Guntram Wolff is director of the Brussels-based economic think tank Bruegel.
The European Commission’s Executive Vice President Margrethe Vestager has been tasked with developing a European strategy on artificial intelligence. For Europe, the crucial question is how to speed up AI uptake in the industrial sector — indeed, only 18 percent of large European companies use AI tools at scale.
So far, however, the European Union has been more interested in writing the rules of AI than in winning the game. That’s unfortunate. As a consequence, there isn’t much European AI ready to be adopted.
When European companies want to adopt AI solutions, they are almost inevitably forced to turn to vendors from the United States or China — the world’s two incontestable leaders when it comes to the technology.
The two countries dominate when it comes to innovation, with about 45 percent of AI-related patent filings made in the U.S., and another 40 perent in China. They also lead on the public research front: Two-thirds of top-filing universities and public research organizations are based in China. And they’re the major players on the commercial front: Four U.S. firms capture about a quarter of the worldwide AI market.
The EU needs to make sure that AI firms in Europe can use data to train their machine learning algorithms.
The EU offering is comparatively poor. Of the top 30 AI-related patent applicants, only four are European. Nor is the future looking more promising. Of the 100 most promising AI startups in the world, only two are from the EU (while six are from the U.K.), and they attract well-below-average funding.
Being so dependent on foreign technology raises at least two concerns.
The first concern is geopolitical.
The adoption of AI is critical for the success of EU businesses. In the industrial sector in particular, AI solutions are becoming ever more important, as companies optimize production processes using internet of things (IoT) devices increasingly powered by 5G connectivity.
For example, car manufacturers can use AI to analyze data from IoT sensors to better predict machinery failures and save costs. AI can also help firms to better set prices of components and other products.
As the use of AI grows, so will the risks that come with depending on a technology that is produced and controlled outside of the EU. The risk that supply chains will be disrupted by economic decoupling has become very real in the current geopolitical climate.
The second concern is that giving access to data can provide long-term advantages to existing AI companies — making the development of European AI ever more difficult. While not all AI applications need large amounts of data to achieve economies of scale and scope, many do.
Only a bold strategy by Vestager and her colleague, Internal Market Commissioner Thierry Breton, can secure AI in Europe.
First, the EU needs to invest in its own AI technologies.
Of course, it makes sense for many companies in Europe to quickly adopt U.S. or Chinese AI solutions. But only if European companies can at least partly master this technology can the EU be sure to have systems that do not undermine its security. And only if the EU has the talent on the ground can AI be productively implemented in firms.
This means substantially more money for research as well as education. Lack of AI-related skills is the most important barrier to adoption in Europe, according to a recent McKinsey study. Without the U.K., the EU has an acute scarcity of data scientists and programmers, and it has difficulties retaining those it has.
Work is also needed to improve the ecosystem for innovative AI startups to strive and scale up. State aid can easily lead to waste, so continued vigilance and rigorous enforcement of competition and state aid control is needed. But conditions for entrepreneurs need to be improved, and industrial policy should be designed to support AI innovation clusters.
Second, the EU needs to make sure that AI firms in Europe can use data to train their machine learning algorithms. Even if EU firms adopt foreign AI technologies, they should retain the rights to their own data and be ready to offer it to either in-house developers or European firms.
Europe leads the world in AI regulation, which is important but insufficient to exploit its potential.
The success of AI applications rides on access to data as much as it does on digital skills and infrastructure. The deployment of IoT in manufacturing is an opportunity for the EU to become a leader in AI if the data management can be governed by European rules.
Finally, the EU needs to create a single market for data, especially data from industrial processes. European AI firms will stand little chance of staying in the game when it comes to machine learning algorithms if they aren’t able to leverage the data of the entire EU market.
The EU regulation on the free flow of non-personal data is an important element for creating a genuine European approach to data. But there are still numerous obstacles to data-sharing across EU countries, for example due to corporate governance requirements.
Europe leads the world in AI regulation, which is important but insufficient to exploit its potential. Now, Vestager and Breton must decide if Europe wishes to become a global player, and not just a global referee.
Julia Anderson, a research assistant at Bruegel, contributed to this article.