AI poses the greatest threat to jobs since the beginning of the Industrial Age. This excellent story is by AFR’s James Eyers.
Prime Minister Anthony Albanese strutted the global stage this week in New York, building a political coalition to regulate artificial intelligence and protect societies from existential risks.
With all the talk about AI spinning out of control, including this week’s revelation that an OpenAI agent had hacked the Services Australia website, most of the attention has been on creating stronger guardrails and introducing more oversight of agents given their unpredictable nature.
“AI could be a risk to humanity as a whole”, Dario Amodei, the chief executive of US frontier AI lab Anthropic, told the United Nations Security Council this week.
His essay, We Must Pace the Frontier, which this month called for a slowdown of model development to allow regulation to catch up – set the tone at the UN. Dario Amodei, CEO of Anthropic, speaks remotely during a Security Council meeting on Wednesday.
However, the big AI companies and national governments face another dilemma, one that could be even more serious than rogue agents penetrating confidential data: AI is advancing so rapidly it could only be a matter of time before economies see widespread disruption of the workforce.
“Over the next 40 years, generative AI will change the economics of information in ways that transform the nature of work,” the Intergenerational Report warns.
“The ways that workplaces, workers and skills systems adopt and respond to AI will have a significant influence on the nature of labour market changes observed in response to AI, globally and in Australia.”
So far, at least, AI’s impact on jobs has been at only the margins. Dire predictions about mass lay-offs have not eventuated.
When ASX-listed Block cut almost half its 10,000-person workforce in February, its chief executive, Jack Dorsey, said AI had made many of the roles unnecessary, and he predicted: “Within the next year, the majority of companies will reach the same conclusion.”
Treasury points to the likelihood that AI will ultimately automate many tasks for which it is a direct substitute for human labour. Of those new jobs in the United States, many are in infrastructure and construction supporting the data centre build; employment is already falling for customer-service workers, secretaries and administrative assistants.
Some economists think that over the coming years, many companies will become trapped in a race to automate. This could ultimately harm economies, by eroding the consumer spending on which their businesses depend. The concept is known as the “AI lay-off trap”.
“If AI displaces human workers faster than the economy can reabsorb them, it risks eroding the very consumer demand firms depend on,” says the paper that introduced the concept, published in
Put another way, AI might be used to produce goods and services more quickly and cheaply than ever, but if people don’t have an income – or are worried they’ll be next on the scrapheap – will they want to buy those goods or services?
Red lights are already flashing as some younger people find it harder to enter the labour market.
The Stanford Digital Economy Lab found employment of US workers aged 22 to 25 in AI-exposed occupations is 19 per cent below where it would be had it kept pace with that of less-exposed peers. Researchers described this as “early, descriptive indicators – canaries in the coal mine”.
The recent Intergenerational Report notes the “wide range of potential future outcomes” and it threw out some thorny situations that economic managers may have to contemplate.
These included the potential for “productivity gains [to] flow through to lower prices rather than higher incomes” and “labour-market transitions [to] increase demand for government support”.
The Intergenerational Report also says AI has the potential to reshape the economy and society more significantly and rapidly than previous technologies, although the scope and timing are uncertain.
“Faster diffusion risks amplifying adjustment pressures, which would require businesses, workers, and governments to adapt more quickly to the changing environment,” the report said.
This raises questions about whether Australia is engaged enough in a subject that could – assuming humanity has not been wiped out – fundamentally change the local national and global economies within just a few years.
Some reckon it is time to consider a new, grand bargain to ensure the right incentives are in place to adopt AI while also looking after workers and reducing the risk the lay-off trap becomes a reality.
“Australia needs to ensure AI improves the lives of Australians, not just of Elon Musk,” says Nick Dyrenfurth, a historian and author who runs the left-leaning John Curtin Research Centre, which is developing AI policy for labour markets.
If AI automation in companies accelerates from here, policy that protects the income tax base may become essential. Deliberate decisions on where AI adds value – and where human expertise remains necessary – will be required.
Discussion will need to consider what sort of AI society Australia wants: the same amount of work being done by fewer workers, or doing more with the workers we have, to expand the economy and ensure enough new jobs are created for the next generation.
Some of the AI labs paint a bleaker picture for jobs than the Intergenerational Report did this week. This month, Anthropic published an analysis, Scenarios for our Economic Future, that warned major economic disruption is possible in the next five years.
One of its versions of the future envisioned a society far richer than we are now as gross domestic product growth soared on the back of AI productivity gains, but with fewer workers doing knowledge work – and unemployment much higher than previous recessionary levels.
“It’s a 10-year shift in a couple of years,” says Nicholas Davis, co-director of the Human Technology Institute at University of Technology Sydney, of the extreme Anthropic scenario based on high AI uptake.
“When you’re talking about AI and Australia, a shift to capital through AI means dollars flowing overseas,” says Davis. “That scenario means Australia’s tax base is quite eroded: we can’t tax that capital if it’s flowing straight offshore.
“There is a discussion that needs to happen. Let’s be clever about this. How do we manage incentives, and how do we generate onshore sources of income the public can benefit from?”
Johanna Weaver, executive director of the Tech Policy Design Institute, a think-tank based in Canberra, says Australia needs clear policy for AI’s impact on jobs.
“With a plan, we can shape workforce transition; without a plan, we will have workforce disruption,” says Weaver. “We have agency to shape how AI affects our economy, productivity and jobs. But we need to be intentional about our choices and prioritisation today.
“Building agency requires more than data centres and models. It also depends on the workforce skills, economic policy and institutional capability needed to ensure the benefits of AI are broadly shared across society.”
The John Curtin Research Centre has made nine recommendations on these issues in it’s For All of Us report that was published in April. These include establishing an AI industrial relations taskforce to assess the Fair Work Act for AI employment risks, and creating an expert panel on AI inside the Fair Work Commission.
“Managed poorly, (AI) risks reproducing the worst patterns of earlier transitions – privatised gains, socialised costs, insecure work and democratic discontent.”
The centre also suggests the federal government engage with the states on payroll tax incentives for businesses that pledge to protect workers’ roles.
Lawyer John Grant and former Treasury executive Barry Thomas suggest another tax-driven mechanism: a sliding-scale corporate tax rate that would reward companies who maintain human staff with lower rates and taxing heavy automation, which they call the Sustainable Future Employment model.
“It’s not a binary ‘AI is going to take all our jobs’ or it isn’t. The risk is that it takes a fair chunk of our jobs, of which are most exposed is the knowledge work,” says Thomas.
“You don’t have to lose all that many jobs to blow a hole through the economy. Treasury should be doing more than they are. It’s about just slowing it down enough, so we can cope.”
The AI companies themselves have plenty of economic policy ideas to offset the impact on societies of expected workforce disruption.
There is a growing debate about a stronger wage insurance market, to help fund a wider social safety net for those displaced.
It is too early to have definitive answers, but policy debate is needed so possible solutions are thrashed out independently of the AI companies themselves. There is certainly a growing sense of urgency that carefully thought-out public policy needs to be developed in this area.
In July, 16 Nobel laureates joined with more than 200 economists and researchers to sign an open letter titled “We Must Act Now”that warned AI may become radically more powerful over the next decade, bringing the risk of large-scale job displacement.
“This could drive an unprecedented transformation of our economy, larger than the industrial revolution, but unfolding over a vastly shorter time frame,” the group said.
“Economists, policymakers and technology leaders must act now to understand the economics of transformative AI and to build the incentives, guardrails, and institutions needed to steer AI in a direction that complements humans and benefits society.”