Recently on my way to the Detroit airport, I noticed a sign on the bus: “Please use earbuds or headphones to further limit noise.”
Not a single person was talking; it was a silent ride, but what struck me was the assumption that the sound of people talking on the phone was unwanted. Had we been on a similar bus in another city or country, perhaps no amount of signage would have prevented loud phone conversations, on speaker no less.
The key word here for me is noise. What is noise in one culture is incidental or environmental in another.
As a scholar of music who teaches a class called “NOISE,” about the evolution of noise in music and culture, I look at this concept differently from many Americans.
To understand the difference between noise, sound, and music, we have to look at history. For hundreds of thousands of years, humans lived in the age of melody—birdsong, human vocals, animal calls. Think 43,000-year-old bone flutes and the Hurian Hymn no. 6, a 4,000-year-old Sumerian clay tablet, the first known notated music.
Melody still reigns. Playing oldies to Alzheimer’s patients helps trigger memories. Earworms proliferate on TikTok. Memory and melody are entwined.
Yet noise persists, even more so today than during the Industrial Revolution, when machines took over the sonic landscape, and it’s divided along class and cultural lines. In 2017, a NIH study found that urban noise pollution like industrial activity or traffic noise is worst in poor and minority neighborhoods, which negatively impacts health outcomes and student academic performance. Noise can also be used as a weapon. Heavy metal music was used as a form of torture in Guantanamo Bay.
But what is noise to some is music to others. Some of my students listen to metal all day. And what many people deem high art—classical music—is used to deter the homeless population in places like Port Authority in New York City.
We need to be talking about noise. Because with the rise in AI, noise is even more important. In fact, every machine learning model needs noise, or else it overfits. That means the model is too tailored on the data it was trained on and performs badly on new, unseen data. Noise is of utmost importance to AI.
The most powerful AI image-generation models are essentially de-noising algorithms. They’re called diffusion models, and they “diffuse” training data with random noise, then learn how to reverse that process to generate new material, effectively “denoising” any new data into something meaningful. It’s what wowed the Internet three years ago with realistic image generation—text-to-image models like DALL-E 2, Midjourney, and Stable Diffusion. When trained on a lot of data, these diffusion models can create something recognizable from seemingly nothing That’s not just useful for generating fantastical memes; diffusion models can also generate detailed MRI scans from sparse data or 3D molecular structures for drug discovery.
Noise can be harmful, but it can also be beneficial. It can be harnessed. I’m not saying that noise should automatically be considered music, nor the other way around. That equivalence requires a certain intention on our part, through active listening. But at the end of the day, things that lend realism are these incidental details: the specific noise profile of a Parisian café with its glass clinks and spoken French “zh”s, or North Seymour Island of the Galapagos with its large population of blue-footed boobies and swallow-tailed gulls whistling and clicking.
Perhaps counterintuitively, we need noise in our political conversations too. Our newsfeeds and social circles are so insulated, we no longer have to wrestle with other points of view that can crystallize our own thoughts and generate new ideas or even compromises. Fittingly, public spaces in which there is the potential for political pluralization, conversation, and demonstration—like parks, town squares, transportation hubs—also happen to be places that are sonically noisiest.
We should remember that podcasts, speeches, and other forms of media are a dialogue. How you receive information, perhaps using difficult and skillful work to parse it, is the muscle we exercise when we learn to appreciate noise.
We should attend to noise. It not only helps us sleep, in the form of white noise machines, or makes for a more interesting bus ride, but it also provides the necessary friction and chaos that make our lives more human.