This policy brief focusses on the institutional and political implications of the authors’ original study, Artificial Intelligence: Key Technologies and Actors. That study is part of a broader framework and addresses various aspects of artificial intelligence, particularly its technological aspects. One of its purposes was to map AI activities and expertise in the eight leading countries in terms of patent applications filed between 1990 and 2017, namely the United States, China, Japan, South Korea, Germany, the United Kingdom, France and Canada, differentiating between the positions of private and public actors within the complex technological field of AI.
Paving the way for innovations that were once out of reach, artificial intelligence (AI) is expected to be a general-purpose technology, just like the steam engine, electricity and electronics in past industrial revolutions. In the space of just a few years, AI has extended into sectors as diverse as transport, telecommunications, healthcare, education, justice and safety.
Since 2010, there has been an interrupted rise in the number of AI patents filed each year. Over the past 30 years, the top countries for AI patent production have been the United States (30%), China (26%), Japan (12%), South Korea (6%), Germany (5%), the United Kingdom (2.5%), France (2.4%) and Canada (1.9%). The US and the Asian powers alone represent nearly three fourths of the AI innovation market. With more than half of global market shares, the United States and China have clearly asserted their dominance. What place is left for France and the rest of Europe in this strategic and ultra-competitive space? Have they achieved sufficient critical mass?
According to the Tortoise Global AI Index 2021, which evaluates nations based on their levels of investment, innovation and implementation of artificial intelligence, the US and China have held on to their leadership positions. Canada moved up to 4th place. It is ranked 1st for government strategy (ahead of China) and 6th for commercial strategy. As for France and Germany, they dropped to 10th and 9th place respectively, right after the Netherlands. However, France rose to 5th place worldwide in terms of government strategy, above the US and Germany among others.
In April 2021, the European Union released its new Coordinated Plan on Artificial Intelligence, built on collaboration between the Commission and Member States. It was based on the 2018 Coordinated Plan on AI. Its key objectives include accelerating investments and aligning AI policy to remove fragmentation. That being said, in a field involving such colossal investment costs, French and European decision makers are subject to ‘path dependence’, meaning their future strategic choices are limited by previous long-term commitment choices. So, they must have a detailed understanding of the comparative advantages enjoyed by their own country and its competitors in AI-related fields in order to target specific investments that will allow them to make the necessary quantitative effort to expand their market share in key sectors.
This policy brief aims to give public and private decision makers some historical perspective on the strategic positions adopted by the leading AI countries over the past 30 years. This perspective is essential to making informed decisions about future investment, organisational and collaboration choices.
In the light of these insights, several recommendations are provided below.
Create the conditions for effective technology transfers between French actors in the public and private sectors, to make France competitive in AI beyond the borders of Europe
In terms of AI specialisations and market share, France never finds itself in a situation of consolidating a dominant global position in any of the sectors along the value chain connecting the science to AI applications. It showed promise in neural networking techniques and in the field of transport but has not gained any market share since the 2010s.
Although ranked 7th for the number of AI patents produced, French private actors have struggled to compete with actors from the US, China and even Germany, as shown by the absence of any French private actors in the list of the top 20 biggest AI patent producers. Still, France is the leader in European public innovation, with six research institutions figuring among the top 10 public AI actors in Europe.
France is characterised by strong public sector research in the field of AI. This means that public sector decision makers have to rely on that research to create the conditions for transferring technology to actors closer to the markets.
Build national AI policies around the particularities of each country’s specific national innovation systems
Contrary to what might be expected, there does not appear to be a clear connection between States’ specialisations and their achievements. Nonetheless, each country is working to specialise in at least one AI discipline. In France’s case, there is a strong specialisation in the field of semantics. It is likely that this comparative advantage will be a real asset for the country when it comes to AI, because semantics is heavily dependent on computer science and expert systems, two areas (one scientific and the other technological) in which France displays an equally significant comparative advantage. The two other functional domains in which France specialises are character recognition and computer vision. However, those fields rely on learning techniques and sciences in which France has not demonstrated any especially strong skill sets.
Coordination between public and private actors in particular is specific to each region of the world and each country. A wide variety of models for interactions have been observed. This becomes all the more obvious when you look at the link between the dominant national companies and the countries’ specialisations. That being said, the organisation of AI innovation is a crucial factor that should guide public policy decisions.
The diversity of national innovation systems calls for caution when it comes to writing policies in support of artificial intelligence that mimic the policies implemented in a benchmark country. In fact, this kind of diversity means that those policies cannot function in a vacuum, without relying on key national actors and tapping into synergies between the scientific, technological and functional domains that are most promising for the country.
Envision and build a European model for AI innovation that is in line with the diversity of its States, to improve Europe’s position on a global scale
Two findings led the authors to this recommendation. The first was that Europe is not at the forefront of global competition in the field of AI. In fact, only two European companies figure among the major actors in AI. Plus, European companies produce far fewer patents than the American and Asian giants. A quantitative effort is necessary at this point.
The second finding was that there are vast differences between national AI innovation systems. This can be seen in the radical divergences in the organisation of innovation in France and in Germany, particularly in terms of coordination between private and public actors and the opposite forms taken by their respective collaboration networks.
To build up its comparative advantages, Europe will need to consider establishing a European model for innovation in the field of AI. But the differences observed, especially between the German and French innovation systems, raise questions about the feasibility and coherence of a potential European model. Would that system be efficient? How would it fortify the specialisations of key European actors, both public and private? It will be up to public sector decision makers to answer these questions and devise more innovative organisational structures, as it is now recognised that being excluded from future developments in artificial intelligence would be synonymous with a loss of international influence and economic independence.
Anticipate the impact that AI might have on innovation activities and the job market
The gains expected from AI, like in business productivity and the creation of new markets, have to be assessed in the light of the expected costs. It is important not to underestimate the challenges posed by the development of digital technologies in general, and AI in particular. The effects of reallocating capital and labour between companies and sectors as a result of AI remain to be assessed, although it is certain that the process will engender substantial adjustment costs for companies and workers alike. The most significant impact of AI is expected to be felt in innovation activities and in companies’ demand for labour.
Companies will need to revise their business models, make the necessary complementary investments and adapt their technical skill sets and human capital. For their part, public authorities will have to update current regulations, provide training to accompany the rise of AI and invest in scientific and technological infrastructure.
