The Label We Currently Demand of AI

What does knowing the tool tell us about the person who used it?

What puzzles me is how much our response can change once we learn that AI was involved. A song may touch a memory or give expression to something we have struggled to say. Then comes the disclosure, and we reconsider what we heard. There may be good reasons for that reconsideration. We may value human performance, care about how the work was produced, or object to the practices behind a particular AI system. Still, I wonder what happens when the label settles the question before we have asked what the person contributed. “Made with AI” labeling is often used to describe any AI involvement in the creative process. That can seem to explain everything while actually telling us very little.

What Earlier Tools Can Teach Us

My background in photography makes the camera a useful place to start. Before photography, someone wanting a recognizable portrait generally needed a person who could draw or paint. A camera changed that requirement. It placed an optical and chemical process between the person making the image and the finished portrait. But creativity remained in the choice of subject, composition, lighting, timing, and what the photographer wanted us to see. Someone could also point the camera carelessly and produce a forgettable picture. Owning the equipment did not settle whether there was anything worth looking at.

Digital photography changed the process again. Software now performs adjustments and combines information that once required considerable darkroom skill, or could not have been done there at all. We generally accept these capabilities without requiring a complete account of the processing behind every image. We do care when a photograph is presented as evidence of something that did not happen. A constructed image offered as an illustration and the same image offered as a news photograph make different claims. In that case, knowing how it was produced helps us judge whether we are being told the truth.

     Music offers similar examples. Multitrack recording allows a finished recording to contain performances made separately, perhaps days apart. Synthesizers produce sounds without the acoustic instruments they may resemble. Sampling brings earlier recordings into new compositions, while pitch correction and digital editing can reshape what a performer originally sang or played. These practices raise their own questions, and some listeners care deeply about the answers. Yet we have found room for a wide range of methods. A recording need not reproduce a single unaltered performance to be meaningful music.

     These comparisons have limits. A printing press reproduces words that someone has already written. A camera records light, even when software later does considerable work with the information. AI can generate the words, propose the melody, and produce the voice singing it. It participates in choices that we associate with imagination and artistic judgment. I cannot reasonably ask people to overlook that difference simply because we have accepted other tools. The history of those tools is useful because it shows that creative practices can change, but it cannot answer every question AI has brought with it.

Where the Person Enters the Work

Consider how differently people already participate in making a song. A lyricist may work with a composer. A singer may interpret words and music written by others. A producer may shape an arrangement, assemble performers, and guide a recording toward a sound the songwriter could not have achieved alone. We can recognize each contribution without pretending that everyone did the same work. In fact, credit becomes more useful when it tells us who did what.

     AI introduces another way of distributing those tasks, although the machine is plainly different from a human collaborator. A person might develop the subject, write and revise the lyrics, listen to many musical versions, reject most of them, and keep working until the song expresses what they hoped to communicate. Another person might type a brief request and publish the first result. Both could receive the same “AI-generated” label. That label tells us something real about their production methods, but it leaves out much of what distinguishes the two efforts. The amount of human thought, revision, and responsibility matters to my understanding of the work.

     The subjects a person wants to explore — faith, love, our relationship with technology, or what it means to remain human — may have occupied that person long before any machine was consulted. AI can help develop and express those concerns, sometimes by offering possibilities the person would not have found alone. Acknowledging that contribution need not make the person's questions, judgment, and purpose incidental. Both parts of the process deserve an honest account.

     There is a difficulty here that we need to keep in view. Caring deeply about a subject does not make every resulting song good, and spending hours choosing among generated versions does not mean someone composed the music in the usual sense. Selecting, directing, writing, arranging, and performing are different activities. The term “songwriter” may need qualification when it conceals those differences. A clear description of the contribution seems more useful than an argument over a title. That same clarity should give the listener a fair chance to recognize whatever human contribution is actually present.

What Disclosure Ought to Tell Us

There are circumstances in which AI disclosure is essential. If a synthetic voice is presented as a particular singer's actual performance, the listener has been misled. If an invented image is offered as documentation, its origin matters. A competition intended to assess human composition has reason to specify which assistance is permitted. People also have legitimate concerns about the creative work used to develop AI systems, the conditions under which it was obtained, and whether its creators were treated fairly. Those concerns cannot be answered merely by saying that the resulting song is beautiful.

     An audience may also value the human act itself. Listening to someone sing can mean more than enjoying a sequence of sounds; it can mean hearing breath, effort, interpretation, and the presence of another person. A handmade object may matter to its owner partly because someone patiently made it. Knowing that a voice was synthesized can therefore change what a listener values in a recording without making that listener narrow-minded. Taking the contribution of an AI-assisted creator seriously should also mean taking the audience’s reasons for valuing human performance seriously.

     What I question is whether a broad label gives them enough information to make that judgment. “Human-written lyrics; AI-generated music and vocals” describes a process much more clearly than “AI-generated song.” So does “Human performance of an AI-assisted composition.” These descriptions distinguish contributions that the larger category puts together. A disclosure can be accurate and still be incomplete. If we want transparency, we ought to want an account that helps us understand the work rather than one that leaves us to fill in the gaps with our assumptions.

     There is an economic concern as well. Writers, musicians, illustrators, and designers have invested years in skills through which they support themselves. A tool that produces a plausible substitute quickly and cheaply can threaten their livelihoods, regardless of what a particular user hoped to accomplish. Capabilities that create opportunities for one person can threaten the livelihood of another. Their objections deserve a fair hearing. Opening creative opportunities to more people is something I welcome; how we do that, and what happens to the people already doing this work, remain serious questions.

What Happens to the Song We Heard?

Suppose a song has already moved us before we discover how it was made. The discovery may change our willingness to support its maker or our admiration for the accomplishment. It may even change how the song feels the next time we hear it. Our experience of art includes what we know about its origin, and that knowledge can deepen or diminish our attachment. But the original encounter also deserves acknowledgment. A memory was awakened. A feeling found expression. Whatever else we conclude, something reached us before we knew the production method.

I think this is part of what makes the subject difficult. We want to know that there is someone behind the work who understands, or at least has lived near, what it communicates. A machine's ability to produce convincing language or music does not establish that kind of experience. Yet a human being using the machine may have it. The question is how that person's experience entered the work and how faithfully the result expresses it. An AI label alone cannot tell us. Neither can the absence of one guarantee that a work has depth, honesty, or much thought behind it.

     For people who have something they want to express but lack the means to produce it, these tools create possibilities worth considering. They do not confer talent automatically. They do allow a person to try, revise, listen, and discover whether an idea can become something another person finds worthwhile. That opportunity matters to me. I suspect it matters to many others who have carried thoughts, stories, or songs without knowing how to give them a form they could share.

     I do not know where our understanding of AI-assisted creativity will eventually settle. We may come to regard some uses as ordinary while continuing to distinguish human performance and largely machine-generated work. Some practices may prove unacceptable for reasons that become clearer with experience. I hope we can make those distinctions with enough care to preserve both honesty and the possibility of discovery. “Was AI involved?” is a useful question. It is the beginning of an inquiry, and I would like us to remain interested in the answers that come after it.

     A creator can explain what they brought to the process and what the technology supplied. Then I hope there is still room to listen. If a song leaves someone thinking about a person they love, or an essay helps them examine a question that matters, I would like to understand that response too. It may tell us something about the human being who made the work, the human being who received it, and what passed between them.

---

A Note on This Essay

This essay began with my questions and concerns about the labeling of AI-assisted creative work. I explored them in depth via dialogue with ChatGPT. I defined my "writing voice," and it assisted in developing the argument and composing much of the text. I edited the resulting completed text. The questions and purpose are mine; the assistance in developing and expressing them was a joint effort between me and AI. This is the kind of collaboration the essay asks us to consider. Knowing that now, I wonder whether it changes how you understand what you have just read. — Glenn Stockton