Boris Katz
Boris Katz is a principal research scientist at MIT CSAIL and the creator of START, an early Web question-answering system. As a Moscow University undergraduate in the 1960s, he taught a computer to read and solve math problems and arrived in Boston from the Soviet Union in 1978. START answered reporters' questions from the JPL press room during the 1989 Voyager 2 Neptune encounter and became the first question-answering system on the Web in 1993. In 2006, his InfoLab technology inspired Siri; by 2011, StartMobile enabled natural-language phone interaction and ideas from START contributed to IBM's Watson. His 2020 work explored grounded language understanding, and his 2025 paper examined guided network learning; as of October 2026, he heads InfoLab and works with PhD student Christopher Wang on language and the brain.
Timeline
-
1960s
A program for math problems
As an undergraduate at Moscow University in the 1960s, he teaches a computer to read and solve math problems.
-
1978
Leaving the Soviet Union
He arrives at Logan International Airport in Boston with his family in 1978 as a Soviet emigrant.
-
1989–1993
START from Voyager to Web
During the Voyager 2 Neptune encounter in 1989, START answers reporters' questions from the JPL press room. In 1993, it becomes the first question-answering system on the Web. Katz is its creator.
-
2006
An inspiration for Siri
In 2006, technology from his InfoLab group becomes a major inspiration for Apple's Siri.
-
2011
StartMobile and Watson
By 2011, StartMobile lets mobile phone users interact with their devices through natural language. Ideas from START also enter IBM's Watson, which defeats Jeopardy! champions in 2011.
-
2020
Compositional networks for language
In 2020, he coauthors “Compositional Networks Enable Systematic Generalization for Grounded Language Understanding,” which helps agents generalize language understanding to commands grounded in vision.
-
2025
Guided learning for neural networks
In 2025, he coauthors “Training the Untrainable: Introducing Inductive Bias via Representational Alignment,” which shows that brief guidance can improve neural networks previously thought untrainable.
-
As of October 2026
MIT CSAIL InfoLab leadership
As of October 2026, he is a principal research scientist and head of MIT CSAIL's InfoLab. He works with PhD student Christopher Wang on models and datasets that map language through the brain.
Career
- Since 2000MIT CSAILPrincipal Research Scientist
- 1987–2026Massachusetts Institute of Technology
Education
- Lomonosov Moscow State UniversityUndergraduate
Most-cited works
- ObjectNet: A large-scale bias-controlled dataset for pushing the limits of object recognition modelsWith Joshua B. Tenenbaum
- Quantitative evaluation of passage retrieval algorithms for question answeringWith Jimmy Lin
- Learning Physical Descriptions From Functional Definitions, Examples, And PrecedentsWith Patrick H. Winston and Thomas O. Binford
- Annotating the world wide web using natural language
- Omnibase: Uniform Access to Heterogeneous Data for Question AnsweringWith Jimmy Lin
Sources and credits
Photo: We have no published photo of Boris Katz yet. Send us a free-licensed photo.
- The man who helped invent virtual assistants thinks they’re doomed without a new AI approach
- Katz Family Re-United After 3 Years | The Harvard Crimson
- Boris Katz's Home Page
- CSAIL's Boris Katz On NPR | MIT CSAIL
- [2008.02742] Compositional Networks Enable Systematic Generalization for Grounded Language Understanding
- Guided learning lets “untrainable” neural networks realize their potential | MIT News | Massachusetts Institute of Technology
- Christopher Wang | CSAIL Alliances



