For the second year in a row, my team and I have surveyed the global AI talent pool to try to piece together a broad view of its size and state. We found that while demand for top-tier AI talent has never been higher, AI specialists with strong educational profiles and experience are still relatively rare.
This should be no news to anyone: we knew from the outset that demand outstripped supply. This was supported by anecdotal accounts from recruiters, who report being unable to find skilled workers to fill AI job vacancies. Several recent surveys have also demonstrated the high demand for top AI talent, including big increases in openings seeking skills in deep learning. As the CEO of Element AI, an active recruiter in the field, I also know from our talent team what a challenge it can be to hire people with the right skills to build and optimize AI products that will help our clients transform their businesses.
What is really hard to get a sense of is: just how scarce is top-tier AI talent?
Overall we found that there were about 22,400 researchers publishing at leading conferences in the field, and of those about 4,000 had published research that had a major impact on the field as measured by citations received in the last two years. In parallel, we also took stock of how people were describing their expertise on LinkedIn and found a total of 36,524 people who were self reporting that they had the education, skills and experience to qualify as an AI expert. This was a 66% increase from our 2018 report. Pretty good, right?
National strategies are making an impact
We expected to see continued U.S. dominance in our survey, and we got it: in absolute numbers, the U.S. led across most metrics. Our research also gave us new insight into China’s momentum. The data showed that China was second only to the U.S. across most categories, including in high-impact research. Some smaller-population countries were also shown to be emerging as leaders in high-impact research in our survey. The United Kingdom, where the private and public sectors last year injected some $1.4 billion in AI research, had the third highest number of researchers publishing impactful work, followed by Australia and Canada. Australia had the highest percentage of its overall pool of conference researchers who contributed high-impact research.
Building an AI World, a 2018 report published by the Canadian Institute for Advanced Research (CIFAR), offers an overview of the major national AI strategies that have been introduced by 18 countries since 2017. Nearly all of the top ten countries that employ AI researchers, according to our research, as well as nearly all of the top ten countries producing the most high-impact research, were among this group of 18 countries. (One notable exception was the U.S., where an executive order on AI was recently signed in February 2019.)
I expect that countries will continue to make major investments in growing their talent, either through training and/or through attractive visa opportunities and work conditions that could lure foreign experts. This is because, in many cases, countries view AI as a leapfrogging opportunity that doesn’t need the same kind of physical infrastructure as other projects. In that sense, with relatively low barrier to entry I see the need for infrastructure flipping to a need for talent. The countries that are investing in growing and nurturing AI talent are likely to reap the benefits to come.
Improving access to AI in the classroom
There are plenty of incentives for students to study AI, and indeed enrollment is up: a recent report from the New York Times noted that from 2013 to 2017, undergraduate registration in computer science programs in the U.S was up more than 100%. Yet universities have not been able to provide enough spots in classes to meet the demand, and there are many reports of students forced to sit on long wait lists or draw from lotteries just to get access to intro-level computer science classes. Other universities squeeze more and more students into the same courses, so that the size of the classes balloons from what it was just a few years ago.
This seems to be because universities are having a hard time recruiting professors: while enrollment in computer science more than doubled over a four-year period, according to the New York Times, computer science faculty increased by just 17%. With tenured researchers and recent doctorates alike getting recruited into the private sector, many computer science departments report being unable to hire qualified tenure-track faculty. Yet it is in universities where the vast majority of machine learning talent is nurtured.
Of course, the definition of AI “talent” isn’t limited only to elite researchers: as a recent article in Harvard Business Review notes, companies need other types of AI talent as well, such as AI engineers to do product development, data experts and business leaders. These are all crucial roles; not everyone will be a fundamental researcher, and the field needs more than elite PhDs to make an impact. But this is still early days for AI, and for now at least the key talent we need and where we see the most demand — at least at my company Element AI — is in machine learning PhDs. I’ll be looking for more and more specialized programs around the world to add talent to meet the demand that I expect will increase even faster than it has so far. Even if we invest big in education and skills development, it could still probably take another decade to see the talent pool mature to the state we need. It’s on all of us to make that happen.
I hope you’ll explore our findings and dive into all the numbers, which are available in the full report. If you can help add more to this picture of the global talent pool, you can reach me with the contact form. Thank you for reading!