The Between Times: Why Australia Must Build AI, Not Just Use It

26 Aug, 2026

There is a moment in the history of every transformative technology when the demonstration has already happened, but the promise has not yet been realised. The gap between those two points is where the most consequential decisions get made, and where the nations and organisations that will shape the future separate themselves from those that will merely inhabit it. 

We are in that moment now with artificial intelligence. And Australia needs to decide.  

We Have Been Here Before 

In 1879, Thomas Edison demonstrated the incandescent light bulb. Twenty years later, only 3% of American households had electricity. Another two decades passed before it reached half the population. Those forty years were not a period of stagnation. They were a period of profound reconfiguration, as industries, cities, and social structures rebuilt themselves around a new capability. This is what Joshua Gans in his book Power and Prediction calls the “between times”. 

The industrial revolution offers an even starker lesson. Carl Benedikt Frey’s landmark work The Technology Trap traces the story from the lamp lighters displaced by gas lighting, through the Luddites, to the full arc of mechanisation. His conclusion is uncomfortable: “the industrial revolution created extraordinary wealth and prosperity over the long run, but the immediate consequences of mechanisation were devastating for large swaths of the population. Middle-income jobs withered, wages stagnated, the labour share of income fell, profits surged, and inequality skyrocketed”. These trends broadly mirror those we are already seeing in the current age of automation. 

The lesson is not that transformative technology is bad. It is that the between times are where the pain is concentrated, and where policy choices determine who bears that pain and who captures the gains. 

AI is entering its between times right now. The technology has been demonstrated beyond any reasonable doubt. The realisation of its full promise across economies and societies is still years away. The decisions made in this window will determine which countries help set the rules, which companies capture the value, and which populations are left to absorb the disruption without a seat at the table where the disruption is being designed. 

This Time Really Is Different 

For those of us who have been in the field of computing long enough to remember the second AI winter in the 1980s and 1990s, AI was sometimes flagged as a dead-end for your research career, funding agencies and corporations dramatically cut-down budgets on AI. Where research continued it was under different labels like machine learning or statistical methods.  Quietly some significant advancements were made in deep learning and computer vision.  

With the arrival of elastic compute fuelled by cloud technology, the rise of big data, and an extraordinary volume of investment from big technology companies, the game changed. The combination produced breakthroughs that the previous generation of researchers had theorised about but could not build. The current wave of AI is not simply a new version of what came before. It is a genuine capability shift, one that touches reasoning, language, image and audio understanding, scientific discovery, and autonomous action in ways that prior waves did not. 

The workforce numbers reflect this. The World Economic Forum’s Future of Jobs Report projects approximately 92 million jobs displaced by 2030 and around 170 million new jobs created, a net gain of roughly 78 million globally. But aggregate numbers obscure the asymmetry underneath them. It is not careers that AI displaces but tasks, and the distribution of which tasks is deeply uneven. Repetitive and routine work is most exposed. Creative, critical, and interpersonal work is less so. The transition will not be painless, and the pain will not be shared equally. 

The Privilege Problem 

Access to AI is not equally distributed, and the gap is widening. Research into what might be called “AI Privilege” reveals a pattern that mirrors the digital divide, but with higher stakes and faster acceleration. Income, education, and age all predict meaningful differences in AI access, capability, and the quality of outcomes people receive from AI systems. 

There is also a subtler risk that deserves more attention: skills erosion. When AI handles the repetitive work, the humans who never learned to do that work lose the foundation from which deeper expertise is built. We are already observing this in education. Globally, enrolments in computing and software engineering are declining. If that trend continues, we risk producing a generation of AI users who cannot evaluate, correct, or improve the systems they depend on. It is a dynamic uncomfortably similar to what happened with Y2K, when market demand caused a rapid deterioration of what it means to be a qualified programmer.  

For Australia specifically, this raises a question not just about individual citizens but about the nation’s relationship with the global AI ecosystem. If the systems Australians rely on are built elsewhere, trained on data from elsewhere, governed by rules shaped elsewhere, and optimised for markets elsewhere, then the AI Privilege gap operates at a national level as much as an individual one.

Australia’s Investment  

The AI opportunity is routinely expressed in the hundreds of billions to trillions of dollars. Australia’s response, measured in public investment, tells a different story. 

Since 2021, Australia has committed approximately 200 to 300 million dollars in dedicated AI investment. This includes the $53.8 million National AI Centre through CSIRO, $24.7 million for next-generation AI graduates, $17 million each for four AI Adopt Centres, $21.6 million for the National AI Centre’s Reshape program, $29 million for an AI Safety Institute, and $53 million for an AI Accelerator Cooperative Research Centre. These are not trivial sums and the programs they fund do real work. 

But when set against what peer nations are committing, and against the scale of the economic opportunity, the picture shifts as highlighted in the Australian Academy for Technological Sciences and Engineering’s recent report on Unleashing Growth: Australia’s AI Investment Blueprint. Beyond the United States and China, the United Kingdom, Canada, France, Germany, and the South Korea are each investing orders of magnitude more in AI research, development, and sovereign infrastructure. The gap is not about effort or intent. It is structural, and it reflects a strategic choice. 

The narrative in Australia has been captured almost entirely by adoption. We are investing in helping businesses use AI, training workers to work alongside AI, and regulating AI as we receive it from elsewhere. These are worthwhile activities. But they are the activities of a consumer, not a creator. And in the AI economy, the returns to creation and the returns to consumption are very different.

The Renter Economy Risk 

Sally McManus, Secretary of the Australian Council of Trade Unions, recently wrote in the Australian Financial Review about the risk of US technology giants “running roughshod” over Australian workplaces. The concern goes beyond labour relations. The deeper risk is that Australia becomes what some have started calling a renter economy of AI: a country that pays to access AI capabilities developed elsewhere, on terms set elsewhere, governed by standards set elsewhere, and aligned with values and interests that may or may not reflect our own. 

This is not a hypothetical. The regulatory frameworks being developed internationally, the safety standards being debated in Brussels and Washington, the norms around data governance and model transparency: these are being shaped by countries with strong domestic AI research ecosystems and by the companies headquartered in those countries. Australia’s voice in those conversations is proportional to our credibility as a player in AI development. A country that only adopts does not get a seat at the table where the rules are made. 

Australia has genuine strengths to build on. We have world-class universities producing excellent AI research. We have an established track record in applied technology and a regulatory environment that, at its best, is both rigorous and pragmatic. The Australian Research Alliance for Enterprise AI, established in 2024 and spanning multiple universities and industry partners, is working to translate that research capability into leadership in enterprise AI applications.  

But infrastructure takes time to build. Research ecosystems take decades to mature. The window in which the foundational choices about AI are being made is not unlimited. The between times are already underway. 

Not at the Table 

The industrial revolution’s lesson was not simply that the technology was disruptive. It was that the countries and institutions that shaped the technology captured the gains, while those that merely adapted to it absorbed the costs. The steam engine was not invented by the countries that eventually adopted it most enthusiastically. But the countries that built the engines, the factories, and the intellectual frameworks around them wrote the rules of the industrial age. 

There is an old principle in international affairs, that captures the stakes precisely.  

If you are not at the table, you are on the menu.

Australia needs to decide which it wants to be. 

Professor Shazia Sadiq FTSE is Director of the Centre for Enterprise AI, Centre for Information Resilience (CIRES), and AI Research Network at The University of Queensland (UQ). Shazia also leads the Australian Research Alliance for Enterprise AI.