It seems a lifetime ago, but it was 2017 when the former NBN CEO Mike Quigley and I wrote a book about the impact of technology on our labour market.
Changing Jobs: The Fair Go in the New Machine Age was our attempt to make sense of what rapid technological change means for Australian workers.
It sprang from a thinkers' circle Andrew Charlton and I convened regularly back then, to consider the biggest most consequential shifts in our economy.
Flicking through the book now makes it very clear the pace of change has been breathtaking since.
The stories of Australian tech companies give a sense of its scale.
In 2017, the cloud design pioneer Canva was valued at $US1 billion - today, more than $US30 billion.
The data centre leader Airtrunk was opening its first 2 centres in Sydney and Melbourne. It now has 6 times that and is backed by one of the world's biggest investors.
We understand a churning and changing world is a source of opportunity but also anxiety for Australians.
While the technology has changed, our goal as leaders hasn't.
The responsibility we embrace is to make Australian workers, businesses and investors beneficiaries, not victims, of that change.
That matters more than ever in a new world of artificial intelligence.
Breakthroughs in 'large language models' (LLMs) - computer programs trained on massive datasets that can understand and respond in human languages - have triggered a booming AI 'hype cycle' and driving a 'cognitive industrial revolution'.
ChatGPT became a household name in a matter of months and has reframed how we think about working, creating and problem‑solving.
LLMs have been adopted 7 times faster than the internet [PDF 1.184 MB] and 20 times faster than electricity. The rapid take‑up has driven the biggest rise in the S&P 500 since the late 1990s.
According to one US estimate, 8 out of 10 workers could use LLMs for at least 10 per cent of their work in future.
Yet businesses are still in the discovery phase, trying to separate hype from reality and determine what AI to build, buy or borrow.
Artificial intelligence will completely transform our economy. Every aspect of life will be impacted.
I'm optimistic that AI will be a force for good, but realistic about the risks.
The Nobel Prize‑winning economist Darren Acemoglu estimates that AI could boost productivity by 0.7 per cent over the next decade, but some private sector estimates are up to 30 times higher.
Goldman Sachs expects AI could drive GDP growth up 7 per cent over the next 10 years, and PwC estimates it could bump up global GDP by $15.7 trillion by 2030.
The wide variation in estimates is partly due to different views on how long it will take to integrate AI into business workflows deeply enough to transform the market size or cost base of industries.
But if some of the predictions prove correct, AI may be the most transformative technology in human history.
At its best it will convert energy into analysis, and more productivity into higher living standards.
It's expected to have at least 2 major economy‑wide impacts.
First, it reduces the cost of information processing.
One example of this is how eBay's AI translation tools have removed language barriers to drive international sales. The increase in cross‑border trade is the equivalent of having buyers and sellers 26 per cent closer to one another - effectively shrinking the distance between Australia and global markets.
This is one reason why the World Trade Organisation forecasts AI will lower trade costs and boost trade volumes by up to 13 per cent.
Second, cheaper analysis accelerates and increases our problem‑solving capacity, which can accelerate innovation by reducing R&D costs and skills bottlenecks.
By making more projects stack up commercially, AI is likely to raise investment and boost GDP, and generate demand for human expertise.
Despite the potential for AI to create more high‑skilled, high‑wage jobs, some are concerned that adoption will lead to big increases in unemployment. The impact of AI on the labour force is uncertain, but there are good reasons to be optimistic.
One study finds that right now, more than half of the use cases of LLMs involve workers iterating back‑and‑forth with the technology, augmenting workers' skills in ways that enable them to achieve more.
Another recent study found that current LLMs often automate only some tasks within roles, freeing up employees to add more value rather than reducing hours worked.
These are some of the reasons many expect the AI transformation to enhance skills and change the nature of work, not cause widespread or long‑term structural unemployment.
Even so, the impact of AI on the nature of work is expected to be significant.
We've seen this play out before - more than half the jobs people do today are in occupations that didn't even exist at the start of the Second World War.
Some economists have suggested AI could increase occupational polarisation - driving a U‑shaped increase in demand for manual roles that are harder to automate and high‑skill roles that leverage technology, but a reduction in demand for medium‑skilled tasks.
But workers in many of these occupations may be able to leverage AI to complete more specialised tasks and take on more productive, higher‑paying roles. In this transition, the middle has the most to gain and the most at stake.
There is also a risk that AI could increase short‑term unemployment if investment in skills does not keep up with the changing nature of work.
Governments have an important role to play here, and a big motivation for our record investment in education is making sure skills keep pace with technological change. But it's also up to business, unions and the broader community to ensure we continue to build the human capital and skills we need to grasp this opportunity.
To be optimistic about AI is not to dismiss the risks, which are not limited to the labour market.
The ability of AI to rapidly collate, create and disseminate information and disinformation makes people more vulnerable to fraud and poses a risk to democracies.
AI technologies are also drastically reducing the cost of surveillance and increasing its effectiveness, with implications for privacy, autonomy at work and in some cases, personal security.
There are questions of ethics, of inequality, of bias in algorithms, and legal responsibility for decision‑making when AI is involved.
These new technologies will also put pressure on resources such as energy, land, water and telecoms infrastructure, with implications for carbon emissions.
But we are well placed to manage the risks and maximise the opportunities. In 2020 Australia was ranked sixth in the world in terms of AI companies and research institutions when accounting for GDP. Our industrial opportunities are vast and varied - from developing AI software to using AI to unlock value in traditional industries.
Markets for AI hardware - particularly chips - and foundational models are quite concentrated. About 70 per cent of the widely‑used foundational models have been developed in the US, and 3 US firms claim 65 per cent of the global cloud computing market.
But further downstream, markets for AI software and services are dynamic, fragmented and more competitive. The Productivity Commission sees potential to develop areas of comparative advantage [PDF 681 KB] in these markets.
Infrastructure is an obvious place to start.
According to the data outfit IDC, global investment in AI infrastructure increased 97 per cent in the first half of 2024 to $US47 billion and is on its way to $US200 billion by 2028. We are among the top 5 global destinations for data centres and a world leader in quantum computing.
Our landmass, renewable energy potential and trusted international partnerships make us an attractive destination for data processing.
Our substantial agenda, from the Capacity Investment Scheme to the Future Made in Australia plan, will be key to this. They are good examples of our strategy to engage and invest, not protect and retreat.
Our intention is to regulate as much as necessary to protect Australians, but as little as possible to encourage innovation.
There is much work already underway: our investment in PsiQuantum, AI Adopt Centres and development of Australia's first voluntary AI safety standard, putting AI on the critical technologies list, a national capability plan and our work on R&D.
Next steps will build on the work of colleagues like Charlton, Tim Ayres and before him Ed Husic, and focus on at least 5 things:
- building confidence in AI to accelerate development and adoption in key sectors
- investing in and encouraging upskilling and reskilling to support our workforce
- helping to attract, streamline, speed‑up and coordinate investment in data infrastructure that's in the national interest, in ways that are cost effective, sustainable and make the most of our advantages
- promoting fair competition in global markets and building demand and capability locally to secure our influence in AI supply chains
- working with Katy Gallagher to deliver safe and better public services using AI.
Artificial intelligence will be a key concern of the economic reform roundtable I'm convening this month because it has major implications for economic resilience, productivity and budget sustainability. I'm setting these thoughts out now, to explain what we'll grapple with and how.
AI is contentious and of course there is a wide spectrum of views, but we are ambitious and optimistic.
We can deploy artificial intelligence in a way consistent with our values if we treat it as an enabler not an enemy, by listening to and training workers to adapt and augment their work.
Because empowering people to use AI well is not just a matter of decency or a choice between prosperity and fairness, it is the only way to get the best out of people and technology at the same time.
It is not beyond us to chart a responsible middle course on AI, which maximises the benefits and manages the risks.
Not by letting it rip, and not by turning back the clock and pretending none of this is happening, but by turning algorithms into opportunities for more Australians so they're beneficiaries not victims of a rapid transformation which is gathering pace.
