What are the limiting factors for AI today?

By Dr Paul Dalby, Business Development Advisor, Australian Institute for Machine Learning, the University of Adelaide.

This article is an extract from Artificial intelligence: your questions answered, a report published in partnership with the Australian Strategic Policy Institute (ASPI).  

The factors that limit the growth and expansion of AI are very different for Australia than for the rest of the world.

Globally, there’s a tidal wave of investment and activity in AI research, start-ups and existing companies. It’s true to say that there’s a stampede towards universities that train the talent needed to grow the AI sector and create the next generation of AI.

Australia’s R&D spend per capita is much lower than for similar-sized OECD economies, and our number of patents filed per million of population, at 14 per year, is way below any other comparable economy (the OECD average is 38 patents per million).1 This exemplifies an overall attitude to innovation that will limit Australian economic development in a Covid-normal world that requires increased self-reliance.

Australia’s comparatively low R&D commitment sets us up to underperform in industries that require investment in innovation, such as AI. Characterised by processes exhibiting high levels of automation, and increasingly AI, both competitive and comparative advantages will belong to the creators of technology, not the consumers. Given the inherently opaque nature of AI, this poses a number of dangers to Australia in allowing technological control to remain with other nations, non-state actors and foreign technology companies.

Image of city

Australia retains some of the best AI research talent in the world, and a number of our universities are ranked in the top 10 globally in various disciplines of AI research. Photo: iStockPhoto, online.

Automation creates both positive and negative outcomes, and if Australia doesn’t actively seek the advantages of automation—including ownership and industry optimisation—then we risk losing control of our future, and all we’ll be left with is the negative effects. We’ve seen this happen in industries such as advertising and transport, where Facebook, Google and Uber have employed high-value AI and software engineers in the US, and Australia has been left with fewer, lower value jobs to do the lower value work. Time is of the essence: numerous commentators, including the Harvard Business Review, warn that companies that wait to adopt AI might never catch up.2

This mismatch in investment is creating a brain drain for Australian AI talent, and many of our best and brightest in the field are being taken up by overseas, and mostly American, AI companies or AI-focused business units. The pace of development in AI makes maintaining the technical currency of staff in industry difficult. Experts can quickly lose currency and relevance if they aren’t constantly keeping up with the latest developments, creating both a potential barrier to global primacy for those who don’t keep up and also a barrier to new competitors for those that do. Without an ability to pay for a large internal team of fundamental AI scientists (like Amazon, Google and Facebook), the only practical mechanism for the Australian AI industry to maintain currency is an ongoing engagement with the research community—because that’s where the leading-edge technological development is occurring outside of the major global technology companies. Fortunately, Australia retains some of the best AI research talent in the world, and a number of our universities are ranked in the top 10 globally in various disciplines of AI research. Australian teams recently ranked 2nd and 3rd in global competitions run by the US Defence Advanced Research Projects Agency (DARPA) and NASA, respectively. This talent is an amazing natural advantage for Australia that must be nurtured.

Despite the rhetoric about AI becoming ubiquitous, there are constraints on the speed of AI development, including the following:

  • Governance. While the AI field is relatively new, effective organisational governance is still relatively primitive for AI. Of particular concern is data governance—who in the organisation is responsible for extracting value from a company’s data, and who is protecting the interests of data contributors (such as staff and customers).
  • Fairness. The debate about fair use of AI is still to be resolved, and new international organisations, such as the Global Partnership for Artificial Intelligence, are looking at those questions. With the rapid expansion in demand for new AI products, and the limited supply of expert practitioners, there’s a strong chance that AI will be built poorly by some organisations. If developers don’t understand the structure of training data, they’re likely to build AI that contains inbuilt biases and errors. That will reduce trust in AI products specifically and in general.
  • Talent shortage. There currently isn’t anywhere near enough human talent to satisfy the expanding demand. This only exacerbates the issues identified above.

In Australia, compared to more advanced nations, there are additional limiting factors for AI.

The general understanding of AI, its role and its impact and benefits is relatively low in Australia (although improving quickly), so confidence in deciding to invest in AI is low. This lack of understanding has several effects that are constraining Australia’s adoption of AI. There’s a lack of understanding of the data that an organisation holds, the value of data to that organisation, and how to extract the enormous social and commercial gains that arise from data. This then limits the sharing and pooling of data across organisations—especially large enterprises and government—that could otherwise yield enormous value to society and the economy.

Australia’s venture capital market is small and immature compared to those of similar countries. Globally, billions of dollars in venture funding is pouring into AI start-ups and scale-ups. The International Data Corporation expects spending on AI technologies to double from 2020 to 2024,3 to reach $110 billion per annum. Australia has only a few AI specialist venture funds, and the funds they’re offering are usually small—one or two orders of magnitude lower than the deals being done internationally.

In general, there’s a relatively strong consensus among AI experts that Australian boards and C-suite4 executives don’t fully grasp these issues and therefore decline to invest in the technology that the rest of the world knows is yielding enormous benefit. Without investment, and the capital that’s required, our AI talent is leaving Australia to follow opportunities being created overseas.

This is already costing Australia. For example, one of the greatest growth areas in AI that has immediate positive impact is health care. The challenge in most countries lies in adopting this technology into healthcare systems that are fragmented, where the teams needed to implement the capabilities aren’t linked up and the technology is driven from a profit motive rather than an outcome focus. Australia has a distinct advantage in adopting AI into health care. We have a unique opportunity to leverage our coordinated national system and data, and world-class AI research expertise, and to build teams of federal and state governments, healthcare operators, researchers, businesses and capital to solve key challenges in our healthcare sector that could drive substantial improvements in cost-effectiveness and individuals’ health outcomes. However, it remains difficult to attract funding for AI research. The Australian Research Council isn’t willing to invest in medical research, and the National Health and Medical Research Council seems hesitant about investing in technology development. There’s limited investment in basic research effort as a result. The consequence is that the Australian public isn’t benefiting from potential improvements in health care, and Australian governments are missing out on potential gains in efficiency.

The limitations on greater adoption of AI in Australia are largely cultural. We haven’t built a culture of investing in innovation, and we don’t have a culture in our business leadership that values data and its potential through the greater adoption of AI. The good news is that there’s nothing inherently limiting in terms of our geographical location or market size. We have amazing natural advantages, including well-managed national datasets and world-class research teams. A concerted effort at cultural change would enable Australia to engage more fully in this new technological revolution and reap the social, health and economic benefits as a result.

(1) Organisation for Economic Development and Co-operation, ‘Gross domestic spending on R&D’, online.

(2) Harvard Business Review, December 2018.

(3) ‘Worldwide spending on artificial intelligence is expected to double in four years, reaching $110 billion in 2024, according to new IDC spending guide’, IDC Corporate USA, 25 August 2020, online.

(4) C-suite = chief executive officers, chief financial officers, chief information officers and so on.

front cover of the report - Artificial intelligence: Your questions answered.

This article is an extract from Artificial intelligence: your questions answered, a report published in partnership with the Australian Strategic Policy Institute (ASPI).  

Tagged in AIML