HENI and Damien Hirst – Generative Algorithm & Machine Learning
The Generative Algorithm
For over a year, Damien Hirst has been collaborating with the HENI data science team to develop a computer program that creates digital artworks in 25 different styles. The program, written in p5.js, a type of JavaScript, uses various techniques including WEBGL shaders, Perlin noise, and trigonometric functions to make paint layers that copy the planned but unpredictable movement of paint on a canvas.
Titles
Artwork titles were created using a modern transformer neural network model. The model was trained on data that was collected from a dependency parser. This parser interprets objects and forms description pairs from a large set of text. The model allows for a wide variety of new and unseen titles that follow the format of Hirst's existing 'Spin Paintings'.
Colour Names
The color names in The Beautiful Paintings generator were created using machine learning models. Over 30,000 names in two color spaces (CIELAB and RGB) were used to fine-tune a GPT model.