ROM’S Curators Weigh In.
This painting would take most artists days, even weeks to make. It took Midjourney—a new artificial intelligence program—less than two minutes.
How? According to PC World, Midjourney “takes user-generated queries, runs them through an AI algorithm, and lets the algorithm pull from its source images and apply various artistic techniques.”
For my image, I wanted to see how well the program mimicked distinct art styles. So, I gave it the prompt “Chinese literati painting of white clouds and red birds.” It gave me back four low-resolution images, from which I chose one (what you see above) to “upscale.”
Even a cursory glance at traditional literati paintings will tell you Midjourney didn’t exactly nail it. The image appears to be made from oils, not inks, and the birds are, well, faceless. Still, the image is beautiful, even gallery-worthy, and the implications of this technology are profound.
Anyone with internet access willing to shell out $10 (USD) a month for a subscription can now create stunning art in minutes, all just with a few keystrokes. (Many of the less powerful programs, like Craiyon, remain free; while others, like DALL·E 2 and Meta’s Make-A-Scene, aren’t yet available to the public.) What does that mean for working artists? For human creativity?
"A portrait of director Jordan Peele by Pablo Picasso” prompted Midjourney to create this image.
These questions are already top-of-mind for many artists and journalists. In June of this year, The Economist claimed to break “new ground” after putting a Midjourney-generated image on its cover. In July, CBC asked if “text-to-image AI would change the way artists work.” And in August, Vice reported that a Midjourney-generated image “won first place at a state fair fine arts competition.”
“To developers and technically minded people, it’s this cool thing,” cartoonist Matt Bors told The Atlantic. “[B]ut to illustrators it’s very upsetting because it feels like you’ve eliminated the need to hire the illustrator."
The obsolescence of human artists, while frightening, is only one potential consequence of AI-generated art. Writing in The New York Times, columnist Kevin Roose wondered “whether we need to worry about a surge in synthetic propaganda, hyper-realistic deepfakes or even nonconsensual pornography.”
However, I wanted to understand how the rise of programs like Midjourney would shape the future of art. So, I put the question to three of ROM’s curators. Here’s what they said:
Silvia Forni, Senior Curator of Global Africa
It is interesting, but derivative. Midjourney can generate cool images, but in my mind, it still does not replace the work of an artist. It is a tool that can “create” only using the visual vocabulary that was developed by artists of the past. It is fun but not moving. It does not present new visual vocabularies. It is a tool. Maybe, in the hand of an artist, it can become surprising and moving, but to be so it would require the creative ability of someone able to instruct the system to do something that the general user cannot achieve, something that pushes the envelope and generates awe. Otherwise, it is just a new accessible fun tool that people will play with until the next trendy program hits the market.
Justin Jennings, Senior Curator, American Archaeology
Machine learning (ML) is already changing the way we study ancient art. Archaeologists encounter the past in fragments worn down and faded from the passage of time. Reconstruction is a leap of faith—Is that a leg or a stick? Was that brown or once red?—aided by years of experience. ML projects are currently training computers on art and architectural styles using massive databases in museums and other institutions. Soon, machines may help fill in the blanks on a dig, “seeing” the most likely vessel form from a tiny fragment or the lost edges of a fresco.
Akiko Takesue, Bishop White Committee Associate Curator of Japanese Art & Culture
I am interested in the way artificial intelligence could co-exist with human creativity, rather than replacing it, by providing a new perception. A Japanese traditional weaving company in Kyoto, Hosoo, for example, has recently used AI to generate new designs for obi sash, through machine learning of over 20,000 uncoloured designs from the past. The new designs that AI created were a result of its different interpretations of traditional patterns as well as of the spaces intentionally left blank. Such is where AI can be best utilized to open a possibility of connecting the past and the future.