PeopleofSuperintelligence

Researcher

Paul Smolensky

Paul Smolensky is a cognitive scientist whose AI work connects neural computation with symbolic accounts of language and reasoning. He studied physics at Harvard and Indiana University, worked with UCSD’s Parallel Distributed Processing group, and taught computer science at Colorado Boulder before joining Johns Hopkins. With Géraldine Legendre and Alan Prince, he developed Harmonic Grammar and Optimality Theory; his Integrated Connectionist/Symbolic architecture linked neural and symbolic accounts of cognition. He received the 2005 Rumelhart Prize for contributions to formal cognitive science, and later papers examine neurocompositional computing, transformer symbol processing, and symbolic structures inside neural networks. As of October 2026, he is a Partner Researcher at Microsoft Research and a part-year Krieger-Eisenhower Professor of Cognitive Science at Johns Hopkins.

Works
290
Citations
18,278
h-index
47

Publication numbers from OpenAlex, as of 14 October 2026.

Photo of Paul Smolensky

Timeline

  1. 1976–1981

    Training in mathematical physics

    At Harvard, he earns an A.B. in physics in 1976. At Indiana University, he completes an M.S. in physics in 1977 and a Ph.D. in mathematical physics in 1981.

  2. 1986

    A group around neural networks

    At UCSD, he is one of the founding members of the Parallel Distributed Processing Research Group, with David Rumelhart, James McClelland and Geoffrey Hinton. His 1986 work shows how neural activation patterns can represent symbolic descriptions. He later teaches computer science at Colorado Boulder before joining Johns Hopkins.

  3. 1990s

    Scoring grammatical well-formedness

    With Géraldine Legendre, he develops Harmonic Grammar, a framework that gives linguistic representations numerical well-formedness values through neural computation.

  4. 1993

    A theory of universal grammar

    With Alan Prince, he develops Optimality Theory. Their manuscript circulates in 1993, and the theory becomes a major approach in phonology.

  5. Since 1994

    A long Hopkins appointment

    He joins Johns Hopkins’ Cognitive Science department in 1994 and later chairs it. From 2006, he holds the Krieger-Eisenhower Professorship.

  6. 2005

    An architecture and a prize

    The Cognitive Science Society awards him the David E. Rumelhart Prize for contributions to the formal analysis of cognition. His Integrated Connectionist/Symbolic architecture is developed in depth in The Harmonic Mind, a two-volume work with Géraldine Legendre.

  7. 2022–2024

    Extending neural computation

    In 2022, he and colleagues describe neurocompositional computing, which combines compositional structure and continuous neural computation. A 2024 paper with coauthors studies mechanisms for symbol processing during transformer in-context learning.

  8. As of October 2026

    A current research role

    An August 2026 preprint with R. Thomas McCoy, Paul Soulos and Tal Linzen examines symbolic structure inside artificial neural networks. As of October 2026, he is a Partner Researcher in Microsoft Research’s Deep Learning Group and a part-year Krieger-Eisenhower Professor of Cognitive Science at Johns Hopkins University.

Sources and credits

Photo: microsoft.com; Paul Smolensky at Microsoft Research; creator not stated. License: Not stated by source. Source: microsoft.com. Cropped to a square and resized. Ask us to remove or replace this photo.

  1. Paul Smolensky – Wikipedia de.wikipedia.org. Checked 14 October 2026.
  2. Paul Smolensky at Microsoft Research microsoft.com. Checked 14 October 2026.
  3. Rumelhart Prize - Cognitive Science Society cognitivesciencesociety.org. Checked 14 October 2026.
  4. [2205.01128] Neurocompositional computing: From the Central Paradox of Cognition to a new generation of AI systems arxiv.org. Checked 14 October 2026.
  5. [2410.17498] Mechanisms of Symbol Processing for In-Context Learning in Transformer Networks arxiv.org. Checked 14 October 2026.
  6. [2608.29530] The Emergent Symbolic Structure of Artificial Neural Networks arxiv.org. Checked 14 October 2026.