Clarifying the scope and activity within the large field of artificial intelligence (AI) can help research leaders, policymakers, funders and investors, and the public navigate AI and understand how it has evolved over time.
This effort can also provide clues to where AI is headed and how policies might be shaped to continue making advances in a responsible way. Elsevier’s new report Artificial Intelligence: How knowledge is created, transferred, and used provides insight into research output, collaboration and mobility for China, Europe and the United States. It's available to download free of charge.
Defining the field of AI
The report tackles the challenge of identifying artificial intelligence research by using AI techniques such as supervised and unsupervised learning. In doing so, it defines a multifaceted field across teaching, research, industry and media perspectives. It reveals seven AI research clusters, and provides insights on growth and regional trends in AI research. A case study offers an illustration of AI education in China, and the report delves into the essential role of ethics in AI.
To learn more about the technical background, please see our methodology document.
The report holds a special focus on delineating the field of AI bottom up by using AI technologies.
AI research clusters
The AI research field clusters around seven main research areas. Keyword clusters and co-occurrences in the AI field, 2017; the color of the keyword represents its originating perspective: Teaching: orange, Research: blue, Industry: green, Media: pink, Multiple perspectives: black. source: Scopus.
Seven interconnected fields of research appear in the analysis. You can investigate which terms appear and connect over time or how they interact differently per region using our interactive co-occurrence charts with time slicer: Global (all terms), China (core terms), Europe (core terms), USA (core terms). You might be interested to find that:
- Most AI topics have been around for more than two decades.
- The subfields of Machine Learning and Probabilistic Reasoning and Neural Networks have intensified in the last 5 years.
- A few topics reach a sufficient number of publications to become visible on the graphs, including Deep Learning and Swarm Intelligence in 2000, Transfer Learning in 2004, and Deep Neural Networks in 2005.
- This is the first step in scoping an emerging field. Further aspects, such as term hierarchies and full AI ontologies, are part of our ongoing research.
The report is a contribution to the wider dialogue around AI and cannot cover all aspects of the discussion. We are open and interested in further exchange with you. You may also be interested in the following seminal AI papers, hand-selected by our AI publisher:
- Schmidhuber, J: Deep learning in neural networks: An overview, Neural Networks (2015)
- Baldi P, Sadowski P: The dropout learning algorithm, Artificial Intelligence (2014) — open access
- Bobadilla, J, Ortega, F, Hernando, A, Gutiérrez: A: Recommender systems survey, Knowledge-Based Systems (2013)
- Ferrucci, D, Levas A, Bagchi S, Gondek D, Mueller E: Watson: Beyond Jeopardy! Artificial Intelligence (2013) — open archive
- Jain, AK: Data clustering: 50 years beyond K-means, Pattern Recognition Letters (2010)
- Hendler, J, Berners-Lee, T: From the Semantic Web to social machines: A research challenge for AI on the World Wide Web, Artificial Intelligence (2010) — open archive
- Salakhutdinov, R, Hinton, G: Semantic hashing, International Journal of Approximate Reasoning (2009)
- Craven, M, Dipasquo, D, Freitag, D, McCallum, A, Mitchell, T, Nigam, K, Slattery, S: Learning to construct knowledge bases from the World Wide Web, Artificial Intelligence (2000) — open archive
- Page, L, Brin, S: The anatomy of a large-scale hypertextual web search engine, Computer Networks (1998)
- Berners-Lee, TJ: The world-wide web, Computer Networks and ISDN Systems (1992)
- Sakai, T, Nagao, M, Fujibayashi, S: Line extraction and pattern detection in a photograph, Pattern Recognition (1969)
Challenges, myths and realities
AI is a very hard field to circumscribe. Yet doing so is essential.
— Raymond Perrault, Senior Technical Advisor, AI Center, SRI International
AI is a very hard field to circumscribe because of the range of its problems of interest, diversity of its technical approaches, and dependence on other scientific areas for fundamental techniques and applications. Yet doing so is essential to understanding AI's technical progress, the resources it consumes and its impact on society. The methodology being developed in this study is a major step in defining its contours.
This era of symbiotic technology will untether us from screens and keyboards.
— Elsevier CTO Dan Olley in Forbes Magazine
In A CIO Hall of Famer’s Approach to Machine Learning, technology author Peter High interviewed Elsevier’s CTO and Executive VP of Product Development Dan Olley about tech trends and their applications at Elsevier and beyond. During the interview, Dan commented: "This era of symbiotic technology – enabled by combining AI, device miniaturization, 5G and an API-rich ecosystem – will untether us from screens and keyboards. As we think about training rather than programming these systems, new problems around bias and ethics will arise that we need to consider carefully."
Read more on Elsevier Connect.
Transferring and using AI are key aspects of the investigation of the field. In the report, we learned about AI as general purpose technology (see interview from Ingrid Ott, report page 80). We saw that its influence in application fields, such as self-driving cars, soccer robots or health applications, is around 30% (share of AI publications in the overall publication field).
For some of the fields, especially in Life Sciences and Health, we are applying AI ourselves to help researchers and clinicians advance science and health. With our partners, we are exploring the advantages, risks and realities of the AI technologies. To get a flavor of our activities please find some stories below:
Big data helping us to understand big ethical questions
Dr. Thomas Steger-Hartmann, VP and Head of Investigational Toxicology at Bayer, partnered with Dr. Matthew Clark of Elsevier’s Life Science Professional Services team to help understand the effectiveness of animal testing during drug development.
Can AI help find cures for rare diseases?
Dr. Jabe Wilson of Elsevier’s Life Science Professional Services team, the Pistoia Alliance and CuresWithinReach present the Rare Disease Drug Repurposing Datathon.
Elsevier launches Entellect
Unlocking the potential of life sciences information with Elsevier’s smart and versatile platform which delivers AI-ready data on the fly.
For AI to truly transform healthcare, we need quality data – now
Dr. Ian Chuang, Dr. Tim Hawkins, Olaf Lodbrok, and Dr. Richard Loomis share how deep information analytics and AI can enable higher quality, personalized healthcare.
Healthcare: Leveraging tech and minimizing data risks
Jan Herzhoff, Managing Director, Asia Pacific at Elsevier shared that machine learning and AI can help revolutionize outcomes by freeing up time for healthcare practitioners to make more informed clinical decisions.
Improving clinical outcomes through Artificial Intelligence
Tim Morris, Product and Partnership Director of Elsevier Health, EMEALAAP, spoke on the potential of using AI and clinical decision support solutions to empower doctors to make more knowledgeable decisions that improve clinical outcomes.