January 9, 2019

What European AI lacks in funding it makes up for in ambition

Kai-fu Lee, ex-president of Google China, said there was no hope for Europe in the global AI race. This is how European AI experts responded.

Carly Minsky

8 min read

In a recent interview with Sifted, Kai-fu Lee, AI-expert and ex-president of Google China, said Europe would struggle to even win third place in the global AI race. “Europe has none of the success factors of the US or China,” he said.

This is how European AI experts, speaking at the European Commission’s policy forum on artificial intelligence, Futurium, responded:

Celestino Alvarez, founder and CEO of social robotics startup Adele Robots, Valencia region, Spain

This line of thinking is something I have heard some more times, most frequently outside Europe.

Although I personally agree with the diagnosis, I do not so much agree with the conclusion. This EU initiative can totally change the board, if we acknowledge the current situation and take the right decisions to head in the right direction.

Fotis Savva, PhD student in machine learning at the University of Glasgow, UK

Indeed the US, Canada and China are leading the way towards building AI systems. This is evident from the number of publications in high-ranking conferences and journals and strong AI research groups residing in those countries. To my knowledge there are no similar groups with available funding and manpower in any European country.

It would be of tremendous help if the EU could announce funding opportunities for research in AI.

Traditionally European universities (mainly in the UK) had strong reputations for training people in AI but those people subsequently left because of the financial incentives given to them by institutes abroad.

However, things are looking brighter due to big technology companies opening up research labs in Europe (Google Brain, Facebook FAIR) but this has nothing to do with advancing AI research as companies are mainly interested in the business value that such AI systems bring and will often not explore research directions which are of no significant business value in the near future.

 It would be of tremendous help if the EU could announce funding opportunities for research in this direction.

Fatos Ismali, data solutions architect in data & AI at Microsoft, London, UK

By definition an AI system will at some level automate a specific decision-making process that humans currently own. Therefore, AI should be trustworthy if it's going to find its place in society.... From that perspective Europe seems to be investing more than other regions in the world.

AI should be trustworthy.... From that perspective Europe seems to be investing more so than other regions.
Darek Saunders, senior research officer for data analytics, Frontex - the European border and coast guard agency, Warsaw, Poland

Activities such as the High-Level Expert Group on Artificial Intelligence show that Europe has decided that it is time to accelerate, otherwise we will drop off the AI race. We should all put as much effort as possible on all AI aspects - standardisation, ethics, research, innovation, etc. There is still time but if we do not put high level sponsorship/support, HR and budget in place asap then it will be getting more and more difficult to be in the front of the peloton.

We need to create the environment to encourage the best people to stay here, return here, or to come here in the first place.
Maik Neubauer, managing director at TSCNET, operational security service-provider for electrical grids, Munich, Germany

Currently the AI activities in Europe are still too slow and fragmented to really cope with the advanced entrepreneurial culture in the US or the huge centrally organised R&D and government-directed AI cluster activities in China. But there are first positive signs in various European countries.

A lot of tiny startups are developing interesting platforms for pilots in various sectors. The flagship study Artificial Intelligence - A European Perspective shows that Europe has acknowledged the urgency to compile the forces to develop an R&D network. 

Europe has to build on its available engineering and technology strengths in manufacturing, automobile, chemical and distributed high-tech industries, but clustering the AI activities will be a key success factor.

Barry O’Sullivan, professor in artificial intelligence at University College Cork (UCC), vice-chair of European Commission’s High-Level Expert Group on AI, president of the European AI Association, Cork, Ireland

The US and China are currently ahead of Europe in the business to consumer setting, but I don't believe this to be true in B2B. Europe produces far more PhD graduates than the US. Some statistics I saw recently stated that there were more AI IPOs in Europe than in the US in 2018. If you look at some of the biggest names in AI (Hinton, Thrun, LeCun, Ng, and others) they are European.


I also don't agree with Kai-Fu Lee's position that the European focus on regulation, especially around the protection of personal data, is a liability. I would argue that it is an advantage. Personal privacy, agency, free democracy, etc., are core values in Europe, and are therefore core to what we expect from technology.

The US and China are currently ahead of Europe in the B2C setting, but I don't believe this to be true in B2B

Europe certainly has challenges, such as keeping its talent pool in Europe (or attracting the diaspora back to Europe), a strong funding culture for innovation, etc., but we can and are addressing those.

Finally, if Europe wasn't a strong market, how does one explain the significant increase in presence of Chinese tech/AI companies in Europe?

Celestino Alvarez, Adele Robots

That those [big names in AI] have been born in Europe, for the matter of this discussion, doesn't mean they are European. They do not pay taxes in Europe, and the revenues they obtain for their companies are reinvested based on decisions taken outside Europe... So that is quite useless...

The statistics about PhD combined with the diaspora, means that Europe is paying for the training of high-qualified workers in the US.

[Chinese AI companies]  come because of the 300 million consumers in Europe. There is no chance for a European AI industry, if foreign companies are already selling AI here.

Europe is paying for the training of high-qualified workers in the US

To me, the point of this discussion is about being conscious that this is a race, that we start from behind (in terms of revenue streams, and so in terms of investment capabilities), and that we cannot keep sitting comfortably on the sofa thinking that we have some of the brightest minds in the world.

To keep our loved core values, we cannot lose this race. US and Chinese companies are better that European companies at creating revenue streams on new technology. We need to fix this.

We both were at a meeting where an "European CERN for AI" was discussed. This is not a proper simile. A "European Airbus for AI" is needed, and the difference is about the focus on the market. The complexity of AI and the need for collaboration that the field requires are the key points that give Europe the chance to succeed.

A "European Airbus for AI" is needed,
Kai Salmela, AI specialist at industry-research innovation centre, Robocoast R&D, Finland

European companies do deliver AI-like systems for special needs, but they do lack the well-known top-notch products that can be used worldwide.

There's no existing standard yet for AI. Europe could see this as an opportunity and participate with full force in [its creation]. GDPR really is a blessing in disguise, and could be used to form European AI with a market-reach outside of this continent. 

There is also a possibility to use the EU-wide AI Digital Innovation Hub’s network for developing truly European AI ( yes we're here, just lacking funding but eager to do wonders.)

There's no existing standard yet for AI. Europe could see this as an opportunity

How about if we develop modular AI where companies could participate by doing just the module they are best at? For example: a camera with machine vision from Leica, smart communication from Nokia and audio modules from Genelec? Users could combine these modules just as they wish and need, so long as all modules are produced within the standard way.

Europe is also multicultural. We could and should use this as an asset. Developing a speech recognition package for every European language and dialect would make European AI the most desirable product around (and when that is done, why stop at these languages?)

So, do not bury the European AI just yet.  Please be innovative and start working.

Barry O’Sullivan

There is a lot of hope for Europe's AI sector. For example, if we look at the Nikkel/Elsevier most-cited AI research institutes in the world, the Top-100 has 32 European institutions, 30 from the US, and 15 from China.  The Times Higher Education ranking for Computer Science has 5 European universities in the top 5, equal with the US.

What we need to do is create the research environment and the commercial environment to encourage the best people to stay here, return here, or to come here in the first place.