AI Music Detection on Unseen Generators
A detector with 0.975 balanced accuracy on the generators it was trained on catches only about half of the clips from a generator it has never seen.

M.S. Data Analytics, University of Central Florida · Applying to Ph.D. programs for Fall 2027
I come to machine learning from music: a B.A. in electronic music at Bard College, an M.A. in music business at NYU, and work in artist management, music publishing and platform operations. Now I study how to tell where a piece of AI-generated music came from.
Was this piece made by AI?
Whose style does it imitate?
Does it copy existing works?
A detector with 0.975 balanced accuracy on the generators it was trained on catches only about half of the clips from a generator it has never seen.
Real Beethoven bars spliced from different works pass as Beethoven 83–92% of the time. A detector that reads the joins between bars separates them from real music.
M.S. Data Analytics (expected May 2027)
M.A. Music Business
B.A. Electronic Music