Thread by @r_zwetsloot

I’m very excited that, after more than half a year of work, we just published our new report: “Keeping Top AI Talent in the United States”

It asks (1) how many int'l AI grad students stay in the US?, and (2) how can the US increase that number?

Here’s what we found. [1/x]

First, we found these questions are actually pretty hard to answer. We did ~9 months of data collection, gathering 2,000 comprehensive career histories of recent US AI PhD grads (& more in the pipeline). We also analyzed four other sources w/ relevant data.

The results hold a lot of good news for the US: retention of top AI grad students is very high, with 90% staying right after graduating and more than 80% still there after five years (when our data coverage ends). [3/x]

None of the data sources we looked at showed any signs of recent decreases in retention rates (which was actually a surprise to us, since it runs counter to a lot of popular press coverage): [4/x]

We also find that retention rates are especially high for Chinese and Indian students, who together make up half of all international grad students in AI. [5/x]

These high retention rates are important, because a large majority (about 2/3) of grad students getting trained in AI-related programs at US universities are not American citizens: [6/x]

In fact, international students are responsible for basically *all* growth in AI-related grad programs at US universities in the past decades.

Or, in other words: there were as many American citizens getting CS/EE grad degrees in 2016 as there were in 1990. [7/x]

Out of the ~10-15% international students who leave, the majority go to US allies and partners and a little under 20% go to China. Some countries hire mostly into the private sector, while others get more academic talent. [8/x]

We find that the main reason for these high retention rates is how professionally attractive the US still is for AI talent relative to other countries.

But will that continue be the case going forward? On this question, our conclusions are a little less rosy for the US. [9/x]

First, numerical caps in the US immigration system have barely changed since the 90s, and backlogs have reached crazy levels: an Indian AI PhD applying for a green card today can expect to wait *50 years* before they actually get permanent residency

(👇via @David_J_Bier) [10/x]

@David_J_Bier Other countries like Canada and China are responding to the US’s immigration issues by becoming more welcoming and explicitly targeting US-based talent. Immigration policy is now an important element of strategic competition. [11/x]

@David_J_Bier At the same time that other countries are opening their doors, they’re also becoming more attractive places for AI professionals. For example, in the past few years the US has lost its virtual monopoly on AI VC activity.

(👇via axios.com/usshare-artifi…) [12/x]

@David_J_Bier These two trends happening in tandem — (1) US immigration issues getting worse while other countries’ immigration systems are getting better, and (2) other countries developing their domestic AI ecosystem — together pose a serious threat to US talent leadership [13/x]

@David_J_Bier In short, the US starts from a position of great strength when it comes to AI grad retention, but that is no reason to be complacent.

I’ll leave options for US policymakers until tomorrow, or you can read about them before then in Ch. 3 here: cset.georgetown.edu/reports/ [14/x]

@David_J_Bier In the meantime, big thanks to co-authors @tinahuang__ , James, & Zach for their awesomeness, and to @CSETGeorgetown for being the kind of place where we get time & resources for doing deep dives like this. Ending the year feeling very blessed with my team & job. [15/x]

@David_J_Bier @tinahuang__ @CSETGeorgetown Thanks also to (among many others) @calebwatney @JoyDantongMa @mattsheehan88 @jjding99 @EBKania @doug_rand @mchorowitz @ChhabraT @RoxanneHeston @flaggster73 @hlntnr for conversations etc. that helped inform the research. [16/x]

@David_J_Bier @tinahuang__ @CSETGeorgetown @calebwatney @JoyDantongMa @mattsheehan88 @jjding99 @EBKania @doug_rand @mchorowitz @ChhabraT @RoxanneHeston @flaggster73 @hlntnr /thread *phew* [17/17]