Scientists split as Europe paves way for ‘Cern of AI’

Scientists are divided over the extent to which European artificial intelligence research should be concentrated in a central facility as Brussels mulls creating a hub equivalent to Cern’s role in nuclear science.

With Europe being outspent by Asia and North America when it comes to AI research, the European Commission has warned that the continent “cannot afford to maintain the current fragmented landscape” in which no research centre has “the scale necessary to compete with the leading institutes globally”.

To remedy that, Brussels wants Europe to have a “lighthouse centre of research, innovation and expertise” that will become a “world reference of excellence in AI” that hosts “the best talents in the field”.

Details are yet to be fleshed out as the commission awaits feedback on its White Paper On Artificial Intelligence – A European Approach to Excellence and Trust.

Nevertheless, Holger Hoos, professor of machine learning at Leiden University, said the document was a step towards creating an AI equivalent of Cern – a goal of the Confederation of Laboratories for Artificial Intelligence Research in Europe (Claire), which has been lobbying policymakers in Brussels to back such a research hub.

“‘Critical mass’ is the word that we really like,” said Professor Hoos, one of Claire’s founders. “You really have to integrate all of Europe.”

Claire envisages a Cern-like institute hosting up to 800 researchers, he explained. “Whenever you’re there, you’re surrounded by the top people,” he said.

But some researchers have argued that a more distributed network would be a better option.

Claire backs a central hub in addition to a Europe-wide network. To compete with the likes of Google, “sometimes you just have to get the right people together”, said Professor Hoos, who views the commission’s White Paper as an endorsement of its strategy.

In contrast, another group, the European Laboratory for Learning and Intelligent Systems (Ellis), wants a “network of excellence”, more akin to the European Molecular Biology Laboratory, which has multiple labs across the continent, explained Nuria Oliver, a computer scientist and Ellis board member.

“Currently, so many needs for innovation are put under the label of AI that there will be no way to address them with a single organisation,” cautioned Daniela Diaconu, Ellis’ scientific coordinator.

One challenge would be deciding where to locate such a centre because European Union member states might fear that a central continental hub would undercut their own national AI research strategies.

Claire’s solution would be to make placements at the hub temporary, Professor Hoos explained. It would not employ permanent academics, instead providing only support staff and computing facilities. In return for sizeable research grants, the continent’s leading AI scholars would agree to spend a proportion of their time at the hub but without fully relocating from their home countries.

Anxious European AI researchers have warned that the continent’s investment is lagging behind the US and China, and that progress is being hampered by a drain of talent to US technology companies such as Google.

According to the commission, Europe invested €3.2 billion (£2.7 billion) in AI research and innovation in 2016, roughly half of Asia’s spend (€6.5 billion) and barely over a quarter of North America’s outlay (€12.1 billion).

david.matthews@timeshighereducation.com