A new age is arising where art, unlike a lot before, is being written, built and performed by AI. Innovators have not just dug up old art forms and created new versions using AI but also constructed completely new magical sculptors, solutions and even robots. 2021 introduced the world with many new artworks built together by artists and AI experts that push the boundaries of what can be called ‘art’ in itself.
Here are the biggest breakthroughs of AI art in 2021:
Beethoven’s 10th Symphony
Ludwig van Beethoven died in 1827 and left behind just some musical sketches of his 10th Symphony. Cut to October 2021, a full recording of Beethoven’s 10th Symphony was released. The Karajan Institute, an Austria-based organisation that promotes music technology, created a team comprising Austrian composer Walter Werzowa; Mark Gotham, a computational music expert; Robert Levin, a musicologist at Harvard University; and Ahmed Elgammal, Professor, Director of the Art & AI Lab, Rutgers University, who made this possible using AI. They had to teach AI how to take a melodic line and harmonise it. The machine had to learn the construction of Beethoven’s Fifth Symphony out of basic four-note motif and develop a longer, complicated musical structure. The AI then learnt to bridge two sections of music together and composed a coda. After the full composition was figured out, the AI had to now orchestrate it. It took 18 months for the team to construct and orchestrate the two entire movements of over 20 minutes a piece.
Idea to Painting
In 2021, notorious videos of celebrities from Elon Musk to Kim Jong-un to even Monalisa singing songs made circles on social media. Yes, these deep fakes were created using AI, and the credit goes to Ben-Zion Benkhin’s Wombo.Ai, an AI-powered lip-sync app. The company also created Wombo Dream – the software that converts text into computer-generated paintings. The company has not revealed the technologies they use, but experts suspect they use the VQGAN+CLIP architecture, where AI converts ideas into paintings.
NVIDIA also announced the latest version of NVIDIA Research’s AI painting demo, GauGAN2. Powered by deep learning, the model consists of a text-to-image feature. This is one of the first AI models that incorporates multiple modalities, semantic segmentation, text, sketch, and style within one GAN network. Earlier this year, NVIDIA released a tool built upon GauGAN, NVIDIA Canvas, which can be used on any NVIDIA RTX GPU. GauGAN2 has been trained on 10 million landscape images using the NVIDIA Selene supercomputer.
Revealed a Pablo Picasso Painting
‘The Lonesome Crouching Nude’ was discovered hidden beneath the surface of a Pablo Picasso painting. It was revealed by Oxia Palus that it uses AI, advanced imaging technology, and 3D printing. The company found this painting under Picasso’s ‘The Blind Man’s Meal’, which is from 1903. Founded by George Cann and Anthony Bourached, the team at Oxia Palus used XRF imaging and image processing to first reveal the outline of the hidden painting and then trained AI to add brushstrokes to the portrait in Picasso style. It then generated a heightmap of the portrait to give it texture. The image was printed onto canvas using 3D printing technology.
Interactive Sculpture
Suchi Reddy and Amazon Web Services produced the piece titled ‘you + me FUTURES’, an interactive sculpture that translates spoken words into lights and colours. It is displayed at Smithsonian’s Arts and Industries Building (AIB), USA. The artists refer to the 25 feet tall and 20 feet wide sculpture as a reflection of “society’s collective conscious” and their thoughts about the future. First, visitors share their thoughts and feelings about the future with (to) the sculpture, after which words are picked up by microphones and sent to the cloud. Here, AWS uses AI to translate these words into corresponding lights and colours that match the theme of the statements. It took two years for Reddy and AWS to build the sculpture, and they used tools like Amazon Comprehend, Amazon Transcribe, and a custom Amazon SageMaker model.
Robot Writing and Performing Poetry
Devised in Oxford by Aidan Meller, the ultra-realistic Ai-Da performed poetry written by her at Oxford’s Ashmolean Museum. While Ai-Da is not the AI who can write poetry, this year, for the first time, it did. The robot was given Dante’s epic three-part narrative poem – the Divine Comedy, and she used her algorithms, data bank of words and speech pattern analysis and produced a reactive work.
Spotify of Paintings
DAISY – an AI that powers KULTURA helps users discover new art, create and share high-resolution collections, and view them in 3D virtual spaces. DAISY is based on a unique dataset of interactions between people and artworks and generates personal recommendations based on behaviour. Using KULTURA.art, an online art platform that hosts over 20,000 artworks from 2,700 artists in over 100,000 3D parallel galleries, users get personalised recommendations on what paintings and artworks they will like – similar to what Spotify and Netflix do for music and movies and shows, respectively.
