Devi Parikh and Abhishek Das

Artist

Devi Parikh and Abhishek Das · stateoftheheart.ai ↗

Practice

Generative art combining algorithmic pattern, AI research, photography, and simulations of physical texture

In the set

Cushions ↗ · 200 editions
Aragnation ↗ · 128 editions

Links

Full bibliography ↗

Devi Parikh and Abhishek Das approach generative art as both a visual practice and a form of inquiry. Virginia Tech records Parikh’s work in artificial intelligence and computer vision, while Das has worked across embodied AI, interpretable machine learning, and computational catalyst discovery.¹² Their shared history extends beyond a single release: the artists report collaborating on dozens of projects since 2015, with Das introducing Parikh to generative art in 2018.³

This background matters because their art does not treat computation as a neutral production tool. It asks what kinds of decisions can be delegated to a system, what remains under artistic direction, and how technical constraints shape aesthetic character. Parikh’s research into interactive generative art found that preferences within a generative form could predict other preferences within that form, suggesting that parameter choices can carry meaningful internal relationships.⁴ Her related work on neuro-symbolic generative art also distinguishes neural generation from symbolic systems designed through artist-authored rules.⁵ The pair’s Art Blocks practice occupies the productive space between these approaches.

Softness through structure

Cushions is Parikh’s long-form generative project of 200 editions. Its central achievement is perceptual: crisp geometry is made to read as padded, folded, or gently inflated. The project’s development account emphasizes repeated experimentation with composition, balance, curves, line thickness, incomplete forms, and color, all directed toward preserving a distinctly soft, cushion-like quality.⁶ A related interactive tool exposes several of these variables, allowing users to alter the grid, curvature, balance, thickness, and completeness of a tiled pattern.⁷

The resulting works turn systematic repetition into an image of touch. Cells align, divide, and mirror, but their curved boundaries resist the severity usually associated with grids. Bright palettes and small disruptions keep the compositions from settling into pure ornament. Variation operates inside a recognizable visual grammar, making the collection legible as a family without reducing its outputs to minor permutations.

Live on Art Blocks: cushions-by-devi-parikh ↗

An AI vocabulary beyond the pixel

Released by Art Blocks on February 6, 2023, Aragnation comprises 128 editions.⁸ The collaboration explicitly challenges the assumption that AI art requires enormous neural networks or must imitate photographs, paintings, and collage. Instead, it uses compact models to construct abstract images from blobs, gradients, and organic textures.³

Its mechanics are unusually transparent. One probabilistic graphical model composes 25 learned prototype shapes and 17 learned colors into interpretations of landscapes, flowers, birds, and urban settings. A second model, a small multilayer perceptron, maps image coordinates to RGB values and controls color saturation in some outputs. The training material includes photographs taken by Das, connecting the generated compositions to his views of nature, landscapes, and cities.³ Paper styles, palettes, textures, reflections, and model-assigned subjects expand the system’s visual range without obscuring its underlying economy.

The title itself is an anagram of “No-GAN AI Art,” using GAN as shorthand for a prominent family of image-generation techniques that the project declines.³ That refusal is not anti-technical. It is an argument for technical plurality, and for AI art whose conceptual stakes can be inspected alongside its surfaces. Research co-authored by Parikh has likewise proposed generative artworks as accessible instruments for discussing AI ethics and the perspectives of different stakeholders.⁹ Here, such accessibility arrives through pleasure: translucent color, simulated paper, compact forms, and recognizable hints of place.

Live on Art Blocks: aragnation-by-devi-parikh-and-abhishek-das ↗

Creativity as a negotiated system

Across both collections, the artists favor systems whose constraints remain aesthetically visible. Cushions translates a parameter space into softness; Aragnation translates learned photographic tendencies into an abstract vocabulary. Parikh has discussed AI and creativity in terms of tools that can support preference prediction, visual journaling, and human creative agency rather than simply automate finished images.¹⁰ The Gradient’s extended conversation with her similarly situates generative art within a wider practice of human-AI collaboration.¹¹

These works therefore resist a simple opposition between hand and machine. Their authorship lies in selecting representations, defining permissible variation, training or structuring models, and judging whether a system consistently produces compelling differences. For Parikh and Das, computation becomes most expressive when its limits are made specific.

Bibliography

The references cited in this essay are listed below. For the complete bibliography see the Devi Parikh and Abhishek Das links page ↗.