Artificial Intelligence, Automation and the Australian Economy
The Coming Failure of Public Policy
Abstract
Given broad consensus Artificial Intelligence (AI) research is progressing steadily (Future of Life Institute, 2016), and with expectations of increasing societal and economic impacts (Stone, et al., 2016), AI policy is an urgent concern (Agrawal, et al., 2016). From increased economic productivity and improved quality of life, to intensified international competition and labour market displacement, the impact of AI and Automation will be profound. AI and Automation has the potential to create tremendous social and economic benefits (Committee for Economic Development Australia, 2015), but has also raised concerns over technological unemployment (Frey & Osborne, 2013), an exacerbation of inequalities and exclusivity of economic opportunities (Lagarde, 2016). Navigating these transformative impacts will require proactive, adaptive and inclusive public policies.
Analysis has identified a lack of Australian public policy addressing the opportunities and challenges of AI and Automation. Drawing on international approaches, this paper provides three high-level recommendations to the Federal Government:
I. Commission a White Paper on Artificial Intelligence and Automation
II. Develop a Federal Artificial Intelligence Strategy
III. Establish a ‘Council on Future Technologies’
AI and Automation present a unique opportunity, but the level of Australian prosperity and economic competitiveness will depend on the Federal Government developing and implementing effective public policy. To realise the full opportunity afforded and address any negative outcomes, this process must begin immediately.
Artificial Intelligence and Automation
Lacking a formal definition, AI can be broadly regarded as a collective set of technologies (including machine learning, computer vision, natural language programming and other sub-fields), which aim to replicate or imitate intelligence. Narrow (or ‘weak’) AI refers to technologies which exceed human intelligence/capabilities within specific domains (Kiron, 2017), whereas general (or ‘strong’) AI meets or exceeds human intelligence in all cognitive domains (Council, National Science and Technology; Office of Science and Technology policy, 2016). Despite concerns over general AI, raised by experts and foundations such as the Future of Life Institute (Future of Life Institute, 2016), this paper focuses on the economic impact of narrow AI.
Not merely an emerging technology, AI is a driving force behind a wide variety of applications and services throughout society today (Adams, 2017). From Netflix and Siri, to autonomous vehicles and search engines, AI is embedded in our daily lives (Adams, 2017). As the proliferation of AI continues, driven by massive private sector investment, AI is expected to have a transformative impact across healthcare, transportation, education, energy, entertainment, employment, government and a wide variety of other sectors and industries (Stone, et al., 2016). Adding up to fifteen trillion USD to the global economy by 2030 (PwC, 2016), AI has the potential to greatly increase the overall wealth of humanity (Brynjolfsson & McAfee, 2014), yet the distribution of this wealth remains a key concern (Gummi, 2017). Progress in AI is also helping drive a larger and more dramatic societal transition, one occurring at ten times the pace and three hundred times the scale of the Industrial Revolution (Dobbs, et al., 2015), presenting significant challenges for policy makers.
Critical to Australian prosperity in the twenty-first century is the ability of the Federal Government to develop effective public policy in response to advancements in AI and Automation; ensuring the capitalisation of opportunities and mitigation of potentially negative outcomes.
Economic Impact of Artificial Intelligence and Automation
Advances in AI and Automation will have a large impact on Australia’s labour market, national competitiveness and economic equality.
Labour Market
The pace and composition of workforce development, re-deployment, and technological unemployment will present considerable policy challenges over upcoming decades. Given the breadth and depth of occupations likely to be impacted by AI, recent debate has re-approached the question of technological unemployment, with numerous international reports indicating an increased effect of AI and Automation on employment (Frey & Osborne, 2013) (McKinsey Global Institute, 2017). The traditional notion technology destroys jobs but also creates jobs, even if correct at a basic level, is too simplistic (Vardi, 2017). In addition to the computerisation of routine manufacturing tasks, Autor and Dom (2013) document a structural shift in the labour market, with workers reallocating their labour supply from middle income manufacturing to low-income service occupations.
While the capitalisation effect of technology has historically outweighed the counter-balancing destruction effect (Frey & Osborne, 2013), our discovery of means to economise labour, particularly in light of advanced technologies, has become more complex (Brynjolfsson & McAfee, 2011). As the real cost of capital (computing) continues to decline, it further incentivises organisations to substitute labour for capital (Frey & Osborne, 2013). Forty-four percent of firms with reduced workforce sizes since 2008, did so because of automation (McKinsey Global Institute, 2011). Multi-methodological forecasts of US employment have identified between five-percent (McKinsey Global Institute, 2017) and forty-seven percent (Frey & Osborne, 2013) of US occupations to be either fully automatable or at high-risk of automation. While Australian research is generally lacking, analysis by CEDA (2015) has concluded forty-percent of Australian jobs face a high-probability of being susceptible to computerisation over the next twenty years. The reason human labour has prevailed historically, has been attributed to an ability to adopt and acquire new skills through education (Goldin & Kats, 2009). However, as computerisation enters more cognitive domains, reskilling through education will become increasingly difficult (Brynjolfsson & McAfee, 2011).
One of the larger challenges for policy makers will be the required large scale re-training and re-deployment of workers resulting from increased automation (Evlin & O’Neill, 2017). Machines have already automated millions of routine working-class jobs in manufacturing, and AI is beginning to automate non-routine and cognitive jobs in administration, transportation, logistics, financial services and more (Vardi, 2017). Workers must be more adaptive to complex and technologically advanced operating environments (Araya & Johal, 2017), and realise a more frequent need to re-skill, re-train and switch occupations (Evlin & O’Neill, 2017). Yet this involves more than technical training, and dependent on the scale and pace of change, this may reflect a foundational change in how citizens plan their lives, highlighting a need to equip citizens with the mechanisms to manage rapid change. More than ever, governments need to distinguish between jobs lost to other countries and jobs lost to the past (Committee for Economic Development Australia, 2015), as well as partaking in frank discussions with the public on managing the transition to a more automated future (Araya & Johal, 2017). The government needs to investigate methods to help facilitate this redeployment of labour, identifying at risk skills, occupations and industries, and ensuring educational frameworks cater to the future economy (Committee for Economic Development Australia, 2015). Public policy must also take more tangible steps toward the re-deployment of labour by assisting in the capability of workers to retrain (West, 2015).
Advances in AI will not only change the way in which Australia’s future workforce is developed, but change the very scope of what a future workforce will look like. As the international competition for AI expertise continues to accelerate (CIFAR, 2017) (Agrawal, et al., 2016), and as AI continues to provide strong economic and societal benefits (Committee for Economic Development Australia, 2015), it becomes imperative Australia develop an AI workforce. Without a highly qualified AI workforce, Australia risks falling behind. Given the increasing pace of change within the labour market, it is important to avoid investing resources in declining industries as opposed to future growth industries (Committee for Economic Development Australia, 2015). Preliminary analysis has identified a critical lack of AI capabilities within Australia and need to reform existing educational approaches.
Economic Competitiveness
Unlocking the potential of AI and Automation could increase Australian GDP by approximately two trillion by 2030 (AlphaBeta, 2017). In response to global advances in AI, the United States (Council, National Science and Technology; Office of Science and Technology policy, 2016), China (New America, 2017), United Kingdom (Hall & Pesenti, 2017) and other nations are prioritising AI investment, research and commercialisation; underscoring the importance of an Australian AI strategy. Despite increasing international resource investment (both financial and intellectual), Australia has been slow to act. With a seventy-percent service-based economy, Australia can either capitalise on AI and Automation to help drive the economy and disrupt international markets, or fail to act and be disrupted (Committee for Economic Development Australia, 2015). Economic competitiveness over upcoming decades will be dependent on Australia transitioning from a strategy of imitation and exploitation, to one of exploration (Committee for Economic Development Australia, 2015), where Australia pushes outward the technological frontier. Critical to this successful transition, and capitalisation of economic opportunities, will be a significant increase in AI research, investment and commercialisation, as well as the promotion of technological adoption (Evlin & O’Neill, 2017).
Economic Equality
Increased productivity, efficiency gains, and increasing returns on capital borne by AI and Automation are likely to substantially increase the amount of wealth throughout society (Brynjolfsson & McAfee, 2014). Yet these effects, coupled with labour market displacement, may further widen income inequalities (Gummi, 2017). This may further exacerbate the declining share of labour income within Australia, which has already decreased from fifty-seven percent of GDP in 1975 to forty-seven percent at near-present day (Committee for Economic Development Australia, 2015). Labelled the ‘great decoupling’, McAfee & Brynjolfsson (2012) further document a trend of employment no longer rising in parallel with productivity. As the decreasing cost of robotics and automation driven technologies continues to expand across industries, and as the share of labour income continues to decrease alongside a corresponding plateau in wage growth (Committee for Economic Development Australia, 2015), more income is flowing to the owners of capital. While there is likely to be offsetting economic effects, such as decreased prices of goods increasing in real income (Frey & Osborne, 2013), it is unclear how impactful this will be. If not managed appropriately, and without sufficient support programs, Australia will see an exacerbation of income inequality and isolation of economic opportunities, both domestically and internationally.
Recommendations
To achieve the desired outcomes identified in this analysis, and in consideration of international comparisons and best practice, the following recommendations provide a critical first step in addressing the absence of Australian public policy on AI & Automation.
Recommendation #1 Commission a White Paper on Artificial Intelligence and Automation
There is a lack of data and economic forecasts relating to AI and Automation within Australia. Critical to the development of proactive and effective public policy is understanding the breadth and depth of impact AI and Automation will have on Australia, and given such lapses of information, a detailed analysis is imperative.
The following areas, in no particular order, are identified here as critical and likely to provide a foundation for future strategy and policy.
I. Australian Labour Market
Utilise various methodologies and modelling techniques to identify tasks, jobs and industries at risk, to inform policy setting on job transition assistance, re-education programs, and investment priorities. Further, the research should, as much as possible, identify areas where demand for workers is likely to be highest, including new roles. Analysis should further identify geographical areas subject to high automation to inform relocation policy.
The Paper should investigate and identify labour market transition mechanisms which support the re-training and redeployment of workers displaced by Automation. Analysis should evaluate both direct and indirect support options, and account for the potential of labour displacement to be gradual, sporadic or ’lump-sum’. Critical to Australian competitiveness will be a highly skilled, adaptive, and flexible workforce. The White Paper should investigate means to identify, attract and retain leading talent, as well identifying any necessary changes to educational frameworks.
II. Economic Analysis
A comprehensive cross-sector analysis on the economic impact of AI and Automation, from manufacturing and energy, to transportation and health, should be conducted. The Paper should further investigate methods to incentivise organisations to align strategies and decisions with socially desirable outcomes (i.e. workforce transparency).
III. Taxation Revenue & Social Support Resilience
Identify how altering employment numbers, and an increased share of income flowing to capital providers, will impact government budgets and funding of social support programs. This Paper should include an analysis of wealth distribution alternatives such as the viability of a Universal Basic Income or tax on robots.
IV. Australia’s AI Industry
Determining the level and scale of Australia’s AI sector can provide significant insight to guide future public policy. This should also include an analysis of existing workforce capabilities and pipeline development. Analysis should further identify both drivers and barriers to AI industry growth.
Recommendation #2 Develop a Federal Artificial Intelligence Strategy
Critical to managing this transition is a federal AI strategy. In line with the economic analysis of the White Paper, the Federal Government should develop a proactive and adaptive strategy with the following considerations and objectives:
I. Increasing Australia’s AI Industry and Workforce
Identify and amplify policy mechanisms which aid the development of Australia’s AI industry and workforce. These mechanisms should include approaches to develop and retain domestic AI expertise and to improve international attractiveness for global specialists. The strategy ought to increase AI development and adoption within start-ups, SME’s and large organisations. To capture this two trillion economic opportunity, Australia must double the rate of investment in, and adoption of, automation technologies (AlphaBeta, 2017).
II. Public Awareness and Engagement
Previous technological advancements have had the propensity to create public pushback and resistance, as illustrated by the decline in US manufacturing (Muro, 2016). Unless citizens accept the ‘verdict’ of the market outcome, decisions to adopt an innovation may be resisted by ‘losers’ through non-market mechanisms and political activism (Frey & Osborne, 2013). To avoid such pushback and to ensure the public is aware of the increasing pace of change and impact of AI, the federal strategy must incorporate an approach for public awareness and engagement.
III. International Engagement and Monitoring
Monitor international advances in AI as well as the development, commercialisation and implementation of AI within other nations. Given the global impact of AI discussed within this paper, an approach for international engagement must be formed.
IV. Reforming Education and Tailoring Social Support Programs
Modernise and customise social support programs, catering to various levels of labour market impacts. This may take form in modernising existing programs and/or include international approaches such as a lifetime learning account. Australia should also identify and form partnerships to provide an allotment of Master’s and PhD qualifications in AI. Further approaches including the use of Massive Online Open Courses (MOOC) and bridging programs should be considered.
V. Financing Mechanisms and Incentives
Australia must bring its funding (direct and indirect) to commensurate levels seen internationally. This could come in the form of grants for AI start-ups, taxation incentives for technological adoption and transparency, and a variety of other mechanisms. The government should evaluate policy mechanisms to bring non-governmental funding to Australian AI development, such as requiring a proportion of superannuation funds to invest in AI commercialisation.
VI. Regulation, Ethics and AI Safety
Provide a platform to review the regulation and classification of AI technologies and their applications. From autonomous vehicles to ‘black-box’ technologies, the Federal Government should begin looking at mechanisms to validate AI security and safety.
VII. Government Intelligentisation
The benefits of implementing AI and Automation within government are vast. Federal policy should involve all departments in developing a set of community standards and best practice. This includes training new and existing staff on AI implications, as well as engagement with technical experts to identify inter-governmental opportunities.
VIII. Data & Digital Trust
Develop policy to improve access to data and work with industry to deliver ‘Data Trusts’. This may include mandating public research provide underlying data in machine-readable format with clear rights, and where possible, open. The strategy must improve digital trust within Australia as it is critical to technological development and adoption.
Recommendation #3 Establish a ‘Council on Future Technologies’
A multi-stakeholder approach to AI policy development is fundamental to capturing opportunities and mitigating any negative effects. The Government needs to understand what support mechanisms and incentives would prove effective. The Council should include industry, academia, financiers, trade unions, public representatives and any other stakeholders likely to be impacted.
The Council should aim to examine and consider 1) increasing the rate of technological adoption within Australian academia, industry and government 2) improved transparency in technological development, adoption and workforce planning 3) the creation of a ‘regulatory sandbox’ to aid AI development 4) mechanisms to remove barriers to AI workforce development; and 5) financing priorities and mechanisms.
Conclusion
It is apparent advances in AI and Automation will continue to have transformative impacts on Australian economic and social prosperity over upcoming decades. Analysis has identified three economic areas AI and Automation will have a large impact, as well as a critical lack of Australian data and analysis. To capitalise on the opportunities afforded, and to mitigate potential negative outcomes, the Federal Government should implement the three recommendations identified within this paper.
References
Adams, R. L., 2017. 10 Powerful Examples of Artificial Intellience in Use Today. Forbes
Available at: https://www.forbes.com/sites/robertadams/2017/01/10/10-powerful-examples-of-artificial-intelligence-in-use-today/#565a60fa420d [Accessed 18 October 2017].
Agrawal, A., Gans, J. & Goldfarb, A., 2016. The Obama Administration’s Roadmap for AI Policy. Harvard Business Review.
AlphaBeta, 2017. The Automation Advantage: How Australia can seize a $2 trillion opportunity from automation and create millions of safer, more meaningful and more valuable jobs, s.l.: AlphaBeta.
Araya, D. & Johal, S., 2017. Work and Social Policy in the Age of Artificial Intelligence. [Online]
Available at: https://www.brookings.edu/blog/techtank/2017/02/28/work-and-social-policy-in-the-age-of-artificial-intelligence/
[Accessed 11 October 2017].
Autor, D. & Dorn, D. 2013. The Growth of low skill service jobs and the polarization of the US labor market. American Economic Review, vol. forthcoming.
Brynjolfsson, E. & McAfee, A., 2011. Race against the machine: How the digital revoluation is accelerating innovation, driving productivity, and irreversibly transforming employment and the economy, Lexington, MA: Digital Frontier Press.
Brynjolfsson, E. & McAfee, A. 2012. Jobs, Productivity & The Great Decoupling. NY Times.
Available at: http://www.nytimes.com/2012/12/12/opinion/global/jobs-productivity-and-the-great-decoupling.html [Accessed 10 October, 2017]
Brynjolfsson, E. & McAfee, A., 2014. The Second Machine Age: Work, Progress, and Prosperity in a Time of Brillian Technologies. New York: Norton & Company.
CEDA, 2015. Australia’s Future Workforce.
Available at: http://www.ceda.com.au/News-and-analysis/Media-releases/More-than-five-million-Aussie-jobs-gone-in-10-to-1
[Accessed 10 October 2017].
CIFAR, 2017. Pan-Canadian Artificial Intelligence Strategy Overview.
Available at: https://www.cifar.ca/assets/pan-canadian-artificial-intelligence-strategy-overview/ [Accessed 10 October 2017].
Commitee for Economic Development Australia, 2015. Australia’s Future Workforce. Committee for Economic Development Australia.
Council, National Science and Technology; Office of Science and Technology policy, 2016. Preparing for the Future of Artificial Intelligence, Washington DC: Executive Office of the President.
Dobbs, R., Manyika, J. & Woetzel, J., 2015. The Four Global Forces Breaking All The Trends.
Available at: https://www.mckinsey.com/business-functions/strategy-and-corporate-finance/our-insights/the-four-global-forces-breaking-all-the-trends [Accessed 13 October 2017].
Evlin, L. & O’Neill, M., 2017. Australia must embrace AI revoluation with Automation set to affect every job, report sats.
Available at: http://www.abc.net.au/news/2017-08-08/australia-must-embrace-ai-revolution-alphabeta-report/8774044
[Accessed 13 October 2017].
Frey, C. B. & Osborne, M. A., 2013. The Future of Employment: How Susceptible Are Jobs to Computerisation?.
Available at: https://www.oxfordmartin.ox.ac.uk/downloads/academic/The_Future_of_Employment.pdf [Accessed 10 October 2017].
Future of Life Institute, 2016. An Open Letter: Research Priorities for Robust & Beneficial Artificial Intelligence.
Available at: https://futureoflife.org/ai-open-letter/ [Accessed 10 October 2017].
Gans, J., 2016. The Obama Administrations Roadmap for AI Policy, s.l.: Harvard Business Review.
Goldin, C. & Kats, L. F., 2009. The race between education and technology, s.l.: Harvard University Press.
Gummi, M., 2017. Artificial Intelligence: The Way Forward for Policy & Regulation. Berkeley Public Policy Journal.
Hall, W. & Pesenti, J., 2017. Growing the Artificial Intelligence Industry in the UK. UK Government.
Kiron, D., 2017. What Managers Need to Know About Artificial Intelligence.
Available at: http://sloanreview.mit.edu/article/what-managers-need-to-know-about-artificial-intelligence/ [Accessed 10 October 2017].
Lagarde, C., 2016. Managing an Inclusive Transition for the Global Economy.
Available at: https://www.imf.org/en/News/Articles/2016/10/07/AM16-SP100716-Managing-an-Inclusive-Transition-for-the-Global-Economy [Accessed 10 October 27].
McKinsey Global Institute, 2011. An Economy that works: Job Creation and America’s Future, s.l.: McKinsey & Company.
McKinsey Global Institute, 2017. A Future that Works: Automation, Employment & Productivity, s.l.: McKinsey & Company.
Muro, M., 2016. Manufacturing Jobs Aren’t Coming Back. MIT Technology Review]
Available at: https://www.technologyreview.com/s/602869/manufacturing-jobs-arent-coming-back/ [Accessed 12 October 2017].
New America, 2017. State Council Notice on the Issuance of the Next Generation Artificial Intelligence Development Plan, s.l.: New America.
PwC, 2016. Sizing the Prize: What’s the real value of AI for your business and how can you capitalise? PwC.
Stone, P. et al., 2016. One Hundred Year Study on Artificial Intelligence 2016 Report.
Available at: https://ai100.stanford.edu/sites/default/files/ai100report10032016fnl_singles.pdf [Accessed 10 October 2017].
Vardi, M., 2017. Artificial Intelligence and the Future of Work: An Interview with Moshe Vardi [Interview] (14 June 2017).
West, D. M., 2015. What happens if robots take the jobs? The impact of emerging technologies on employment and public policy. Centre for Technology Innovation at Brookings, October.
