Artificial intelligence and machine learning — where machines perform routines and tasks once performed only by humans — are becoming ubiquitous. These deep technologies are finding applications in every industry, from national defense to agriculture — and everywhere in between.
AI requires a combination of talent and technology that are usually found only in the world’s best research labs and classrooms. Using Microsoft Academia, here is a list of institutions that are leading in research output so far in 2021.
Harvard University is excels not just at AI research, but applying that research in a range of interdisciplinary and multidisciplinary pursuits. This year, the university’s researchers have landed research studies in leading journals. “Using graph convolutional neural networks to learn a representation for glycans” landed in Cell Reports while “Protein design and variant prediction using autoregressive generative models” was published in Nature. Another strength of Harvard researchers, beyond applying AI applications to certain areas and problems, is improving on the technology of AI, itself. Harvard researchers shared, “HyperMorph: Amortized Hyperparameter Learning for Image Registration” on ArXiv.
Harvard’s top-rated student body and its commitment to entrepreneurial ventures of its faculty and staff are other reasons for the university’s global leadership in AI.
Stanford University is located the cradle of the high technology sector. This heritage has extended to artificial intelligence and machine learning. The Stanford Artificial Intelligence Laboratory (SAIL), in fact, has promoted teaching and research into AI since its in 1962. That’s almost six decades. The university’s recent research output shows that AI research tradition is continuing.
Some recent research studies include CheXtransfer: Performance and Parameter Efficiency of ImageNet Models for Chest X-Ray interpretation, which appeared recently in Proceedings of the Conference on Health, Inference, and Learning. Stanford econometrics researchers are using AI to study the economy and finance, including this study: “Using Wasserstein Generative Adversarial Networks for the Design of Monte Carlo Simulations,” which appeared in the Journal of Econometrics. Neural engineering is a deep tech area where AI is absolutely necessary. Stanford researchers have made progress in using AI tools to study the brain, like this new study: A Review of Classification Algorithms for EEG-based Brain–Computer Interfaces, which appeared in the Journal of Neural Engineering.
Massachusetts Institute of Technology
“For more than 50 years, CSAIL—the MIT Computer Science and Artificial Intelligence Laboratory—has pioneered new approaches to computing that improve how people work, play, and learn. Now we stand on the verge of an exciting new era ready to make great new contributions,” writes Daniela Rus Director, CSAIL Andrew (1956) and Erna Viterbi Professor EECS, MIT.
The 2021 research output numbers bears out Dr. Rus’s vision — and backs up MIT’s reputation as an AI pioneer.
Recently, MIT researchers have published studies on AI, including “Grounding Physical Concepts of Objects and Events Through Dynamic Visual Reasoning,” which appeared in ArXiv. The university’s AI scientists were also involved in work, such as “Deep Learning: A Statistical Viewpoint,” also in ArXiv. “Roadmap on Emerging Hardware and Technology for Machine Learning,” an MIT-led study, appeared in Nanotechnology.
Chinese Academy of Sciences
China has an array of academic institutions and companies with the firepower to stay on the cutting edge of AI, including Tsinghua University, Peking University, the Chinese Academy of Sciences, Baidu, and Xiaomi. These universities and companies all participated in the Beijing Academy of Artificial Intelligence, which created Wudao 2.0. — NikkeiAsia
The first non-American entry on the list of top AI research institutions is the Chinese Academy of Sciences, a global leader in research in general, but has lately become one of the world’s premier research hubs for artificial intelligence and machine learning. The heart of that hub is The School of Artificial Intelligence of the University, which received its first group of students in 2017.
The university’s researchers contributed to recent work, including AI includes “Fully Convolutional Networks for Panoptic Segmentation,” which appeared in ArXiv and “Improving Proton Dose Calculation Accuracy by Using Deep Learning,” which was presented at 2021 Machine Learning: Science and Technology. “More Diverse Means Better: Multimodal Deep Learning Meets Remote Sensing Imagery Classification,” also included Chinese Academy of Sciences contributions — and is available at IEEE.
University of California, Berkeley
Pedigree should give you some idea of the research excellence into artificial intelligence and machine learning at the University of California, Berkeley. Peter Norvig, director of research at Google, received his doctoral degree at Berkeley. And the list goes on.
Another list that goes on is the list of recent AI advances of Berkeley researchers. Those papers include: “A Generalizable and Accessible Approach to Machine Learning with Global Satellite Imagery,” which appeared in Nature Communications; “Provable Meta-Learning of Linear Representations,” which was published in IEEE and “Probabilistic Harmonization and Annotation of Single-Cell Transcriptomics Data with Deep Generative Models,” which appeared in Molecular Systems Biology.
University of Oxford
Mind Foundry and DiffBlue are just two examples of AI-based startups that were nurtured by the University of Oxford. The AI startup ecosystem at Oxford is a result of how the university blends education, inspiration, research and entrepreneurship.
Oxford scholars have contributed to major AI research projects this year, including: “Analysing factorizations of action-value networks for cooperative multi-agent reinforcement learning,” published in Autonomous Agents and Multi-Agent Systems; “Using Neural Network and Random Forest Algorithmic Approaches to Predicting Particulate Emissions from a Highly Boosted GDI Engine,” which appeared in SAE and “Multi-scale Driver Behavior Modeling Based on Deep Spatial-temporal Representation for Intelligent Vehicles,” which the researchers published in Transportation Research.
Google is the first commercial research institution to make the list. There is one reason that this company is listed among large, powerful research institutions with centuries of scientific legacy and decades of AI mastery: It’s very existence depends on AI.
Google needs continual advances to power its traditional products, such as search, but also expand the bounds of its experimental products, such as voice recognition.
Another reasons is Google partners with many of the leading institutions, including ones in this list.
Here are a few AI-related projects that Google researchers are involved in: “Thinking Fast and Slow: Efficient Text-to-Visual Retrieval With Transformers,” presented at Computer Vision and Pattern Recognition and published on ArXiv; “Machine Learning-accelerated Computational Fluid Dynamics, published in PNAS and “NeRF in the Wild: Neural Radiance Fields for Unconstrained Photo Collections,” presented at Computer Visions and Pattern Recognition and published in ArXiv.
University of Washington
With Microsoft, Amazon and AWS in its backyard, the University of Washington has become a breakout academic leader in technological research. Artificial Intelligence is increasingly a focus. UW, in fact, is leading an NSF institute for AI and the Paul G. Allen School of Computer Science and Engineering is now considered an elite center for AI research. UW researchers have recently worked on AI- and machine learning-related projects, including: On “Nonconvex Optimization for Machine Learning: Gradients, Stochasticity, and Saddle Points,” published in the Journal of ACM; “Multicenter, Head-to-Head, Real-World Validation Study of Seven Automated Artificial Intelligence Diabetic Retinopathy Screening Systems,” in Diabetes Care and “Selecting Directors Using Machine Learning,” a working paper published in Review of Financial Studies.
Max Planck Society
Max Planck Society is one of Europe’s most respected research institutions in multiple fields and disciplines. One of those is certainly artificial intelligence and associated technologies, such as robotics. The Max Planck Institute for Intelligent Systems, for example, is one of the 86 research institutes of the Max Planck Society.The center has two campuses, one in Stuttgart and the other in Tübingen.
This year, so far, Max Planck Society researchers have participated in published and presented work, including “Toward Causal Representation Learning,” published in IEEE; Spatially Structured Recurrent Modules, presented at the International Conference on Learning Representation and Real-time Deep Dynamic Characters, published in ACM Transactions on Graphics.
University of California, Los Angeles
At the UCLA Samueli School of Engineering, we know the power engineering has to change lives. And that transformational power comes in many forms. — Jayathi Murthy
Ronald and Valerie Sugar Dean, UCLA
This year, it’s researchers contributed to numerous projects, published papers and conference presentations. Many of those papers were published in leading journals and presented at the nation’s most prestigious academic conferences.
UCLA AI scientists worked on the following research projects: “Deep Learning for Person Re-identification: A Survey and Outlook,” in IEEE; “Bio-JOIE: Joint Representation Learning of Biological Knowledge Bases,” presented at BCB and published in ArXiv and “Universal Representation Learning of Knowledge Bases by Jointly Embedding Instances and Ontological Concepts,” also available in ArXiv.
This list, drawn from Microsoft Academia as a primary source, should be seen as a snapshot in time and not as a static, set-in-stone list — scientists are continually releasing new work and publishing and conference calendars often dictate the pace of research. This list can change — and will change.