Extract from ABC News
Analysis
Let's have the debate on how to distribute AI on Australia's terms, says Amelia Lester.
The 2025 Boyer Lectures will be broadcast on ABC Radio National and ABC News Channel on Sundays from October 19.
The sci-fi author William Gibson has said the future is here, it's just not very evenly distributed. Well, I'm here to say the future is in San Francisco, where a modern-day gold rush is unfolding. Because it's in that city that artificial intelligence is already a part of everyday life.
We've all heard the hype. The most important technology of any lifetime, Eric Schmidt says. The most transformative technology since electricity, according to Stephen Hawking. Sundar Pichai, the chief executive of Alphabet, says AI will be more important than fire. But it wasn't until I got into a self-driving Jaguar in San Francisco — a Waymo — that I finally got the enormity of what is coming, and how fundamentally it will change everything.
At first I was nervous, but as I watched the wheel steer itself — gradually and almost imperceptibly, just as a series of driving instructors have failed to teach me — I came to actually prefer the experience to a regular taxi. All the human variables were gone. Did the driver sleep well last night? What if he goes on a political rant? I actually hate this song. Please don't take the long way around. It felt both safer, and easier.
While in the car I passed a billboard for an AI startup. It read: STOP GIVING HUMANS JOBS THEY HATE. Could AI in fact usher in a post-capitalist Utopia, as the ad suggested? Since that trip I've been swinging between bouts of wild optimism and assuming the robots will eventually kill us all.
It wasn't until I got into a self-driving car in San Francisco — a Waymo — that I finally got the enormity of what is coming. (Reuters: Paresh Dave)
These are clearly two extreme scenarios, but I'd wager the thinking is probably familiar to you. Overall, polling shows the public feels more pessimistic than optimistic about AI — that we're afraid of it, even. There is a sense of inevitability about the technology's growth and eventual ubiquity. Why do we collectively feel this lack of agency about AI? I think a big part of this is that it's really difficult to talk about.
Before I suggest some reasons for this, I want to pause and quickly establish what I mean by artificial intelligence here. AI is a grab bag name for technologies that are supposed to mimic human capabilities. Most of us are familiar with at least one type of AI — the generative kind, think ChatGPT — but there are other types, too, such as machine learning.
So far, since its beginnings in the 1940s, AI has contributed many "narrow" systems which can do useful individual tasks in particular circumstances. Yet these systems consistently fall short of ambitions to make something that "has it all". It's possible that large language models — a type of generative AI — might be different, but there's a lot of warranted scepticism about them. This scepticism has only increased as new models have been released and have proven underwhelming — to the point now where there's some talk of AI as a bubble.
This is maybe the first reason why it feels difficult to talk about AI — its trajectory is shrouded in uncertainty.
This brings me to a second, related reason: a handful of big tech companies control what we know about it. Unlike the atom bomb, the internet, radar, or space exploration, artificial intelligence is really the first big technology of modern times not developed by governments. It's been siloed in the private sector. And because these companies want to consolidate oligopoly control over the AI ecosystem, we're constantly having to parse what's factual and what's hype.
But just because AI's hard to talk about, that doesn't mean we have to resign ourselves to it, or any technology, being harmful to humanity. It's the job of the tech companies to articulate to consumers what AI can do for us. And then it's our job as citizens to forge a shared public vision about its use.
It feels like Australia, especially, has been caught a little bit unawares by all this — not because we don't have world-leading scientists and technologists working on it, we do, but because it's not yet consuming the public debate in a way that it probably should.
For democracies like Australia, AI poses a lot of risks. It could degrade our information environment further. It has the potential to impinge on our civil liberties. It undermines the notion of wage labour. It even threatens to destabilise or call into question the essence of what makes us human. In some ways, AI itself could be described as a technology with illiberal tendencies — one which favours autocracies and maybe even has the potential to create them.
Australia is small. We're never going to win on compute power alone. For one, we're not able to spend money on AI like the superpowers can. The United States is spending $230 billion a year on it. But here in Australia, there are fewer billionaires, and they're not yet calling all the shots. We remain a robust democracy, and we believe in a fair go for everyone.
That's why Australia has the chance to become a beacon. We can be clear-eyed about the risks and rewards of AI, especially when it comes to the world of work. And then, rather than pursue maximum profit for a few, we can ensure artificial intelligence is integrated safely and humanely — on our terms.
Before I get to those points, though, I need to first zoom out and give you the big picture. I need to tell you about the geopolitical contest for AI supremacy that's very much under way.
A new world order
There's a "new world order" which will be forged depending on whether the US or China wins on AI. That's according to a former OpenAI developer named Leopold Aschenbrenner. Aschenbrenner's view about the stakes of this technology has been very influential — it's now held by many American technologists as well as by the US government.
President Donald Trump talks a lot about "beating" China in the "AI race". It's an odd way of putting things, since we don't know what the finish line is, but Trump's meaning is clear — he sees AI as a blunt tool of geopolitical competition. In fact, US Interior Secretary Doug Burgum went so far as to say recently we should forget about climate change — it's not important. Instead, Burgum said, the US was in an AI "arms race" with China to "save the planet".
So to save the planet, and win that race with China, the Trump administration has set about dismantling any regulations put on the AI industry by President Biden.
Getting rid of guardrails is precisely what Amazon, Google, Microsoft and Meta have lobbied for. It's why their chief executives went to Washington to celebrate Trump's inauguration. They want acceleration by any means possible — forget about the harms to the public.
US Interior Secretary Doug Burgum recently said the US was in an AI "arms race" with China to "save the planet". (Reuters: Brendan McDermid)
Of course, the binary between innovation and regulation is a false one. AI has a trust problem with consumers around the world. Elon Musk's Grok "hallucinating" support for neo-Nazis was a particularly egregious example, but all of the generative AI models are prone to unreliable and bizarre slop. This distrust keeps consumers from embracing AI even as we're constantly told how great it is.
We've seen this dynamic before — the idea that we should just trust the people selling us the shiny new thing. The internet and its development is a cautionary tale. Tech entrepreneurs "moved fast and broke things" and then the rest of us had to figure out how to fix them.
Take Australia's law prohibiting children from accessing social media platforms. It's a world first, and I think we should be proud of it. But the complexities around that rollout demonstrate how hard it is to retrofit existing systems with societal safeguards. The doom loop — that tendency for us to seek out and be served only the information we want, not the information we need — is potentially even stronger with AI than social media. That's why we need to get ahead of it now, before it's too late.
Social media is an example of something else, too — how Australia and other countries have typically exhibited a sense of powerlessness, as "tech takers". Because there's been a neoliberal open markets approach, Australia passively accepted consumer technologies like social media — and then discovered how hard it is to put the genie back in the bottle even with regulatory interventions. If Australia can discard that passivity and actively seek to shape the market in AI, it can demonstrate a different model.
To recap, the US has decided it's in an existential competition with China on AI. It's clearing the way to win by whatever means possible. But here in Australia, things look different. We need to be clear-eyed about capabilities and dangers and avoid the mistakes we made with social media and the internet. What are the risks of AI? What do we need to regulate against and avoid? It's easy to get caught up in worst-case scenarios. But what is already happening to our society, and to states, is what I want to show you today.
A unique threat to democracy
Most of us interact with AI through large language models. The "compute" needed to power these models is immense. Whenever we ask Perplexity or ChatGPT anything, the answer the model spits out uses up five drops of water, and nearly 10 times as much electricity as a Google search. The environmental costs of running the data centres which power AI will require massive amounts of cheap green energy.
Then there's the mineral extraction required, and the emissions. It's estimated that by 2026, data centre power demand will be roughly equivalent to the electricity consumption of Japan. These environmental costs will impact everyone, everywhere.
For democracies, though, how AI will be applied poses some additional risks. Depending on how models are used, and if they're left ungoverned, AI threatens to undermine democratic practice, norms and the rule of law.
For one, there's what AI will do to the information landscape. Experts warn AI propaganda factories are already here — with cheap, downloadable models, anyone can build a full-fledged persuasive framework from their bedroom. This is unprecedented, and dangerously destabilising. It also can't be easily turned off.
On the other hand, AI has the potential for creating a more informed citizenry. Australia's chief political scientist Dr Susan Foley has noted how the invention of the printing press in the 15th century was accompanied by very similar concerns — and that wound up starting a social revolution and the democratisation of education.
Even if the information dangers turn out to be overblown, the unique threat to democracy posed by AI systems goes much further.
AI is creating a power imbalance, redistributing what used to belong to individuals, and giving it potentially to the state, but also and perhaps more significantly, unelected tech executives. An increasingly concentrated group of corporate players control data, algorithms and processing infrastructure. They make their employees sign sweeping non-disclosure agreements and use aggressive silencing techniques against anyone who leaves.
This secrecy is meant to stop us asking too many questions about what we're told is inevitable. It also distracts us from the soup of machine learning we're already swimming in. Chatbots are everywhere. They're your banker, your therapist, even; occasionally, your child's educator. And they're collecting information on you all the time.
We are already living in an age of "surveillance capitalism" — Harvard's Shoshana Zuboff describes this as a "global architecture of behaviour modification". She writes that just as industrial capitalism disfigured the natural world in the 20th century, AI threatens human nature in the 21st.
You actually don't even need to go to an Ivy League professor to hear this. AI's biggest booster, the powerful venture capitalist, Peter Thiel, suggested in a 2019 talk that AI might be inherently authoritarian. Thiel called the technology "communist" and mused about how AI could bring about a world where "a few large computers control everything".
Sir Zelman Cowen warned of a growing threat to rights of the individual in Australia, including from "American listening devices". (The Australian Jewish News: Peter Haskin)
Key to this premise of a bloodless, synthetic future is how AI models can impinge on privacy. In his prescient 1969 Boyer lecture entitled The Private Man, Sir Zelman Cowen talked about a growing threat to rights of the individual in Australia, including from — wait for it — "American listening devices".
Cowen said Australians had tended not to guard their civil liberties as zealously as they should and now must. He went on to make the case for why privacy is a core value of a civilised society — and he outlined four components to it: one, it protects an individual's solitude; two, his intimacy in various groups of his own choosing; three, his anonymity; and four, his reserve — his ability to be cut off from unwanted intrusion.
All of these freedoms have been eroded by what we can broadly call "the internet". But they have the potential to be obliterated by artificial intelligence. The most obvious implication is in law enforcement. AI is an ubiquitous police officer, able to sift through unimaginable amounts of data with the potential for incessant enforcement. That's not something our current laws are designed for. And as the surveillance state gets really good, the human-ness of our justice system will decrease.
Our human-ness is being threatened in other ways, too. AI is right now a catastrophe for the traditional university, and maybe for high school too. It threatens that nexus of reading and writing and thinking which constitutes our concept of education. For artists and writers, there are similarly profound concerns. The writer Anna Funder talks about the ethical problem of generative AI — in that it plunders artistic creations without credit or compensation.
And when it comes to human relationships, there's a new story every day about AI's destabilising capabilities — models being trained to clone children's voices in order to scam their grandparents, or generating deepfake images without a subject's consent. We even now hear of chatbots convincing people to leave their families.
Things are not hopeless. There are solutions being floated to the information problem — like watermarks, or laws to regulate facial recognition technologies. We can legislate safeguards against both malicious actors and corporate interests which threaten our individual identity and safety.
But these practical remedies aren't enough. We have to think bigger. Not only do we need to protect individuals from AI overreach, we also need to empower them — give them access to AI tools and democratise the technology. After all — AI should work for us, not the other way around.
The idea of AI as augmenting human labour rather than replacing it is a good principle to keep in mind when we consider the topic of AI and the labour force. How we work may change beyond recognition in the next decade or so, and it's going to reorganise our society in the process. Let's take a look at what's already happening with AI at work, and then at what might happen next.
AI at work
AI models have arrived in the workplace and are already shaking things up. These models have the potential to make employees' work easier and less tedious, but they are also shifting the power balance towards employers. One way in which we see this is that employers are now able to assemble massive amounts of information about their employees. Australian unions are actively seeking tougher regulations on this point.
On a more fundamental level, AI models can perform jobs previously done by humans — just like I discovered first-hand in that San Francisco Waymo. As this process continues, it will cause enormous disruptions to the way our economy is structured — what jobs are available, and what jobs simply disappear. There's a fuzziness around what things will look like even in the near future, but that's a feature, not a bug, for employers — look at how they are already seizing on AI's nebulous potential to trim their workforces.
Nobel prize-winning economist Daron Acemoglu has looked into the implications of AI in the workforce for society writ large. His top line is that AI is likely to increase inequality. Crucially, he argues it will widen the gap between capital and labour income. There is some evidence this is what big investors in AI even want. Venture capitalist Marc Andreessen has surfaced a belief, prevalent in the Bay Area, that we only have a few years until the value of labour totally collapses.
Nobel prize-winning economist Daron Acemoglu's top line is that AI in the workforce is likely to increase inequality. (Reuters: Murad Sezer)
The tipping of the scales in favour of capital should ring alarm bells for Australians. This is, after all, a country in which salaried workers have long been on the backfoot to landlords. Australians are already being ripped off by our wealth-friendly, labour-hostile tax system.
But what about efficiency, I hear the Productivity Commission say. That body has warned the government that it should not use a heavy hand in regulating AI, because doing so would stifle billions of dollars in economic potential, and risk Australia "falling behind the curve".
To this I would counter that Australia is precisely the country that needs to take a stand on AI and its use in the workplace, because of its pioneering role in workers' rights. This is the nation that, thanks to an early and strong union movement, established vital labour protections — like the weekend, minimum wage, and landmark anti-discrimination laws. Australia was a trailblazer on these, just as we were with occupational health and safety reforms and the recently introduced right to disconnect law.
As Hugh Mackay has described it, the Australian work culture is about efficiency without rush — it prioritises both productivity and a laid-back lifestyle. It's a unique balance to have struck. What happens, though, when and if the notion of labour itself is undermined by machines which can do everything better, and faster, and cheaper? Computers do not care about the weekend.
In the short term, telling young people that machines have taken all the entry-level jobs, or are about to, will breed understandable nihilism and alienation. This is arguably already having a corrosive effect on our politics, and if young people feel their futures are doomed, it could potentially lead to social unrest.
Over the next 20 years, McKinsey estimates about half of all jobs could be automated. What happens to our social safety net, and our cohesion, if this comes to pass, and if those jobs aren't replaced with new ones? And what is the purpose of work anyway? Maybe we need to be asking what responsibilities governments have to citizens when many of them are no longer able to assume paid employment and file taxes. Radical ideas — including, especially, the concept of Universal Basic Income, or UBI — start to look very relevant.
The future is probably going to require a new social contract. Still, there's no need to succumb to fatalism — we can decide its terms. Let's talk about how to do that here in Australia. But though we can assert our digital sovereignty, we're not going to be able to do it alone. We'll need allies — to find them — let's zoom out again, to the big geopolitical picture.
It's up to us
Now, the obvious question to ask is, can we team up with our old friends the United States? In May, OpenAI chief executive Sam Altman told Congress: "We want to make sure democratic AI wins over authoritarian AI." Altman seems to be saying that American technology is inherently, almost definitionally, aligned with democratic values. With what's happening in the US right now, I'm not so sure we can count on this being true.
There's also a degree of state involvement in the American tech industry which should make us wary. The US government has taken a stake in Intel, the chip manufacturer, while US vice president JD Vance has said he will not accept foreign powers "tightening the screws" on US tech companies. Vance made that remark last year at the Paris AI Action Summit. And at that summit, the US notably refused to sign onto an agreement to develop AI that is safe, secure and trustworthy.
In some ways, China is taking quite a different approach. It goes to show that the binary between transparent democracies and opaque authoritarian states is no longer as clear cut as it once was. China, for instance, is promoting international cooperation on AI research — unlike the United States. It is also pushing for the equitable distribution of AI infrastructure worldwide.
Gathering for the ABC's Boyer Lectures are (from left) James Curran, Larissa Behrendt, Julia Baird, Justin Wolfers, Amelia Lester and John Anderson
The task now for Australia is to look for middle powers from Europe to Asia who want democracy, and safer artificial intelligence. With the power of multilateral institutions on the decline, we can't hold our breath for international accords. That means there is a unique responsibility for countries like Australia to team up with like-minded states and articulate the rules of the road.
Europe, which has led the way on internet regulation, is also making inroads with AI rules. Big tech is not happy with the European Union's new code of practice on AI — a sure sign they're on the right track. And while some say it doesn't go far enough, proponents say the code is a useful playbook for democratically making decisions about AI.
Beyond Brussels, and as with other elements of our foreign policy, we will want to find ways of forming sympathetic groupings in our region. Japan's Hiroshima AI Process is a global framework for AI governance. ASEAN has adopted an innovation-friendly approach to AI regulation, with an eye to preventing dominance in the market by either China or the US. And Vietnam is an interesting case study in a third way. It's developing a stack — the collection of software and hardware and tools to run AI applications — that draws on elements from the US, China, Canada and Japan.
For Australia, just as we led the world with our labour movement, we can do the same in developing safe artificial intelligence that works for us — and preserves how we want to work at the same time. This is not just an opportunity for a democracy like ours; ever since our superpower ally checked out of its responsibilities on this front, it's an obligation.
So let's finish on a high note. AI does have the potential to transform our society in positive ways. I've been concerned here with the specific dangers posed to democracies by AI, but there are also reasons to feel hopeful about it. We could build AI foundation models for science to drive innovation and progress. We could cure more diseases, extend lifespans, preserve Indigenous languages, increase civic engagement. And we could get rid of jobs no one likes doing – find new ways to use our gloriously capacious brains.
But AI will also undoubtedly test Australia on many levels: as a democracy; as a liberal democracy; as a country which values work, but which doesn't think work should be everything. Most of all, it will test Australia as a country which enshrines equality as its core value.
It's up to us, not unelected tech titans, to figure out how to govern and implement this technology fairly. When William Gibson wrote the future may not be evenly distributed, he could have been talking about how AI will transform our society. Let's have the debate on how to distribute AI on Australia's terms. It's time for us to get in the driver's seat.
Amelia Lester is deputy editor at Foreign Policy magazine in Washington. She was previously managing editor of The New Yorker and editor of the Good Weekend magazine.
No comments:
Post a Comment