欧洲的人工智能劳动力来自哪里？

执行摘要  

全球对人工智能 (AI) 人才的竞争日益激烈。各司法管辖区和企业都意识到，拥有人力资源来开发、实施和控制这项我们时代在经济、社会和政治上最具变革性的技术之一至关重要。从美国到欧盟，再到中国，各国都在努力增强其人工智能人才库。例如，今年 10 月，总统乔·拜登签署了一项行政命令，旨在放宽移民规定，允许更多人工智能专家在美国学习和工作。 ¹ 同样，中国计划在北京和上海建立人工智能学院来吸引人才 ² 谷歌、微软等公司已表示支持这一举措。与此同时，微软、Aleph Alpha、腾讯等企业也在争夺顶尖人工智能人才，以保持竞争优势。  

优先发展人工智能专业知识也已成为欧洲政策制定者近期声明中反复出现的主题。刚刚连任的欧盟委员会主席乌尔苏拉·冯德莱恩倡导齐心协力应对劳动力市场挑战，并强调技能和劳动力短缺等关键问题。冯德莱恩为欧盟委员会制定的政治指导方针旨在通过提供必要的基础设施和公私合作伙伴关系来支持研究人员，从而培养欧洲的人工智能人才。欧盟委员会将通过加强学术界和产业界的合作来吸引和留住顶尖人才，并将提出一项战略计划，以改善STEM教育并提高女性在这些领域的参与度。 ³ 近日，法国总统提出了“通过吸引顶尖人才，将法国打造成人工智能强国”的指导方针。 ⁴ 无论从哪个角度看，政策制定者都在做出切实努力，扩大国内人工智能人才库，吸引和留住人工智能专家。

尽管全球对人工智能人才的竞争愈演愈烈，欧洲政策制定者和相关研究人员仍然缺乏该地区所需专家库的关键信息。这使得他们制定适当政策的任务变得更加困难，甚至有可能忽视劳动力市场的实际动态。因此，本数据简报旨在作为了解欧洲及其他地区人工智能人才格局的资源。其中，我们回答了以下问题：

• 欧洲的人工智能人才库来自哪里？

• 欧洲的人才流失到了哪些国家？

• 各国人工智能人才库中国际人才占比有多少？

• AI人才的教育背景是怎样的？

在制定有效的政策来培养、吸引和留住欧洲的人工智能人才之前，我们需要回答这些问题以及其他问题。分析一个地区的人工智能人才时，可以采用两种不同的方法。第一种方法侧重于吸引人工智能专业人士到该地区，而第二种方法则强调培养和发展本地人才的重要性。这两种方法的总体目标都是留住这些人才，从而增强该地区的竞争优势。本数据简报重点关注第一种方法，旨在深入探讨欧洲吸引的人工智能人才在地理和教育背景的多样性。

本报告揭示了欧洲人工智能人才格局的几个重要见解。欧洲国家正在流失大量国内外人工智能人才，这些人才流向美国。印度正在成为欧洲人工智能人才的主要来源地，爱尔兰和英国等国的许多人工智能专业人士都在印度获得了本科学位。与南欧和东欧相比，北欧和西欧国家的国际人工智能人才比例更高，卢森堡和瑞士尤其以其多元化的人才库而闻名。欧洲人工智能专业人士往往比其他地区的专业人士拥有更高的学位，其中相当一部分至少拥有硕士学位。这一趋势与美国和印度形成了鲜明对比，这两个国家更多的人工智能专业人士只拥有学士学位。

虽然我们的数据可以让我们对这些新兴趋势背后的原因进行一些推测，但我们的目标是通过定性研究来跟进本报告，以了解这些模式背后的原因。

作者衷心感谢数据科学家 Laurenz Hemmen，他严谨的数据分析和统计洞见巩固了本研究的基础。本文是“德国和欧洲人工智能生态系统的优势与劣势——以人才为重点”项目的一部分。该项目由卡尔蔡司基金会资助。 

---

介绍  

The rapid advancements in Artificial Intelligence (AI)—evident in technological breakthroughs, widespread adoption, and significant impacts on daily life—mirror the unprecedented pace of technological progress and serve as a vital metric for assessing a country or region's innovative capabilities, economic strength, and geopolitical influence. The European Union (EU), with its access to a highly educated AI talent pool and its regulatory leadership, is strategically positioned to lead in responsible AI innovation, especially following the recent passage of the EU AI Act. Experts agree that the success of AI innovation heavily relies on the availability of skilled talent. Europe possesses diverse AI talent from various geographical and educational backgrounds, uniquely positioning the region ⁵ to excel in AI innovation and regulation. At the same time, the global demand for AI talent has increased fivefold since 2015 ⁶ , intensifying competition for well-trained AI experts worldwide. Consequently, prioritizing AI expertise has become a recurring theme in European policymakers' recent statements.

Of course, Europe is not starting from scratch when it comes to AI talent. It boasts a per-capita concentration of AI experts that surpasses that of the United States by 30% and nearly triples that of China ⁷ . This impressive statistic underscores Europe's pivotal contribution to the AI domain, highlighting the substantial pool of talent already present within its borders. And yet, even if Europe has a lot to show in terms of AI talent, the narrative does not end here. Two critical challenges emerge that require strategic attention to strengthen Europe’s capabilities in AI innovation. The first challenge is the escalating global demand for AI talent. This demand has intensified, yet the talent pool remains concentrated in specific cities in the United States and India, with very few European cities emerging as AI talent hotspots. This surge in demand underscores a pressing need for Europe to preserve its existing pool of AI experts and expand it significantly. The European nations must be prepared to engage in fierce competition, adopting strategies that are both attractive to potential talent abroad and conducive to nurturing homegrown expertise.

 Simultaneously, the second challenge revolves around the depth of understanding among policymakers of Europe’s own AI talent ecosystem. This is a rapidly evolving sector, characterized by a constant influx of new talent and high migration rates. Mapping migration within EU countries is particularly challenging due to variations in labour market measurements across member states. Furthermore, AI talent is often subsumed under the broader category of STEM (Science, Technology, Engineering, and Mathematics)  talent, rather than being recognized as a distinct and specialized field. Despite having a substantial talent pool, a noticeable knowledge gap exists concerning the dynamics of talent flow, development, and retention within the region. This gap makes it more challenging to formulate and implement effective policies to enhance the attractiveness of Europe as a destination for AI professionals and ensure the retention of homegrown talent. Or, to put it another way: If one wants AI specialists to come or to stay, we need to know what the factors are that determine their careers in the AI sector. To get to the bottom of these factors, we first need to collect, understand, and classify the relevant data.

Two distinct perspectives can be employed when analyzing the AI talent in a region. The first approach centers on attracting AI professionals to the area, while the second stresses the importance of nurturing and developing local talent. The overarching goal of both approaches is to retain such talent, thereby enhancing the region's competitive edge. This data brief focuses on the first strategy, aiming to explore in depth the diversity in geographical and educational backgrounds of the AI talent attracted to Europe. It builds on our previous analysis, which examined empirical evidence showcasing talent flows into and out of top AI PhD programs in Germany. 

Our analysis offers insights into AI talent flows, their geographical diversity, and educational backgrounds. The following sections will present the data we have gathered and analyzed through interactive visualizations. We address questions such as:

• Where is AI talent migrating from? What proportion of the talent consists of skilled foreign nationals?

• How much of the talent comes from outside the EU?

• What percentage of the countries’ AI talent pool is international talent?

• What is the highest degree of education attained by AI talent in various countries?

Determining the exact causes of these talent flows from quantitative data alone is challenging; our report lays a strong groundwork for further exploration. Our insights are potentially helpful for various purposes, whether for research, gaining a deeper understanding of industry opportunities, or developing and implementing evidence-based strategies and policies. They build a strong baseline understanding of the AI landscape in Europe for stakeholders influencing AI policy directly or indirectly. This data brief concludes with open questions and aspects that have emerged from the data analysis, which provide starting points for further research projects.

Dataset 

This data brief maps the landscape of AI Talent, focusing on geographical and educational backgrounds in the EU and beyond. The data and preliminary analysis for this report has been provided by LinkedIn.

LinkedIn used its own definition of AI talent based on a combination of skills and job roles one reports on their LinkedIn profiles. It classifies members with at least 2 AI skills on their profile and/or in an AI occupation as AI talent. Charts 1-4 of this data brief are based on LinkedIn data, where we explore the geographical origin of AI talent based on where they attained their undergraduate degrees as well as the highest degree attained, from sub-bachelor's to Ph.D.

To explore the exact ways we have analyzed this data, please look at the detailed methodology section at the end of this data brief.

Findings 

Chart 1: AI Talent Migration 

What you see:  

These charts show migration rates based on the locations people shared in their LinkedIn profiles. The right bar shows where AI talent is immigrating from, and the left bar shows where AI talent is emigrating to. In this chart, the countries listed on the y-axis represent the previous working locations of individuals, not their nationalities. This is a key distinction between this chart and the following one (Chart 2). In this chart, we focus exclusively on previous working locations, whereas Chart 2 examines the origin based on the locations where individuals obtained their undergraduate degrees.

The data covers AI talent movement from January 1st, 2024 to June 5th, 2024. The immigration and emigration rates are calculated per 10,000 LinkedIn members in each country. For example, relative immigration from India to Germany of 18 means that for every 10,000 LinkedIn members in Germany, 18 AI professionals moved there from India during this period. This method allows for easier comparison between countries with different population sizes.

For each country, we display the top ten countries by migration flow, which sums the absolute numbers of inflow and outflow. On top of each graph, you can see the net flow of AI talent in each country.

What it means:

This chart highlights the dynamic movement of AI talent among major players in the AI landscape, with transitions occurring at various career stages. Switzerland and Germany have the highest immigration of AI talent from other countries. Only France, among the six countries studied, is experiencing a brain drain with more emigration than immigration.

Switzerland emerges as a significant attractor of international AI talent, predominantly from European countries. Notably, it draws substantial talent from France, Germany, and Italy, which is most likely related to the official languages of Switzerland as well as to geographical proximity.

Germany is experiencing a notable outflow of AI talent to the UK, Switzerland, and especially the US.

There is a lot of movement of AI talent from India to Germany, the UK, Ireland, and the US. To explore this movement further, we’ve added an additional chart in the appendix that looks at the distribution of Indian AI talent in Europe. However, this trend is not mirrored in Switzerland and France, potentially due to language barriers and professional networks that are less accessible to Indian talent. Conversely, India experiences a significant inflow of AI professionals from Germany, the UK, Ireland, and the US. This reverse migration is likely driven by Indian professionals returning after completing their studies or switching jobs, motivated by India's booming tech sector and strong cultural and familial ties. Since we are looking at immigration and emigration based on locations shared on LinkedIn, it is difficult to distinguish between Indian nationals returning from other countries and foreign nationals moving to India for work. This is an important distinction to keep in mind while analyzing this chart. 

The United States stands out as a primary destination for AI talent from Europe, probably attracted by its leading tech companies, renowned universities, and extensive research facilities. The UK, in particular, sees a substantial outflow of its AI talent to the US.

Pakistan, while a significant supplier of AI talent to the US, the UK, and Germany, struggles to attract much talent itself, at least from the countries we have studied.

Chart 2: Origin of AI Talent by Country 

(Based on the location of their Undergraduate Degree) 

Note: This interactive chart allows you to compare data for 50 countries, both within the European Union and globally, as well as aggregated data for the EU.

What you see:  

This chart offers insights into the origins of the AI workforce across various countries, based on the assumption that the location of their undergraduate degrees represents their country of origin. This assumption is generally reliable, as most individuals pursue early education in their home country. OECD data supports this, indicating that only 8% of bachelor's degree students in OECD countries are international. However, notable exceptions exist, which are addressed in detail in this paper's methodology section. Compared to Chart 1, where we looked at migration in the workforce in 2024, the focus here is on the educational background.

What it means:

The dependence on international talent within the AI sector shows remarkable variation across different countries, despite the largest group of AI talent in most analyzed countries being “homegrown”. For example, nearly half of Germany's AI workforce attained their undergraduate degrees from foreign institutions, highlighting a significant reliance on international education, and the ability to attract foreign talent. In contrast, France benefits from a more domestically oriented talent pool, with nearly 70% of its AI professionals having completed their undergraduate studies within the country.

The migration trends of AI talent do not mirror broader demographic movements in countries like Germany. In 2021, 1.3 million people moved to the country, mostly from other European Union countries, with Romania, Poland, and Bulgaria being the top sources. However, these countries are not significant sources of AI talent. When looking only at labour migration, a pattern similar to the one in our chart emerges. Germany's labour and migration policies have long focused on attracting academic professionals, exemplified by the Blue CardEU introduced in 2012 for highly skilled non-EU workers. By the end of 2023, the country had issued 113,000 Blue CardsEU, a 26% increase from the previous year. The majority of Blue CardEU holders were from India (33,000), Russia (10,000), and Turkey (8,000). To qualify, individuals need a university degree and a job offer with a minimum salary, ⁸ whereas EU nationals can migrate with minimum administrative hurdles.

In France, on the other hand, the AI workforce significantly consists of individuals from French-speaking former colonies such as Algeria, Morocco, and Lebanon. The direct comparison of undergraduate degrees as an indicator of origin might be less effective here due to the closely linked higher education systems between these countries and France, characterized by unique educational pathways like "prepas" and "Grande Ecoles" that are uncommon elsewhere. This streamlined acceptance of their qualifications might attract more international talent from these countries to France at the undergraduate level.  By highlighting these examples, we demonstrate that the migration of AI talent cannot be simplified to general migration patterns. Various factors, including education systems and language, significantly impact migration trends. 

A noteworthy aspect is the significant contribution of Indian-educated professionals to the AI talent pool in many European countries. Ireland stands out in this context, with 28% of its AI workforce having completed their undergraduate studies in India. Ireland's appeal could be further magnified by its status as the biggest country with English as one of its official state language in the EU post-Brexit, coupled with a thriving tech sector, with companies such as OpenAI opening their first EU office in Dublin last year. Similarly, 14% of the United Kingdom’s AI talent also comes from India.

Once again, language seems to play an important role in attracting AI talent as native English-speaking countries like Ireland, the United Kingdom, and the United States attract the largest share of Indian talent. Similarly, Switzerland (with its official languages including German, French and Italian) attracts foreign AI talent from Germany, France and Italy. It would be interesting to explore what attracts so many Indians to Germany and, in general, the factors other than common language that attract AI talent to some countries or cities.

Chart 3: International Representation in AI Workforces  

What you see:  

This chart shows the proportion of AI professionals in various countries who received their undergraduate education abroad, suggesting the extent of international talent within each country's AI workforce, as we assume that most people get their undergraduate degree in their country of origin. The (EU) average is calculated by taking into account the population size of each member country, ensuring that larger countries have a proportionally greater impact on the average. However, this calculation excludes Malta as the LinkedIn data did not include this country for data quality standards. Therefore, the EU average is derived using data from all other member states.

What it means:

This chart shows the makeup of AI workforces in different countries, revealing insights into international talent migration and its impact on the AI sector. It highlights how countries vary in their reliance on international talent, influenced by education systems, immigration policies, and global demand for AI skills.

Countries in Northern and Western Europe display a higher proportion of international AI talent compared to Southern Europe. These differences may reflect varying economic opportunities, sectoral demands, and the overall appeal of different regions as destinations for international professionals.
Luxembourg, within the EU, and Switzerland, outside the EU, stand out for their diverse talent pools. Luxembourg relies almost entirely on talent with foreign degrees. Over 47% of its entire population of 660.000 are foreign nationals ⁹ . For a small country like Luxembourg, it is also likely that a large share of nationals get their degree internationally and return later. Italy, Israel and India rely almost exclusively on their national talent for their AI workforce. Population sizes could be playing an important role in the trends in these statistics.

Liberal immigration policies and attractive work environments, such as in Ireland and Switzerland, might contribute to successfully drawing a significant share of their AI workforce from abroad. This success might be linked to policies that facilitate the integration of international professionals, including visa arrangements and work permits that are favorable to highly skilled workers. Other factors could include the presence of companies that hire international talent, and well-paid job opportunities.

In total, the AI talent in the EU is slightly more international than in the United States. Looking at Europe as a region, including the United Kingdom and Switzerland, the observation is even stronger. The five countries with the highest share of foreign talent are all relatively small countries that either have one or more highly-ranked universities and/or are home to big AI and tech companies. Estonia stands out for its efforts to attract talent, potentially through its e-residencies, digital nomad visas, and double taxation agreements with over 60 countries, including France, Germany, Spain, and the UK, ensuring e-residents are not taxed twice on the same income. Approximately 21% of Estonian adults hold a master’s degree. Over 78% speak at least one foreign language, and 35% speak at least two, with English being the most commonly spoken foreign language. Additionally, Estonia is highly digitally literate. IT is a compulsory subject in schools, with children starting to learn programming at age seven. According to Haridussilm, a government body reporting on education statistics, 28% of Estonian graduates study STEM subjects ¹⁰ .

Comparing the share of foreign talent in AI to the share of foreign talent in the general workforce, our data (not mentioned in the charts) indicates that AI talent is much more international. For instance, in Germany in June 2023, around 15% of all employees were foreign nationals ¹¹ , whereas our data shows that 45% of AI talent had a foreign undergraduate degree. We acknowledge that the assumption we are making in this paper (the country where someone earns their undergraduate degree is their country of origin), does not always hold true. Despite that, the large disparity between these numbers—15% versus 45%—is too significant to be attributed solely to this difference. This suggests that AI talent in Germany is indeed much more international.

Similarly, in the UK in 2022, the share of foreign-born workers has been 19% ¹² , whereas 47% of the AI talent has foreign undergraduate degrees. The EU-27 average in 2020 was around 13% foreign-born employees overall ¹³ , significantly lower than our number of 37% of AI talent with foreign undergraduate degrees. This substantial gap further supports the trend that AI talent is significantly more international compared to the general workforce in Europe. 

 The chart could further hint at the role of educational systems in attracting international students who later join the local AI workforce. For instance, the high share of international talent in the AI sectors of countries with prestigious universities and research institutions might reflect the effectiveness of these institutions in attracting and retaining global talent. We investigate this further in Chart 5. 

 These observations raise important questions for further research: How do national policies, industry needs, and educational systems converge to influence the composition of AI talent within a country? What lessons can be learned from countries like Ireland in attracting international talent, and how can other nations adapt these strategies to bolster their own AI sectors? Additionally, exploring the long-term impacts of such diverse talent compositions on innovation, economic growth, and technological leadership in the AI domain would provide valuable insights into the strategic development of the global AI industry. 

Chart 4: Highest degree held by AI talent in different countries 

What you see:  

This chart shows the highest academic degree attained by AI professionals across different countries, with significant variations in the level of education. Members without self-reported degrees are excluded. Sub-bachelor’s degrees include high-school diplomas, apprenticeship degrees, and associate’s degrees exclusively. The countries are sorted by the share of AI talent whose highest degree is a Bachelor’s, from highest to lowest.

What it means:

The composition of AI workforces in terms of educational qualifications reveals significant variations across countries, shedding light on the diverse pathways into the AI sector and the potential implications for innovation, industry needs, and policymaking.

AI talent in Europe, on average, holds more advanced degrees with more people having at least a Master's degree, compared to other regions. Other reports confirm this, as a report by Sequoia states that seven in ten AI talent in Europe have a master’s degree or PhD and slightly more have over a decade of experience—well above the figures for engineers as a whole. ¹⁴

There’s an interesting trend in our data where countries such as Israel, Japan, South Korea, and the United States simultaneously have above average undergraduate and PhD holding AI talent. This means that relatively few people in the AI workforce hold a Master's degree. This could be because, in the US, for example, people often start PhDs directly after their undergraduate degree and don’t attend a separate Master’s program.

Other countries, like India, have a majority of their AI workforce with only a Bachelor's degree. The United States and Singapore also have a significant share of their AI workforce with just bachelor's degrees. This significant reliance on Bachelor's degrees could hint at a more practical orientation within the AI sectors of these nations, valuing on-the-job learning and applied skills over advanced academic credentials. Such a trend underscores the importance of practical experience and possibly reflects a cultural inclination towards "learning by doing."

This observation raises an interesting point about the duration and nature of undergraduate education across different countries. For example, engineering and certain other undergraduate programs typically span four years in India and the United States, which is longer than the duration of similar programs in some other countries. This extended period of undergraduate education might not just provide a depth of academic knowledge but also allow for internships, projects, and other practical learning opportunities that are highly relevant to the AI industry. Therefore, it's plausible that the sector's preference for Bachelor's degrees in these countries is not solely a matter of academic level but also relates to the comprehensive nature and length of the educational programs.

Contrastingly, European talent’s academic qualifications highlight the emphasis on advanced research capabilities. France presents an intriguing case with the lowest share of Bachelor's degrees and one of the highest of sub-bachelor's qualifications among the countries in the chart. The French AI talent pipeline is distinctively shaped by the Grande Ecoles system, offering education equivalent to a Master's degree without a preceding Bachelor's degree. This system is preceded by preparatory classes (prepas), potentially recognized as sub-Bachelor's degrees, illustrating an alternative, highly specialized pathway into AI careers.

Poland emerges as another interesting case where most of the AI talent has a master's degree (74%), more than any other country in our data. This warrants further exploration of AI education in the country as some countries offer subject degrees only at master's levels.

It is important to keep in mind that some differences could be attributed to different norms of self-reporting degrees or special kinds of degrees that might not be matched by LinkedIn. These countries' educational systems and job market demands might influence these trends.

These findings underscore the diversity of educational backgrounds in the AI workforce, reflecting not just national educational philosophies but also the evolving demands of the AI industry. They suggest that there is no one-size-fits-all approach to preparing individuals for careers in AI, with different countries leveraging their unique educational strengths to meet the needs of this dynamic sector.

 The data invites further exploration into how different educational systems and policies influence the development of AI talent. For instance, examining the impact of Ph.D.-level education on innovation within the AI sector or how bachelor-focused pathways support the rapid expansion of the AI workforce can provide valuable insights. Additionally, understanding the role of cultural and systemic factors in shaping these educational trajectories offers crucial lessons for policymakers and educators worldwide aiming to cultivate a robust and innovative AI talent pool with specialized pathway into AI careers.

Conclusion  

After analyzing the inflow and outflow of AI talent in Europe, and examining the educational backgrounds, and geographical origins, one thing is clear: attracting and retaining this highly sought-after talent is a global challenge. In this tug-of-war of AI talent, countries are looking to hire talent from abroad, as it is faster than training new talent.

 In light of this fast-moving competition, this data brief serves as a resource for European policymakers and academics, providing them with valuable insights to develop effective strategies for attracting, retaining, and optimizing AI talent, thereby contributing to the EU's position as a leader in the global AI landscape. Various critical insights emerge from this data brief:

• European countries are experiencing a significant loss of both national AI talent and the international talent they nurture to the United States. This trend underscores the need for Europe to implement stronger retention strategies to keep its AI professionals within the continent.

• India is rapidly emerging as a primary source of AI talent for European countries and globally. A substantial number of AI professionals in Europe have received their undergraduate degrees in India, particularly in countries like Ireland and the United Kingdom. This trend highlights the role of global talent pipelines in bolstering the AI workforce in Europe.

• Countries in Northern and Western Europe display a higher proportion of international AI talent compared to Southern and Eastern Europe. Luxembourg and Switzerland stand out for their diverse talent pools, with Luxembourg relying almost entirely on talent with foreign degrees.

• On average, the AI talent in European countries hold more advanced degrees than in other world regions, with a significant proportion holding at least a Master's degree. This contrasts with other regions like the United States and India, where a larger share of AI professionals hold only a Bachelor's degree.
  

While these findings provide valuable insights, it is important to recognize that the data does not establish causal relationships. However, it suggests several factors that influence the attractiveness of regions for AI talent:

• Strong Educational Institutions: Having renowned educational institutions with robust research facilities attracts significant international talent. This is evident in the case of Cambridge, which has become a hotspot for AI talent despite its small size.

• Language: Language is a critical factor. Countries where English is an official language or widely spoken, such as the United States and the United Kingdom, attract a considerable amount of global talent.

• Tech Ecosystems: Strong tech ecosystems, characterized by a concentration of tech companies, startups, and research institutions, also draw international talent. The presence of a vibrant tech industry creates an environment conducive to innovation and professional growth.

• Cultural and Political Histories: Cultural and political histories play a role in talent attraction. For instance, France continues to attract a significant amount of AI talent from countries that gained their independence from France during the last century, likely due to similar education systems and shared language.
  

Building on the insights from this data brief, our future research will focus on understanding the factors that make certain cities and countries hotspots for AI talent. We will not only explore the attraction of AI talent but also look into retention strategies for sustained innovation in Europe.

We hope that this data brief will guide the development of targeted, well-informed policy frameworks addressing the evolving needs of the AI sector and enable stakeholders to make strategic choices that bolster the growth and sustainability of Europe's AI landscape.

Methodology  

This section details the methodology of how we have used the LinkedIn dataset (May 2024).

Definition of AI talent:

A LinkedIn member is considered AI talent if they have explicitly added at least two AI skills to their profile and/or are employed in an AI job.

Members self-report skills on LinkedIn. LinkedIn tracks more than 41,000 distinct, standardized skills, which are coded and classified into 250 skill groupings by expert taxonomists. The top skills that make up the AI skill grouping are machine learning, natural language processing, data structures, artificial intelligence, computer vision, image processing, deep learning, TensorFlow, Pandas (software), and OpenCV, among others. Any person with these skills is considered AI talent.

Alternatively, people can be classified as AI talent based on their job titles. LinkedIn standardizes members' titles into around 15,000 occupations. These occupations are further grouped into about 3,600 categories called "occupation representatives," which serve as umbrella terms for roles with a common specialty, regardless of seniority.

An "AI" job, or AI occupation representative in LinkedIn’s terms, is a category that requires high competency in AI skills. LinkedIn determines whether AI skills are common in a specific occupation representative through skills penetration. Examples of such occupations include (but are not limited to): Machine Learning Engineer, Artificial Intelligence Specialist, Data Scientist, Computer Vision Engineer, etc.

Country Selection:  

For Chart 1, LinkedIn provided data on the six countries that are shown. In Chart 3 and 4 we chose a subset of countries that were most relevant.

LinkedIn’s data on migration based on undergraduate degrees for Charts 2 and 3 includes all EU countries except Bulgaria and Slovakia, and all additional OECD countries except Colombia. We also have data on Argentina, Brazil, Costa Rica, India, Indonesia, Saudi Arabia, Singapore, South Africa, the United Arab Emirates, and Uruguay.

For Chart 4, LinkedIn data does not include Malta but has data for Bulgaria and Slovakia.

LinkedIn applies minimum thresholds for labour force coverage, total membership size, and the number of monthly AI hires to exclude countries where the data may not be representative.

General Limitations  

Using Undergraduate Degrees as Origin Proxies

The country of an individual’s undergraduate degree serves as a proxy for their origin, based on the assumption that most people pursue their early education in their country of origin. This method is generally reliable; for instance, OECD data ¹⁵ cases. Nonetheless, we acknowledge outliers like Australia and New Zealand, where the share of international undergraduates peaks at 28%, which could be due to English being their official language. Further evidence supports this hypothesis: the UK (18%) and Austria (17%) have high percentages of international undergraduates, contrasting with lower percentages in Southern European countries like Italy (4%) and Spain (2%). A case that challenges this hypothesis is the United States, where, despite the English language being the medium of education, only 4% of students at the undergraduate level are international students. For detailed statistics, please refer to the OECD report "Education at a Glance 2020." ¹⁶ Not all countries of origin in our dataset are OECD countries. In our analysis, we verified a low percentage of international students from the most relevant countries. In India, for example, the percentage of international students at the undergraduate level is less than 0.5%. ¹⁷

When interpreting our findings, we take into account the variances in international student percentages across different countries. These variances can influence the perceived origin of AI talent. For example, our analysis may reveal that a percentage of AI professionals working in Ireland received their degrees in the UK. Given the UK's high proportion of international students (18%), a fraction of this group might actually originate from other countries. We believe that these effects do not substantially alter the primary insights regarding key source countries for AI talent.

Self-Reporting

LinkedIn 数据主要依赖于用户自述信息。这种方法有两个固有的局限性：个人可能不存在于特定的职业网络中，或者可能使用该网络但提供的信息不完整或不准确。例如，在图 2 中，我们仅显示已提供学士学位的 LinkedIn 会员。这些局限性可能会引入选择效应，从而扭曲我们的结果。即使我们选择覆盖范围广泛的国家/地区，拥有个人资料的可能性仍会因原籍国或最高学位而异。例如，如果在法国从事人工智能工作的中国专业人士创建个人资料的可能性低于在法国工作的法国公民，那么他们可能被低估。

此外，不同背景的人在技能自我报告方面可能存在差异，这可能会影响AI人才的识别。尽管存在这些限制，但领英（LinkedIn）在全球拥有超过10亿会员，提供了最全面的全球职业发展路径数据集，可用于分析欧洲及其他地区的AI人才流动。

鉴于本文主要考察的是全球劳动力中技术水平更高的群体，这一局限性在本文的背景下得到了进一步缓解。这些人通常具备数字素养和必要的设备，能够在专业求职网站上创建和维护个人资料。他们通常受益于自我报告技能。

本文是“德国及欧洲人工智能生态系统的优势与劣势——以人才为重点”项目的一部分。该项目由卡尔蔡司基金会资助。