The News
Wednesday 29 of May 2024

Caedmon: A Machine that Dreams of Painting

An untitled image produced by Caedmon from a painting by Italian futurist painter and composer Luigi Russolo,photo: Courtesy of Maurizio Fusillo
An untitled image produced by Caedmon from a painting by Italian futurist painter and composer Luigi Russolo,photo: Courtesy of Maurizio Fusillo
Italian computer artist Maurizio Fusillo coded an art-generating program in hopes of achieving his goal of building a bona fide creative machine. Did he accomplish his purpose?

Maurizio Fusillo is not an artist, but an accountant. Or so he states on his own website. Nevertheless, he holds a couple of strong opinions on the nature of art itself. “For the biggest part of the last century,” he writes on the page for Caedmon, one of his latest projects, “avant-garde [artists] struggled in creating an art which could be defined as ‘impersonal,’ freed from the feelings of its maker … Who better than a machine to completely free art from this limit and create a totally impersonal art, free from all human feeling?”

Fusillo — though his site states otherwise — is a computer programmer and digital artist known for his work on what has been termed computer art. While most would consider computer art to be art created by humans through computers, Fusillo’s approach is more literal. “Photographs are made by photographers, paintings by painters and sculptures by sculptors; computer art should be made by computers,” he writes. If computer art is to be regarded as such, computers must be more than mere tools for its creation; they must become the creators.

To test his proposition, Fusillo spent six months coding Caedmon, a computer program that aims at probing the limits of what art is and what it is not. The concept is relatively simple: Caedmon will create pieces of its own out of existing artworks from several places, styles and eras; it will then post them on social media, where the public will judge which qualify as art pieces. Caedmon will then evaluate results and start the process all over again, this time working not with artwork from the past, but with pieces it made. This will, in theory, help Caedmon establish a general definition of what constitute a work of art and what doesn’t, pushing it’s creations ever closer to the real thing.

Many would be skeptical of Caedmon’s creative capabilities. Can a machine really make art? Can creativity really be achieved without emotion and imagination, without consciousness and personhood? Can a human being really build not the ghost of an artist but a bona fide creative mind?

Answering such questions is, for Fusillo, less important than the fact of them being formulated and put out there. “[Caedmon] is meant to be just a proposal, a manifesto,” he writes. And Caedmon is, indeed, more than an artistic project. It is a question that unfolds into more questions. Some of them are perhaps unanswerable, while others could be answered sooner than expected.


In the English literary tradition, Caedmon is the earliest poet known by name. His story, written in St. Bede’s “Ecclesiastical History of the English People,” tells of a seventh century monk who tended animals at a monastery. This monk, ignorant in the arts of poetry and music, dreamed one night of a pressence that asked him to praise the glories of God and creation in a song. Caedmon dreamed himself singning, and just like that the monk had been granted the gift of musical and poetic composition.

Fusillo’s Caedmon bears the monk’s name because creativity is supposed to be beyond its grasp. “As him, machines today are considered incapable of creating art,” Fusillo writes. “Since our aim is to transform a machine into an artist, Caedmon looked like the right name for this project.” Also, Fusillo thought giving it a human name would push Caedmon closer to the traditional concept of the artist as person. “[A person’s name] makes this project closer to our understanding of what an artist is: someone with his own will, skills and identity.”

Caedmon (the computer) is no dreamer, but it was granted the power of creation by its creator. Fusillo provided it with a training data set, a collection of around 10,000 paintings and illustrations extracted in batch from several sources. The training data set gives Caedmon a vast outlook on artistic traditions throughout several ages, cultures and schools of thought: from European and Asian art to baroque and expressionist paintings. Caedmon is programmed to sift through its collection, distilling the statistical structure of every picture — their DNA, so to speak — and pair them to create new pictures. This process results in bizarre images resembling glitched projections from a monitor, though one can sometimes see patterns of form or color and at times even glimpses of the artworks that fathered Caedmon’s new pieces.

An untitled image produced by Caedmon from a painting by Italian sculptor and painter Osvaldo Bot. Photo: Courtesy of Maurizio Fusillo

The new artwork is posted on Caedmon’s Facebook and Twitter profiles and also uploaded to the gallery on its webpage. There, people are encouraged to like or share whatever picture they deem satisfying attempts at artistic creation. This helps Caedmon identify which ones are worthy of being kept around, while the rest are discarded. Pictures with a higher-than-average rate of likes or shares are selected to form a new data set, earning the right to “pass on their genes.”

If the process sounds roughly familiar it’s because it is. In his search for a creative system that could be mimicked with today’s limited technology, Fusillo looked at nature itself for inspiration and made Caedmon into “a small emulator of natural selection,”as he puts it. “The only reference model we have of a working mechanism through which a ‘stupid’ force can create, and by create I mean the actual creation of something totally new, that did not exist before, is nature,” he writes.

Caedmon replicates this “stupid force,” mindless and blind in its drive towards the new, the unseen, yet undeniably effective in its results.


From “Tradition and the Individual Talent,” an essay by T.S. Eliot: “Poetry is not a turning loose of emotion, but an escape from emotion; it is not the expression of personality, but an escape from personality. But, of course, only those who have personality and emotions know what it means to want to escape from these things.”

As explained by literary critic C.S. Sukla (“Theory of Impersonal Art”), Eliot belonged to an anti-romantic movement which rejected the involvement of emotion and personal experience as part of artistic work. Poetry, as seen by Eliot, was an equation of sorts, and poets were only elements in that equation. “The poet, though an individual, is not isolated,” writes Sukla. “It follows then that the material used for his poetry [emotion] is not truly his; it belongs to the whole of tradition, to which he belongs.”

Fusillo believes Caedmon to be an inheritor of this concept of the impersonal artist, though filtered through the lenses of twentieth century painting, sculpture and conceptual art. “The most frequent objection to machines making art is that computers don’t have feelings, but a great part of cutting edge research in the last century revolved around the opportunity to remove, or strictly limit the part human feelings play in the creation of a work of art,” he says. “In this framework, is not unbelievable that computers will be able to create art.”

If we were to present Eliot with Caedmon’s concept, he would probably reject its validity highlighting that though the real poet needs to dispose of the personal, removing it completely would result not in a creator but an engine. There would be creation without thought, automatic production and reproduction with no meaning to convey.

Caedmon’s potential is inhibited by the limits of today’s technology. Its pairing process is randomized. Its combinations don’t follow any criteria aside from giving higher-rated images more opportunities to be paired. Caedmon is a machine that generates picture after picture after picture in seemingly mindless fashion. Yet, what if Caedmon isn’t following the human creative tradition? What if it is working outside humanity’s need for feelings and cultural constructs, imitating the proceedings of a force that isn’t quite human?

“Caedmon is mimicking, on purpose, the natural selection mechanism. This because evolution is the only known process which can generate a ‘best fitting solution’ in a totally context-agnostic way,” Fusillo says. “Even not having any idea, the ‘blind watchmaker’ managed to create such beautiful things and such perfect machinery, including the human brain, which no man-made creation can equal.”

Perhaps Fusillo’s creation isn’t the rightful heir of Kandinksy, Rembrandt or Gaugin. Perhaps it shouldn’t be regarded as a true creator. Perhaps Fusillo, who stated no human creation could compare to those of the “blind watchmaker,” has built a blind watchmaker himself.


Maurizio Fusillo was born in San Severo, a city in southeastern Italy. He learned to program when he was 13 using a manual handed down to him by his uncle. The first programming language he learned was Visual Basic 6.

Fusillo is a 90s kids. His interest in machines started when he was three years old, after receiving a “game station,” a knock-off of Nintendo’s SNES gaming console. Later, when he was eight, Fusillo had his first encounter with an actual computer. His uncle had brought home a machine with a 4GB hard drive and Windows 95. Fusillo, too young then to handle such a complex piece of hardware, saw his dad play around with colors and figures in a graphics editor, either Photoshop or CorelDRAW. A year later, Fusillo would be allowed to play with that same machine.

Things have come a long way. Today Fusillo lives in Luxembourg City, working as a software developer. He has a background in graphic design, plus his programming experience and a couple years of accounting knowledge. His work has been recognized nationally by Italy’s Premio Nazionale delle Arti and exhibited both in Italy and Luxembourg.

Though his résumé would tell of his accomplishments as an artist, Fusillo is not very comfortable with being labeled as one. “There is this idea of the artist as a undisciplined genius, driven by intuition,” he says. “Intuition is important, as well as creativity, but constant research and strong discipline in the creation process are very important as well.”

Caedmon reflects Fusillo’s outlook on artistic creation. Effortless as it may seem, Caedmon’s creative quest demands a never-ending cycle of examination, abstraction, combination and evaluation. Fusillo has come to refer to this process as a form or research. “Research, in art, is very important,” he says. “There is not true innovation if you ignore what happened before you. Creativity for the sake of being creative is a pretty sterile exercise.”

Fusillo has done his research too. When asked about the currents that have influenced him —and Caedmon — he mentions surrealism’s idea on automatic creation, futurism’s cult of the mechanical, pop’s mass-produced art and dada’s constant inquiries into the nature of art itself. “But this isn’t something planned,” he says. “It’s just that you start having better ideas.”

Another big influence of his is Joseph Kosuth, a conceptual artist from the United States. In one of his essays, titled “Art After Philosophy,” Kosuth writes that the artist’s mission is not to create objects of beauty and admiration, but rather to un-define the definition of art, stretching it beyond its current standing. “Being an artist now means to question the nature of art,” he says. “A work of art is a kind of proposition presented within the context of art as a comment on art.”

Under this viewpoint one has to wonder: who’s the artist here? Is it Caedmon, realizing a mode of artistic creation that falls outside of humanity’s capabilities? Or is it Fusillo, who devised an artifact that is creative but not human nor conscious?

“Caedmon is my opera, because I created him and gave him a meaning, a concept,” Fusillo says in response to the question. “I don’t have any control on what he does, on what he creates. He is somehow alive, he does basically whatever he wants. My role was to create him, but now he is free to pursue his research. I’m just a spectator, as you are.”


Caedmon’s Twitter profile has 4,178 followers and 27,300 likes. “I’m very satisfied with Caedmon’s results,” Fusillo says, “and apparently people like it, which is very rewarding.”

Its audience, according to Fusillo, seems to be reacting more positively to pieces derived from abstract artwork. This could be due to the fact that Caedmon is better at handling abstract imagery. “Caedmon’s algorithm works better with those images compared to figurative ones because abstract paintings are, actually, easier to ‘abstract,’ and therefore Caedmon can better extract a pattern from them,” Fusillo explains. “Also, those works tend to be more colorful, which is something people like.”

The plan is for Caedmon’s creations to become more polished with every generation produced. If Fusillo’s method proves correct, Caedmon’s struggle will culminate in what he has dubbed “the definitive art piece,” something that many human artists seek to achieve too.

All for argument’s sake, of course. Fusillo is aware of Caedmon’s limits as a piece of technology and also of how conservative his approach would have to be in order to achieve this so-called definitive art piece. The point is to define accomplishments in accurate, almost scientific terms, thus giving him a suitable point of reference for Caedmon’s results. Also, in this way “[the artwork]  would be easily accepted by the general public, opening then the door to a new generation of machine/artists,” he explains.

But Caedmon won’t be the last word on the possibility of creative machines. Fusillo regards it a stepping stone, the first utterances of a grander enunciation. “Technology constantly evolves, as new tools emerge, things which today we can only dream about become normal,” he says.