($) Tracking Global AI Talent Flow
Is sovereign AI keeping more talents at home?
Back in January, I published a rather provocative post, “Does China Want Generative AI?”, where I dove deep into China’s lukewarm attitude towards generative AI (at least so far), and how this attitude is causing a brain drain of AI talent. In the post, I cited a 2019 analysis done by the think tank, MacroPolo, which mapped the AI talent flow from countries to countries.
After it was published, I soon received a ping from the MacroPolo team telling me that they are working on an updated version of that analysis. Since then, I’ve been eagerly awaiting its arrival. This week, they finally released the fresh new analysis, and it does not disappoint.
By sampling and tracking the authors of accepted papers to the December 2022 NeurIPS conference – the most prestigious AI academic conference, where anyone who is anyone in AI would go to scout out talent to hire – MacroPolo produced a chock full of charts and graphics, like this wonderful data visualization mapping the talent flow from undergraduate, to graduate, to post-graduate work locations of AI talent everywhere.
There are sights you can glean from just staring at this video for a few seconds. The MacroPolo team also shared some topline “key takeaways” that I highly recommend. One obvious observation that jumps off the page is that China appears to be producing the most AI talent at the undergraduate level (~47%) of any country, up from its proportion in 2019 (~29%). Another obvious one along the same thread is that a bit more than half of these Chinese undergraduate talent (~51%) stay in China to do graduate level work, while a good number of them (~36%) advance their study in the US, and the vast majority of that cohort stay in the US to work.
It’s worth scrutinizing how an “AI talent” is defined. The MarcoPolo team is transparent and honest with their definition and its limitations, so I encourage any reader to examine the footnotes and methodology section before drawing any sweeping conclusions.
With all this in mind, I will do a deeper dive into a few additional insights that jumped out at me, particularly as they relate to the new buzzy term of the day, sovereign AI, plus some recommendations of how this type of talent tracking analysis, a valuable exercise, could be expanded.