The R2 – AI R&D score indicator assesses the level of involvement of different geographical areas in AI R&D activities, by considering the weight of the activities developed by AI economic players and therefore taking into account the relative importance of players. To avoid double counting, the indicator considers co-participation (e.g., a patent filed by two players is counted half for each of them). Therefore R2 is calculated as the sum of the fractional count of the activities that the players located in each geographical area develop.
As the AI landscape involves different types of AI activities, the indicator is first calculated separately for each of them and normalised in the interval [0,1]. This enables us to overcome the limitations concerning the aggregation of types of activities of different natures and economic implications.
The R2 indicator facilitates a finer analysis of the importance that different geographic areas have in the different types of R&D activities assessed, i.e., scientific publications in frontier research (the black bars) and patent applications (the blue bars).
It is possible to observe that the level of patent applications filed by players is highly concentrated in China, which dominates the landscape in this respect. However, this evidence should be considered in light of the subsidies that the Chinese government has provided for the developments of AI patents since 2015. In 2021, the Chinese government announced that, in order to boost the country’s high-quality intellectual property services, this policy should be reconsidered by the end of the year. Therefore in following years this could lead to a decrease in the number of patents Chinese players are filing on a yearly basis, but also an increase of their overall quality. On the contrary, we could also expect an increase in the number of AI patent applications, as the revision of patent examination guidelines in China are meant to protect and support AI intellectual property generation in the country. After China, the countries with highest AI patenting activity scores are the US, South Korea and Japan. The EU comes fifth, which highlights its modest performance in terms of development of applied innovations in AI.
In frontier research the US is the global leader, with activity that is twice as intense as that of the EU. Although the gap is considerable, the EU holds the second position, showing remarkable involvement in terms of scientific research in the domain. Given that AI has a strong component related to the improvement of algorithms and software, scientific publications have a more relevant role in applied AI innovation compared with other domains. China and the UK follow the EU, and then the remaining geographical areas present relatively similar values.
Overall, the US appears to have a key role in AI R&D, as its level of activity is outstanding in terms of both patents and frontier research. The EU presents a low level of patenting activity (including if considered in comparison with the US), but the position in frontier research is indeed relevant and appears to be a crucial point for future innovativeness and competitiveness in this domain. Nonetheless, the low level of patenting activity observed should be considered.
The European Union
We find several insights regarding AI R&D at the Member State level. Firstly, the EU Member States are mostly involved in AI-related EC-funded projects. However, it is important to note that the Member State that gets the highest score, i.e., Spain, is not the country which has the largest involvement of EC-funded projects, as its score is not equal to 1. Indeed, the country leading the ranking (of all countries worldwide) is the UK.
This is important, as it highlights the role that the UK has had so far in the EU AI landscape. Indeed, after Brexit the involvement of the UK in EC-funded projects will rapidly decrease, probably causing further modifications in the amount and distribution of research funds among Member States. Secondly, the gap between involvement in EC-funded projects and other R&D activities is considerable. In terms of scientific publications, some Member States show considerable activity scores (such as Germany and France) and this enables the EU (when all Member States are considered jointly) to have a relevant overall role in this type of R&D activity. By contrast, in terms of patents no Member State shows some any substantial involvement at all.