Global AI Talent Report 2018

 

 

 

Summary

For further commentary beyond the report, see the accompanying blog post. 
For a table of the full list of countries and their numbers, or to submit information about the talent pool in your region, send a message using the contact form.

The demand for AI experts has grown exponentially over the last few years. As companies increasingly adopt AI solutions for their businesses, the need for highly experienced, PhD-educated, and technically-adept talent shows no signs of stopping anytime soon.

This report summarizes our research into the scope and breadth of the worldwide AI talent pool. Although these data visualizations map the distribution of worldwide talent at the start of 2018, we want to acknowledge that this is a predominantly Western-centric model of AI expertise.

We are submitting our work amidst similar, though much broader, reports such as Tencent’s recent “2017 Global AI Talent White Paper,” which focused primarily on China in comparison to the United States. Tencent’s research found that currently “200,000 of the 300,000 active researchers and practitioners” are already employed in the industry, while some 100,000 are researching or studying in academia. Their number far exceeds the high-end of our measure at 22,000, primarily because it includes the entire technical teams and not just the specially-trained experts. Our report, however, focuses on finding out where the relatively small number of "AI experts" currently reside around the world.

We drew on two popular data sources for this line of inquiry. First, we used the results from several LinkedIn searches, which showed us the total number of profiles according to our own specialized parameters. Second, for an even more advanced subset, we captured the names of leading AI conference presenters who we consider to be influential experts in their respective fields of AI. Finally, we relied on other reports and anecdotes from the global community to put our numbers in greater context and see how the picture may develop in the near future.

Even though we relied primarily on English-language data sources, our view of the talent pool provides a good global representation of the best experts that the field currently has to offer. For this reason, the second half of the report focuses on a qualitative assessment of talent and funding in Asia and Africa, where the reliability of our numbers drops off significantly and does not match the industry or academic output of these hotspots.

According to our broadest LinkedIn measures, we have found that there are roughly 22,000 PhD-educated researchers in the entire world who are capable of working in AI research and applications, with only 3,074 candidates currently looking for work. In the smaller, more advanced subset, we have found that there are currently 5,400 AI experts in the world who are publishing and presenting at leading AI conferences across the globe and who are well-versed enough in the technology to work with teams taking to take it from research to application.  

 

 

 

How We Defined “Talent”

Building transformative AI applications for enterprises requires teams of people who have proven technical competency in Machine Learning/Deep Learning, several years of work experience, and can collaborate and thrive in an interdisciplinary environment.

The critical shortage of “talent” in the current AI job market suggests that there are currently not enough people with the strong grasp of academic research and applied software development required to mediate the worlds of business, science, and engineering.

The teams that need to be filled should be able to identify a problem that can be solved with modern machine learning techniques, build and implement that solution from scratch, and then optimizing the solution to work efficiently.

In our search, our hypothetical expert must be either highly talented or very experienced in order to capture the most elite leaders, seniors, and top juniors who would be able to work on such an effort. We used two different approaches to accurately size the pool of people in the world: LinkedIn searches and identifying participants in academic conferences.

LinkedIn

Using LinkedIn, we broke down these search criteria to capture a broad view of what it means to be an AI specialist.1 These search parameters were built to find candidates who were awarded a PhD no later than 2015, to account for several years of work experience.2 Although a PhD is not technically required to be considered an AI expert (since experience applying AI solutions in a real-world setting is more important than a degree), we’ve nonetheless found that having a PhD is a good proxy for assessing the technical ability of the talent pool across different nations.

To qualify for this subset, these profiles must also have mentioned "AI" or Artificial Intelligence in addition to one or more advanced concepts, such as deep learning, artificial neural networks, machine learning, computer vision, natural language processing, or robotics. These candidates must also be technically adept: we filtered our numbers to include only people who have a solid grasp of either Python, Tensorflow, or Theano, to make sure they have some experience developing real-world applications. Using these very broad parameters, we identified a total of 22,064 experts.

We also ran a more advanced subset that did not include the ("AI" OR "Artificial Intelligence") qualifier, omitted “python”, and included more specific AI-only frameworks.3 The idea behind this search was to capture candidates who listed very specific frameworks that we typically employ in our own work (these include torch, caffe, and nltk) and omit candidates who are using “AI” as a buzzword. In this search, we identified a number that comes very close to that of our conference presenter numbers: we found 6,138 experts using this search, with 1,735 indicating that they are available for work.

For our visualization, we decided to plot the less conservative estimate, in order to capture talent with potentially interdisciplinary skills.

Academic Conferences

In addition to these narrower LinkedIn searches, we counted authors of published papers or posters to estimate other high-level influencers and "rising stars" in the field. In theory, these candidates are required to apply AI theories established in controlled environments to messier real-world settings. In this talent pool, we found 5,400 experts who have presented a research paper in the last few years.

The following conferences were prioritized in our research: the Conference on Neural Information Processing Systems (NIPS), the International Machine Learning Society (ICML) conference, and finally the International Conference on Learning Representations (ICLR). We scraped researcher names from these conferences and filled out their location, experience, and education profiles using Mechanical Turk.

Dataset Biases

According to our data, European and Asian countries have significantly fewer researchers than the US, the UK, or Canada, but we are the first to acknowledge that this is most likely due to LinkedIn being a predominantly Western platform. Our searches found 413 candidates in China, 291 in Singapore, 204 in Japan, and 147 in Korea.

A of LinkedIn users by country done by Meenakshi Chaudhary points out a large discrepancy in LinkedIn user penetration rates, even among developed countries. Chaudhary mentions that “after [the] US, India, Brazil, Great Britain, and Canada have the highest number of LinkedIn users,” which suggests that LinkedIn’s adoption in certain countries and markets heavily skews the representations within our sample. To that effect, while the quantities of LinkedIn experts found in Asia are much lower than in North America or Europe, these numbers are still very high given the fact that LinkedIn’s penetration rates are lower in Asia.

The same goes for the careful examination of presentations at academic conferences. By limiting our search to several English-speaking conferences in the Western world, we risk missing other institutions where AI research and development is done: research centres, private labs, think tanks, smaller universities and institutes, independent researchers and consultants. These people, although experts and domain-leaders, might not be engaging with the global community when they are working at a smaller scale or privately.

  1. (“Deep learning” OR “artificial neural networks” OR “machine learning” OR “neural networks” OR “speech recognition” OR “computer vision” OR “image processing” OR “natural language processing” OR “natural language understanding” OR “robotics”) AND (“python” OR “tensorflow” OR “Keras” OR “theano”) AND ("AI" OR "Artificial intelligence")

  2. With no adjustment for the 2015 graduation date, the eligible pool of candidates grew to some 90,000 people, which begins to match Tencent’s numbers, which also showed a large proportion of self-identified talent to be lacking real experience or deep know-how.

  3. (“Deep learning” OR “artificial neural networks” OR “machine learning” OR “neural networks” OR “speech recognition” OR “computer vision” OR “image processing” OR “natural language processing” OR “natural language understanding” OR “robotics”) AND (“tensorflow” OR “Keras” OR “theano” OR “pytorch” OR “torch” OR “caffe”)

AI Talent Hotspots Across the Globe

North America

Out of our 22,000 LinkedIn profiles, almost half of all candidates (9,010) are living and working in the United States. Most of the LinkedIn experts listed their field of study as either Computer Science (12,856) or Computer Engineering (3,879) –– less common fields of study included Mathematics (2,592), Physics (2,157), and IT (1,175). A substantial portion of these experts have worked, at some point, for either Google (756), Microsoft (357), or IBM (265), and have anywhere between three and 10 years of experience working.

The dominance of the U.S. in the AI talent markets is not at all surprising. Paysa, in a recent study of artificial intelligence talent, found that nearly $650 million is slated to be spent in the United States on annual AI-related salaries alone, with several U.S. companies, having raised an additional $1 billion to fund AI development, making it hard for smaller countries to compete with the U.S.

Nonetheless, Canada came in third place for the number of researchers in our LinkedIn and conference presenter searches, making it a viable competitor to the U.S., with 1,154 high-level profiles, which is high given Canada’s small population and GDP. The Canadian AI talent pool has been refilling with former students and new international researchers alike, with leading the charge (Facebook, Google, Uber, Samsung, DeepMind have all set up labs there, among others) and Toronto, Edmonton and Vancouver close in tow.

 

Europe