Revered symbolist painter Gustav Klimt once said, “art is a line around your thoughts.” Is it possible, a century later, that art is no longer human-exclusive? These days, the sort of artistic representations described by Klimt’s thought-bounding-boxes are also being produced by generative adversarial networks, whose outputs both fool and please human evaluators.
AI has become increasingly capable of generating impressive artworks across a wide range of styles and forms, from abstract painting to prose writing, film scores, and even operas.
Many researchers spent much of 2020 at home, where, apparently, many explored AI’s creative potential. As part of our year-end series, Synced highlights 10 AI-powered art projects that inspired and entertained us in 2020.
San Francisco-based AI company OpenAI in May unveiled its transformer-based language model GPT-3,
which drew attention and made headlines both inside and outside the machine learning community. Many innovative new applications — from recipe creation to writing philosophical essays — have leveraged the GPT-3 API since its release.
Toronto-based machine learning engineer Aditya Joshi has curated a list of GPT-3 powered applications that includes producing podcast episodes, writing creative fiction and generating financial statements, to name just to name a few.
The AI-driven Verse by Verse tool provides a shortcut to becoming a poet. Users can choose up to three American poets from a list that includes the likes of Emily Dickinson, Walt Whitman and Edgar Allen Poe, then pick a structure for their poem. After the first line of verse is entered, the tool will suggest the following lines. Users can accept, reject or tweak suggested lines from a list, and end the poem at any point.
In styles ranging from Japanese-style anime characters (Waifu) to oil portraits (AI-Art Tokyo) and Disney-style animation (Toonify), AI researchers and developers this year have come up with many ways to artify profile photos.
According to National Geographic, there are an estimated 8.7 million different species on Earth — and so far we’ve only identified about 1.6 million of them. You might venture into a jungle or peer into a microscope in search of the unknown, or you could simply make your own new species — who would know? “Chimera Painter” is a CreatureGAN based model that gives you the chance to “create” fantastical creatures by simply drawing out basic shapes and letting the AI do the rest.
Stanislas Chaillou, an architect, artist, and AI researcher from Paris, applied AI to botany for his “Bio Morphosis” project. Inspired by the work of German photographer Karl Blossfeldt, Chaillou created a series of AI-generated animations that blend different plant species together in a dynamic herbarium.
Blob Opera is a machine learning experiment by David Li in collaboration with Google Arts & Culture that enables users to create their own opera-inspired songs — no previous music skills required!
In a bid to raise awareness of the threats posed by climate change, the Mila team this year published a paper using GANs to generate images of how climate events may impact our environments — with a particular focus on floods.
Real data on climate impacts such as floods in urban environments is scarce and, when available, often lacks key information. The Mila study explored the potential of using images from simulated 3D environments and domain adaptation tasks.
This summer, SIGGRAPH 2020, the world’s leading computer graphics and interactive techniques conference, presented a special online program showcasing innovations in digital art. The selections include 13 virtual installations featured in the Art Gallery and 15 research advancements presented in the Art Papers program. The Art Gallery’s Best in Show award was given to “Cacophonic Choir.”
Cacophonic Choir, an interactive art installation, uses machine learning to explore how sexual assault survivors’ experiences can be distorted by digital and mass media and its effect. The installation is composed of distributed agents that respond by becoming visually bright, semantically coherent, and sonically clear, revealing the original testimonies of survivors.
München-based web developer Valentin Vieriu launched an AI-generated art project this year to explore the “infinite” creative possibilities of an AI artist trained on a carefully selected set of cubist artworks.
Paris-based ML researcher and Google Arts & Culture Lab resident Emil Wallner curated a collection of eclectic ML experiment projects, such as training a StyleGAN2-ADA on myths and interpolating it with music, a robot that writes poems on the beach, and Neural Radiance Fields “applied to a rotating butterfly.”
AI and art may seem an odd couple, but the Christie’s auction sale two years ago of the AI-created painting Portrait of Edmond de Belamy for US$432,500 has resulted in the pair being taken more seriously. Questions for the future will include ownership of the generated artistic properties, but in the short-term, as these projects illustrate, AI and art have few bounds.