The Art of AI: An Interview with Christine McLeavey Payne

This piece marks the twenty-third installment in a series of dialogues with various explorers, thinkers, artists, and activists from across the globe. These individuals embody the core values of our founding principles.

Today, we highlight Christine McLeavey Payne, a member of the Technical Staff at OpenAI, where she has recently contributed to the development of MuseNet and Jukebox. Christine’s academic background includes studies in physics at Princeton and neuroscience and medicine at Stanford, alongside her training as a classical pianist and concert musician.

In our conversation, Christine and I delved into the current and future implications of AI on society and the workplace.

GP: You have dedicated recent years to employing AI algorithms for music creation. As a trained classical pianist, do you think these algorithms are developing their own unique styles? How can human musicians collaborate with these composing algorithms?

CMP: Tools like MuseNet and Jukebox are still in their infancy—akin to conservatory students learning to emulate the great composers of history. I find it fascinating when MuseNet proposes a fresh take on a Mozart piece or when Jukebox produces a sample reminiscent of Ella Fitzgerald. However, the real excitement will come when future models can create entirely new, emotionally resonant sounds and styles that move us in unexpected ways. I envision a moment where I am truly astonished by an original idea that I never anticipated. Just as we have witnessed superhuman performances in chess and Go, I ponder whether superhuman art could also emerge.

I’ve collaborated with several composers who have experimented with MuseNet. Currently, my preferred approach is to train the model on a composer’s previous works. Many composers express surprise when the model reflects their unique voice, creating compositions that feel familiar yet novel. They might initiate a piece, and MuseNet can propose various continuations, fostering a dynamic exchange of ideas. These models excel at generating a multitude of options quickly, making them ideal brainstorming partners.

GP: Do you believe machines possess authentic creativity? How long until we see AI-generated masterpieces gaining popularity? What are the implications for creative ownership? Should you, as the algorithm's creator, receive credit for a remarkable musical piece?

CMP: I wouldn't be shocked if we see an AI-generated song become a hit in the near future, largely due to the music industry's formulaic approach to crafting major pop hits. However, it may take longer for us to encounter an AI-generated masterpiece that is entirely innovative and surprising, rather than just a fusion of existing human styles.

The issue of attribution is complex. We must be cautious not to grant full copyright protection to AI compositions. Otherwise, it could lead to situations where someone uses a machine to generate countless songs and claims copyright over the entire collection. Nonetheless, the algorithm's creator should still receive recognition for the music produced.

At OpenAI, our music research is part of a broader mission. We’ve made MuseNet and Jukebox available for public use, requesting only that users acknowledge our contributions. However, if an independent developer dedicates years to creating an algorithm that produces a significant hit, we will need to ensure they are justly compensated.

GP: Has the current climate accelerated the development of educational AI? Will future learning primarily occur through dialogues with algorithms? Who determines the curriculum for these AIs, and why is this significant?

CMP: I am genuinely enthusiastic about the potential of educational AI—it could be one of the earliest ways for AI to positively impact society. During the pandemic, I found myself spending more time teaching my children. While we are fortunate to have excellent teachers in our school district, I’ve noticed how much more efficiently my kids learn with constant one-on-one attention. We can explore topics they are genuinely interested in—right now, it’s Minecraft math!

I am acutely aware of how privileged we are to have jobs that allow us to work from home during COVID, enabling us to dedicate time to our children. I often reflect on how AI could scale this individualized attention.

Imagine personalized AI tutors—supportive, patient, and tailored to each student’s skills and interests. Currently, when I want to acquire a new skill, I search for blogs, textbooks, or online courses. There’s usually something close but never perfectly aligned with my existing knowledge.

What if I could engage directly with a tutor that fills in my gaps in understanding? It could recognize my confusion or my readiness to advance, suggesting breaks or transforming lessons into engaging educational games. Picture a world where all children have access to such customized, enjoyable, and supportive learning experiences. (Of course, we still face significant challenges with access to technology, but I am hopeful this will improve over time.)

I do wonder, however, who will determine the curriculum these AIs teach. Societal debates are already contentious regarding educational standards—should we include creationism, anti-vaccination arguments, or focus solely on national history versus a global perspective? Our educational diversity helps safeguard against the homogenization of thought.

As we move toward AI-based tutoring systems, who will decide what knowledge is imparted? Will it be a small group of startup executives, or major corporations like Google or Facebook? Even if we are still a ways off from effective teaching AI, these discussions are essential to have now.

GP: In 1930, economist John Maynard Keynes predicted that his grandchildren would enjoy a 15-hour work week. Clearly, that hasn't happened. What do you envision as our most fulfilling future regarding automation? Could we finally achieve something akin to the 15-hour work week?

CMP: I love that concept! I hadn’t heard that prediction before. I frequently notice that my productivity peaks when I concentrate on completing three exceptional hours of work each day, rather than laboring for long hours. I would welcome AI to facilitate this by taking over monotonous tasks and helping us focus during our most productive moments.

However, a cultural shift will be necessary—especially in the U.S., where there’s a pervasive expectation of constant busyness. We also have significant economic issues to address. For instance, I’m unsure how we would convince companies to provide full benefits for a 15-hour work week; this would likely require societal changes.

Nonetheless, this is a fantastic goal. I don’t believe people will become unmotivated without the financial pressure to work excessively long hours. People are inherently driven to feel useful and respected.

GP: You and I were introduced through Renaissance Weekend, thanks to our mutual friend and Executive Director, Ali Gelles. What draws you and Steve back to this community?

CMP: The most compelling aspect of Renaissance Weekend is the strong sense of service. It’s a privilege to choose a career based on passion rather than necessity, and with that privilege comes the responsibility to engage in meaningful work.

I often leave RW gatherings contemplating new global challenges I had never considered, yet reassured by the presence of capable individuals dedicated to addressing these issues.

GP: Have the recent tumultuous events dampened your enthusiasm? Are there any evident silver linings in the changes we’ve experienced since March? How are these struggles manifesting for families versus single individuals?

CMP: This year has been nothing short of extraordinary! More than ever, it feels like we are divided by contrasting challenges.

In the workplace, there’s a clear distinction between single individuals and those with families. Singles often grapple with feelings of isolation but tend to be highly productive at work. In contrast, those with children may not experience loneliness, but are overwhelmed and fatigued. I empathize with teachers who understandably feel unsafe returning to in-person classrooms, as well as parents whose jobs cannot be done remotely and who need safe spaces for their children to learn. I also worry about my musician friends facing uncertainty regarding their future work.

The most significant silver lining for me has been the quality time spent with my children and witnessing the close bond they’ve developed as brothers. I’ve also enjoyed participating in small Zoom classes on various topics from around the world. For example, during the spring, I joined several Italian cooking classes with families globally, sharing a few hours cooking delicious pasta together. I used to struggle in the kitchen, but this time has bolstered my confidence. Recently, I’ve been practicing French and Spanish through online meetups with people eager to converse. My kids are even taking Capoeira Zoom classes from a teacher in a different time zone.

While I eagerly anticipate reconnecting with friends in person, I hope we can maintain the sense that long-distance relationships can feel just as intimate as those nearby. I cherish the new tradition of staying connected with friends across the globe through Zoom dinners, morning workouts, and virtual happy hours, and I aspire to continue this practice.

The first video provides insights into MuseNet, exploring how OpenAI and deep learning research are reshaping the music landscape.

The second video features Christine McLeavey discussing the breakthroughs and challenges in AI's role in music and education.