A major new study of Australia’s AI industry points to a gap between industry achievement and aspirations.
“Australia’s AI innovation is emerging organically from existing industrial capabilities rather
than developing in isolation,” Alexandra Bratanova and colleagues write in a new analysis for the feds.
It’s a pragmatic approach, “a distinctive hybrid positioning as a developed ‘AI‑taker’ and a developing ‘AI‑maker’. It balances “global technology adoption with targeted domestic innovation in areas of competitive advantage,” they write in an information-rich, cogent and concise assessment of AI in Australia.
While they offer no opinion on the techno-patriot position, the authors state it is especially important that Australia must develop – and not merely import – AI systems. Their evidence makes plain that while industry is geographically neutral when it comes to adopting what works, participants want an all-Australian AI plan.
Thus, the authors report industry-opinion that:
“developing sovereign AI capability has become essential, as control over these technologies increasingly defines economic and geopolitical power. They indicated strategic investment must balance immediate commercial applications with longer term foundation‑building to ensure Australia can shape AI development pathways aligned with national values and interests.”
Even so, on the ground, smallish businesses are taking what they need from where they can find it.
“A growing number of companies and research teams are developing proprietary AI tools, though much of the ecosystem remains reliant on globally-developed foundation models.”
This may be because of scale, with 85% of private AI businesses having fewer than 50 people, making for a, “vibrant but potentially fragile innovation ecosystem.”
And an AI-for-Australia approach will need more connections between lab and market, pointing to “a disconnect between research activity and commercial outcomes”, on the basis of 93,000 AI publications 2015-24,4,075 patents and 1.88% of global AI publications, Australia accounts for a tiny portion of the world’s activity – only 0.18% of global AI patents.
Overall, they identify “key characteristics” of the state of the AI nation, including;
- “Dual-track ecosystem:” new (2023-’24) sector-specific private companies, in healthcare, logistics, “creative industries. Established public companies integrating AI – in energy, mining, finance, healthcare
- “Organic clustering and geographic diversity”: 68 per cent (858) of AI companies in 25 geographic clusters, generally in capital cities but also in regions, Gold Coast, Sunshine Coast, Newcastle.” Transcends simple urbanisation, representing organic innovation-district formation.
- AI aligning with traditional industries: “this creates a distinctive national AI profile where technical innovation enhances rather than disrupts established capabilities”
- Imbalance between discovery research and product innovation: 1.88 per cent of global AI publications but 0.18 per cent of patents. It is “particularly acute” in deep learning and reinforcement learning, “where Australia shows strong research output but limited patent activity.” “These patterns suggest structural barriers in the innovation pipeline between academic research and industry applications, potentially limiting the economic impact of Australia's considerable AI knowledge production.”
- Distinct public and private sector focus: Public providers adopt more than develop, in energy, raw materials, utilities, healthcare, IT infrastructure. Private companies specialise, notably in business process, IT infrastructure and media/marketing
- Robust growth: continues but not as strongly as rest of world
- Broad-based research: including (who would have thought) vet science and the humanities
- Recruitment for new skillsets: moving from general implementation skills to specific applications requiring specialist knowledge.“ It also highlights the need for education systems to produce graduates capable of applying AU in specific industry contexts
- Culture-shift needed: “A national culture shift – underpinned by strategic leadership, public trust and robust public storytelling – is essential to convert societal apprehension into confidence”
Can calls for a national AI commissioner be far away?