Dr. Pablo Samuel Castro
Senior Research Software Developer, Google
Pablo was born and raised in Quito, Ecuador, and moved to Montreal after high school to study at McGill. He obtained his undergraduate, masters, and PhD at McGill, focusing on Reinforcement Learning. After his PhD Pablo did a 10-month postdoc in Paris before moving to Pittsburgh to join Google. He has worked at Google for almost 6 years, and is currently a research Software Developer in Google Brain in Montreal, focusing on fundamental Reinforcement Learning research, as well as Machine Learning and Music. Aside from his interest in coding/AI/math, Pablo is an active musician (https://www.psctrio.com), loves running (5 marathons so far, including Boston!), and discussing politics and activism.
Learning the Underlying Structure of Music | Tuesday, October 3 at 2:30 PM
Structure plays an essential role in the creative process of musicians, and artists in general. Whether it is the structure relating melody to harmony, the rigid rules of counterpoint, what renders a song “blues”, or what makes a rap lyric hip, musicians weave different structural threads into the tapestry of their art. But finding good, and lasting, structures is hard, often taking many iterations over multiple centuries. In a sense, as a musical species, we have been sifting through centuries of “data” trying to extract the underlying patterns that represent the type of music that works best.
Extracting patterns from very large datasets is exactly what machine learning (and deep learning in particular) are best at. At Google Brain, we are exploring different ways we can leverage deep learning to uncover some of the patterns in the music around us; this work ranges from learning to complete Bach chorales, to generate virtuoso-sounding piano performances, to producing interesting lyrics. Our research aims to gain a better scientific understanding for how humans produce and consume music, as well as to build intelligent tools to assist in the creative process. We want to empower musicians, not replace them!