I make paintings based on machine learning, and sometimes I write about A.I., art, and code.

Why A.I. Art

I've always been fascinated by machine learning, statistics, and AI. The first time I went out of my way to understand recommender systems I was in awe at how simply from aggregating the individual behavior of many users, an algorithm could find similarities and patterns that felt so human.

Machine learning has become a focus of my professional life in parallel to my developing interest in painting, so naturally I grew curious about the interplay between machine learning and art.

As I continue to study machine learning, it continues to fascinate me how the rate of technology progresses. Incresingly difficult problems suddenly become tractible in a matter of months. I'm excited by the prospect of new technological capabilities as they become available, but I'm also apprehensive. I wonder about how humanity will handle a new technology with the power to change large parts of the labor market through automation, and how we'll be able to trust what we see and hear as computers learn to speak in our voices and render videos of lifelike faces speaking.

Since showing my friends and family the math equations and code hasn't been the best way to share this mix of apprehension and wonder with the people in my life, the visual artifacts produced as a byproduct of debugging machine learning systems have proven to be accessible (and sometimes disturbing). The most fascinating aspect of this for me is how machine learning models are still inherently human. Models trained to recognize cats don't have any special feline-specific knowledge, they just managed to tap into a small part of humanity's collective knowledge by looking at a few hundred thousand photos of cats. When I see a model dreaming of cats, I appreciate the mathematical beauty inherent in the model as well as the fact that what I'm seeing is the "prototypical" cat, as defined by the psyches of the thousands of human beings who labeled the training data.

My Background

I studied Computer Science at UCLA and then started a company with some friends to make relevant an useful mobile ads, before joining Pinterest. Currently, I work full-time on building the Taste Graph, with the goal of building a deeper understanding of the particular tastes of hundreds of millions of users.

As a first-time startup CTO working 100-hour weeks, I found a relaxing hobby in teaching myself to paint. After days of trying to balance coding duties with hiring and managing an engineering team, the calm process of learning by trial and error was the only thing that took my mind off of code.