Inferno by Helena Sarin

The software engineer Sarin, formerly of Bell Labs, began her experiments with GANs in early 2018. This Twitter post from March 2018 shows an example of possibly her earliest public work. Just a year later, Sarin had already began standing out with work suggesting the surreal. By the end of 2018, Sarin spoke to Jason Bailey about her practice on his Artnome blog.

2019 saw Sarin’s first exhibition, which included this work, part of her #latentDoodles series. It was showcased at the Computer Vision Art Gallery during ICCV 2019 in Seoul, Korea. The work explores the human-like quality of hand-drawn marks with the distinct patterns of computational creation. Using SNGAN trained on her sketches and doodles, Sarin upscaled and colorized the image with cycleGAN for added texture. Sarin enhances the abstract images with carefully crafted captions, adding a verbal layer that deepens the folk-like yet computationally intricate nature of her work. The work exemplifies Sarin’s master from an early of coaxing the traditional and human out of modern AI techniques.