DeepMind dominates European AI research: What does this mean for researchers?
AI’s steady impact on the academic and research community is measurable through citation metrics, essentially showing how many times a study has directly influenced subsequent research. A recent analysis of AI-related citations showed beyond doubt the impact of AI. It also revealed another noteworthy statistic: Google’s DeepMind made up just under half of all AI-related citations from 2020-2024.
The company’s dominance is undoubtedly a testament to the importance of its work -- but it also serves as a starting point from which to probe further into the research landscape in Europe and how it’s been impacted by AI. Concentrated influences in academia and research can have long-lasting effects on funding distributions, channels of collaboration, and ultimately the potential for innovation. Add to this the powerful and relatively new tool of AI, and suddenly the future trajectory of scientific research on the continent looks a lot less clear.
DeepMind’s Leading Role in AI Citations
DeepMind’s influence in AI research can’t be overstated. Its citation count in Europe exceeds that of its next seven closest competitors combined -- making up almost half of the continent’s citations in the field between 2020 and 2024 (15,213 out of 33,450). There’s no secret as to why -- DeepMind’s research is of extremely high quality and importance.
The company’s achievements are worth laying out in detail. AlphaFold DB (built in partnership with the European Molecular Biology Laboratory’s Bioinformatics Institute) is among the most impressive. This open-source database details almost all known protein sequences and has become an indispensable resource for researchers across the globe.
Indeed, AlphaFold research constitutes the bulk of DeepMind’s citations. A 2021 paper from the company allowed the prediction of three-dimensional protein structures from their amino acid sequences. Solving this (one of the oldest problems in computational biology) puts DeepMind’s research among the most significant breakthroughs ever made in biotechnology.
In light of all this, DeepMind’s dominance is unsurprising. However, there’s no escaping the fact that when one entity holds such enormous influence, questions must be asked about the health of the ecosystem as a whole.
The Consequences of Citation Dominance
One question that has dogged researchers for decades is a fundamental one: how useful is citation count as a metric to measure research impact? It’s a difficult subject to unpick. On the one hand, of course, the most impactful research will attract the most citations -- making it a valid measure. However, at a certain point, it can reach a critical mass and become a vicious cycle. Researchers cite DeepMind because they’re the best known… due to the number of citations! This is when we start to see citation bias emerge. Researchers become more likely to cite papers from the largest institutions, when in fact there’s no reason to suspect that their research is of higher quality than lesser-known organizations. This in turn bolsters citation figures further.
This would be a shame if it simply meant that smaller institutions’ research was underrepresented compared to bodies like DeepMind. However, this has much further reaching implications. Recognition and prestige translate into funding opportunities. When citation bias takes hold, not only do we lose the value of potentially excellent research, but it becomes materially more difficult for less well-known research institutions to undertake further research in underrepresented fields. Ultimately, this risks a narrowing of the focus of European AI research compared to other regions.
To avoid this, a new means of evaluating research is needed. Raw citation figures can only provide a guide. What’s needed is a system that’s broader in scope and addresses the actual impact of research across the whole continent.
Impacts on European AI Research
There are other existing imbalances revealed by the data which may be exacerbated by excessive concentration of influence in a few institutions. The data shows that the majority of AI research citations come from a small cluster of Western European countries. The UK and Germany take the lead, with the Netherlands, Spain and Switzerland also performing well. Eastern Europe, by contrast, scarcely registers. This is not because all the most important research is being undertaken in a handful of countries, but due to a self-reinforcing cycle which functions in much the same way as the citation bias described above.
There are a number of ways of addressing this issue. Higher levels of collaboration would both address the underrepresentation of Eastern European entities in AI research and enable original insights to emerge from a more diverse range of views. In fact, Europe has the opportunity to leapfrog other nations on co-authorship. The U.S.A. and China both display fairly low levels of collaboration between researchers from different institutions, so a renewed focus on international co-operation in Europe may give the continent a global competitive edge.
Other challenges are even harder to overcome. As mentioned, citation biases can result in funding allocations being concentrated in a few locations, fields, or institutions. Policymakers are naturally highly sensitive to perceptions of how public money is spent, so are more likely to fund research with an easily measurable impact. Citation figures provide a straightforward statistic, making it easier to justify funding decisions.
Even here, though, there are opportunities. Policymakers are also concerned with ensuring taxpayer-funded research will produce tangible benefits for the public good. This is, generally, good news for AI researchers. Some of the top-cited publications in AI research in recent years have been medically focused. Research into the recent COVID-19 pandemic and protein structures suggests a bright future for collaboration between biomedical sciences and AI.
A boom in other areas has prevented biological applications from completely dominating the space. Responsible and transparent AI is also a prominent topic for policymakers and industry leaders alike. Recent research from Spanish institution Tecnalia has brought in a range of experts to tackle the issue. Again, we see the benefits of collaboration -- this time between AI researchers and the social sciences, bringing legal, political and sociological knowledge to bear on the technology.
Clearly, collaboration is a powerful tool to counteract biases and a concentration of influence in scientific research. Indeed, projects like AlphaFold DB are both a symptom and an enabler of increased cross-border collaboration. Europe’s diversity is a major competitive advantage in research, which it has yet to leverage fully. As open science practices gather pace, European AI researchers are set to have an even greater impact on global research.
Strategic Approaches for Researchers
As we’ve seen, international and interdisciplinary collaborations are among the best ways for European AI researchers to maximize their impact. DeepMind’s success is itself a prime example of this. But whilst powerful, academic collaboration alone is not enough. Industry partnerships provide another avenue for discoveries, allowing access to specialized equipment and insights into practical uses of AI. Many institutions are increasingly choosing to specialize to overcome size and geographical disadvantages. Focusing on region-specific problems or niche challenges can allow research teams to differentiate themselves through their work.
In this vein, there are countless sectors which might be transformed by AI given the right application. Healthcare currently leads the way, but areas like material sciences and climate research are fertile ground for AI researchers. Initiatives dealing with sustainable energy or extreme weather mapping, for example, are by their nature global in scope. As such, involvement in these areas could offer a gateway for Europe’s AI ecosystem to have its impact felt across the world.
DeepMind’s grip on European AI research serves to highlight all of these challenges and opportunities. The company’s work has been rightly identified as genuinely groundbreaking, and projects like AlphaFold DB offer a glimpse of how AI researchers might be able to overcome the obstacles in their way. Through interdisciplinary and international co-operation, diversified areas of study, and open science, a more diverse AI ecosystem is possible. It’s up to Europe’s AI researchers to chart this course and maximize the community’s impact across the globe.
Anita Schjøll Abildgaard is co-founder and CEO, Iris.ai.