At Rise of AI 2022 in Berlin, a panel with Florian Schütz (K.I. Park), Joanna Bryson (Hertie School), André Loesekrug-Pietri (JEDI) and Vanessa Cann (German AI Association) discusses the state of the European AI ecosystem. Schütz opens by arguing that ecosystems are not needed to develop AI technology itself, but are essential to turn technology into real societal impact, since that requires values, regulation, funding, infrastructure, data, talent and users working together. Bryson presents patent and market-capitalisation data from a paper written with a colleague at DG Competition, arguing the EU and China are closer in AI patents and market capitalisation than commonly assumed, and that the US, not Europe, is the outlier. Loesekrug-Pietri describes building JEDI, a foundation modelled on DARPA, and criticises European fragmentation, weak deep-tech scale-up and dependency on foreign infrastructure, citing examples from space, chips and vaccines. Cann outlines Germany's strong AI research base and established companies, but points to weak corporate-startup collaboration, low research-to-product transfer, and underfunded startups as the main obstacles to scaling the ecosystem.
Key takeaways
- Florian Schütz argues ecosystems are not needed to build AI technology, only to turn it into societal impact.
- Joanna Bryson's patent and market-cap research finds the EU roughly on par with China in AI, with the US as the real outlier.
- André Loesekrug-Pietri set up JEDI as a foundation, not a government body, modelled on DARPA, after concluding state-led efforts move too slowly.
- Loesekrug-Pietri cites Europe's falling cloud market share and reliance on Intel and Chinese solar/battery supply chains as signs of dependency.
- Vanessa Cann says German AI startups struggle less with quality than with corporate collaboration, research-to-product transfer and investment.
Detailed notes
Why ecosystems matter for impact, not just technology
Florian Schütz, managing director of the newly founded K.I. Park ecosystem, opens by arguing that ecosystems are not required to develop AI technology: a single researcher, startup or large company can build new technology alone. What ecosystems are needed for is turning technology into real impact, for example using AI for future healthcare, mobility or sustainable production, because that requires accounting for societal values, user habits, political regulation and funding, business models, infrastructure, data, capital, talent and skills, and end users in industry and society.
Bryson: patent and market-cap data on the EU, China and the US
Joanna Bryson, professor of ethics and technology at the Hertie School, presents research done with Helena Malikova, then at DG Competition, published as a paper called "the AI Cold War." Using the WIPO international patent database for AI-related categories rather than a narrow list of top platforms, they found the EU and the EEA were roughly on par with China in patents and market capitalisation, while the US far outstripped the rest of the world on both measures, and the rest of the world outside the EU, China and the US combined had more patents and market capitalisation than the EU and China together. An updated analysis by a master's student, using companies with even one patent, shows the EU ahead of China on both measures in 2019, with unusual shifts for Japan and China by 2021. Bryson concludes Europe should not feel as threatened by the US narrative on regulation holding it back, and questions why large firms like Google, Microsoft and Facebook resist disaggregation when HP increased shareholder value by 40 percent after splitting.
Loesekrug-Pietri: building a European DARPA and naming dependencies
André Loesekrug-Pietri describes running JEDI, an initiative built as the European equivalent of DARPA and also the largest deep-tech ecosystem in Europe, set up as a foundation rather than a government body because he concluded government-led efforts are too slow. He lists examples of European dependency and fragmentation: SpaceX had 23 launches last year versus three for Ariane; European cloud market share fell from 30 to 15 percent over five years; the EU is subsidising an Intel chip plant near Berlin despite Intel not yet mastering 5-nanometer chips; 99.7 percent of solar panels and 60 percent of batteries are made in China; and Berlin has the second-highest number of CCTV cameras per inhabitant in Europe, after London. He argues Europe reacts to events (the pandemic, Ukraine) instead of shaping them, that national AI strategies fragment effort across 27 ecosystems, and that Europe should compete on scale rather than ideology. He proposes concrete cross-border projects, for example privacy-respecting facial recognition without centralised databases, batteries without lithium, and rotating civil servants and entrepreneurs between government, academia and startups, and notes the EU is investing 750 billion euros in "Next Generation Europe" without precise KPIs beyond spending targets.
Cann: Germany's research strength versus weak scale-up
Vanessa Cann, managing director of the German AI Association, which represents about 400 AI companies with at least 70 percent of their activity in AI and a German seat, describes Germany's strengths as a strong research base and established corporates and SMEs that became more open to AI during the pandemic. The main challenges she names are that corporates and SMEs struggle to collaborate, too few startups are founded on German research results, and startups struggle to find investors, pushing German AI companies toward quality and "ethics by design" rather than scale due to underfunding. She argues no single actor (politics, academia or industry) is responsible for keeping talent in Europe; building venues where research, product and business people meet is a shared task.
Q&A highlights
Bryson pushes back on Loesekrug-Pietri's framing of European vaccine exports as weak, noting the EU exported half its vaccine supply while keeping half, and questions whether Tesla's market capitalisation reflects real competitive advantage over traditional carmakers. Loesekrug-Pietri responds that market capitalisation, while only one indicator, is self-fulfilling because it lets companies acquire firms and attract talent, and cites COVAX data suggesting the US, not the EU, donated the highest share of vaccines relative to its own use, crediting the US Operation Warp Speed's parallel-funded vaccine projects. On what would most change the ecosystem, Bryson stresses diversity of approaches and engaging with people who want startups to succeed rather than working in isolation; Loesekrug-Pietri argues for concrete cross-border projects over abstract frameworks and rotating people between government, academia and business; and Cann argues for federal-level collaboration around shared "moonshot" projects instead of German states or European countries each working alone.
Quotes
AI strongly depends on ecosystems to develop real impact.
Florian Schütz
I don't think we should be feeling so threatened by America.
Joanna Bryson
Europe should not be in an ideology but it should be the scale that we need for technology.
André Loesekrug-Pietri
Questions answered in this talk
Are the EU and China really far apart in AI development?
According to Joanna Bryson at Rise of AI 2022, patent and market-capitalisation data from the WIPO database shows the EU and China roughly on par with each other, while the US far outstrips both.
Why did André Loesekrug-Pietri set up JEDI as a foundation instead of a government agency?
At Rise of AI 2022, Loesekrug-Pietri explains he concluded government-led initiatives move too slowly, so JEDI, Europe's DARPA-style initiative, was set up as a foundation to act as a common good outside government control.
What are the main obstacles to scaling Germany's AI startups?
According to Vanessa Cann of the German AI Association at Rise of AI 2022, the obstacles are weak collaboration between corporates and SMEs, too few startups founded on research results, and difficulty finding investors.