Image: University of Southampton
How a music researcher scooped a fellowship from the Alan Turing Institute
Many interdisciplinary artificial intelligence schemes are open to arts and humanities researchers, yet science, technology, engineering and maths applicants still dominate the grantee lists. Thomas Irvine (pictured), a musicologist at the University of Southampton, bucked this trend in 2019 when he won an Alan Turing Institute fellowship for his investigation of AI programs that compose jazz.
His project, Jazz as a Social Machine, started in October and will run until the end of January 2021, with Irvine working on it one day a week. The Turing institute supports several types of fellowship, some of them, like Irvine’s, via partnership arrangements with universities.
What sparked your proposal?
I was listening to some jazz composed by computers and thought: ‘Why is this so bad?’ I began talking to people in my field about music and data. The more I learned about how music composition algorithms worked, the more I thought there was a problem. There’s clearly a misunderstanding between people who understand jazz and those who write algorithms for it.
What is your aim?
My project is not about creating algorithms or new AI jazz—instead, I’m trying to comment on the problem and say why it’s not working. I saw the call for pilots and decided to take the risk.
Has it been a steep learning curve for you?
I knew as much about AI as the average layperson at the start, so I’ve learned all sorts over the past few months. Putting the proposal together forced me to think through someone else’s area, and that’s really hard. If you’re an established academic, usually that’s because you do one thing very well—I definitely needed to take a deep breath and trust myself on this.
What was your approach?
I started with a really simple book on AI—a book with no equations in it—and sat down one morning and read it cover to cover. I just needed to understand the subject. If you’re doing this, don’t be afraid to start at the beginning.
Did you enlist people to supplement your skill set?
Definitely. I could not have done this otherwise. One partner is at Academia Sinica in Taiwan, which has an AI music programme. I’m also working closely with Texas Tech University, where hundreds of future Texas school teachers in music are learning to improvise.
Who was in the core team when you drafted the bid?
I have a formal co-investigator who is a social scientist—we wrote the proposal together. I also collaborate with a senior colleague in music at Texas Tech, and there’s a research fellow working with us. She’s a composer and is creating artistic outputs to go with the project, including a site-specific installation.
Did you lean on them in writing the proposal?
Yes, the four of us worked closely on the bid in Google Docs. The funding proposal was straightforward—especially compared with research council applications. It reminded me of funding applications to the British Academy. It was a few pages with a one-page budget
Do you think you benefited from humanities applications being rare for this scheme?
Yes. If I were giving advice to my younger self, I’d ask myself why I didn’t apply to these things more often.
What else would you tell your younger self?
Be fearless about what you can contribute from your area of expertise. But at the same time, be modest about what you know and don’t know. Not trying to pretend I understood parts of these domains that I didn’t understand was a big part of my winning, I think. For these applications, where you’re unsure of the field, you shouldn’t promise something you are not sure you can deliver just because it’s what you think the funder wants to hear. Just focus on your contribution and be clear about what that is. In my case, I framed a lot of the application as: ‘I want to understand this better.’ In that sense, it was a very humanities-style application—it’s about understanding, whereas STEM is more often about explaining phenomena.
How have you found the experience so far?
From an arts and humanities perspective, I look at the AI landscape and think: ‘Wow, there are some really big opportunities here.’ It’s very exciting. But also, the grant has allowed me to set up a team—an unusual experience for someone in my field. It’s been eye-opening and positive for me to work that way. It’s also a considerably less lonely experience.
Have you felt out of your depth at all during the project?
Oh yes. I was at a lecture recently where a gentleman gave a presentation on maths problems that can arise in machine learning, and the slides in the lecture were impenetrable to me. I asked a question and felt like I’d been exposed as not getting it. But later I turned to a colleague who is a fairly well-known statistician and he said: “Don’t worry, I don’t get it either.”
This is an extract from an article in Research Professional’s Funding Insight service. To subscribe contact sales@researchresearch.com.