CS + Linguistics Major Gives Students a Front-Row Seat to the Future of AI

Mary Osborne isn’t afraid of AI in the classroom — quite the contrary.   

The lecturing fellow in Linguistics teaches the two cornerstone classes of Duke’s interdepartmental major (IDM) in Linguistics and Computer Science. She worked with Edna Andrews, Nancy & Jeffrey Marcus Professor of Linguistics & Cultural Anthropology, to shape the major, aiming to connect the theory and insights of linguistics with the problem-solving and innovation found in computer science. Launched in 2020, this is one of only three such majors in the country. 

Almost immediately after launch, the major faced a huge challenge as the coronavirus pandemic forced faculty to quickly move their classes online. A couple of years later, right as students and faculty were regaining a sense of normalcy, ChatGPT was rolled out. Where others would have seen another challenge, or even a threat, Osborne saw an opportunity.  

“Everybody else was saying, ‘ChatGPT is evil. It's terrible. It's going to ruin education,’” she said. “And then students come in my class and I'm telling them to sign up for a ChatGPT account.” 

Osborne brings a different perspective to the classroom: On top of her teaching role in the Linguistics department, she works full time at SAS in Cary — a software company that specializes in data analytics, from marketing to working with non-profits. Osborne has taken on many roles at SAS, and is currently the product manager for their natural language processing portfolio.  

Natural language processing helps computers understand and generate human language. This is the part of AI that powers chatbots, translation services and speech recognition like Siri or Alexa. It allows computers to not only recognize speech, but also determine sentiment, summarize text and produce human-like responses — just like ChatGPT does. Rather than fear it, Osborne wants students to make it better.  

She says there's a natural harmony between linguistics and computer science, which is an intersection she's long been drawn to, and that, when it comes to AI, using cross-disciplinary approaches helps boost innovation.  

“It’s important to understand that the early pioneers in these areas didn't lock themselves into their silos,” she said. “They weren’t just computer scientists, mathematicians or linguists, they were cross-disciplinary.” 

This perspective matches Duke’s emphasis on interdisciplinarity. “One of the most wonderful things about Duke’s curriculum is that we’re able to support students who are deeply interested in two different things,” said Andrews.  

Andrews said the key to having the major approved and seeing it succeed laid in finding someone able to meld linguistics with computer science. “We couldn't just have students take linguistics and computer science separately,” she said. “We needed someone who could combine the two subjects into our core courses.”  

That wasn’t an easy task.  

It was then-department chair Jun Yang, Knut Schmidt Nielsen Distinguished Professor of Computer Science, who brought Osborne aboard the new shared major.  

“Jun Yang sent her to me, I interviewed her, and I immediately agreed with him,” said Andrews. “I haven't looked back since.”  

And the feeling is mutual. 

Osborne said she never pictured herself teaching at Duke — a job she associated with having a Ph.D. But both departments told her that her industry experience and her expertise in the field were her Ph.D. 

“Mary is an extraordinary educator who brings to life computational approaches to natural language processing in a way that is both deeply technical and profoundly human-centered,” said Jian Pei, Arthur S. Pearse Distinguished Professor and current chair of Computer Science. “Her visionary, interdisciplinary teaching really enriches our program.” 

And teaching Duke students has become a great love for Osborne. “It feels great when students come back to me after they’ve gotten jobs and tell me how they’re using the things I taught them in the real world.” 

Osborne’s goal is to help students create systems that utilize a variety of tools in the natural language processing toolbox, including explicit, human-defined rules

 about language (like grammar and syntax), to process or analyze text, rather than solely relying on machine learning or statistical models. That’s where linguistics comes in. 

“The models we can build using linguistic rules can run 9,000 times faster than machine learning models, so there's a huge benefit, especially in the age of agentic AI,” she said. 

Because of that practical application, Osborne wants her classes to be accessible to people who aren’t necessarily interested in being engineers and doesn’t want students to avoid the class just because they're afraid they're not the best programmers. 

“There's a place for all types of people and all types of backgrounds,” she said. “There are so many people who are very interested in understanding how generative AI works, and that's a big piece of this class.” 

“Mary has so much energy, and is so very patient,” said Andrews. “She bends over backwards so that everybody understands what's happening.” 

Osborne said what she does at Duke and at SAS complement each other beautifully, and she hopes to keep doing both jobs for as long as she can.  

“I hope I’ll be one of the lucky ones who can retire from SAS, and then I’ll keep teaching, because it fills a bucket for me like nothing else ever has.”