AI is coming. Prepare for the worst

Its rapid growth and increasing sophistication threatens jobs across the economy – from radiologists and lawyers to middle managers and financial analysts.

The AI revolution could be different to – and bigger than – earlier technological advances that primarily displaced lower-skilled manufacturing, agricultural, and administration workers.

Some experts are sounding the alarm bell, warning of mass unemployment, professional workers having to move into lower paid roles and the collapse of the tax base that funds our public services.

Economist Daron Acemoglu warns that many companies are mainly using AI to replace workers to reduce costs.

These doomsday scenarios are unlikely, but they are not inconceivable.

While the Industrial Revolution lifted living standards in the long run, initially real wages stagnated for decades and working conditions deteriorated.

Australia should be alert, not alarmed, but should start preparing for disruption now. Our policymakers should be doing scenario analyses and practical contingency planning.

The balance of evidence suggests AI is more likely to augment than replace workers at scale, at least in the near term – but mass displacement is a scenario that policy makers can and should plan for.

The right policy response will depend on the extent of the AI shock and how it affects workers. Governments can also draw on lessons from the responses to past shocks.

Having policy options ready can enable a quick, effective response, potentially limiting the damage caused to individuals and the economy by long-term unemployment.

In the event of significant disruption, the federal government may need to consider how Australia’s safety net and retraining systems can be strengthened if more people need them for longer.

In a world of mass unemployment, the punitively low JobSeeker payment will not be enough.

The government could consider whether there is a role for unemployment insurance – a higher but time-limited payment, typically linked to a person’s previous income and used in many advanced economies – as a way of providing more support to people as they search for a new job or retrain.

Another option to support people through an extended period of unemployment could be to allow them early access to their superannuation.

Currently, only people under “severe financial hardship” – those who have received government income support payments for at least six months and are not able to meet reasonable and immediate family living expenses – can access their super early.

Policymakers could revisit the conditions under which early access to superannuation would be appropriate.

There are more radical options, such as Denmark’s “flexicurity” labour market system. It allows businesses to fire workers more easily, but in combination with a strong safety net (including unemployment insurance) and active education, retraining, and support programs.

This type of approach could help workers move more quickly into growing sectors.

Unions have also floated the idea of a shorter work week.

This idea is unlikely to gain traction in today’s tight labour market, but in a world in which there were big productivity gains from AI but less work to go around, a shorter work week could potentially share the benefits of work more widely, while increasing leisure time and minimising job losses.

Policymakers should also be prepared to rethink how we educate and train workers.

For example, education providers might need to offer more intensive programs that can retrain displaced workers quickly, building on their existing skills, so they can transition into growing sectors such as healthcare and clean energy.

Another idea could be to combine or require greater collaboration between universities and vocational training institutions, which could help speed up the retraining process and improve the quality of training.

People could fund their retraining via lifelong learning accounts, which are individual training accounts that workers can draw on throughout their careers. These could be funded through a combination of government contributions, employer levies, and worker savings.

Funding an expanded retraining system could be a major challenge. But there is also the potential that AI drastically reduces the cost of delivering education – for example through personalised AI tutors or adaptive learning platforms.

Hopefully these worst-case scenarios of major economic and social disruption never eventuate. But even in the more likely scenarios – where AI complements and augments human capabilities and creates new types of productive work – there will inevitably be some job losses as firms integrate AI into their operations, or laggard firms close. 

AI may simultaneously represent our greatest economic opportunity and our greatest social challenge. We don’t know exactly what’s coming, but our governments can and should prepare for the worst.