When a singer’s pitch is corrected by Auto-Tune, did Auto-Tune cheat?
The question sounds slightly ridiculous because the grammar gives away the problem. Auto-Tune did not decide to correct the performance. It did not decide how much correction to apply, whether the effect should be audible, or what should finally be released. It was used.
That seems obvious now.
It was not always obvious with the microphone.
When electrical amplification changed popular singing, critics did not simply object to a new sound. They objected to what the machine appeared to permit. Before amplification, a popular singer needed a voice capable of projecting into a theater. The microphone changed the physical requirement. A singer could come close, lower the voice, and sing with an intimacy that would previously have disappeared before reaching the audience.
Crooning followed.
The reaction was fierce. The New York Singing Teachers Association publicly condemned the new vocal form. Cardinal William O’Connell of Boston called crooning base, degenerate, and un-American. Even “crooner,” when applied to white male singers, began as an insult, carrying suggestions of artlessness and emasculation.
The equipment seemed to be doing something the singer could not.
The microphone had its Auto-Tune period.
Then the microphone became ordinary.
The technology did not stop affecting the performance. It still determined what could be heard. It still changed distance, intimacy, volume, texture, and the kinds of voices that could succeed. What changed was our ability to place it. The microphone became part of the process rather than the culprit in it.
Auto-Tune arrived at another boundary.
Andy Hildebrand had worked with signal-processing techniques used to interpret seismic data in the search for oil. The same mathematical approach could be applied to the pitch of a human voice. In 1997, Antares released Auto-Tune, initially as a way of correcting pitch unobtrusively.
Then, in 1998, Cher released “Believe.”
The correction became an effect. Instead of disappearing into the recording, the technology announced itself. The voice snapped between pitches in a way that immediately sounded new.
And almost immediately, something revealing happened.
The producers did not initially tell people how they had made the sound. They attributed it to another kind of processor, protecting the technique they had discovered.
That gives us two different acts.
Auto-Tune altered the voice.
Someone decided what to say about how the voice had been altered.
Only one of those statements needs a moral verb.
Did Auto-Tune cheat?
No.
Could Auto-Tune be used in a performance someone considers deceptive? Of course.
But those are not the same proposition.
The distinction becomes harder to maintain as the tool becomes more capable. A microphone amplifies a performance. Auto-Tune can alter one. Generative AI can produce substantial parts of the finished object itself. As the contribution of the tool grows, the temptation grows with it to make the tool the grammatical actor.
The machine wrote.
The machine copied.
The machine imitated.
The machine stole.
Perhaps some of those sentences are useful descriptions of process. But when they become accusations, something has happened to the location of agency.
There is a user.
That word matters because it does not require us to decide what kind of entity occupies the position. “User” is not another word for person. It describes a relationship: something is using something else toward an end.
That is enough for the distinction.
When an AI-generated song or video appears, the sophistication of the technology can make the machine seem like the obvious place to locate the act. But the existence of a powerful tool does not make the user disappear. Someone used it to make this object. Someone chose this result rather than another. Someone placed it where we encountered it.
We do not need to solve every question about AI to notice that.
We do not need a theory of copyright.
We do not need to settle how models should be trained.
We do not need to decide who should be compensated, what should be disclosed, or how responsibility should be divided among everyone involved.
Those may be important questions. They are simply later questions.
This one comes first.
When we encounter something made with a new technology and dislike what has been done, where do we put the verb?
We once put extraordinary weight on the microphone. Eventually we learned to hear the singer through the equipment rather than treating the equipment as the singer.
Auto-Tune made the distinction difficult again. It could alter the performance rather than merely transmit it. Yet even there, the distinction survives. Auto-Tune can change a note. It cannot decide to cheat.
AI makes the distinction harder still because the tool can contribute so much more to the finished object. It can make the instrument look remarkably like the actor.
But complexity of process does not require confusion of roles.
Before blaming the tool, locate the user.
WE&P by: EZorrillaMc&Co.
Sources
- Allison McCracken, Real Men Don’t Sing: Crooning in American Culture — on the microphone, crooning, masculinity, and the cultural backlash against the new vocal style.
- “Crooning Comes by Nature,” The New York Times, February 24, 1932 — contemporary account of Cardinal William O’Connell’s attack on crooning and the New York Singing Teachers’ Association’s condemnation of it.
- Antares, “What Is Auto-Tune?” — history of Andy Hildebrand’s seismic signal-processing work and the 1997 release of Auto-Tune.
- Sue Sillitoe, “Recording Cher’s ‘Believe,’” Sound On Sound, February 1999 — on the production of “Believe,” the deliberately audible vocal effect, and the producers’ initial attribution of the sound to a Digitech Talker rather than Auto-Tune.
Travel in Life 