The Australian Government’s Digital Economy Strategy sets its sights on Australia being a top 10 digital economy and society by 2030. One of the strategies to reach this outcome is to position Australia as a global leader in artificial intelligence (AI) technology, with the Federal Government investing $124.1 million in its Artificial Intelligence Action Plan.
What can we learn from the UK? A recently published report commissioned by UK Department for Digital, Culture, Media and Sport’s (DCMS) identified a number of critical considerations for the long-term success and productivity of AI R&D commercialisation.
Australia vs UK
Global ‘league charts’ are tricky things, but that said, a comparison between Australia and the UK is illustrative of the challenges Australia faces.
The 2021 Global Innovation Index prepared by the World Intellectual Property Organisation ranks countries by their capacity for, and success in, innovation. Countries are measured against 7 ‘pillars’ of innovation, with 5 pillars relating to innovation inputs (i.e. components of the economy that enable innovation) and 2 pillars relating to innovation outputs (i.e. products and services reaching the market).
The UK ranks 4th overall while Australia ranks 25th overall. But beneath those headline rankings, the comparative UK and Australian rankings against the 7 pillars of innovation are more troubling.
We do well in – and our overall performance is pulled up by – the general inputs needed to drive innovation: human capital, market sophistication and the strength of our institutions. But that does not translate into creative outputs. In other words, we seem to be “sitting on” our natural talent, not doing too much with it.
Pathways to commercialisation
The DCMS report categorises the pathways to commercialisation of R&D as follows:
- direct commercialisation – including spin outs, from university researchers and existing companies;
- knowledge exchange – including joint collaboration with academics and technology firms, conferences and publications; and
- formal and de facto standards and IP – development of industry standards and IP protection.
Of these avenues, the DCMS report considered that the academic ecosystem presents a natural competitive advantage for the UK, given their global ranking. This is reflected in the pivotal role UK universities currently play in the commercialisation of AI, particularly by spinning out R&D into start-ups. Partly this is due to the sheer volume of research being conducted at universities (although the important role of large technology firms and research and technology organisations was recognised), but the clincher is the concentration of AI talent, as distinct from general software and programming talent, at top universities.
Australian universities also rank well globally. The UK has seven universities in the top 100 and so too does Australia, although UK universities such as Oxford rank in the top 10 whereas Australia’s top ranking university, Melbourne, ranks 32nd. Australian universities also have geared up their focus on AI research: the ARC Centre of Excellence for Automated Decision-Making and Society based at RMIT is a Commonwealth funded research centre which “brings together universities, industry, government and the community to support the development of responsible, ethical and inclusive automated decision making”.
Academic and industry silos hold back AI development
The exponential growth of AI and its potential to transform the digital landscape is what has almost every sector buzzing with excitement. However, the DCMS report observes that this is paradoxically one of the difficulties in commercialising AI because, without a sector-specific focus, the broad applicability of the technology across a range of sectors otherwise renders R&D as “a solution in search of a problem”. Which is a polite way of the DCMS saying that academic researchers need to be more focused on translation of their research into the real world.
More fluidity between academia and industry will promote a commercial understanding of market forces, regulations and end-users within university institutions, which is of critical importance given the core value of AI comes from its application to specific challenges. The DCMS report recommends that this be achieved by borrowing from the US academic/business world experiences:
- academics simultaneously holding positions at universities and in commercial businesses. This serves a dual purpose by:
- allowing top AI talent, who ultimately become founders of university spin-outs to work in and alongside sectors on applied AI projects and develop an adequate understanding of the markets they are building applications for; and
- mitigating the risk of universities losing top AI R&D talent to bigger technology firms who offer more competitive salaries and protecting the next generation of AI talent from being orphaned;
“I think down the line, it could lead to a reduction of companies appearing from the UK because if all the most talented people are going to these tech giants, we won’t have many people to work on developing startups” - University
- drawing on the success of industrial-focused labs in US universities such as CalTech and MIT, which legitimises industry careers as a pathway to academic careers. Students are able to join a firm spun-out from these university labs and are not restricted from returning to university research at a later time; and
- commercial AI fellowships, connecting technology entrepreneurs focused on a specific market application with AI researchers to build out the R&D, and a providing direct path to commercialisation.
More fluidity between academia and industry will promote a commercial understanding of market forces, regulations and end-users within university institutions, which is of critical importance given the core value of AI comes from its application to specific challenges.
Reliance on funding
Most AI firms have used private or public funding to commercialise AI R&D through the development of new products, services or lines of business, which indicates the importance of funding for successfully commercialising AI R&D. However, the DCMS Report considered that existing barriers to funding are currently limiting the effectiveness of commercialisation, including:
- administrative burdens for grant funding, which is the primary source of funding for universities. Specifically, grants are time consuming, causing researchers to spend more time writing business proposals than researching or scaling R&D, with no guarantee of funding and often do not incentivise start-ups to align with market needs.
“And I think down the line, the more the risk that the government funding ends up taking you away from the concept of commercialisation by accident, because you have to swim to where the money is. The government is designing schemes that are not for the market.” – Coadec, Startup Association
- UK investors being unambitious compared to their US and Chinese counterparts, and prioritising short-term profits over long-term gains. This means that key innovations which have high-capital costs are not being pursued; and
- universities taking too high an equity share in spin-out companies. The DCMS report encourages a shift to lower equity share demands from universities to incentivise private investors which has been demonstrated by the US as a financially sustainable model by increasing total spin-outs.
One of the recommendations made by the DCMS report to mitigate the challenges faced by public and private finding is for the Government to be a ‘first customer’ and offer procurement contracts rather than grants. Procurement contracts create a domino effect as they signal market confidence and the value of start-ups.
“…When the government becomes a customer of a new tech company, it creates a unicorn. It doesn’t take a very big contract to do so…” - Mind Foundry, Spinout
IP ownership and protection
The DCMS report considered the difficulties of the UK IP regime when it came to the protection and patentability of software, and the impacts on realising commercial value from AI software inventions. Unlike the United States, the UK and Australia do not provide for the patentability of business processes which limits the ability to realise commercial value from software inventions.
Australia had been hoping for clarification and reform to permit patent protection of software inventions. However, the High Court’s recent decision in Aristocrat Technologies Australia Pty Ltd v Commissioner of Patents  HCA 29 has reinforced the difficulty of obtaining protection for software inventions. Until the Federal Parliament prioritises legislative reform, it is unlikely that Australia will see any improvement in this area and this will continue to have a critical role in its ability to extract value from AI R&D and commercialisation.
How does this apply here?
Australia’s ecosystem echoes many of the difficulties identified in the UK including a lack of fluidity between academia and industry, an uncertain IP regime and conservative funding, particularly given weak public trust in AI.
The recently enacted Artificial Intelligence Action Plan addresses some of the concerns identified in the DCMS report:
- the Next Generation AI Graduates Program has been established, with universities and industry co-funding scholarships for honours and doctorate programmes. As part of the scholarship, students participate in industry-led research projects and placements to build job‑ready skills setting the groundwork for the possibility of future joint-tenure programs; and
- the AI Solutions to Build a Stronger Australia program, which addresses the availability of public funding by providing grant funding to businesses to pilot AI projects focused on developing AI-based solutions to national challenges and that also have benefits for job creation, economic recovery and other social benefits.
But there are other more fundamental issues identified by the DCMS Report that need some thinking here – including more fluidity between universities and business and legislative reform to maximise the value of AI R&D commercialisation in Australia.
Read more: Blueprint for an AI Bill of Rights