BYU lab works to improve artificial intelligence communication

335
(From left)  Berkeley Andrus, Nathaniel Robinson, Jay Cui, Ben Cullen, Nathaniel Carlson, Professor Nancy Fulda, Hazar Handal and Nathan Tibbetts are all members of the DRAGN Labs research team. The research team specializes in language communication within artificial intelligence. (Allison McArthur)

BYU research group DRAGN Labs is making big progress in its efforts to better artificial intelligence. 

DRAGN is an acronym for Deep Representations and Architectures for Generative systems and Natural language understanding. The team’s projects focus on artificial intelligence applications, research and language processing.

BYU professor Nancy Fulda started DRAGN Labs in August 2019 with only a handful of students. Since then the team has seen progress during its research of conversational artificial intelligence.

Computer science graduate student Berkeley Andrus and undergraduate applied and computational mathematics student Nate Robinson have worked under Fulda since the beginning of DRAGN Labs.

“We’re trying to make computers better at understanding what people say and write, then also be able to speak and write back to us,” Andrus said. 

DRAGN Lab students meet either in their teams or as an entire group once a week. Andrus said he sees many different backgrounds in the lab, with students majoring in math, computer science and even genetics, and values being able to know who has what specialty and collaborate with them.

A notable project during his time at DRAGN labs was working on natural language understanding. He focused on how video games figured out user speech (what people were saying as they played) and how the computer could respond.

The DRAGN Labs logo was inspired by Fulda’s love of science fiction and dragons. (Nancy Fulda)

The biggest project from Robinson’s team has been creating a new algorithm to control biased language generated by artificial intelligence programs. These programs can generate huge amounts of text and have human-level fluency. 

“A lot of the time, these programs’ text is biased or just talks about whatever it wants to, so we created a new algorithm to control what it can or cannot say,” he said.

Robinson is currently working on a machine translation project that explores different methods and combinations of un-studied languages.

“Sometimes when you finish a project, the end product is really cool to sit back and look at. I think to myself, ‘I made this and nobody understands it better than I do.’ Some projects take over a year, so it’s really satisfying to see the final reward,” Robinson said.  

Andrus said he wishes he could tell students who might be interested in the data or computer science field that BYU is a great environment for trying lots of things.

When starting his major, he said it was difficult to see how research success manifested differently than it might for other career fields. “It takes a lot of time, but it’s really fulfilling.”

Print Friendly, PDF & Email