In Diane Ackerman’s 2014 book The Human Age,* she discussed AI with roboticist Hod Lipson.
Professor Lipson is the director of Columbia’s Creative Machines Lab, as well as a faculty member since 2015 at Columbia University.
In a 3/30/2023 Columbia News article “Will ChatGPT and AI Help or Harm Us,” he argues that use of ChatGPT, and its’ “artificial cousins,” should be encouraged by educators, and that professors should teach students to use the new AI tools, or be left behind.
Ackerman is more cautious.
She discusses “robotic delinquents,” and envisions problems if bots were used to "man" crisis hotlines.
(Think of the “Parce Domine” episode of Westworld in which Caleb Nichols isn’t sure whether he’s talking to a human being, or a bot.)
Ackerman warns that although robots do learn, “even robo-tots will need good parenting.”
In another paragraph of The Human Age, Ackerman mentions how Lipson’s Creative Machines Lab nearly finagled a robot-created painting into a Yale Art Museum exhibit.
Ultimately, the painting wasn’t displayed, but this story leads us to the subject of AI-created artwork.
Stable Diffusion, Diffusion Bee, Lensa, Starryai, DALL-E.2, Craiyon, Dream by Wombo, StyleGAN, and Midjourney are some of the programs that can be used to generate digital artworks.
The copyright status of AI-created art is hazy.
In some programs (like Dream by Wombo), the fine print says that the software owns all the creations.
The contract for DALL-E.2 (an AI-powered synthesizer created by OpenAI) says that the “artist” owns the work, but DALL-E.2 must be credited (by retaining the watermark).
In Starryai, the “artist” only owns the work if they own all the elements used in the work.
According to the U.S. Copyright office, any images produced by AI-generated art cannot be copyrighted; however, the artist may own the art.
In 2022, an artist named Jason M. Allen won the “Digital Art/Digitally Manipulated Photography” prize ($300) at the Colorado State Fair by using Midjourney.
(Allen actually paid to have his Midjourney image printed on a canvas!)
Just as ChatGBT is “built from” 300 billion words taken from Wikipedia (and other material on the open web), programs like Midjourney are “built” from billions of images taken from the open web—many of which are watermarked, and in copyright.
The AI-art generators take the images, and use them to construct algorithms to generate new images.
I imagine the engineers who create the software think that if one uses billions of images—rather than just use one—you’re free and clear.
Aren’t there enough public domain, and creative common images to build an image library?
Why are images being culled from the web?
I was once asked to obtain the credit line for an image, to be used in a book I was producing.
The editors wanted to use a particular drawing.
However, after I found the artist, the editors didn’t want to pay the very reasonable usage fee ($150) that the artist wanted to charge.
Their argument was that other companies were using the image uncredited on the web.
Why pay anything?
We ended up not using an image in the book, because the editors didn’t like any of the public domain, or Creative Common, substitutes that I had found.
Nowadays, they would have insisted that I use AI software to draw a similar illustration for them (as long as we could own the copyright).
Another issue about AI-created images is the quality of the work.
I’ve drawn ever since I was a small child, and my BFA is in painting and drawing.
One thing I wonder is how well AI programs draw in perspective.
Is the software extracting image data from created drawings and paintings plus photographs, or mainly from photographs?
That would make quite a difference in how the software dealt with perspective.
We began to hear about art created by animals in the 1950’s.
Today, you can find: the Elephant Art Gallery (images available on pillows, as well as canvas); elephant foot prints and “kisses” that help to support an elephant preserve in Texas; paintings by a pig named Pigcasso; and (of course) paintings by the famous gorilla Koko, and the chimpanzee Congo.
The London Institute of Contemporary Art exhibited Congo’s paintings in 1957, and two works were purchased by Picasso and Miro.
(“It Seems Art is Indeed Monkey Business” by Sarah Boxer, 11/8/1997, N.Y. Times.)
I feel about people choosing to put art created by animals on their walls, much the same way as I feel about people choosing to place AI-created artwork on their walls.
To each their own.
However, it is upsetting that illustrators will lose work to word-people using AI—especially since their art may have been used as raw material.
Publishers will continue to need illustrators (unless they are content with hack work thought up by editors).
Medical illustrators have long studied the work of Dr. Frank H. Netter (1906-1991), the medical doctor and great medical illustrator.
(He might have been a little weak on women’s faces, but none could draw internal organs like he could.)
Netter is so great because he merged a scientist's understanding of anatomical structures with an artist’s skills.
His work is appreciated (and copied) for how well it helps us to understand medical concepts and anatomical structures, as well as its’ aesthetic value and accuracy.
When I was the Design Director for an encyclopedia and yearbook company, I hired artists to draw everything from plants and birds, to scientific diagrams, to comedic scenes for feature articles.
Every artist brings a different skill set, so deciding which artist to use is an important decision.
Selecting the wrong artist for a project could lead to disaster.
Perhaps, I could have used an AI program to draw a comedic scene, but I needed to hire an ornithologist, or a scientific illustrator, to draw a bird.
*The Human Age: The World Shaped by Us, by Diane Ackerman, published by W.W. Norton, Ltd. 2014, Chapters: “When Robots Weep, Who will Comfort Them?” and “Robots on a Date.”
