Home ›› 21 Oct 2022 ›› Back
AI-generated artwork, a new genre of art created by artificial intelligence-based algorithms by writing some text descriptions in English, has caused one of the greatest stirs among artist community.
Though platforms researching AI text-to-image projects are slowly opening their services to public, every platform is running out of subscription capacity within a month and sometimes a week. Millions of AI-based images and artworks are generated every day.
The internet is flooded with mind-blowing concepts of generating artwork just by writing some simple texts in the prompt (the input field in a text-based user interface screen for an operating system (OS) or program).
This has created an opportunity for non-artist and amateurs to create mind-blowing artworks of any concept and genre without any experience. People are now creating images that were not possible before for everyone.
Traditional artists create art spending hours while AI takes minutes to output similar results and some experts are calling it ‘Plagiarism by machine’. However, AI does not create an image out of nowhere, rather it analyzes the existing database, learns from it and generates image. We, humans similarly make art using our past experience, philosophy of life, imagination and creativity.
The concept of generating images and art is not something new but within the last six months, almost all artists and tech enthusiasts are using commercial platforms like Open AI’s DALL-E or open-source platforms like Midjourney, StyleGAN and Stable Diffusion.
One of the first generations of AI image generators AARON was developed by Harold Cohen, an English artist in the late 1960s. Later IBM, Microsoft and Google tried to improve the idea but DALL-E, Midjourney and Stable Diffusion became the major players in the market.
Art has always been very much subjective and relied upon the opinion of the viewers but these platforms use millions of images found on online with their descriptive information generating images which captivate millions of people.
Artists and amateurs are testing their creativity, generating new ideas from AI outputs to create their work. AI can generate future project plans for artists, framing ideas for photographers, and characters for animation, video games and fiction. This way AI is also enhancing human creativity while saving time.
Though the outputs are not perfect in the current state, the way its popularity is growing and how major tech giants are investing in these fields also raises a question - will AI replace human artists?
At the 2022 Colorado State Fair’s annual art competition, an image generated with Modjourney AI won the first prize over other artists who spent hours creating digital artworks. Yet, comprehension and appreciation of art are widely considered exclusively human capabilities. This instance has made artists concerned about their future.
However, enthusiasts and artists working or testing these platforms are saying AI is revolutionizing digital art but it won’t affect the existing art mediums. “Did photography replace traditional art? So won’t AI,” questioned one on a social media platform.
The way these tools work is quite fascinating. For example, DALL-E derived from the names of famous Pixar character Wall-E and Spanish surrealist artist Salvador Dalí creates images seen never before and artwork by using OpenAI models called CLIP (Contrastive Language-Image Pre-training) and Diffusion Model.
CLIP trains the model using hundreds of million images and associated captions to recognize the similarity between the image and its caption. This process helps the AI to learn which English text is associated with which object or style.
And then comes the Diffusion model, which works differently and intricately. To make it simple, we can think of a solved puzzle mixed up and then solve again by AI. This way, the model takes a perfect image, adds noise and randomly scrambles its pixels and then it denoises the image again pixel by pixel. This way it learns to generate new images.
DALL-E also has a text encoding to image encoding process called prior which takes the text input and maps them to process the initial command. Midjouney AI and Stable Diffusion also use similar models but outputs are different because of the Different parameters they use.
DALL-E trains its models using images found on the internet whereas Midjouney AI trains using images from DeviantArt. On the other hand, Stable Diffusion is trained using 256 Nvidia A100 GPUs at Amazon Web Service and details are proprietary to the developers.
However, these models are not perfect at this moment but the way it is gaining popularity and how the major tech giants are investing make experts think about the commercial aspects of these platforms.
Developers working in NFT and Crypto market are developing tools that incorporate AI-generated NFTs and sell them using cryptocurrency.
In 2018, Christie’s, a British auction house, auctioned the first AI art titled ‘Portrait of Edmond Belamy’ for $432,500. In 2022, many auction houses and online platforms are selling AI arts on regular basis like the UnRealArt exhibition.
XANA, a platform built for Metaverse 10,000 AI-based Japan anime-inspired unique avatars that they sold as NFTs. These NFTs also provide special privileges for the Metaverse. Some other artists are selling AI-generated artworks as NFT on independent platforms. Developers are jumping to grab the market.
LinkedIn founder and tech entrepreneur Reid Hoffman released a small collection of AI-generated art for sale as a part of market research and one piece of ‘art’ called mångata was sold on an NFT marketplace for $24,000.
At this time, no one can precisely say anything about the future of AI art platforms but we can predict the momentum as AI is making an impact on our day-to-day life. As AI is taking over every aspect of technology, we cannot but have to live in an AI-powered world.
Though art was considered a window of a person’s soul, now we are creating art looking through a computer’s window. We have to wait to find out whether AI replaces human creativity or boosts human creativity.
