Grants support new research on role of AI in the arts

Binghamton University has announced the latest recipients of the Provost Awards for Research Grants, providing seed funding to explore the intersection of artificial intelligence and the creative arts. The initiative supports four interdisciplinary projects aimed at understanding how machine learning affects artistic agency, musical improvisation, and literary interpretation. This research is significant for the AI sector as it addresses ethical integration and the technical limitations of generative models in understanding complex human social knowledge and creativity.

The university has allocated funding to projects that position AI as a collaborative partner rather than a replacement for human creators. One project, led by James Budinich and Gregory Evans with $12,530 in funding, utilizes sensory percussion and machine learning to create a generative system that responds to live musical performances in real time. This research explores how AI can change musical identity and documentation through live improvisation. Similarly, a $37,500 grant was awarded to Magdalena Bermudez and Jason Bernagozzi to develop a responsive machine learning system for cinema, focusing on maintaining artist agency while integrating historical and contemporary cinematic technologies with modern AI tools.

Another key area of investigation involves the limitations of computer vision and the concept of "irreducibly human" art. A collaborative team including Christopher Swift, Alexandros Skouras, Ruth Carpenter, and Gregory Hallenbeck received $32,000 to study how algorithmic interpretations handle various physical artworks. By using printmaking techniques to create images that are unclassifiable or incomprehensible to machines, the researchers aim to identify the specific boundaries between human perception and machine vision. This project specifically tests what types of visual information trip up current AI algorithms, providing insights into the technical gaps of automated detection systems.

The research also extends into the linguistic and social capabilities of large language models through a $17,500 project led by Junting Huang, Sujoy Sikdar, and William Hayes. This team is designing a Retrieval-Augmented Generation (RAG) benchmark to test how generative AI interprets literary fiction and social knowledge. By testing multiple RAG configurations against a curated set of 30 global literary works and 150 annotated question-answer pairs, the study seeks to determine if AI can grasp the moral conundrums and empathy-building aspects of literature. These projects, funded over an 18-month period through a New York state matching fund, represent a concerted effort to define the ethical and functional role of AI within the humanities.