🔬 Research Summary by Sulamaan Rahim, Partnerships Associate at Oxford insights, where he leads on bid writing and partnerships.
[Original report by Oxford Insights]
Artificial Intelligence (AI) represents an important acceleration in the digitalisation of our professional, personal, and economic lives. We can collect, store, and marshal data on scales previously unthought — with applications spanning sectors from healthcare to transport to energy. As more and more public services become digitalised, governments are turning to AI to improve their citizens’ experience and the functioning of these services. To do this effectively, there are myriad factors governments must consider. The Government AI Readiness Index seeks to understand how this potential AI uptake is playing out across the globe.
Since its inception in 2017, we have seen huge growth in the number of countries adopting AI strategies and putting digital transformation at the forefront of their policy aims. Our primary research question for the 2021 Index remains the same: how ready is a given government to implement AI in the delivery of public services to their citizens?
The 2021 Index provides an in-depth look at the state of AI readiness globally, ranking 160 countries using 42 indicators across 10 dimensions. These 10 dimensions fall into three pillars, which reflect three distinct hypotheses about the way in which AI is mobilised within a state:
- The Government pillar: a government should have a strategic vision for how it develops and manages AI, supported by appropriate regulation and attention to ethical problems (governance & ethics). Moreover, it needs to have strong internal digital capacity, including the skills and practises that support its adaptability in the face of new technologies.
- The Technology Sector pillar: a government depends on a good supply of AI tools from the country’s technology sector, which needs to be competitive and dynamic (size). The sector should have high innovation capacity, underpinned by a business environment that supports entrepreneurship and a good flow of R&D spending. The skills and education of the people working in this sector are also crucial (human capital).
- The Data and Infrastructure pillar: AI tools need lots of high-quality data (data availability) which, to avoid bias and error, should also be representative of the citizens in a given country (data representativeness). Finally, this data’s potential cannot be realised without the infrastructure necessary to power AI tools and deliver them to citizens.
Taking into account all three pillars gives us a comprehensive overview of the state of play of AI within a particular country and the key trends emerging globally.
The index sees the USA come out on top again, in large part due to their uniquely developed, runaway technology sector. It is unrivalled in its size and maturity, laying claim to the most technology unicorns of any country. This provides fertile ground for dynamic and innovative use of AI within government, providing a deep pool of technological resources and human capital from which the government can draw.
Singapore comes in second, due largely to its institutional strength. This is reflected in its leading score in the Government pillar, 94.88 — 6.4 points clear of second place (the USA). With a strong relationship between the government and the private sector, Singapore is well on its way to achieving its goal of becoming a “Smart Nation” , with exceptional digital services. Setting out its ambitions in its Digital Government Blueprint , Singapore puts AI adoption as central to this.
This year, there was also a rise in the number of East Asian countries claiming top spots in the index, with one quarter of top 20 countries hailing from the region: Singapore (2nd), South Korea (10th), Japan (12th), China (15th), and Taiwan (18th). Other than Taiwan, these countries all score significantly above the global average in indicators measuring the levels of human capital and technological infrastructure within a country. These scores point to the region’s global success in AI research and its advanced computing power. Despite these great strides at the global level from the top-scoring countries of the region, East Asia remains a highly unequal region. Several countries in the region score well below the global average of 47.42 out of 100.
This reflects a broader global trend. We found there to be clear inter-regional and intra-regional inequalities in AI readiness across all three pillars of our methodology. The average scores of the two highest ranked (North America and Western Europe) regions is more than double that of the two lowest-scoring regions (Sub-Saharan Africa and Central & South Asia): 76.75 compared to 36.27. Within regions, we see similarly stark differences. Across all regions (excluding North America and the Pacific), the three top-ranked countries score on average 1.95 times more than the three bottom ranked countries. 
This inequality is present across all three pillars of the index.  Most of the business activity measured in our Technology Sector pillar is taking place in the developed world and in the biggest economies. For instance, the US and China host 52 AI unicorns, which is roughly three times more than the rest of the world. This feeds into the inequality we find in our Data & Infrastructure pillar. The levels of infrastructure and data availability required to support AI development is concentrated in the same countries. Low levels of technological infrastructure in developing countries contribute to Digital Capacity’s status as one of the lowest-scoring dimensions of the Government pillar across developing countries. Arguably, with governments in some developing nations having limited capacity to build and deliver AI in their public services, few are at the stage in AI readiness to publish their vision for AI.
Our Vision indicator in the index looks at whether or not a country has (or is in the process of creating) a national AI strategy. Its position in the Government pillar indicates the importance we place on such a strategy to allow a country to present a clear, unified approach to dealing with the cross-sector, cross-departmental opportunities and challenges that AI presents. As one might expect, these national strategies remain concentrated in the Global North, contributing to the wider inequality in AI readiness scores globally.
The focus on official data and structures (such as national AI strategies) presents a methodological challenge for certain regions. Our sub-Saharan Africa expert, Abdijabar Mohamed, emphasised this point. There is a paucity of official government data alongside a lack of robust governance structures and political conflict in the region and so countries in sub-Saharan Africa score poorly.
Citizens, in spite of any lack of structured government support, can nonetheless develop their own initiatives to ensure they are ready for AI. Our regional analysis for sub-Saharan Africa highlights some of these initiatives in the region. However, the capacity to collect adequate data — something hampered by political conflict — and use these to train and refine AI tools is imperative. Without such capacity these kinds of initiatives will face large barriers.
The foregoing highlights the need for strong collaboration across the public and private sector when it comes to AI readiness: robust, open government data (collected, stored, and used in a way that is ethical and respects privacy) alongside a facilitative political and economic climate can prove invaluable to how AI technology firms develop and train tools they build.
As useful as public sector data can be for the private sector, private sector developments can also have huge implications for government and public sector advancements. Within the Technology Sector pillar, we can see a boom in the size of large public technology firms and number of private technology unicorns. The total market value of large technology firms as measured in our index has also risen with the top 7 largest public technology firms adding $3.4 trillion in value in 2020.  Not only can these growing firms provide jobs to an economy, but their developments can also be marshalled to improve public services.
There are many ways governments can harness these developments. They can use procurement practices (especially with the rise of GovTech startups) to digitalise and improve public services. They can also generate in-house AI skills and capacity through strategic hires from the private sector, or by creating knowledge-sharing practices between the private and public sector via initiatives like AI Councils that provide multi-sectoral expert advice to governments. To be truly AI ready, there must be strong collaboration between private and public sectors with an unwavering focus on improving the lives of all citizens and the services provided to them.
Infrastructure generated by these private sector developments is crucial to AI readiness. Our Data and Infrastructure pillar looks to measure this, analysing access to the internet, mobile devices, 5G infrastructure, and internet bandwidth (amongst other things), as well as the demographics of those who access them and open data policies within a state. All of these are integral to AI readiness and development in both the public and private sectors.
As stated above, much technological development and market activity takes place in the Global North; the subsequent infrastructure is thus not distributed globally. At a very basic level, despite ever-rising numbers of users, a 2021 UN report found that 2.9 billion people have still never used the internet — 96% of whom live in developing countries.  Within populations, our indicators continue to show differences in access to the internet between genders and socioeconomic groups. Though we do not measure it specifically in indicators, the total percentage of internet users in a country also likely masks urban-rural divides in internet access. We know that tools developed from data that excludes or overlooks certain groups are liable to systematically disprivilege and discriminate against them. Governments must reflect on this and understand the barriers that unrepresentative data has on the potential to become truly AI ready.
It would be an oversight to complete this piece without a final look at the ethical implications of AI readiness. Our index looks broadly at AI readiness, assessing the overall policy and technology developments to understand how ready governments are to use AI when delivering public services. We are of course sensitive to the ethical implications and trustworthy use of AI and have included important indicators in the 2021 iteration to measure things like cybersecurity, data protection legislation, and national AI ethics frameworks.
But this is not the totality of trustworthy AI.
AI should be about improving all citizens’ lives; these technologies should be harnessed for the public good. Our 2021 Trustworthy AI index (formerly the Responsible AI sub-Index from 2020) is set to be published in early March. It analyses in detail the ethical dimensions of government AI use, ranking 50 countries using 20 indicators across six pillars: Policies, Accountability, Transparency, Inclusivity, Privacy, and Human Centred Values. This will act as a further annex to our 2021 Government AI Readiness Index and will provide important context to how well-positioned AI “ready” countries might be to advance that AI readiness in a trustworthy manner.
AI is set to revolutionise our lives; governments must be AI ready and we must be active, informed participants to help hold them to account. We hope our 2021 Index goes some way to advancing both of these aims.
 More detail can be found here: https://www.smartnation.gov.sg/
 The full policy can be found here: https://www.tech.gov.sg/digital-government-blueprint/
 What I mean by this is that I calculated the average score of the top and bottom three countries in each region, calculated the ratio between the two for each region, and then averaged these ratios. I excluded North America and the Pacific as both regions have fewer than six countries and so it was not possible to take a top and bottom three average.
 A similar measure to the foregoing — a ratio of the top quintile’s average to the bottom quintile’s average — reveals that the highest-ranking fifth of countries score on average 2.87 times more than the lowest-ranking fifth across each dimension. Ratios across the individual pillars are as follows: Government — 3.06; Technology Sector — 3.04; Data and Infrastructure — 2.51.