Master of Unexpected Melodies

Exploring the harmonies of music and machine learning

Stanton Stephens

Intentional and life-changing were the nine years that Eric P. Nichols (Ph.D. ’12, Cognitive Science) spent preparing his Ph.D. at Indiana University. Intentional because it was his goal to work under the direction of Pulitzer Prize-winning professor Douglas Hofstadter. Life-changing because those years have largely set the tone for almost every step of his career since.

That’s my realization as I talk with Eric at his Microsoft office in Redmond, Wash. Eric currently works as a principal science lead at Microsoft — a position that, as he describes it, focuses on “projects in audio understanding with machine learning.”

It’s late August, and Eric’s office building — one of many in Microsoft’s expansive campus — is almost half empty. It’s summer, after all, and Pacific Northwesterners cherish these few months of sunshine and balmy weather. Eric and I sit down in a small meeting room and begin to talk about his work and life. His voice is deep and confident, and his smile has all the warmth of a kind teacher.

“Doug was one of my heroes in undergrad, because I read his book Gödel, Escher, Bach,” he tells me. “I hadn’t thought of ever studying with him. I just loved that book.”

Eric was researching options for his Ph.D. when he met with Steve Larson, a professor at the University of Oregon and a friend of Hofstadter. When Larson realized Eric’s fascination with Hofstadter’s work, he offered Eric a startling realization: “Oh, you can just go be his student.”

Nichols and his wife, Andréia, relax in their Seattle-area home.

Even now, Eric feels that he owes it to Larson for connecting him with Hofstadter and guiding him to IU.

“That conversation with Steve was really huge,” he says.

So huge that Eric dedicated his dissertation, which is on music listening, to Larson and Helga Keller, Hofstadter’s administrative assistant. Keller helped arrange Eric’s first phone call with Hofstadter and, once he arrived in town, she helped him get settled in and feel at home in Bloomington.

And indeed, to Eric, Bloomington felt like home almost right away.

“The city was beautiful,” he tells me now, “and despite being a small town, it had everything I needed: an intellectual community, like-minded graduate student friends, and an incredible music department. I felt like I belonged there, like it was my natural habitat.”

But his nine years at IU weren’t the only factor that set the tone for his future career. Eric’s decision to take on summer internships in what he calls “the industry” was a game-changer, as well, though at the time, those pursuits were mostly financially driven.

“As a grad student, you run out of money,” he says, “especially when you’re there for nine years.”

His two summer internships — one with Google Research and another with Microsoft Research — were both related to music, and these pushed him to pursue a career path away from academia.

“I originally planned to go straight into academia after I finished my Ph.D.,” Eric says. “That’s why people do Ph.D.s usually, right?”

Music has been a central element of Nichols' life since he was a child, and much of his work with Microsoft and Google has dealt with machine learning as it applies to music recognition.

But academia wasn’t in the cards for Eric, at least not academia of the tenure-track kind. After finishing his Ph.D., he spent about four years with Google Research and YouTube in San Francisco, and then he moved onto his current job at Microsoft in September 2017.

For nearly a year since Eric returned to Microsoft, he’s focused on constructing audio models and studying machine learning. To cite just one example of his work, he tells me about a project recently pursued in partnership with the German elevator company, Thyssenkrupp, in which his team developed audio models to assess the efficiency of saws during their operation to cut metal.

“They wanted to know how often the saws are actually being used effectively, and how often they're sitting idle,” Eric says. “We actually realized we could detect that really easily just by listening to the saw spinning.”

Although it’s far from music, being part of developing machine learning and audio modeling excites Eric. This type of work is one of the reasons why he enjoys working in “the industry.”

Still, he hasn’t given up completely on academia. At the University of Washington, for instance, he’s created a 10-week graduate certificate course on deep learning, and he attends the International Society of Music Information Retrieval Conference (ISMIR) with regularity. Eric even met his wife Andréia through a joint friend during his attendance of ISMIR in Brazil.

“I want to keep the other path [in academia] open because I think of myself, at heart, as an academic first,” Eric says, pausing for a moment, “and maybe a musician second.”

Being identified as a musician goes back to the time when Eric worked with Hofstadter at IU. When meeting new colleagues or people around town, he recalls that he would often introduce himself as a student of computer science and cognitive science, but one day Hofstadter told him, “Eric, you’re a musician. You should start by telling people that.”

Eric’s passion for music goes back to his childhood. He learned to play the piano from his grandmother at age five and, in high school, he played the cello before switching to playing bass.

“I was last-chair cello,” he says. “The worst in the section. Our orchestra teacher asked us if anybody wanted to play the bass. I replied, ‘Sure, I'll play the bass.’ So, I went from last-chair cellist to first-chair bassist,” he adds with a laugh.

But in spite of this early start in music, Eric’s goal was never to become a professional musician. As much as the music itself, Eric valued the social aspect of music, as well — that opportunity to connect with friends and work as a uniform group.

“I want to keep the other path [in academia] open because I think of myself, at heart, as an academic first and maybe a musician second.”

Eric is the type of person who instinctively sees the multisided nature of things. With a bachelor’s degree in mathematics, his interests span mathematical finance and the stock market to music and creativity — and he sees a connection between these two worlds.

“There’s one aspect that connects these two seemingly unrelated fields,” he tells me. “Both involve events unfolding in time. Music listening involves pattern recognition and structure, whereas in stock analysis people often fall into the trap of falsely attributing structure to what are essentially random sequences. Nevertheless, perhaps the fact that both fields involve time makes them both interesting to me.”

Much of Eric’s research has to do with “melodic expectation,” where a listener hears a piece of music up until a certain point and then wonders what’s going to come next.

“Music plays with these expectations,” he says. “Sometimes, a melody will go the way you expect, and you’re content that it did what you expected. It feels natural. But other times it goes in a different direction, which is surprising, but interesting. This surprise can also be pleasing.”

In a way, Eric’s overall career seems similar to those unexpected melodies. He’s pursued paths that may have surprised him, but ultimately those paths have been pleasant and harmonious all the same.

“I realize there’s a sort of discrepancy between my desire to work on academic topics involving music and the reality of me working in the computing industry,” Eric says. “But I’m actually really enjoying my time in the industry, especially the machine-learning and deep-learning work I did both at Google previously and now here at Microsoft.”

“It’s been a really exciting time for me,” he says with a smile.

Rania Efthemes

Rania Efthemes is a Seattle-based, award-winning writer, content strategist, and editor with more than 20 years of print and digital-media experience. She reports and writes on various topics, including business, finance, money management, career, commercial and residential real estate, and banking. Efthemes is a recognized journalist and leader in digital media, and her work has won numerous editorial awards from the Society of Professional Journalists, the American Society of Business Publication Editors, the National Association of Real Estate Editors, and the Western Publishing Association. She holds a B.A. in English from Ain Shams University in Cairo, Egypt. Reach her rania.efthemes@gmail.com.