Stop reassuring yourself with history: Why the AI disruption pattern is different

For over 250 years, labour markets have recovered from every technological disruption through four key mechanisms - time, cognitive higher ground, capital diffusion, and state action. But this time around it's different.

Employment and Workplace Relations Minister Amanda Rishworth opened her address to the AFR Workforce Summit recently by linking the current approaching wave of AI transformation with the Luddite opposition to textile mechanisation. It's a well worn analogy and seems to be emblematic of the federal government's position on AI as "just the next industrial revolution". In their minds this is a well worn pattern of misplaced fear that history has repeatedly shown to be irrational, because the labour market always recovers, and aggregate employment numbers have remained steady for generations.

But what they are missing is there is something structurally different this time around.

The historical pattern is real and that's why it's so seductive

Every major technological shock for the last 250 years has produced a confident wave of predictions that work was about to end. Thomas Mortimer, the late-1700s economist who was credited as the first to write about "bulls and bears", argued the power loom would permanently remove men from the labour force. David Ricardo, one of the most influential classical economists, initially defended mechanisation during the industrial revolution, but then later recanted having decided that it would displace labour at a scale that threatened the working class. In 2013 Oxford economists Carl Benedikt Frey and Michael Osborne suggested that 47% of US jobs were at high risk of automation within two decades.

Every one of those predictions overshot.

Indeed, history is littered with these sorts of predictions about workforce displacement that never eventuate. The handloom weavers were displaced, but the textile sector grew. The microprocessor displaced clerks but turned bookkeepers into analysts. The internet flattened bricks-and-mortar retail but created entire new categories of work. Two centuries on, the labour market is still running near full employment in most developed economies.

So why did they all get it so wrong and how is it that the labour market always recovers back to the same steady levels of employment?

The recovery pattern

Drawing on Daron Acemoglu's task-displacement framework, Goldin and Katz on the race between education and technology, and the broader economic history of past shocks, there are four clearly identifiable mechanisms that have quietly done the recovery work every time we hit some sort of technological shock.

• Time - change happened over decades, long enough for institutions to catch up.

• Cognitive higher ground - humans always had a higher rung of cognitive work to escape into.

• Capital diffusion - the technology eventually got cheap and democratised.

• State action - governments stepped in with retraining, education expansion, labour protections, and demand stimulus.

The question no one seems to be asking is will these tried and true mechanisms be there to save us this time?

Time is definitely not on our side

Demis Hassabis, the DeepMind CEO and Nobel laureate, recently claimed AGI is "going to be something like 10 times the impact of the Industrial Revolution, but happening at 10 times the speed".

Engels' Pause - the lag between productivity rising and real wages catching up was 80 years for the power loom and 30 years for the microprocessor. AI capability is doubling roughly every six months. Universities, professional bodies and governments operate on five to ten year cycles.

When the disruption curve is steeper than the recovery curve, there's very little time for any sort of recovery to happen.

There are no more rungs for us to climb

Prior shocks saw technology substituted for muscle (steam, the assembly line), then precision (Bessemer, robotics), then calculation (microprocessor). Through all of these humans migrated up the cognitive stack, and our ability for synthesis, judgment, and original creative work was never challenged.

AI is the first technology that climbs the synthesis layer itself.

Jobs and Skills Australia commissioner Barney Glover recently said that AI-driven job losses can be expected to exceed the current 600,000 positions forecast, which is 4% of the national workforce, with substantial repercussions for banking, law and accounting. Those are the cognitive higher-ground professions humans have previously escaped into and there aren't a swathe of new data centre jobs or AI engineering jobs on the horizon - data centres once built require minimal staff and Ai is beginning to engineer itself.

Capital is concentrating, not diffusing

In past technology shocks the tech in question was quickly The loom, the car and the PC all got cheap because the technology was democratised - and so were the productivity gains.

AI is being built by four or five hyperscalers requiring billions in compute, and the moat is widening. When the technology sits in the hands of a few, the financial gains concentrate there too.

Productivity gains may arrive sharply this time, but they get captured on the corporate balance sheet rather than flowing through to wages, prices or new sectors. As Andrew Yang relayed on Scott Galloway's podcast said, it can be summed up with the statement that "capital displaces labour with the help of AI".

State's are pouring fuel on the fire

Past recoveries leaned heavily on government intervention - the Wagner Act, the GI Bill, retraining and stimulus when the labour market lagged. This time the political machinery has been bought before it can act. AI labs hold hundreds of billions of market cap with direct lobbying lines into Washington.

Meanwhile no western government wants to hamstring its own AI labs while China builds theirs as fast as it can. The geopolitical race has neutered the political will to regulate or intervene at all and for the most part AI development and use is regulation free.

So what happens next?

For centuries these recovery mechanisms have held fast, working together to help the workforce absorb every technological shock thrown at it. This time it looks different though, the stars aren’t aligning and it is steering us towards a significant, permanent, structural change in employment levels.

It won't happen overnight of course, and it will come in waves.

White collar jobs are clearly first - LLMs already eat cognitive entry-level work and are climbing the synthesis layer fast which we can already see in US data. In time, blue collar will follow as AI-powered robotics catches up to the dexterity needed for the trades, manufacturing floors, and service work that has so far sat behind the analogue moat.

Our political and business leaders need to recognise the risks we are confronting and start preparing for a labour market that doesn't bounce back this time. Unfortunately, at present, there doesn't seem to be any pressure on them to do so, indeed it may be quite the opposite.