You are on page 1of 3

AI is Blurring the Definition of Art

1 Christie’s recently sold its first piece of auctioned AI art—a blurred face titled “Portrait
2 of Edmond Belamy”—for $432,500.The piece sold at Christie’s is part of a new wave of AI art
3 created via machine learning. Paris-based artists Hugo Caselles-Dupré, Pierre Fautrel, and
4 Gauthier Vernier fed thousands of portraits into an algorithm, “teaching” it the aesthetics of past
5 examples of portraiture. The algorithm then created “Portrait of Edmond Belamy.”The painting
6 is “not the product of a human mind,” Christie’s noted in its preview. “It was created by artificial
7 intelligence, an algorithm defined by [an] algebraic formula.”
8 If artificial intelligence is used to create images, can the final product really be thought of
9 as art? Should there be a threshold of influence over the final product that an artist needs to
10 wield? To create AI art, artists write algorithms not to follow a set of rules, but to “learn” a
11 specific aesthetic by analyzing thousands of images. The algorithm then tries to generate new
12 images in adherence to the aesthetics it has learned.

13 Most of the AI artworks that have emerged over the past few years have used a class of
14 algorithms called generative adversarial networks (GANs). First introduced by computer
15 scientist Ian Goodfellow in 2014, these algorithms are called “adversarial” because there are two
16 sides to them: One generates random images; the other has been taught, via the input, how to
17 judge these images and deem which best align with the input. One part of the system generates
18 new images (based on the input of old images); the other part of the system judges whether those
19 images are successful—meaning enough in line with the old images.In other words, AI is trying
20 to make images that fit the established historical precedent. It’s making images that we think are
21 art—not images that it thinks is art. For example, an artist could feed portraits from the past 500
22 years into a generative AI algorithm. The algorithms then tries to imitate these inputs, producing
23 a range of output images. The artist must sift through the output images and select those he or
24 she wishes to use.Throughout this process, the artist maintains an active hand: He or she is very
25 involved in pre- and post-curation, and might also tweak the algorithm as needed to generate the
26 desired outputs.

27 The generative algorithm can produce images that surprise even the artist presiding over
28 the process. For example, a GAN being fed portraits could end up producing a series of
29 deformed faces. What should we make of this? Psychologist Daniel E. Berlyne has studied the
30 psychology of aesthetics for several decades. He found that novelty, surprise, complexity,
31 ambiguity, and eccentricity tend to be the most powerful stimuli in works of art.The generated
32 portraits from the GAN—with all of the deformed faces—are certainly novel, surprising, and
33 eccentric.They also evoke British figurative painter Francis Bacon’s famous deformed portraits,
34 such as “Three Studies for a Portrait of Henrietta Moraes.” But there’s something missing in the
35 deformed, machine-made faces: intent. Although it was Bacon’s intent to make his faces
36 deformed, the deformed faces we see in the example of AI art aren’t necessarily the goal of the
37 artist or the machine. What we are looking at are instances in which the machine has failed to
38 properly imitate a human face, and has instead spat out some surprising deformities.
39 This is exactly the sort of image that Christie’s auctioned.
40 Another example is Mario Klingemann’s “The Butcher’s Son,” a nude portrait that was
41 generated by feeding the algorithm images of stick figures and images of pornography. Seeing
42 such works, many are skeptical of AI art. Pulitzer Prize–winning art critic Jerry Saltz has said he
43 finds the art produced by AI artists boring and dull, including “The Butcher’s Son.”In the
44 deformed portraits, for example, you could argue that the resulting images aren’t all that
45 interesting: They’re really just imitations—with a twist—of pre-curated inputs.
46 Finally, for a piece called “This Sculpture Doesn’t Exist,” artist Matteo Rattini used a
47 GAN to create contemporary art sculptures based on Instagram’s suggestions for popular
48 contemporary art. Specifically, Rattini created an Instagram account to follow contemporary art
49 accounts, as well as a bot to engage with the art, and then he used the images shown to him via
50 Instagram’s curated list to feed his GAN.His GAN then created new images, based on the data
51 fed into it, until it ended on “pictures of minimal modernist sculptures showcased inside an
52 aseptic white cube.” Those images were renderings of sculptures that the GAN determined were
53 art by human standards.“[AI] reflects how our brain works in a very simple way,” Rattini says
54 “You feed it a lot of information on trees, for example, and when it comes time to draw a tree,
55 not only is your idea of a tree informed by all the trees you’ve seen, but while you are drawing it,
56 your brain is constantly judging it through your eyes and giving real-time feedback on how to
57 compensate and adjust.”

58 If artificial intelligence is used to create images, can the final product really be thought of
59 as art? Many have been wrestling with this question—specifically, the point at which the artist
60 should cede credit to the machine. Some argue it’s not just about the final image. It’s about the
61 creative process—one that involves an artist and a machine collaborating to explore new visual
62 forms in revolutionary ways.

63 What then if a machine were programmed to create art on its own, with little to no human
64 involvement? AICAN (artificial intelligence creative adversarial network), is a program that
65 could be thought of as a nearly autonomous artist that has learned existing styles and aesthetics
66 and can generate innovate images of its own. AICAN, has adhered to a theory proposed by
67 psychologist Colin Martindale. He hypothesized that many artists will seek to make their works
68 appealing by rejecting existing forms, subjects, and styles that the public has become accustomed
69 to. Artists seem to intuitively understand that they’re more likely to arouse viewers and capture
70 their attention by doing something new. In other words, novelty reigns. When programming
71 AICAN, the algorithm called the creative adversarial network, compels AICAN to contend with
72 two opposing forces. On one end, it tries to learn the aesthetics of existing works of art. On the
73 other, it will be penalized if, when creating a work of its own, it too closely emulates an
74 established style.At the same time, AICAN adheres to the “least effort” principle, in which too
75 much novelty will turn off viewers. This careful combination ensures that the art generated will
76 be novel but won’t depart too far from what’s considered acceptable. Ideally, it will create
77 something new that builds off what already exists.The AICAN algorithm was fed 80,000 images
78 that represent the Western art canon over the previous five centuries. It’s somewhat like an artist
79 taking an art history survey course, with no particular focus on a style or genre.At the click of a
80 button, the machine creates an image that can then be printed. The works will often surprise us in
81 their range, sophistication, and variation.
82 Still, there’s something missing in AICAN’s artistic process: The algorithm might create
83 appealing images, but it lives in an isolated creative space that lacks social context. Human
84 artists, on the other hand, are inspired by people, places, and politics. They create art to tell
85 stories and make sense of the world. So the question remains, is the result AICAN produces a
86 bona fide work of art? Is AI genuinely creative? Moreover, do we want AI to be used in the
87 making of artwork at all?
88

You might also like