Theory

With the astonishing successes of artificial intelligence in recent years—recognizing faces, translating languages, generating cogent text, driving cars—CBS scientists have increasingly turned to thorny problems of how our brains achieve intelligent cognition.

Advances in AI are proceeding at an historic pace, yet it is widely acknowledged that AI falls far short of the flexible intelligence evident in humans and animals. Animal brains—­­­­and human brains in particular—are nature’s existence proof that physical mechanisms can give rise to general intelligence.

The CBS effort to discover the bases of intelligent cognition draws together neuroscience, cognitive science, and computer science. As an essential first step towards this goal, we are building a coherent theoretical community aimed at understanding natural and machine intelligence. This group is closely allied with experimentalists studying the neural bases of intelligent cognition in humans and experimental animals. Theorists—with backgrounds in computer science, physics, statistics, mathematics, and psychology—engage in frequent interactions to build an intellectual framework for understanding the complexities of intelligence.

Our NeuroTheory Initiative is expanding the research and education at Harvard in the multidisciplinary field of theoretical neuroscience-cognition-AI.

Thanks to generous support from Dean of Science Christopher Stubbs, we have begun with a set of shared, interactive activities. We are building on faculty strength in these fields, which is spread across schools and departments but is focused in the FAS and SEAS. A robust Visitors program is establishing close ties to industry, where rapid advances in AI are being made.

We launched the NeuroTheory Initiative in the spring of 2022 with a kickoff event featuring posters, a panel discussion, and talks by George Alvarez, Nada Amin, Demba Ba, Boaz Barak, Jonathan Frankle, Sam Gershman, Sham Kakade, Talia Konkle, Hima Lakkaraju, David Nelson, Cengiz Pehlevan, Hanspeter Pfister, and Tomer Ullman.

Our colleagues from NTT Research at Harvard, Hidenori Tanaka and Gautam Reddy, host visiting graduate students and interns throughout the summer and continuing into the academic year. In August 2022, CBS hosted a 2-day workshop on Reinforcement Learning, featuring talks, posters, panel discussions, introductory tutorials, and many informal conversations. Speakers included Sam Gershman, Petros Koumoutsakos, Poornima Kumar, Lucy Lai, Jack Lindsey, Susan Murphy, Gautam Reddy, Sandra Romero Pinto, Maurice Smith, Anna Trella, Naoshige Uchida, and John Vastola.

We have assembled a group of Harvard scientists with a key research interest aimed at developing a theoretical understanding of intelligence. This group of faculty (listed below), with their trainees, forms the core of the NeuroTheory Initiative. They are drawn from Applied Math, Computer Science, Electrical Engineering, Mathematics, Neurobiology, Physics, Psychology, and Statistics. As in fields like physics with a mature theory-experiment tradition, the CBS NeuroTheory Initiative encourages close interactions with experimental neuroscientists and cognitive scientists.