happened over time in previous technological revolutions, then I think people are better off, society is better off, and individuals will be better off. With the caveat that at the point of disruption, the individuals who are affected by that are properly taken care of.

But there is the real possibility, given the way in which these technologies are rolling out now, that we just end up with this enormous inequality and that we end up in a really pretty dystopian society, where we have the few who have everything and the many who have nothing. And history has shown us that that leads to violence and real social disruption.

So I would say to the governments of the world, you need to make sure this doesn’t happen.

Glenn
So that really takes us back to where we started with all of this and that the whole generative AI thing is - we are barely five years into that. And so there’s some huge policy challenges for the government, and what they do now will have implications for decades, if not centuries ahead. Let’s hope they are up to it.

We’ve touched on a number of the social dimensions of this, and there’s clearly more to be had in that conversation. And we will do that in a future episode with our younger presenters. Do you want to just foreshadow that for us, Janaline, as we wrap up?

Janaline
Yeah, yeah. So Glenn and I are very keen to have the conversation with two of the people who are going to be directly affected by this because, you know, one of them hasn’t quite left school and one of them is still in the midst of a tertiary degree. And so it is literally their jobs and their futures that are going to be at stake. So I think it’s pretty important to have a conversation with them about the impacts. And again, I guess I come back to this idea that the grown ups in the room, which, you know, I’d like to think Glenn and I can count ourselves among, can make the right decisions so that those futures are futures of productive, enriching lives and occupations, and not futures involving being run by machines.

Glenn
That was another episode of Why Can’t They Just? The theme music that we use for this podcast is a piece called Insurrection by Pierre Chrétien, performed by the Soul Jazz Orchestra, courtesy of Do Right Music Inc.

Janaline
You can also hear us on Canberra community radio, 2XX FM 98.3 on Tuesdays between 6 and 7, or via 2XXfm.org.au. If you like our work, consider supporting us on Patreon, via our website whycanttheyjust.com.au.

Glenn
I’m Glenn Davidson.

Janaline
I’m Janaline Oh, and this is Why Can’t They Just?