Abstract
The Artificial Intelligence research works and patented developments in the last five years are shaping the future, having direct impact on the global economy and society. This paper provides a comprehensive literacy analysis and metrics sourced from the Lens.org platform. We focused the analysis on the cur-rent state of AI innovations, captured from research papers and patents. Since1978, the total number of registered research production is 1,262,789 pa-pers (198,163 from 2019 to 2024), 2023being the most productive year (74,426 articles). The analysis underscores Asia, notably China (CH), as the primary hub for artificial intelligence research. Institutional affiliations reside in Asia, with the Chinese Academy of Sciences leading with 5,512 articles. Wei Wang emerg-es as a prominent figure and the thematic dominate cluster is Computer Vision. On the other hand, in the last five years of production there were 316,883 patents. US leads the applications, followed by CH. Samsung Electronics Co. Ltd. is the first in the east and IBM leads the west region. The top inventors are found at Qualcomm Inc. and LG Electronics. The main cluster is Computing arrangement based on specific computational models, such as machine learning, learning methods, combinations of networks, backpropagation, recurrent networks, prob-abilistic graphical models and inference or reasoning models. This analysis sheds light to researchers, policymakers, investors, and inventors on the current land-scape of AI research and development, showing as its main insight that there is a wide gap between industry and academy.
Over more than 65 years, scientists and engineers have been continuously leveraging the study of artificial intelligence (AI). Machines created by humans can do more than just labor-intensive work; they can also develop human-like intelligence. Whether we are aware of it or not, AI has penetrated our daily lives, playing novel roles in industry, healthcare, transportation, education, and many other fields, thus getting closer to all of us [1].
Artificial intelligence has emerged as a pivotal field, driving innovation across various domains. The dispersion of information not only does not facilitate the analysis of today´s status, but also reveals the different objectives pursued by academia and industry. These are two distinct worlds, with immense potential if they work together.
As research in AI continues to rapidly evolve, it becomes imperative to comprehensively analyze scholarly works in this domain. Our aim is to bridge the gap between the industry and academia by presenting the present status and spotlighting the top AI players worldwide. To achieve this, we have conducted an analysis of the trends in AI from January 2019 to March 2024, utilizing the freely available open knowledge repository of academic literature and patent metadata on Lens.org. Lens Lab, an initiative of Cambia and the Knowledge Futures Group at MIT, serves as a data collaborative for open innovation data and related analytics, tools, and metrics. It includes global patent datasets, citation graphs among patents and scholarship, and metrics derived from these datasets. The important partners collaborating with Lens Lab provide insight into the quality of its results [2].
Research in this discipline encompasses a wide range of topics and interests, shaping the trajectory of technological innovation and societal transformation. In this context, our analysis aims to address several key research questions: a. Which countries exhibit the highest production of AI research papers? b. Which institutions are the most relevant contributors to AI research? c. Who are the authors with the greatest impact, based on research paper publications, citations, and registered patents? d. Which research topics are being developed in AI?
Additionally, for patents, the questions are a. Which countries exhibit the highest production of AI patents? b. Which applicants are the most relevant contributors to AI patent development? c. Who are the inventors with the greatest impact, based on registered patents? Which patent types are the most relevant in AI?
This paper outlines the analysis of related research papers and patents, highlighting their transformative impact across various sectors. The methodology used allows us to understand the current state and future directions of AI. It seeks to answer key research questions regarding AI production, contributors, impact, research topics, and patent types. Future research directions may involve deeper analyses of specific research topics, tracking emerging trends, exploring interdisciplinary collaborations, and monitoring the evolving landscape of AI patents and scholarly works.
AI exhibits a wide range of applications. However, there is a significant gap in the literature covering AI comprehensively. While numerous studies focus on specific topics within the realm of AI, few endeavor to encapsulate AI. Some research has been conducted on patents related to specific AI topics.
One such comprehensive survey [3] introduces the evolution of artificial intelligence and subsequently analyzes key AI technologies, including computer vision, machine learning, natural language processing, human-computer interaction, virtual reality, and biometric recognition. Additionally, it outlines the developmental trajectory of AI technology. Another compelling study provides a systematic exploration of innovative research avenues surrounding AI, unraveling the backdrop and significant implications of AI within the innovation landscape. It identifies technological, social, and economic factors driving companies to adopt AI for innovation. Furthermore, it identifies product innovation, process innovation, business model innovation, and social innovation as key consequences of AI deployment [4].
Another interesting article analyzes AI journals from the previous decade (2008-2018) [5]. The Web of Science database was chosen, and the 10% of the articles with more citations were selected (12,301 articles). They proposed the five most related topics, which are perception intelligence (1), simulated intelligence of the human mind (2), machine learning based on classical models (3), intelligence inspired by biology (4), and intelligence based on big data (5). See the discussion section for comparison of results.
Only one paper addresses both aspects: patents and research articles. It compares the development of AI in China (CH) and the EU based on a survey of papers and patents [6]. The study spans from 2015 to 2019, predating the current article, and focuses on specific AI subjects, including artificial intelligence, neural networks, expert systems, genetic algorithms, support vector machines, and machine learning. In both countries, the number of papers increases over time, with CH experiencing a threefold increase. Regarding patents, the search was conducted using the Derwent Innovations Index. The researchers observe that CH and the US collectively hold approximately 89% of all patents worldwide. They further note that in recent years, the number of patents issued in CH has been on the rise, along with its share of the global patent market.
The search strategy employed for analyzing the last five years of research work production focused on specific criteria to ensure relevance and comprehensiveness. The strategy included: a. Publication Date: the period was set from January 1, 2019, to March 1, 2024, to capture recent publications and developments in the field; b. Publication Type: three primary publication types were considered: book chapters, journal articles, and conference proceedings articles, which ensured a comprehensive coverage of scholarly literature across various formats; c. Field of Study: the search was narrowed down to the field of artificial intelligence, ensuring that only relevant publications within this specific domain were included in the analysis. By utilizing these key words and filters, the search strategy aimed to gather a focused and up-to-date collection of scholarly works in artificial intelligence published within the defined period and across various publication types.
The overall AI research works during 2019-2024 were 15.52% of the article production (198,163) of the scholarly AI research. Fig. 1 shows the world distribution of the research work. Only 3,136 research articles cite patents.
Asia clearly leads the ranking; the countries that exhibit the highest production of AI research papers are CH, with 50,225 articles (25.36%), followed by United States with 22,287 (11.25%). The other countries in the ranking are India 10,380 (5.24%), UK 8,003 (4.04%), Germany 5,501 (2.78%), Australia 5,193 (2.62%), Republic of Korea 4,386 (2.21%), Italy 4,337 (2.19%), Canada 4,280 (2.16%) and Japan 3,704 (1.87%).
Analyzing institutional collaborations and author´s affiliations, we may conclude that leading universities play a crucial role in advancing AI research globally. Nine of the top ten most relevant institutions that have made notable contributions to AI literature are across Asia (9) and one is in North America (1). The latter, Harvard University, is also the only private research university.
Fig. 1. World distribution of research work (left) and patents (right)
The first institution, the Chinese Academy of Sciences is a National Public university with 74 years of existence, 69,000 faculty members and 59,985 students. The academic staff/student’s ratio they have equals 1.15. Harvard University is the oldest (388 years old) with 2,400 faculty members and 21,613 students. The top 10 university ranking is shown in Table 1.
Table 1. University ranking
| # | Universities | Papers | Students | Faculty | Ratio | Age | Type |
| 1 | Chinese Academy of Sciences | 5,112 | 59,985 | 69,000 | 1.15 | 74 | National Public |
| 2 | Tsinghua University | 1,739 | 50,390 | 3,565 | 0.07 | 113 | Public |
| 3 | Shanghai Jiao Tong University | 1,598 | 18,004 | 3,061 | 0.17 | 128 | Public |
| 4 | Zhejiang University | 1,558 | 65,821 | 4,557 | 0.07 | 97 | Public |
| 5 | Huazhong University of Science and Technology | 1,443 | 55,000 | 3,448 | 0.06 | 71 | National Public |
| 6 | Peking University | 1,278 | 44,730 | 11,337 | 0.25 | 126 | Public |
| 7 | Harvard University | 1,151 | 21,613 | 2,400 | 0.11 | 388 | Private research |
| 8 | Beihang University | 1,138 | 37,706 | 3,359 | 0.09 | 72 | Public research |
| 9 | Harbin Institute of Technology | 1,094 | 29,368 | 2,957 | 0.10 | 104 | Public |
| 10 | University of Electronic Science and Technology of China | 1,073 | 40,000 | 3,800 | 0.10 | 68 | Public |
As the formula to calculate the prominent author profile rank is not explicitly provided by Lens.org, we decided to build an average rank based on the number of research paper publications, citations, and registered patents by author, using a proprietary algorithm to calculate it. The ranking, led by Asia, is shown in Table 2.
Table 2. Author’s name ranking
| # | Author’s name | Papers | Total Citations in Scholar Works | Total Citations in patents | Region |
| 1 | Wei Wang | 292 | 2808 | 19 | Asia |
| 2 | Yang Liu | 253 | 2388 | 13 | Asia |
| 3 | Yang Yang | 214 | 2854 | 13 | Asia |
| 4 | Wei Li | 242 | 4222 | 5 | Asia |
| 5 | Wei Zhang | 204 | 1722 | 2 | Asia |
| 6 | Xin Wang | 196 | 1527 | 6 | Asia |
| 7 | Lei Wang | 187 | 1387 | 9 | Asia |
| 8 | Hao Wang | 182 | 1127 | 2 | Asia |
| 9 | Wei Liu | 172 | 1936 | 10 | Asia |
| 10 | Jun Wang | 191 | 2214 | 2 | Asia |
Many Chinese authors share the same name: “Wei Wang”, without ORCID, so what we can say is that this name leads the ranking in two categories, because of writing in 292 papers and being cited in19 patents, followed by name Yang Liu with 253 written papers and cited in 13 patents and Yang Yang with 214 written papers, and cited in 2854 research papers and 13 patents. Wei Li leads the cited research papers category with 4222 cites, holding the fourth place.
Fig. 2 highlights the distribution of research publications across key fields within the realm of Artificial Intelligence (AI), AI boasting a substantial 198,163 publications, indicating the significant role it plays. Computer science follows closely behind, with 189,538 publications, emphasizing the close relationship between AI and computer science disciplines. This highlights the significant overlap and mutual influence between the two fields, as AI innovations often drive advancements in computer science and vice versa. Machine learning and Deep learning are also prominent areas of study, with 64,631 and 29,561 publications, respectively. Other notable fields include «Artificial neural network,» «Computer vision,» «Data mining,» and «Pattern recognition (psychology),» each contributing significantly to the breadth and depth of AI research. Furthermore, the inclusion of fields such as «Biology,» «Engineering,» «Medicine,» «Philosophy,» and «Physics», among others, underscores the interdisciplinary nature of AI research. This interdisciplinary approach reflects the multifaceted challenges and opportunities presented by AI, as researchers draw upon insights and methodologies from various domains to address complex problems and drive innovation.
Fig. 2. Field of study word cloud
The top ten research paper field of study ranking is: Convolutional Neural Networks (1), Computer vision (3), Generative AI(Acoustics) (1), Generative Neural Networks (1), Adversarial systems (1), Facial Recognition (1), Explainable AI (1) and Machine Learning decision trees (Health) (1). The top ten cases of study are 2D pose estimation, object detection, face recognition, graph neural networks, interpretable models, and explainable AI for trees.
Intellectual property encompasses creations originating from the mind, including inventions, literary and artistic works, designs, symbols, as well as names and images utilized in commercial contexts. Legal protection such as patents, copyrights, and trademarks safeguard intellectual property, granting individuals the opportunity to gain recognition or financial rewards for their inventions or creations. An invention refers to a novel product or method that offers a fresh approach to accomplishing a task, and it must be revealed to the public through a patent application.
Essentially, having a patent grants the owner exclusive rights to control the commercial use of their invention. This means that others cannot produce, use, sell, distribute, or import the patented invention without obtaining permission from the patent holder.
Patents are limited to specific territories. Typically, the exclusive rights provided by a patent apply only within the country or region where the patent has been officially filed and approved, in accordance with the respective laws of that authority.
Our search was conducted using the following search stream: “»artificial intelligence» OR «inteligencia artificial» OR «intelligence artificielle» OR «künstliche Intelligenz» OR «intelligenza artificiale» OR 인공지능 OR 人工智能 OR 人工知能 OR «искусственный интеллект» OR «inteligência artificial».” The topic “artificial intelligence” was included in the most important languages, namely Spanish, French, German, Italian, Korean, Chinese, Japanese, Russian and Portuguese. Arabic was not included as there were no patents registered in that language in the database. Acronyms, such as IA, were excluded as they did not improve the search accuracy. The date search interval ranged from January 2019 to March 2024. Only two legal statuses were included: «active,» indicating that a patent has been granted and is still valid, and «pending,» indicating that the examination of a patent application is ongoing.
The Lens application identified 316,883 patent records across all districts. The overall active and pending patents during 2019-2024 were 34,146,538 patents, which represents 0.92% of patent applications. The US boasts the highest number of patents at 180,085, representing 56% of the total. CH follows in the second place, with 63,503 patents.
The World Intellectual Property Organization (WIPO) has 43,522 patents, which are not included in the counts of the two countries. European Patents account for 20,359 patents valid in European countries.
In the eastern part of the world, three countries lead the ranking after China: South Korea with 8,736 patents, Japan with 715, and Taiwan with 707 patents. In the western hemisphere, the US and Canada lead with 180,085 and 509 patents respectively, followed by the United Kingdom with 244. Australia has 507 patents. Latin America is led by Mexico, with 81 patents.
Multinational and international companies lead the artificial intelligence (AI) market. Samsung Electronics Co. Ltd. holds the highest number of patents globally, with 17,089 patents. It is a multinational electronics and information technology company headquartered in Samsung Town, Seoul, South Korea, and it is the main subsidiary of the Samsung Group.
The second company is IBM, with 9,286 patents. IBM produces and sells computer hardware and software, providing infrastructure, internet hosting, and consulting services in various computing-related fields, spanning from mainframe computers to nanotechnology. It holds more patents than any other US technology company.
The third company is LG Electronics Inc., with 8,913 patents. It develops technological advances in electronics, mobile communications, and household appliances. With headquarters in Seoul, South Korea, it is one of the largest electronic conglomerates globally.
Four of the top ten companies have headquarters in CH. Tencent Tech Shenzhen Co. Ltd. holds 7,155 patents and is a world-leading internet and technology company, founded in 1998, with headquarters in Shenzhen. Beijing Baidu Netcom Sci & Tech Co. Ltd. holds 6,604 patents and provides internet access and internet-based services.
Huawei Technologies Co. Ltd. holds 5,466 patents and is a leading global provider of information and communications technology (ICT) infrastructure and smart devices. It is headquartered in Shenzhen, Guangdong.
Ping An Tech Shenzhen Co. Ltd. holds 3,906 patents and is the core technology arm of Ping An Group, using AI, cloud, and other innovative technologies to develop and operate mission-critical platforms and services that support financial services, medical health, automotive services, real estate services, and smart cities. It is headquartered in Guangdong Sheng.
Three of the top ten companies are headquartered in the USA. Qualcomm Inc., which holds 6,380 patents, is a multinational corporation based in San Diego, California, which specializes in the development of semiconductors, software, and services associated with wireless technology. Intel Corp. holds 5,010 patents and is a manufacturer of integrated circuits. AT&T Inc. holds 4,296 patents and is a multinational telecommunications holding company based at Whitacre Tower in Downtown Dallas, Dallas, Texas.
We do not know much about inventors profiles’, when compared to those of researchers’, because there is limited available data about the former. See in Fig. 3 inventors ranking. It has only been possible to identify their relationship with their workplace.
Two companies house these top inventors. At Qualcomm INC, Dr. Tao Luo serves as the Senior Director of Engineering, having joined in 2004 [7]. Dr. Li Junyi has held the position of Vice President of Engineering at Qualcomm since 2006 [8], and Zhou Yan has been a staff engineer since 2006 [9].
At LG Electronics, Lee Seungmin is a principal inventor [10], Hanbyul Seo is a Senior Research Fellow who has been working since 2008 [11], and Kim Seonwook has been a Research Engineer since 2014 [12]. Additionally, Yang Suck-chel has several patents with LG Electronics .
Moreover, Paul Shala Henry has developed patents at AT&T Intellectual Property I, L.P. [13].
Fig. 3 Inventors ranking
The Cooperative Patent Classification (CPC) was selected to analyze the top types of patents. The European Patent Office (EPO) and the US Patent and Trademark Office jointly manage it. Fig. 4 shows the five clusters of patents we identified: a. Computing arrangements based on specific computational models (G06N), composed by 39,218 patents of Machine Learning, 39,089 of Learning methods, 31,693 of Combinations of Networks, 12,128 of Backpropagation, 10,913 of Recurrent networks, 9,884 of Architecture, 8,555 of Probabilistic graphical models, and 7,810 of Inference or reasoning models patents. b. Image data processing or generation (G06T), composed by 12,055 Training Learning and 11,664 Artificial Neural networks patents. c. Image or video recognition or understanding (G06V), using 13,534 Neural networks, and 7,709 Classification patents. d. Electric digital data processing (G06F), composed by 11,109 Generating training patterns Bootstrap methods, and 9,989 Semantic analysis patents. e. Healthcare informatics (G16H), composed by 11,958 Computer-aided diagnosis patents.
We note that the number of patents (316,883) is bigger that the number of papers (198,163); only 3,136 patents cite research articles (1.58%) and 6,511 articles cite patents (3.28%). This indicates that there could be more collaboration between academia and industry to enhance the societal impact of research. While maintaining innovation is essential for research agendas, there might be missed opportunities for resource acquisition if there is insufficient focus on product development.
Eastern countries are prominent in the AI field, with many leading researchers and inventors having names of Eastern origin. However, this does not necessarily mean they are affiliated with or located in Eastern countries. As AI continues to reshape society, Western countries and universities must consider how to motivate and prepare students for AI-related education. A solid foundation in mathematics is crucial to cultivate capable individuals in this context.
The data illustrates that CH has been leading research since the previous decade (2008-2018). During the same period, the US held the second position, but India, previously ranked sixth, has now risen to the third place, surpassing countries like the UK, Australia, and Spain. Additionally, the Chinese Academy of Sciences has been at the forefront of universities since 2008. All other Chinese universities listed in Table 1 were among the top 20 universities of the previous decade, indicating an increase in their research output. The author Juan Wang is also ranked among the top ten authors in this bibliometric study and in ours. Even though the referred paper has a different classification criterion, the fields recognized as prominent in our study are also included in the Cluster analysis of high-frequency keywords, which evidences the persistent interest in the area.
US leads in patent ownership. However, there are factors such as country policies, strategies regarding intellectual property, regulatory frameworks, and market value which differ between the east and the west, which may influence this scenario. For example, Huawei, a company headquartered in CH, holds more patents in US (1,690) than in CH (1,263), while IBM holds 91% of its patents in US. Additionally, globalization and the presence of international or regional entities holding patents, which are indeed valuable initiatives, serve as instruments to facilitate the internationalization of patents.
AI presents a broad spectrum of applications, and research innovations or patents are often classified into multiple categories. The top-ranked types of patents could guide academic research efforts, directing more attention towards them. «Computer science» or «computing arrangements based on specific computational models» are particularly significant, as they form the foundation of AI developments. Interest in computer vision and medicine is evident in both academia and industry. «Electric digital data processing» is not ranked as a top subject in academia. Further analysis per subject is required to clarify finer gaps between academia and industry.
Lens is a valuable tool for analyzing research articles and patents on a particular issue. It is an open initiative of MIT, freely available for research purposes, which allows users to explore related works from academia and industry. We believe it is important to train professors and students to incorporate these two aspects into their related works sections when applicable. Lens collects a wide range of data from research and industry sources, including major editorials and patent sources.
The triple helix innovation model [14], proposed as a basis for scientific and technological policy, outlines the relations among governments, universities that provide the knowledge resulting from technological research and companies that provide funds to obtain such knowledge to transfer it for the benefit of their productive structures. Like everything that is human, this model represents an ecosystem whose purpose is to generate benefits and allow social growth. Moreover, it needs the joint effort of all parties to achieve that common good. The results clearly show the weak cohesion between the objectives of the propeller blades. This, unsurprisingly, creates mutual obstacles in trying to achieve a solid boost in research and patent development of artificial intelligence technologies. However, faced with the new challenges posed by our current world and, especially, innovative technologies such as AI, the simple proper functioning of each propeller is insufficient. In her book, entitled «Technology and virtues» [15], the philosopher Shannon Vallor invites us to be aware of the power that is currently in the hands of humanity, and the consequent need to nurture personally and socially the virtues that command and channel technology to achieve common happiness.
Moreover, it is important to point out that we only used the Lens platform to extract the data about research papers and patents. The research paper tool allows to refine the search by selecting “Field of Study”, on the contrary, the patent tool does not have this option, which may introduce an error in the total number of patents when it is compared to the total number of research papers, but this does not invalidate the conclusions and the percentages informed. This study must be validated with equivalent searches in other knowledge platforms such Scopus (limited to research papers), Espacenet [16] (limited to patents, which is supported by the European Patent Register). However, it may be possible to fully validate the results of this study in the case of patents, if we use other tools such as Google Patents [17] or Questel [18] or patsnap [19], although these two last tools have the limitation that cannot be freely accessed. In addition, it is important to note that this document provides the overview of the last 5 years, as our goal is to understand the current status of this technology. One aspect to consider when analyzing the results is that Lens uses ML techniques for the association of data from diverse sources and the elaboration of different rankings. We understand that this aspect can generate some differences in the possible comparison with the results of other knowledge bases.
In summary, our analysis reveals Asia, particularly CH, is the leading producer of artificial intelligence research articles, accounting for 37.30% of the total publications. Following closely is North America, primarily the US (13.41%), and Europe (9%). Institutional affiliations of authors lie in Asia, with the Chinese Academy of Sciences leading the list with 5,112 articles, while Harvard University stands as the sole North American institution in the top rankings, at the 7th position. Notably, Wei Wang emerges as the top author, with 292 papers cited by 2,808 research works and 19 patents. Thematic clusters such as Computer Vision, Convolutional and Neural Networks, Adversarial Systems, Facial Recognition, Explainable AI, and ML Decision Trees dominate the research landscape. From the analysis, it emerges that a sizable portion of articles (40%) are Open Access (OA), with 80% of them receiving funding from governments, private companies (Nvidia, Google, Microsoft, and InsightFace), and educational institutions.
On the other hand, US leads the patent market and CH follows in the second place. Also, the organizations that hold international patents as WIPO or European Patents hold a significant number of them. Multinational and international companies lead the artificial intelligence (AI) market, the first ones being Samsung Electronics Co. Ltd. in the east and IBM in the west. At the same time, LG Electronics Inc. and several companies of CH have an important participation in world market. The top inventors are gathered at Qualcomm Inc. and at LG Electronics. Machine Learning, learning methods, Combinations of Networks, Backpropagation, Recurrent networks, Probabilistic graphical models and Inference or reasoning models are the top topics for patents. Also, Electrical digital and image data processing, Healthcare Informatics and image or video recognition are subjects that companies are interested in.
Research articles that cite patents are only 3.28% of the total number of research articles. This shows that Academia and Industry must work together more often, to generate more value for society.
Another aspect that may be included in future works is the identification of research papers and patents that have more impact in academy and industry. For that we plan to do a network connection analysis to better understand the topography of this ecosystem (academy, industry, and government). We are planning to detect main actors, fields of interest and government strategic areas of development related with AI. Also, it is important to replicate this study per region to understand the evolution of AI in each one and lead their resources to world interest, to unify efforts.
To conclude, these findings offer critical insights into the global dynamics of AI research and collaboration, guiding future endeavors in this rapidly evolving field.
Acknowledgments. This work is supported by a research grant from Engineering School, Universidad Austral, Argentina.
Carreras de Grado: info@austral.edu.ar
Posgrados: posgradosfi@austral.edu.ar
