Confluence is a generative art project by Devi Parikh on the BrainDrops platform. The series is part AI, part human-simulated art, and part human digital art.
Purpose and Background: The project aims to explore how AI tools can enhance human creative expression. Artist Devi Parikh's physical sketchbooks contain sketches, doodles, paper cutouts, mandalas, Zentangle drawings, and more, often with brief handwritten descriptions. The motivation stems from curiosity: observing how AI models consume sketchbook art that was not made for coherence or an audience; observing what form appears when AI models reproduce art—which features are recognizable as coming from the artist's sketches, which appear unfamiliar, and which are a fusion of both; and what opportunities the AI model's generations provide for the artist to reinject her personal voice.
Main Capabilities and Creative Process:
- Generate works using a neural generative model (GAN, generative adversarial network) trained on the artist's physical sketchbooks.
- Re-render the generations using image processing techniques.
- Recolor the works into one of 39 hand-designed palettes.
- Each work is reinterpreted by an AI model, similar to the Rorschach inkblot test or the phenomenon of pareidolia in AI models.
Work Features (five features in total per work):
- Palette: Indicates which palette the colors come from, with 39 values in total: Magritte, Birthday, Ethnic, Love, Activity, Cartoon, Neon, Surprise, Cotton-candy, Truth, Safe, Home, Wood, Mighty, Insomnia, Munch, Monet, Gold, Grown-up, Pop, Lichtenstein, Statement, Scheme, Meadow, Cake, Journal, Warhol, Gradient, Salient, Potential, Dali, Iconic, Gauguin, Icing, Deep, Authentic, Bright, Business, Gogh.
- Colorful: Indicates how many colors the work contains, with 4 values in total: Mono (monochrome), Di (two-color), Tri (three-color), Tetra (four-color).
- Marks: Indicates the number of black marks in the work, with 4 values in total: Minimal (very few), Sparse (sparse), Healthy (healthy), Many (many).
- AI-Pareidolia (AI pareidolia): indicates what the AI model sees in the work, with 52 values in total: Skateboard, Fish, Cake, Surfboard, Train, Skateboarder, Face, Woman, Cartoon, Vase, Flower, Scissors, Elephant, Fire-hydrant, Items, Dog, Tree, Kite, Striped, Flag, Collage, Shirt, Paint, Stuffed-animal, Graffiti, Air, People, Bird, Water, Rose, Table, Banana, Cow, Animal, Penguin, Checkered, Bear, Photo, Kitty, Horse, Toy, Sign, Painting, Airplane, Butterfly, Beach, Plant, Person, Girl, Wall, Cat, Man.
- Format (format): indicates whether the work is an image or a short video, with 2 values in total: png, mp4.
Among the 1000 works, the number of works with each feature is listed. In addition, several different methods for calculating the overall rarity score of the works are also provided.
Artist bio: Devi Parikh is an AI researcher and generative artist. She also makes Macrame, origami, and maintains a physical sketchbook containing cute doodles, Zentangle, mandalas, paper cutting, and sketches. Her generative art tends toward vibrant colors, geometric patterns, symmetry, and clear textures coexisting with organic shapes. Sources of inspiration are diverse - some ordinary, some exotic; some traditional, some unexpected; some concrete, some diffuse - including colors in rare stones, snakeskin textures, graffiti, wall art, digital illustration, acrylic pouring, mandalas, ethnic fabrics, looking out the window in a long taxi ride, Indian culture, or feedback from trusted friends. She maintains a growing list of ideas - from small adjustments to existing projects to entirely new projects or styles. Her process includes: starting from an idea, making a core prototype to see whether it feels right, and then iterating, iterating, iterating. The joy and mystery lie in seeing where the iterations lead. She gets energy from creative expression - partly through art (digital and analog), partly through AI research. In fact, the two intersect - part of her work is developing AI that can enhance human creativity and provide people with new tools for creative expression.
She was born in Tulsa, Oklahoma, USA, has lived in Riyadh, Saudi Arabia, grew up in Ahmedabad, India, and currently lives in San Francisco, USA. She has an extremely poor sense of space and direction, and an excellent sense of time. She hates small talk and loves deep connections with people. She has appeared in Forbes' list of 20 "Incredible Women Advancing AI Research."
Throughout her life she has dabbled in sketching, painting, origami, dance, and music - some of it quite good, the rest not so much. She first encountered AI in 2003 during her third year of university, and has been working in AI ever since. She was first introduced to generative art by a friend in 2018, when her day job offered fewer opportunities to write code by hand. Generative art satisfied the desire to write code and explore AI creativity.
More about Devi's art can be found at http://stateoftheheart.ai/ . She is also on Twitter, Instagram, Foundation, and fxhash.
