Jakub Zavrel, Zeta Alpha, 13 December 2021.
AI is widely seen as a strategically important ingredient for future economic prosperity and technological dominance. This leads countries and companies to compete for the best AI talent and to heavily invest in accelerating AI innovation. Who is leading the global AI technology race? At Zeta Alpha, we are focused on analyzing and organizing most publications in AI, so with the yearly flagship conference NeurIPS (Neural Information Processing Systems) behind us, we can draw up the score for 2021.
NeurIPS is highly prestigious and the competition to get a paper accepted is very tough (26% acceptance rate). Most large R&D organizations are participating in this open research community, as that makes them more attractive to the best talent in the field. Hence, the number of publications at NeurIPS can be used as a proxy for the technological quality of innovation in AI. No wonder NeurIPS sometimes feels like the Olympics of AI.
The US leads far and wide by number of publications, but China is a good second and catching up fast. Does it even matter? Even though a lot is at stake in leading the pack, AI research fortunately remains an open global intellectual and scientific dialog, as the NeurIPS program shows. As global geo-political tensions rise, it could start to matter more in 2022. In fact, the current leading position of the West might be a large overestimation, as close to 50% of the researchers in the US (and in Europe) are from China, India and other ‘competing’ countries. This reflects well on US and EU ability to attract global AI talent, but one can hardly expect to keep that talent and expertise in the West forever.
When we look at the Year-on-Year growth in the larger countries, and we take into account that NeurIPS has in total 22% more papers in 2021 than in the year before, we see that with 63%, China is indeed far outpacing the average growth, and moving forward at a very rapid pace. Germany is also accelerating growth with an impressive plus of 83% over last year. Canada (+58%), South Korea (+50%), Singapore (+77%), and The Netherlands (71%) are all seeing very significant growth. Quite notable is the rise of Austria (+77%) in this ranking, which can largely be attributed to the investments the country has made in basic research via the Institute of Science and Technology Austria. A smaller, though similar effect is seen in the growth of Saudi Arabia (+50%), attributable to the growing AI group at King Abdullah University of Science and Technology. Note that both of these research institutes were founded as recently as 2009. Russia, though severely impacted by brain drain, which causes it to punch far below its weight, does see a decent growth level (+58%) here, primarily contributed by research groups at the Higher School of Economics and Yandex.
When we turn to the rankings by company, we see that Google (incl. Deepmind) is the largest organization in terms of contributed papers (see below). If Google were a country, it would rank fourth after the US, China, and the UK.
In terms of growth percentage, Meta (formerly Facebook AI Research; +52%), Huawei (+105%) and Amazon (+106%) lead the pack, and Huawei will likely soon overtake the old technology behemoth IBM and land in the top-5. Unfortunately, this ranking does not do full justice to R&D in the world of Big Tech, with Apple and Tesla notably missing in the charts, due to their notoriously secretive stance on what’s cooking in their R&D labs. Another interesting missing actor in this chart is OpenAI, for reasons we can only speculate on. The same might be true for various military-oriented labs. For a brighter perspective on 2022, let’s hope that and NeurIPS rankings will continue to provide a fair and fun spectacle to watch, just like the Olympics.