A Different Take on the Turtle Pond

Turtle Pond (2019), Tomas Aquino

In Turtle Pond (2019), a Caltech landmark gets a modernist makeover with a little help from artificial intelligence. Tomas Aquino, a graduate student in the lab of Fletcher Jones Professor of Psychology John O’Doherty, used a machine learning platform called PyTorch to create this piece of art. The system works by assessing the essentials of just one example of a particular style of art, after which it can transfer those characteristics onto another image.

Aquino and colleague Sanghyun Yi, a graduate student in social science, applied the essential look of the work of Brazilian artist Tarsila do Amaral to a photograph of Throop Pond, Caltech’s iconic water feature. The resulting image was displayed as part of the 2019 Caltech Art of Science exhibit. “I am originally from São Paulo, where Tarsila established herself as one of the most influential Brazilian artists ever,” Aquino says, “so her style was a natural choice for me.”

In a recent study, Aquino, Yi, and their fellow researchers in the O’Doherty laboratory used a similar system to demonstrate that artificial intelligence can not only determine the essential elements of a visual style—be it impressionism, realism, or abstract—but also can predict whether a person will like a particular painting based on their previous preferences.