{"id":841,"date":"2026-10-01T10:13:19","date_gmt":"2026-10-01T10:13:19","guid":{"rendered":"https:\/\/gedespi.co\/1\/?p=841"},"modified":"2026-10-01T10:13:19","modified_gmt":"2026-10-01T10:13:19","slug":"a-brief-history-of-ai-art-from-algorithms-to-masterpieces","status":"publish","type":"post","link":"https:\/\/gedespi.co\/1\/a-brief-history-of-ai-art-from-algorithms-to-masterpieces\/","title":{"rendered":"A Brief History of AI Art: From Algorithms to Masterpieces"},"content":{"rendered":"\r\n

The history of AI art does not begin with a chat window and a typed prompt. It begins with programmers who treated the computer as a drawing instrument decades before anyone settled on a name for machine creativity. The line from those first experiments to a robot portrait of Alan Turing<\/a> selling at auction is shorter than most people assume, and the arguments it raised are still being had.<\/p>\r\n

What follows is a walk through the main milestones: the code-based experiments of the 1960s, Harold Cohen’s AARON, the systems that learned from enormous collections of existing paintings, the text-to-image tools<\/a> that arrived in 2022, and the auction and gallery moments that pushed AI art into public argument.<\/p>\r\n

Code as a drawing tool in the 1960s<\/h2>\r\n

Code-based art has been around since the 1960s, when early adopters first programmed computers to draw wonky shapes and abstract forms. The results were often simple by modern standards: lines, repetitions and geometric patterns produced by rules the artist had written.<\/p>\r\n

What mattered was the arrangement. The human set the conditions, the machine produced variations, and the finished piece emerged from the interaction between the two. That basic split, human intent plus machine execution<\/a>, still describes most AI art made today, even when the machine is doing far more of the visual heavy lifting.<\/p>\r\n

Harold Cohen and AARON in 1973<\/h2>\r\n

The commonly cited starting point for AI art as a deliberate practice<\/a> is 1973, when a computer scientist and artist named Harold Cohen created the first known AI art creation program. The system became known as AARON, and Cohen is the figure most often credited with inventing AI art.<\/p>\r\n

His importance lies less in any single image than in the idea he committed to: that art-making could be described as a process and then encoded. Once drawing could be expressed as instructions, the question stopped being whether a machine could make marks at all and became a question about authorship, intention and judgement.<\/p>\r\n

Teaching a machine to read art history<\/h2>\r\n

By January 2021, the field had moved well beyond making marks. Led by Dr Ahmed Elgammal, an AI designed by a group of programmers had already sorted 77,000 Wikiart images of paintings into historical order. The work was discussed as a kind of brushstroke biography, in which the history of a style can be reconstructed from the evidence inside the images themselves.<\/p>\r\n

The purpose was categorisation rather than creation, and the argument behind it was practical. Frameworks can justify choices and give a level of transparency that the art world often lacks, which matters when decisions about who gets shown, funded and reviewed are being made. Not everyone welcomes that. Many artists and curators regard systems that categorise art as restricting freedom, and that tension between useful structure and imposed order runs through the whole debate.<\/p>\r\n

\"art\r\n
Photo by Bjorn Pierre<\/a> on Pexels<\/a><\/figcaption>\r\n<\/figure>\r\n

GANs and the move away from hand-written rules<\/h2>\r\n

Ian Goodfellow’s GANs marked another shift in how these systems were built. Rather than following rules that a programmer wrote by hand, later systems learned from large collections of existing images, absorbing patterns of colour, composition and style from the data they were given.<\/p>\r\n

This is the point at which AI art stops looking like a set of instructions and starts looking like a set of influences. It is also where the most persistent criticism begins, because a system trained on other people’s pictures carries something of those pictures inside it.<\/p>\r\n

2022: prompts replace programming<\/h2>\r\n

AI image generators<\/a> are text-to-image tools designed to turn written prompts into pictures, and by 2022 they had become accessible enough to change who could make them. Tools such as DALL\u00b7E, Midjourney and Stable Diffusion brought generative images into ordinary conversations, and the volume of output rose sharply.<\/p>\r\n

Commentary from September 2022 identified the awkward core of the moment. As it was put at the time, this AI art could not exist if it had not been trained on copyrighted images and the work of artists. Both things are true at once: the results can be striking, and the means of producing them depend on material that someone else made.<\/p>\r\n

Ai-Da, Turing and the Sotheby’s sale<\/h2>\r\n

In November 2024, a painting created by a robot called Ai-Da, depicting the pioneering computer scientist Alan Turing, sold at Sotheby’s for more than \u00a31m. Whatever anyone thinks of the picture itself, the sale moved AI art out of online arguments and into the commercial machinery of the art market.<\/p>\r\n

The London Art Fair responded in kind. On Friday 24 January 2025, a panel of prominent art critics, journalists and authors gathered in the Talks Theatre at the Business Design Centre in Islington to debate the rise of AI and digital art against traditional media such as painting and sculpture, framed by the question of whether AI is the new avant-garde.<\/p>\r\n

\"computer\r\n
Photo by Al Nahian<\/a> on Pexels<\/a><\/figcaption>\r\n<\/figure>\r\n

Audiences, values and rights<\/h2>\r\n

The conversation has also turned towards the people looking at the work. An April 2025 exhibition, titled A.I. in Art: Perceptions, Values, & Rights and listed by Georgetown University Library, invited viewers to contribute a third-party perspective by sharing their reflections on their own experience of AI involvement in art.<\/p>\r\n

That is a meaningful change of emphasis. Instead of asking only whether machines can be creative, the question becomes what audiences value, what they are willing to accept, and which rights artists and viewers feel are being affected.<\/p>\r\n

A short timeline of AI art<\/h2>\r\n\r\n
\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Period<\/th>\r\nDevelopment<\/th>\r\n<\/tr>\r\n<\/thead>\r\n
1960s<\/td>\r\nEarly adopters program computers to draw abstract shapes and forms<\/td>\r\n<\/tr>\r\n
1973<\/td>\r\nHarold Cohen creates the first known AI art creation program, later known as AARON<\/td>\r\n<\/tr>\r\n
2021<\/td>\r\nAn AI led by Dr Ahmed Elgammal sorts 77,000 Wikiart images into historical order<\/td>\r\n<\/tr>\r\n
2022<\/td>\r\nText-to-image generators such as DALL\u00b7E, Midjourney and Stable Diffusion reach a wide audience<\/td>\r\n<\/tr>\r\n
October 2023<\/td>\r\nGenerative art described by GQ as the next frontier for the NFT market<\/td>\r\n<\/tr>\r\n
November 2024<\/td>\r\nAi-Da’s portrait of Alan Turing sells at Sotheby’s for more than \u00a31m<\/td>\r\n<\/tr>\r\n
January 2025<\/td>\r\nLondon Art Fair panel debates whether AI is the new avant-garde<\/td>\r\n<\/tr>\r\n
April 2025<\/td>\r\nExhibition on AI in art invites viewers to share their own reflections<\/td>\r\n<\/tr>\r\n<\/tbody>\r\n<\/table>\r\n<\/figure>\r\n\r\n
\"robot\r\n
Photo by Freek Wolsink<\/a> on Pexels<\/a><\/figcaption>\r\n<\/figure>\r\n

What older technologies can tell us<\/h2>\r\n

Machines have taken over manual labour before, and the reaction has followed a familiar pattern. Spinning jennies, steam engines and washing machines were all controversial technologies in their time, and each of them displaced work that people had previously done by hand.<\/p>\r\n

Nesta’s prediction series framed the artistic version of this shift as a partnership rather than a replacement, under the heading Artist + AI = Creativity\u00b2: Humans and machines will create prize-winning art. Whether that reads as optimism or a warning depends on where you sit, but it is a fair description of how the practice has developed. The artists making interesting work with these systems are not choosing between the machine and their own judgement; they are using one to extend the other.<\/p>\r\n

Where the history leaves us<\/h2>\r\n

The story runs from wonky shapes drawn by code in the 1960s to a seven-figure auction result in 2024, and it is not finished. Generative art<\/a> has already had a commercial moment in the NFT market, which GQ described in October 2023 as its next frontier, and each new generation of tools resets the argument about training data and consent.<\/p>\r\n

What remains constant is the human decision about what is worth making. Every stage of the history of AI art has depended on someone choosing to point the technology at something. The algorithms supply variation and speed, while taste, purpose and meaning still come from the person at the keyboard. The AI-assisted pieces collected on this blog sit within that same line of experimentation, and the same questions follow them into the gallery.<\/p>\r\n

Frequently Asked Questions<\/h2>\r\n

When did AI art begin?<\/h3>\r\n

AI art has two starting points worth knowing. Code-based art dates to the 1960s, when programmers first used computers to draw abstract shapes and forms. The first known AI art creation program arrived in 1973, when Harold Cohen built the system known as AARON. Most histories treat that later date as the moment AI art became a deliberate practice rather than a technical curiosity.<\/p>\r\n

Who invented AI art?<\/h3>\r\n

Harold Cohen, a computer scientist who was also an artist, is usually credited. In 1973 he created the first known AI art creation program, which became known as AARON. Ian Goodfellow’s GANs later changed how these systems were built, and the text-to-image generators of 2022 broadened who could make AI images. No single person invented the whole field.<\/p>\r\n

Why do some artists object to AI art?<\/h3>\r\n

The main objection concerns training data. As commentators noted in September 2022, this kind of AI art could not exist if the systems had not been trained on copyrighted images and the work of artists. Alongside that, many artists and curators regard systems that categorise art as restricting freedom, and worry about how such frameworks shape who receives opportunities.<\/p>\r\n

Can you tell whether a picture was made by AI?<\/h3>\r\n

Often, no, not reliably from the image alone. AI image generators are text-to-image tools that turn written prompts into pictures, and their output can imitate traditional media closely. Public debate has shifted towards audiences instead, with a 2025 exhibition inviting viewers to share their own reflections on AI involvement in art rather than relying on technical detection.<\/p>\r\n

How much did Ai-Da’s painting sell for?<\/h3>\r\n

In November 2024, a painting created by a robot called Ai-Da, depicting the pioneering computer scientist Alan Turing, sold at Sotheby’s for more than \u00a31m. The result was widely discussed because it placed machine-made work inside the same auction system that handles traditional painting and sculpture, and because it gave a clear figure to an argument that had previously been mostly theoretical.<\/p>","protected":false},"excerpt":{"rendered":"

The history of AI art does not begin with a chat window and a typed prompt. It begins with programmers who treated the computer as a drawing instrument decades before anyone settled on a...<\/p>\n","protected":false},"author":1,"featured_media":840,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":"","jetpack_publicize_message":"","jetpack_publicize_feature_enabled":true,"jetpack_social_post_already_shared":true,"jetpack_social_options":{"image_generator_settings":{"template":"highway","default_image_id":0,"font":"","enabled":false},"version":2}},"categories":[3],"tags":[],"class_list":["post-841","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-posts"],"jetpack_publicize_connections":[],"jetpack_sharing_enabled":true,"jetpack_likes_enabled":true,"jetpack_featured_media_url":"https:\/\/gedespi.co\/1\/wp-content\/uploads\/2026\/10\/featured-30784971.jpg","_links":{"self":[{"href":"https:\/\/gedespi.co\/1\/wp-json\/wp\/v2\/posts\/841","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/gedespi.co\/1\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/gedespi.co\/1\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/gedespi.co\/1\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/gedespi.co\/1\/wp-json\/wp\/v2\/comments?post=841"}],"version-history":[{"count":1,"href":"https:\/\/gedespi.co\/1\/wp-json\/wp\/v2\/posts\/841\/revisions"}],"predecessor-version":[{"id":843,"href":"https:\/\/gedespi.co\/1\/wp-json\/wp\/v2\/posts\/841\/revisions\/843"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/gedespi.co\/1\/wp-json\/wp\/v2\/media\/840"}],"wp:attachment":[{"href":"https:\/\/gedespi.co\/1\/wp-json\/wp\/v2\/media?parent=841"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/gedespi.co\/1\/wp-json\/wp\/v2\/categories?post=841"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/gedespi.co\/1\/wp-json\/wp\/v2\/tags?post=841"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}