Credit: Kiel Mutschelknaus/Columbia Engineering
New York, NY—May 4, 2023—The National Science Foundation (NSF) announced today that it is awarding $20 million to establish the AI Institute for ARtificial and Natural Intelligence (ARNI), an interdisciplinary center led by Columbia University that will draw together top researchers across the country to focus on a national priority: connecting the major progress made in artificial intelligence (AI) systems to the revolution in our understanding of the brain.
ARNI is a collaboration between Columbia, Baylor College of Medicine, City University of New York, Harvard, Princeton, Howard Hughes Medical Institute, Mila Quebec AI Institute, Tuskegee University, the University of Pennsylvania, and UTHealth Houston. Industry partners include Amazon, DeepMind, Google, IBM, and Meta, and outreach partners include the Neuromatch Academy and the New York Hall of Science. In addition to receiving NSF funding, ARNI is funded by a partnership between NSF and the Office of the Under Secretary of Defense for Intelligence and Security (R&E).
By bringing together the amazing progress being made in AI systems and our growing understanding of the brain, ARNI will ignite advances in both neuroscience and AI, and transform our world in the next decade.
Provost, Columbia University
“The National Science Foundation has long been a strong supporter of research at Columbia University and we are very excited about this new collaboration,” said Mary Boyce, Provost, Columbia University. “The AI Institute for Artificial and Natural Intelligence draws not only on our interdisciplinary strengths throughout the University but also our partnerships — both old and new — across the country. By bringing together the amazing progress being made in AI systems and our growing understanding of the brain, ARNI will ignite advances in both neuroscience and AI, and transform our world in the next decade.”
Revolution in neuroscience, cognitive science, and AI research
The past 10 years have seen spectacular progress in interrogating neural activity, circuitry, and learning, yet our neuroscience insights have so far informed AI only superficially. Conversely, our rapidly advancing AI methods and systems based on massive amounts of data have only begun to impact neuroscience. ARNI will meet the urgent need for new paradigms of interdisciplinary research between neuroscience, cognitive science, and AI. This will accelerate progress in all three fields and broaden the transformative impact on society in the next decade.
“ARNI is an ambitious plan that requires a dedicated effort across institutions, and we have assembled one of the strongest groups of investigators in theoretical neuroscience and foundational machine learning in the world,” said Jeannette Wing, Executive Vice President for Research, Columbia University. “Our PIs are building on existing, and often tightly interacting, neural and AI groups at Columbia, Baylor, Penn, together with Janelia, MILA, Google/DeepMind, and Meta. At the same time, we are building new bridges to Tuskegee, CUNY, Yale, IBM, and beyond. Our track record is already strong and now, thanks to the National Science Foundation, we expect ARNI to meet the urgent need for new paradigms of interdisciplinary research between neuroscience and AI.”
ARNI will be led by Principal Investigators (PIs) Richard Zemel, Kathleen McKeown, and Christos Papadimitriou (Computer Science, Columbia Engineering), Liam Paninski (Zuckerman Institute and Statistics and Neuroscience Departments, Columbia University), and Xaq Pitkow (Baylor College of Medicine, Rice University). These PIs bring together expertise from a wide variety of disciplines, including artificial intelligence, theoretical computer science, statistics, neuroscience, physics, and cognitive science. They will work with a large team of researchers to tackle the limitations and challenges of current machine learning systems, including learning with limited data, reasoning about causality and uncertainty, and lifelong learning–all hallmarks of biological systems–while also pushing the boundaries of our understanding of how brains compute and learn.
Bridging the gaps between artificial and biological networks
ARNI will bridge the current significant gaps between artificial and biological networks and make room for a broad, diverse range of applications, from the industrial sector, such as robust, interpretable medical decisions and smarter home assistants; to societal applications, such as better social safety nets and assistive multimodal systems to help the vulnerable; to scientific discoveries such as providing hypotheses about brain function and creating powerful tools for extracting insights from massive data.
“Thanks to new AI algorithms, our knowledge of neuroscience and cognitive science expands every day,” said Shih-Fu Chang, Dean of Columbia Engineering. “And with our growing knowledge of the brain and cognitive science, we have better AI algorithms, making progress on important applications that impact our world. ARNI aims to overcome current limitations in AI while also introducing modern AI into neuroscience, foundational machine learning, and cognitive science. Engineers are pivotal for applying scientific insights to real-world problems, and we look forward to the groundbreaking discoveries that will come from this exciting large-scale collaboration. We are grateful to the National Science Foundation for helping us create this modern cross-disciplinary arsenal, converging to generate new insights and advance this very important, emerging field.”
Richard Zemel, the Director of ARNI and the Trianthe Dakolias Professor of Engineering and Applied Science at Columbia Engineering, has been integral in the development of AI technology, most recently as the co-founder and Research Director of the Vector Institute for Artificial Intelligence. His research spans machine learning and its interaction with neuroscience and cognitive science, as well as robust and fair machine learning. He noted that robust and fair machine learning is critical for using these new AI tools to improve society.
“A key characteristic of our approach is a focus on developing interpretable models, often based on causal approaches, that are cognitively grounded, given our research on the brain,” Zemel said. “This will lead to the development of trustworthy systems that can explain their reasoning to end users in terms they understand. This is critical in high-stakes applications such as healthcare, law, and in support of vulnerable populations.”
Education and outreach
The institute will provide educational and research opportunities for undergraduate and graduate students, as well as postdoctoral trainees, within and at the interface of AI, neuroscience, and cognitive science. Outreach partners, including the Neuromatch Academy and the New York Hall of Science, will help inform the public of these new developments and teach critical skills to the next generation of students.
Since 1864, the Fu Foundation School of Engineering and Applied Science at Columbia University has been a resource to the world for major advances in human progress. Today, Columbia Engineering is a leading engineering school and a nexus for high-impact research. Embedded in New York City, the School convenes more than 250 faculty members and more than 6,000 undergraduate and graduate students from around the globe to push the frontiers of knowledge and solve humanity’s most pressing problems.
In collaboration with Columbia’s Zuckerman Institute, the ARNI team includes leading senior investigators and visionaries in the field of theoretical and cognitive neuroscience. The Zuckerman Institute brings together diverse researchers whose expertise spans a wide range of interdisciplinary neuroscience research areas, providing an unsurpassed intellectual environment, multi-level support, and opportunities for interaction.
The U.S. National Science Foundation propels the nation forward by advancing fundamental research in all fields of science and engineering. NSF supports research and people by providing facilities, instruments, and funding to support their ingenuity and sustain the U.S. as a global leader in research and innovation. With a fiscal year 2022 budget of $8.8 billion, NSF funds reach all 50 states through grants to nearly 2,000 colleges, universities, and institutions. Each year, NSF receives more than 40,000 competitive proposals and makes about 11,000 new awards. Those awards include support for cooperative research with industry, Arctic and Antarctic research and operations, and U.S. participation in international scientific efforts. www.nsf.gov