Brown CS News

Brown CS is mourning the loss of University Professor Emeritus of Computer Science and Cognitive Science Eugene Charniak, one of our founding faculty members. He passed away on June 13, just a few days after his seventy-seventh birthday. A pioneer in the field of natural language processing, Eugene joined the department in 1978, served as Department Chair from 1991 to 1997, was recognized with an endowed professorship in 2006, and retired last year. 

“I’m very sorry to hear of Eugene’s passing,” says An Wang Professor Emeritus of Computer Science John Savage. “He played a critical role in the department as our first AI hire, Chair, and distinguished colleague. He will be missed.”

“Eugene was an amazing mentor to me, and it has been an honor to serve in the department he helped found,” says Brown CS faculty member Stefanie Tellex. “As a researcher, he saw tremendous changes in his field, from his early work on symbolic AI, to his career in statistical language processing, to his world-renowned work in deep learning that achieved human-level performance on the Penn Treebank. I will never forget him telling me that he was first going to implement backward propagation in C++ to understand it, then learn to use the latest toolkit, and then his student DK's work using deep learning and dataset augmentation, to solve the Penn Treebank. He was always at the forefront of his field, and I learned so much from him. I am very sad to learn of his passing. ‘If I have seen further,’ Isaac Newton wrote in a 1675 letter to fellow scientist Robert Hooke, ‘it is by standing on the shoulders of giants.’ Eugene was truly a giant.”

Born in Chicago in 1946, Eugene initially pursued biology and genetics at the University of Chicago. After taking a programming course in FORTRAN, spending a summer with the Argonne National Laboratory’s high-energy physics group, and reading about Arthur Samuel’s now-legendary computer program that could play checkers against a computer opponent, he abandoned them in favor of artificial intelligence, opting for a PhD at Massachusetts Institute of Technology.

Studying under Marvin Minsky, one of the field’s pioneers, Eugene took an interest in language processing and met his future wife, Lynette. After graduating from MIT, Charniak took an offer at a research institute in Lugano, Switzerland. He was there for two years, then spent another two in Geneva, then held a one-year visiting position at Yale. Eugene came to Brown in 1978, co-founding the Department of Computer Science with Tom Doeppner, Steve Reiss, John Savage, Robert Sedgewick, Andy van Dam, and Peter Wegner. He received tenure in 1981, the year the couple’s first child was born.

Brown CS faculty member Amy Greenwald says that her husband went to University of Chicago as an undergraduate and credits an inspiring talk from Eugene as being part of the reason why he went on to pursue a PhD in CS. 

“As for myself,” she says, “I remember being interviewed by Eugene before coming to Brown. He had such a rich selection of books on his shelves, and we ended up talking about much more than just computer science, or even AI or NLP: cognitive science, philosophy of language, rationality, and so on. But in spite of how heady/intellectual he was, he taught me that not only my research, but even my teaching, should be motivated by applications; my impression was that he didn't think it was worth developing or lecturing about methods if I didn't have in mind a plan by which to put them to use. For years, he gave a guest lecture in my introductory AI class: the topic was (of course) statistical NLP, which he taught as a means to an end, namely machine translation.”

Starting around 1990, Eugene made a remarkable career turn, abandoning the heuristic programming tradition of AI research for a statistical approach. At the age of 47, it was a complete paradigm shift. 

University Professor of Computer Science Michael Littman met Eugene as a prospective PhD student in 1992. “I had been working on statistical methods for natural language processing,” he remembers, “and wondered if I could collaborate with him on that topic. He replied that he was not interested in statistical methods, only probabilistic methods, and educated me about the distinctions. I ended up joining the department a year later, by which time Eugene was, in fact, applying statistical methods to natural language processing. Moreover, he was teaching a class on the topic and was in the process of writing a textbook. (I took the class and ultimately got to co-author a paper with him about the term project.) When I came back to Brown as a faculty member in 2012, I saw Eugene do the same trick again – teaching a class and writing a book about the new ‘deep learning’ approach to AI. I very much admired his ability to throw himself into a new area and become one of the world experts in a matter of months. His energy and insight will be sorely missed.”

The third major phase of Eugene’s career came with his embrace of continuous representation, resulting in his publication of Introduction to Deep Learning in 2018. “I personally believe,” he once said of the continuous model, “it must be telling us quite a bit about how the brain works, to a better approximation than any other mathematical thing I can state….It’s the best idea we have for a theory of mind.”

Eugene’s other books included Computational Semantics, with Yorick Wilks (1976); Artificial Intelligence Programming, with Chris Riesbeck, Drew McDermott, and James Meehan (1980, 1987); Introduction to Artificial Intelligence, with Drew McDermott (1985); and Statistical Language Learning (1993). He was a Fellow of the Association for Computational Linguistics (ACL) and the American Association of Artificial Intelligence and was also a Councilor of the latter. In 2011, he received ACL’s Lifetime Achievement Award. 

“Since Eugene’s retirement,” says Thomas J. Watson, Jr. University Professor of Technology and Education and Professor of Computer Science Andy van Dam, “I enjoyed having regular lunch with him, mostly in my office with takeout from Kabob & Curry or Haruki, and (mostly) talking about hype and reality in AI. Recently, we had been talking about his latest book draft, a partially technical, partially first-hand historical review of AI progress and personalities since the start of the field. He remained an unflappable ‘show me the data’ kind of AI enthusiast, who felt the recent fears of what generative AI might do to our society were largely overblown, and that we were nowhere near achieving Artificial General Intelligence. His favorite example was how difficult it would continue to be for even the most sophisticated robot to become a decent brick layer. I pleaded with him to turn the last few chapters of the draft into a Scientific American-style, amply illustrated small book on how machine learning and large language models work, what transformers are all about, and other basic ideas behind generative AI that he’d patiently explain to me. I am still in shock about this sudden, devastating loss, and will miss his common sense and friendship more than I can express in a few words.”

Even in recent days, faced with repeated doomsday AI prophecies, Eugene described himself as an optimist. “Society will be transformed [by artificial intelligence],” he said, “but very gradually, and I think to the better.” 

This retrospective, written on the occasion of Eugene’s retirement, includes highlights of his career and personal life. Brown CS is currently creating a page that will compile tributes shared by Eugene’s colleagues and friends. To have yours added, please email us.