Dartmouth Events

Abstract

Throughout history, new tools—from the paintbrush to the camera—have transformed the possibilities of creative expression. Today, machine learning models capable of generating complex media offer tremendous power but also challenge our very notion of creativity. Will these tools devalue human creative work, or can they unlock unprecedented opportunities for human creators? In this talk, I will discuss an interdisciplinary approach that combines insights from applied machine learning, human-computer interaction, and application domains like music, multimedia, and writing to scaffold and accelerate creative work. I will showcase projects that (1) design new computational machinery for content creation, such as simulating room acoustics from photographs and synthesizing sounds from descriptions, (2) propose models for interactive systems that leverage computational tools for creative and technical writing, and (3) study mechanisms of integration between generative models and creative processes through eliciting mental models and observing patterns of behavior. This work offers a blueprint for fruitful future relationships between human creators and generative AI tools, which draws on the complementary advantages of each to achieve results neither could accomplish alone.

Bio

Nikhil Singh is a PhD candidate at the MIT Media Lab, working in the Opera of the Future group. His research asks how we can leverage computation to scaffold and accelerate creative work. To investigate this, he develops systems that synergistically combine methods from AI and machine learning with human interaction across modalities like sound, vision, and language. His contributions have appeared in selective computer science and interdisciplinary publication venues including TOCHI, ICCV, IUI, and PNAS. During his graduate studies, Nikhil was awarded the LEGO Papert fellowship and also conducted research at leading organizations, including the Allen Institute for AI and Netflix Research. As an artist and musician, he has co-created musical and multimedia works presented at venues like Moogfest, Mass MoCA, Festival Multiplicidade in Rio de Janeiro, the international space station, and NeurIPS' inaugural creative track. Nikhil holds a master’s degree from MIT and was previously an instructor at the Berklee College of Music in 2018, where he earned a bachelor's degree focusing on music composition, jazz, and computer music in 2017. While at Berklee, he received the Max Mathews and Earle Brown awards for computer music and composition.