"My art is scenic, and I always try to capture a mood," explains Sanchit Sawaria, a graphic designer from New Delhi, India. With a portfolio that screams with energy and vibrance, it's no wonder he refers to his work this way.
From uniquely angular (and functional) typography to textural identities and lucid, shiny 3D letters, Sanchit's work is experimental as it broadens our understanding of what can be achieved in the medium of design.
Having worked on a mix of projects for the likes of Bombay, Plenty and Stompy via &Walsh (he's recently gone freelance), it's his most recent personal endeavour that further exemplifies this fact. Named Gan Experiments, the designer first stumbled across a tool called GauGau by NVIDIA, which "is still in its beta stage," he says. "It fascinated me in terms of how it creates images with the help of machine learning." In this sense, there are two "neural networks" that GAN works with – one, a generator that makes the output, and the other, a "discriminator that compares the output with the actual image from the dataset." The final image, then, is made from the discriminator, and the viewer cannot tell which one is real or generated.
So in the context of these experiments, Sanchit started by building his own landscapes – inspired by the buildings and environments in GauGau. He says they ended up looking like "magic" and were "very realistic", and we couldn't agree more. The tool itself allows the user – in this case Sanchit – to draw or paint with colours, then change that colour to look like grass, sea, sky, rocks or a house and such.
"After spending some time with it, I tried drawing shapes that wouldn't exist in nature and then playing with the colour coding," he explains. "This led to some unexpectedly beautiful results. It was a way of breaking the discriminator neural network to a point where there is no point of comparison from a real image."
Gan Experiments is the type of project that makes you reassess our relationship with technology. But rather than freaking us all out about how much control a bot might have in the future, it shows the artistic wonders of what can be achieved through the careful balance between human and machine. "It sits somewhere between recent AI tools and manual making," he shares. "Gaugan is equal parts control and equal parts serendipity."
Having experimented with the Gaugan tool, Sanchit has now moved on to using Midjourney, an AI-powered tool that creates images from prompts. "I saw a lot of beautiful concept art and pop culture references being created with it," he notes, citing how he's been experimenting by creating fashion and jewellery with Midjourney alongside his work with &Walsh. There are many great things to come for this designer!