This dimension of the AI Watch Index investigates those activities that contribute to the development of the technological domain.
From a socio-economic point of view these activities are crucial in order to make progress happen, and so to improve technology maturity, and to reach (or maintain) economic competitiveness.
Research and development (R&D) activities considered in this work are of three types:
- frontier research activities (i.e., scientific publications in top international conferences),
- filing of patent applications,
- participation in EC-funded projects.
To avoid a biased vision of AI R&D, the latter are considered only when addressing the EU Member States. This dimension considers several network indicators, as information exchange and interactions between multiple actors are essential for the blooming of innovation.
At worldwide level, the AI Watch Index analysis of R&D relies on data sources about patent applications and publications in AI international conferences. The former are used to address innovation capacity, while the latter are used as a proxy for involvement in frontier research. When the analysis is done at the EU level, participation in EC-funded projects is also considered.
The indicators show that in terms of AI R&D, the three worldwide leading regions are the US, China and the EU. The US presents the most consolidated position, with a remarkable level of activity in terms of frontier publications, in which it leads in terms of the number of players involved, the number of publications and the strategic position of US players in the network of collaborations – a score that provides a metric of players’ capacity to act as connecting bridges between other players. It is important to note that the role of scientific publications is more relevant for innovation in a domain such AI than it is in technologies developed during the third industrial revolution (e.g., semiconductor materials, automation, computers). Indeed, a large part of AI research is on physical supports enabling increasingly faster and more distributed computation. Additionally, AI is to a significant degree about algorithmic and software-related improvements, without forgetting interaction between humans and machines and its related ethical considerations. Since software innovations are typically not patentable, this makes scientific publications more appropriate than patents as a measure of the techno-scientific progress in the AI domain.
The US also performs well in patents, where it ranks second by number of applications, just after China. What seems to be the common feature of US patenting activity and US frontier research is the outstanding involvement of firms in both. This insight suggests that the US private sector is very active in AI R&D, and thus it is building its future competitiveness in the domain. This may also reflect the shift from academia to industry that is increasingly observable in the AI domain, as shown by the diversity indices analysed in the societal aspects dimension. China also has a prominent role in R&D. Its competitiveness in this respect mainly comes from an extensive involvement in patenting activity. However, what observed with regard to scientific publications is modest.
The EU holds a significant position in R&D, although its patenting activity is quite limited when compared to China and the US. Nevertheless, patenting involves a network of collaborations that makes the EU the third geographic area in terms of strategic position in that network, above Japan and other Asian countries (among the others). Only China and the US have a better strategic position than the EU. With regard to AI frontier research, the EU has basically the same number of players as the US, and it is second in terms of overall number of activities developed. These two elements enable it to be the second geographic area in terms of strategic position in the network of frontier research collaborations. It is important to note that, especially in R&D activities, a favourable position in terms of collaborations is expected to result in a future advantage in competitiveness. Indeed, as innovations typically emerge from the accumulation of knowledge that follows interactions and exchange of information, it is of major importance to have and maintain a strong set of connections and collaborations.
EC-funded projects stimulate AI R&D in the EU. Thanks to them, the number of players involved in the AI domain has gone up considerably in every country of the Union, and it is thanks to these projects that EU Member States can establish connections across multiple countries. Although aimed at the development of research projects, it is important to note that EC-funded projects not only substantially stimulate the joint work of research centres of the EU, they also give an important push to collaborations between firms located in different Member States.
The R1 – AI players in AI R&D indicator shows the presence of players active in AI R&D activities, to assess how distinct geographical areas are involved in the technical development of AI.
This 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.
Network of collaborations
With this indicator we investigate the EU 27 Member States’ propensity to structure collaborations with many different countries.
This indicator measures how many collaborations are developed by AI players from each geographical area (or Member State) by type of R&D activity: patent applications, frontier research publications and, when analysing the EU, EC-funded projects.
This indicator provides a metric for players’ capacity to act as connecting bridges between other players, aggregated at the geographical level, to assess the influence that a geographical area can exert on other areas thanks to the structure of collaborations in which they are involved.