Broken Horizon: In Greece, research in automation fails to find applications
Greek researchers led hundreds of research projects that involved automation, but very few found their way into real-life uses. Opinions differ on where the innovation pipeline is clogged.
This story is part of AlgorithmWatch's upcoming report Automating Society 2020, to be published later this year. Subscribe to our newsletter to be alerted when the report is out.
“Roborder” is a European research project. If its name sounds fancy or dystopian, wait for the tagline: “Autonomous swarms of heterogeneous robots for border surveillance”.
Roborder is endowed by Horizon 2020. The 80-billion-euro fund is “the biggest EU Research and Innovation programme ever” according to the European Commission, which oversees the program. Horizon 2020 is divided into thematic “challenges”. Roborder belongs to the “Secure Societies” challenge, which aspires to protect Europeans against a series of threats: Crime and terrorism, natural and man-made disasters, border security and cyber-security. (The European Commission makes little secret that “Secure Societies” is about preventing asylum seekers from reaching Europe, according to research by Mark Akkerman, a researcher in border militarization.)
Almost all Horizon 2020 consortia consist of a coalition of research centers, large and small companies, and public bodies, which the European Commission calls “end users”. Roborder’s consortium is made of 26 partners from 12 countries; its coordinator is CERTH, a Greek research center in Thessaloniki. This should come as no surprise.
Greek champions
Greece is only the 9th-largest EU country by population, but Greek researchers rank first in winning Horizon 2020 grants in the field of security, according to data from Horizon 2020’s dashboard. “This is a huge breakthrough – and evidence that the security research ecosystem in Greece is blossoming” says Dimitris Kyriazanos, a leading researcher for Greece’s top research organization, NCSR Demokritos, in Athens. Mr Kyriazanos has been Greece’s delegate to the “programme committee” for Secure Societies, an advisory body to the European Commission, since 2019.
Greek organizations were part of several Horizon 2020 projects that developed automated decision-making systems. Thessaloniki's CERTH is part of Anita, a program to spot criminals active in online marketplaces, and of Tensor, which aims at “retrieving and analyzing heterogeneous online content for terrorist activity recognition.” Kemea received half a million euros to develop Foldout, a project where machine learning shall be put to use to detect asylum-seekers when they enter the EU overland. Spirit, where the Greek police is an end-user, “will take a novel approach in the development of a new scalable privacy preserving intelligence analysis for resolving identities system prototype (sic).” (The Greek police did not answer requests for clarification.) There are many more.
Despite a flurry of research projects where Greek research institutions or end-users play a major role, very little evidence of automation could be found in Greece. If Greece is such a prolific producer of AI research, why does it have so little to show for it?
Not more that meets the AI
George Vardoulias, a researcher at the American College of Greece, with extensive research experience both in academia and in the industry, deplores the resistance to embrace automation: “Often automation is not compatible with the end-users’ wishes and stumbles on human prejudice. In Greece, we tend to politicize everything, labeling things as right-wing or left-wing, looking out who is to blame for the tech, who is behind it. In the past ten years, irrationalism has grown globally, not just in Greece. More and more people see science and engineering as a tool of control and oppression.”
We challenged various stakeholders of the Greek research ecosystem, from research centers to end-users, to provide examples of Horizon 2020 research projects that were transformed into products or have been applied in real life. Almost all our interlocutors had little or nothing to say.
Only one official from the ministry of Maritime Affairs, one of Horizon 2020’s main end-users, did point to one emblematic ‘success-story’ involving AI: “Optinet”, which was funded by the European Commission’s regional development fund. Optinet is “an innovative tool to optimize the network of coastal navigation” according to Athina Foka, the head of the EU Structural Funds service at the ministry. But Optinet started in 2018 and is scheduled to be delivered in 2021. Currently, it does not even have a website, and it is unclear what parts of the project involve automated decision-making.
Mr Vardoulias, the researcher at the American College of Greece, argues that “major breakthroughs are rare in research by definition. In most cases, EU-funded research does not directly lead to products; but it makes procedures better, it changes the way of thinking, it enhances cross-border collaborations. The closest you can get to a tangible effect is the impact research has on the rationale of designing the products in the next five to ten years. And the only way to achieve this is in the ‘casual’ environment of research.”
No blue skies in Greece
But another Greek researcher, who asked to be identified only as “Artemis” and who has worked on several Horizon 2020 projects, strongly disagrees: “This is not ‘blue skies’ research [research projects with no implementation goals]; this is supposed to be applied science. Projects are funded in order to lead to products or platforms, to serve real-life needs.”
Mr Artemis is vindicated by Horizon 2020’s official motto: “Taking great ideas from the lab to the market”. Even the European Commission acknowledged this shortcoming. In the “interim evaluation” of the Horizon 2020 programme, among the “lessons learned” was the need for more “impact” and “breakthrough innovation”.
If Horizon 2020 does not lead to impact or breakthrough innovation, then where is the added value?
Mr Vardoulias lays out its implicit rationale: “Since direct funding of companies is prohibited in the EU, the alternative was to find an indirect way to provide funding for the companies so that they remain competitive on a global scale. The US and China support their companies, so the EU should do the same. This is a very competitive process and funds are very limited. But it can provide both an initial boost for start-ups and an incentive for innovation for larger companies.”
Greece’s research on life support
The benefits are even more critical for a country such as Greece. Since 2010, hundreds of thousands of highly-educated graduates have emigrated. Austerity, imposed by foreign creditors including the EU Commission, exhausted the last drops of public funding to research. Mr Vardoulias points out that “during the crisis years, one of the only imports of capital was via the research programmes. They still fulfill an existential need for university graduates active in AI research, which is extremely difficult in a country with very limited industry.”
Mr Kyriazanos from Demokritos agrees: “Currently, many highly-skilled executives in Greek companies of the security sector originate from the research ecosystem. This would not have been possible without EU research programmes and the contribution of research institutions.”
Research programs are not without flaws, says Mr Vardoulias. “The research projects’ mantra is innovation. But calls are not always compatible with reality, with what is needed and what is feasible. In some cases, this along with the fierce competition leads the applicants to promise far-fetched things. Proposals that sound cool may get promoted at the expense of others that may be better fit for the reality imposed by budget and time constraints.”
Short life spans
Mr Kyriazanos says: “At the end of each project, there will be some tangible results that could be placed on track for commercialisation. However, if there are no national grants, support or calls to assist the researcher, they will move on to ensure funding for the next project. If there are no favorable conditions for the ‘next logical step’ in innovation, results either remain on a low maturity prototype or worse: they are shelved. This means lots of lost opportunities.”
In other words, Greece might only lack a way to take automated decision-making from the lab to the streets. But not all agree.
In 2018, a wildfire at Mati in the outskirts of Athens cost 102 lives. It was the second-deadliest forest fire in the 21st century after the 2009 bushfires in Australia. Mr Kyriazanos says: “It makes you wonder if things could have turned out differently. In 2014 we ran a big drill in forest firefighting in the framework of AF3” (Advance Forest Fire Fighting), a research project financed by the European Commission under Horizon 2020’s predecessor, called FP7. AF3 included several automated systems to predict the evolution of fires and evacuate communities. “And this has been hanging over me, Mr Kyriazanos added. It was never used – actually, even if we offered it for free to the Fire Service, they still would not use it.”
In the aftermath of the wildfires, it was reported in the Greek press that several projects and platforms that provided an automated response to such incidents and that had been developed at a cost of dozens of millions had been cast aside.
“Researchers should get out of their labs”
One of the main end-users of Horizon 2020 projects is the Greek Coast Guard. Ms Foka, of the ministry of Maritime Affairs, overviews the European research projects. She is critical of the work done by research centers, and not just in Greece. “Researchers should get out of their labs and see the real needs. If the deliverable is not tailored on the needs of the end-user, then nobody’s going to use it.”
But Mr Artemis, who has dealt with the Coast Guard within Horizon 2020 projects, tells a completely different story. “By design, the end-user is tasked to define specifications for the project. The Coast Guard rarely does so, and in the end, researchers themselves have to define the specifications.”
Almost two years after the end of a Horizon 2020 project in the field of security, he received a call from a Coast Guard officer. The officer asked the researcher on which workstation the software had been installed. Mr Artemis realized that for almost two years the so-called ‘deliverable’ had been standing idle in one of the ministry’s computers. The ministry’s own staff probably did not know where it had been installed, and certainly did not use it.
Erasmus for grown ups?
Mr Artemis says: “The sad reality is that behind a significant proportion of Horizon 2020 projects on AI, there is nothing but hot air. The acronyms are cool, the wording too but what’s really cool is the lifestyle! European research institutions and companies are in for the money. As for individuals, both in the research institutions and the end-users, it’s about a lifestyle of free travel, food and socializing around Europe. It’s an ecosystem addicted to EU funding; without it, it would die.”
Mr Vardoulias rejects this approach as over-simplifying but he asserts that, for some European companies, funding depending on EU research grants becomes an end in itself, as if this were the product: “It’s a Europe-wide problem. There are many start-ups across the continent that exist only to attract EU funds. In reality, they have no marketable product. What is their activity? European research funding.”
KEMEA, a rookie that tops the charts
In Greece, research has historically mostly been done by publicly-funded research centres such as NCSR Demokritos in Athens, the Centre for Research and Technology Hellas (CERTH) in Thessaloniki and the Foundation for Research and Technology-Hellas (FORTH) in Crete. But in the last few years, there’s a new entry in the scene: KEMEA, the Centre for Security Studies.
Despite being a rookie, KEMEA has grown to become the biggest recipient in total funding for the entire Horizon 2020 Secure Societies programme. For the years 2018-2020, it has received more than €24 million when the second European recipient lags behind with €11 million. (KEMEA’s share is so big that it still ranks first even when earlier years are taken into account.)
KEMEA is a hybrid. From an administration point of view, it is fully controlled by the ministry of Citizen Protection (the equivalent of the ministry of the Interior, also responsible for the police) which appoints its 7-member board of directors and its executive director. From a legal point of view, KEMEA is a private law structure, which frees it of some public scrutiny. Although it answers directly to the government, KEMEA is a member of the European Organisation for Security, the European lobby representing corporations such as Airbus, Thales and G4S.
KEMEA has participated in more than 100 European research programmes. To what avail?
We tried to approach KEMEA for months in order to understand the impact of the many projects it contributed to in the field of security and automated decision-making. Despite communications via phone and email and a preliminary meeting with its executive director, KEMEA declined to answer questions.
We also addressed questions to three of the most important end-users in Greece: the ministry of Defense, the police and the coast guard. Only the latter responded – and not in writing but over a casual phone conversation.
Did you like this story?
Every two weeks, our newsletter Automated Society delves into the unreported ways automated systems affect society and the world around you. Subscribe now to receive the next issue in your inbox!