Signal to Noise (2025)
SWIM, seen at Signal to Noise, 2025. Photo by Phoebe Powell, courtesy of NCM.
Signal to Noise explores how artists work with, challenge, or complicate the relationship between signals and noise—disruptions, glitches or interference—in communication technologies and the messages they send. These technologies include the internet, telephones, radio and television, artificial intelligence, social media algorithms, and even the sounds of the natural world.
Artists are the buzz in the radio, the data that brings AI to a glitching halt. Instead of seeing noise as something to block out, artists in the exhibition reframe noise as a signal in itself— an opportunity for creative engagement.
Through international artworks, new commissions and technology collections, the political, social, philosophical and aesthetic possibilities for noise are examined: revealing the limits of technology’s capacity to contain noise, while embracing the inevitable and productive friction noise makes possible.
With works from:
Nam June Paik
George Brecht
Lillian Schwartz
JODI
elekhlekha อีเหละเขละขละ
Craftwork
Rowan Savage
Mimi Ọnụọha
Machine Listening
Eryk Salvaggio
Internet Dream, Nam June Paik. Seen at Signal to Noise, 2025. Photo by Phoebe Powell, courtesy of NCM.
In the age of computation, the role of noise — or at least, the close-enough approximation through pseudorandom numbers — has been the bulwark against the rigid and orderly logic of mathematics and prediction. Leslei Mezei, writing for Jasia Reichardt’s “Cybernetics, Art, and Ideas,” puts forward a theory of the computer-generated image that would be just as accurate for describing generative AI, with a few million more pixels:
“... Constraints impose the unity, the overall structure, or in the terminology of ‘information aesthetics,’ the macro-aesthetics of the picture. Were we to repeat the experiment any number of times, we would get a larger number of variations, some more interesting than others, but the overall structure would remain recognizably the same. The program embodies the generating rules, or the algorithm, with which random selection is capable of producing an infinite variety of similar pictures. (165)”
This seemingly paradoxical point of “an infinite variety of similar pictures” is all too often lost on us in discussing generative AI. When Lev Manovich writes that “We can now, in principle, create an infinite museum filled with endless AI-generated images that simulate artworks from every period in history for all genres, artistic techniques, and media,” he is, in principle, wrong. The generative AI system is just as constrained by data as Leslie Mezei’s computer was in 1971. There is just more data through which to constrain possibility.
Generative AI is, after all, a noise reduction technology. It just so happens to produce an image at the end of that process. It dissolves training data by introducing noise in a sequence of steps, storing the path of this noise to memory. When prompted, it plays these steps in reverse, applied to random noise, in a bid to denoise it into a statistically centered outcome based on accumulations of images related to the prompt. This reliance on noise is what produces the illusion of creativity in the system, a means of creating variety. But it is always constrained by data, by historical reference, by language.
It is this structure of generation from which Signal to Noise takes its cues as a “post-AI” art exhibition. It is structured as a historical lens on artists who worked with the unique affordances of noise in given communication technologies of their era, embracing the breakdowns that pushed computational rigidity to its limits. In this way, we look at AI in hindsight, rather than the forward-looking lens of utopian hype or doomer panic. Instead, Signal to Noise looks at AI as one more media format designed to reduce noise — that is, unpredictable and unwanted intrusions into “clean” media signals. And artists have been introducing noise into channels for as long as we have had technology in which to intervene.
George Brecht, Universal Machine. Seen at Signal to Noise, 2025. Photo by Phoebe Powell, courtesy of NCM.
Consider Universal Machine, George Brecht’s 1965 take on the rigid mathematics of computational systems. Brecht places his assemblage of images and text — cut out from magazines — into a cardboard box. The instructions tell us to shake this box to generate new, random patterns (though bound, as always, by the constraint of his pre-selected vocabulary). Arguably, Brecht’s box is just as generative as any AI system: a scattering of data points, activated by the shaking of a “black box” to propose new structures for whatever the shaker has intended. Brecht’s instructions even seem to echo the hype of Silicon Valley: “Need a friend? Shake the box.” The box can even write novels: shake the box, open it, and that’s chapter one. Do it again for chapter two. Use Number 18: “Travel Itinerary.”
Nam June Paik’s “Internet Dream” embraces the noise of information overload, with Paik himself pioneering the use of magnets onto televisions to distort the carefully calibrated signals of cathode ray tubes. In JODI’s My%Desktop, artists insert friction into the friction-reducing site of the computer interface. The graphic user interface of desktop computing was designed to constrain the options of computation while making them more readily accessible (just think if you had to write a command line query for every email). JODI’s work inverts this, turning the desktop interface into a site of contention and error, estranging us from its simplicity and use for “productivity.”
Machine Sees More Than it Says, from artist Mimi Onuoha, hints at the origins of computer vision and image generation, “a sketch of a computer's imagining of itself and the courses of development which formed it, hinting at processes of resource extraction, transport, technical advancement, interfaces and labor.” The archival footage is a document of how we have allowed computers to produce images of themselves: each scene is the result of a computer’s existence as a product (the computers and screens being filmed) or political purposes (the reasons they are being filmed to begin with). As such, the images in this piece evoke a powerful arrangement of order and control that ellides the chaotic origins, histories and purposes to which they were meant to be used.
Elekhlekha อีเหละเขละขละ present “The beautiful sounds are gathered together in one cluster,” in Signal to Noise, 2025. Photo by Phoebe Powell, courtesy of NCM.
In contrast to this order, Thai / New York artist duo elekhlekha อีเหละเขละขละ present “The beautiful sounds are gathered together in one cluster,” a work which embraces the shifting tonalities and relationships that emerge when a listener attends to noise. This is a mix of 32 speakers suspended from the ceiling and transducers triggering assorted aluminium objects. As samples of percussive gongs are triggered, they blur into a meditative drone, depending on one’s position and willingness to hear. The auditory and political message come to overlap: the work is a challenge to the idea that the music of gongs in the artist’s home, Thailand, is described by Western observers as “noise.” Position in relation to noise is what defines it, far more than any inherent tonal or communicative quality. The question of order and stability is relative to the desire for change and transformation. What fits into our patterns is order, what resists them is noise.
From Australian artist Rowan Savage, the transformation comes about in another way: the boundaries between the sounds of human and natural language. Using a home-trained AI system to convert samples of Savage’s voice and field recordings of crows, the sonic texture of crows transforming into human vocals creates a distorted audio track punctuated by natural crow calls and the question Savage repeats on loop: Who am I? For me this speaks to the internal dissonance of noise in our heads, the cognitive friction that emerges from our desire to separate from the natural world (or to abandon claims to be separate, and the friction that rises in attempts to annihilate our ego and become something more fluidly connected).
There is a moment in the show — just after “The beautiful sounds are gathered together in one cluster,” that one can stop and look back at the path the visitor has just traveled. And from there, the “noise” is revealed as, paradoxically, a set of isolated, structured zones: Onuoha, JODI, and Paik, each work complex up close, now isolated, by distance, in a grid. That these zone are only apparent after you move through them reflects the psychology of noise, or moving through noise into order. Order appears in hindsight — a result of sense-making. The patterns become clear, the sounds in the din have been isolated.
Craftwork, Ancient Futures, in Signal to Noise, 2025. Photo by Phoebe Powell, courtesy of NCM.
And while it is the first piece in the show, it is perhaps worth discussing Craftworks “Ancient Futures” as the endpoint of this attempt at sense-making. Craftwork, another New York based duo, creates textiles entwined with fiber optic cables that capture colored light. Visitors are also offered a telephone to speak through, and voices are analyzed for key words and tones that come to be represented as colors in the cables. As more people contribute over the duration of the exhibition, the colored lights of past recordings will be replayed, weaving these stories into the textiles.
Ancient Futures anchors us in the most basic form of communication, that which takes place when people gather together to weave together, to chat, to share stories. The textile (itself evocative of computation’s origins in looms) connects this basic interpersonal communication to the escalating abstractions of communication technologies: the telephone, AI. The voice shapes the signal, and yet, the voice dissolves into the system. Noise can be directed in all kinds of paths.
As you exit the cavern of textiles, you will encounter the exposed threaded tapestry of telephone patch cabling hidden behind the removed back panel of a telephone switchboard from the 1920s. Operators, too, weaved communications through wires — a fabric capped by plugs to form circuits in a constant shuffle by human operators, connecting voice to voice in all directions. Traveling through these wires would have been the incessant din — here, of Australian voices dialing through switchboard operators.
Between the two is a quote from Cecile Malaspina’s Epistemologies of Noise:
“We will someday come to see noise as an inescapable freedom of choice.”
It’s an echo of Claude Shannon, whose 1948 Shannon-Weaver model for communication systems occupies a wall opposite JODI’s glitched desktops. The lines of the message are concentrated by the signal source, dissipating as they move from the source to the intended receiver. By the end of the process, our “message” is a scattered array of dots, the message obliterated in the media mess.
“Information is, we must steadily remember, a measure of one’s freedom of choice in selecting a message. The greater this freedom of choice, and hence the greater the Information, the greater is the uncertainty that the message actually selected is some particular one. Thus greater freedom of choice, greater uncertainty, greater information, go hand in hand.”
More information, more noise. More noise, more freedom to choose. And then what?
Signal to Noise runs at the National Communication Museum, Melbourne, from April 12 to September 11.