Gradation Descent

© private Nirav Beni is a South African-born engineer and new media artist. He has a BSc degree in mechatronics engineering from the University of Cape Town, and is currently completing an MA in information experience design at the Royal College of Art in London. He positions himself at the intersection of art and technology, where he can incorporate AI, machine learning and other new media into his practice. He aims to create audio-visual, immersive and interactive installations that focus on shared and evocative experiences that touch on themes such as the human-machine dynamic.

The work “Gradation Descent” by Nirav Beni demonstrates a good sense of visual tempo. By adding a rhythmic sensation to the machine-“reading” procedure, it creates an interesting projection of the gaze of a machine – although it doesn’t entirely imitate how a machine would work in real life. Step by step, Alexandre Cabanel’s “Fallen Angel” is algorithmically reworked; and in the checkerboard-like texture the gaze leaves behind, we see the discrete mechanisms of the algorithms that can never analyze and synthesize an image as a whole but can only do so in separate parts.

This work is a short audio-visual video that depicts a conceptual imitation or visual reinterpretation of machine-learning, image-processing algorithms acting on Alexandre Cabanel’s painting “Fallen Angel”.

The moving images creating the video come from the Generative Adversarial Network (GAN). As GAN is trained on “real” images, it produces its own “fake,” almost identical, but slightly inaccurate versions. The algorithm slides across the original image and processes each block of the original image, leaving in its place an AI-generated frame. In this way, it creates a “fake” collage as the output image, which emphasizes similarities and discrepancies between original and replacement. The sliding motion is triggered by audio signals from a melancholic piano playing a piece expressly produced to accompany the teary-eyed visuals of the fallen angel.