What the Global Innovation Index tells us about Artificial Intelligence Efficiency

Monday, May 10, 2021

Ekkehard Ernst, International Labour Organization (ILO) and Saurabh Mishra, Taiyo

May 2021

 

Artificial Intelligence (AI) promises to be the break-through technology for the 21st century. New products and services powered by advancements in AI have started to revolutionize the way we work, consume and interact. The latest promise lies in developing medication and synthetic chemical products faster and with less waste, thanks to the capacity of AI to simulate the impact of new drugs and proteins. Countries that are at the forefront in catalysing AI could expect to experience significant gains in productivity and living standards. Policy makers, therefore, face the challenge to identify how to grow and strengthen their national innovation systems around AI.

 

To understand policy priorities in developing AI ecosystems, we propose a new AI efficiency index.
In order to get a detailed overview of the AI landscape in different economies, we source data on AI investment, AI research output and impact, AI hiring and AI education. Using these indicators, an economies’ efficiency is estimated in AI production, AI implementation and AI diffusion. The efficiency index does not measure the overall output in AI products and services or its quality. Rather it measures how efficiently AI is generated relative to available inputs. Higher AI efficiency – both overall and for its subcomponents – is highly correlated with higher concentration of AI start-ups, more AI investment, relatively higher scientific output and impact in AI and faster growth in AI hiring across the entire AI value chain. Moreover, AI efficiency is also positively associated with economy-wide performance indicators such as the Economic Complexity Index.

 

We match our AI efficiency estimates with WIPO’s Global Innovation Index (GII) and its subcomponents to better understand how well countries convert their AI inputs (investment in AI start-ups, their AI students and the AI experts hired by companies) to innovation output (patents, creating start-ups and publishing well-cited papers). What does our analysis suggest? First, the GII shows a strong link with the overall efficiency of a country’s AI innovation system; the better an economy ranks in the GII, the better it also ranks in the AI efficiency index. A more detailed analysis shows that individual components of the GII are related to specific parts of the AI value chain: strong protection of intellectual property rights, access to foreign markets, high-tech FDI or exchange of foreign knowledge, are particularly important to ensure efficient production of AI (patents). Government provision of digital infrastructure and procurement of digital services can strengthen the efficiency of the AI start-up scene and growing such services to scale. Knowledge diffusion – the third component of our AI efficiency index – can be strengthened through domestic innovation linkages (industry-university links and research collaborations), as well as foreign R&D expenditure and is positively related to international trade of ICT goods and services. All significant correlations between the GII and the AI index can be seen in the table.

 

Our research also shows that different forms of government interventions can strengthen the AI ecosystem: A key finding of our analysis points to market barriers and weak (intellectual) property rights as major obstacles for enhanced AI efficiency. Product market frictions such as public ownership in the telecommunications industry, for instance, are pernicious to patenting efficiency in the field of AI. Tax subsidies can remedy such barriers and enhance start-up investment activity but improvements in the overall institutional framework are more effective and efficient. Our results also highlight the importance of differentiating policy interventions along the stages of the AI production chain: openness to (foreign) knowledge exchange is important for efficiency in AI production and diffusion whereas policies supporting vibrant high-tech tradable services could help drive efficiency in the AI ecosystem.

 

Table: Correlations between Global Innovation Index Components and the AI Efficiency Indices

 

Note: The table summarizes the correlation between different characteristics of a country’s institutions and innovation system. Estimates are based on a sample of 24 countries for the period 2015 to 2018. Green-colored cells indicate that AI efficiency yields a statistically significant improvement of the corresponding GII (sub-)index at the 5% level. White cells refer to statistically insignificant coefficients. Source: WIPO, Global Innovation Index, Years 2015-2018; World Governance Indicators; own calculations. “AI inv.” Stands for AI investment.

 

To access the full paper and dataset, please visit:

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3800783

You may also like

View all blog posts