• Analysis of most-cited ‘deep learning’ research papers since 2006 finds London, Oxford and Cambridge in top 20 regions in the world
• Singapore, China and Japan have experienced fastest growth in volume of research
• AI research in the UK is becoming more concentrated in a small number of locations

The figures, which compare volumes of research papers related to deep learning between 2006-2011 and 2012-2018, show a 77-times increase in the volume of research from Singapore, 46-times from China, and 41-times increase from Japan. This compares to a 21-times increase in research from the UK which ranks 9th in the world for growth of deep learning research.

Authors in the UK were responsible for 7.5% of the total research, coming in third behind the US and China for the total volume published. The US and UK lead the world when it comes to the 25% most cited papers, with the UK host to three of the 20 most-cited regions worldwide - London, Cambridge and Oxford.

Within the the UK itself the distribution of research is becoming more concentrated, In 2010 the top five locations - London, Oxford, Cambridge, Edinburgh and Surrey accounted for less than 40% of the total deep learning research in the UK but by 2018 those same five locations accounted for more than 60% of research activity. Oxford in particular has seen a dramatic increase in the percentage of research it accounts for.

The findings have been produced by the innovation mapping team at Nesta, the innovation charity. The team uses new data sources and methods to monitor the emergence of new technologies and inform innovation policy. The team are launching a report ‘Innovation Mapping Now’ at an event with the European Commission and the OECD on Thursday 21 March in London.

Juan Mateos-Garcia, Nesta’s Director of Innovation Mapping, said:

“Innovation maps like this allow us to measure closer to real time how new technologies and industries develop. This is precisely the kind of information that innovation policy-makers need in order to ensure that new technologies promote growth and wellbeing.

“For example, while some UK regions are world leaders in deep learning research, and this is to be celebrated, we also find that research is becoming ever more concentrated in a small number of locations across the country. Interventions to develop strong AI research capabilities in more parts of the UK will require relevant, inclusive and timely data like this.”

Ends

Press contact - Will Hoyles, 07812 362714, [email protected]

Full copies of the research are available on request and researchers are available for interview.

Notes:

- Deep Learning papers were identified using topic modelling, a text mining methodology.

- In all cases, the rates of growth in Deep Learning activity where much higher than in overall activity in arXiv, suggesting that the results are not driven by the increasing popularity of pre-prints as a medium to disseminate research.

- We use data from arXiv, a popular pre-print platform that is widely used to disseminate research findings in the AI community.

About Nesta

Nesta is a global innovation foundation. We back new ideas to tackle the big challenges of our time, making use of our knowledge, networks, funding and skills. We work in partnership with others, including governments, businesses and charities. We are a UK charity that works all over the world, supported by a financial endowment. To find out more visit www.nesta.org.uk