The history of Artificial Intelligence in art goes back further than many might imagine - including myself. AI art can trace it roots to the 1960s and even all the way back to visions of mechanized creativity in ancient egypt. Early AI art experiments, sparked by the broader fascination with artificial creativity, set the stage for the digital art boom we see today.

The idea of machines capable of creating art isn’t new. Early visionaries like the ancient Greeks and figures such as Daedalus and Hero of Alexandria imagined mechanical creativity, and inventors throughout history aimed to make these dreams a reality. 

The 20th Century: Foundational AI Art Machines

The mid-20th century brought a surge of machine-driven art experimentation, starting with artists and inventors like Gordon Pasks, who created the MusiColour machine that responded to musical inputs with light displays. 

Jean Tinguely followed with kinetic sculptures called "painting machines," where users selected colors, and the machine created unique art. These inventions laid the groundwork for AI art's evolution.

The 1960s and 70s marked the beginning of modern AI art with Harold Cohen’s AARON. Created to explore machine drawing, AARON produced endless abstract images using coded patterns. Exhibited in 1972, AARON initially made simple black-and-white drawings, but Cohen spent decades enhancing its capabilities, allowing it to paint in color and even make brush selections independently. Although AARON could only work within Cohen’s unique style, it demonstrated the potential for AI-driven creativity.

In 2014, Ian Goodfellow and colleagues at Université de Montréal created a type of deep neural network that could learn to mimic the statistical distribution of input data including images. Named generative adversarial network (GAN), it comprised a 'generator' to create new images and a 'discriminator' to decide which of these images was successful. This represented a major evolution from algorithmic art because the network could learn to create any given style by analysing a big enough dataset.

A major contributor to the interest in GAN models was Google's DeepDream. Emerging in 2015, it used a convolutional neural network to identify and enhance patterns in images using algorithmic pareidolia, resulting in images with a dream-like look.

As coding became more accessible in the 2000s, artists began exploring generative art in new ways. Karl Sims used AI in the early 90's to create evolutionary animations, winning awards and inspiring others. Scott Draves’s project, *Electric Sheep*, allowed users to influence evolving fractal animations. With the creation of public databases like ImageNet, artists and developers could train AI models to catalog visual data, leading to advanced machine-learning techniques that power today’s AI art generators.

In 2014, the introduction of Generative Adversarial Networks (GANs) represented a pivotal step. Developed by Ian Goodfellow, GANs consist of a generator creating images and a discriminator evaluating them. GANs enabled the recreation of specific artistic styles based on dataset analysis. 

For example, the *Next Rembrandt* project used GANs to mimic Rembrandt’s style, generating a new painting as if by the artist himself. GANs also led to *DeepDream* by Google, creating surreal, dreamlike visuals.

'Electric Sheep' by Scott Draves on YouTube

Medium / @Dutch Digital, 'Bringing Back the Old Master'

'The Next Rembrandt' website

'The Next Rembrandt' on YouTube

'GANs' definition via Google

'Harold Cohen AARON Exhibit' at Whitney Museum on YouTube

Mark Derho is a seasoned expert in the Internet industry with over 25 years of experience in NYC's software development, digital marketing, and advertising sectors. A certified Google Partner, Mark specializes in content creation, AI chatbot development, open-source software, modern website design, and SEO/SEM marketing. He leads PR Website Agency and lives in Puerto Rico with his dog, Luno.