Can AI Address The Melody Shortage?

The 12 notes of the standard musical octave means there are a finite number of combinations available for melodies, which means it’s increasingly difficult for musicians to write new songs without inadvertently copying older ones.

Such cases of “accidental infringement” are more common than you might think. A musician might imagine a tune and not be consciously aware that it’s a remembered melody, or it might simply pop into existence multiple times with no causal link to prior versions whatsoever.

Therefore, allowing artists to copyright their melodies means that the threat and lived reality of lawsuits amounts to a stranglehold on new creative expression.

So, two smart guys had a novel theory: Why not use AI to create every conceivable melody, “fix” them to disk and thereby assert copyright, and then dump the disk into the public domain so that nobody can ever claim ownership of them again.

Their initial stab at it, delivered by an LLC called All the Music, amounted to 68.7 billion 12-note progressions. 

Wait, so is copyright a bad thing?

Copyright, like patents and other intellectual property laws, is intended to protect creators’ work and thereby incentivize them to create. Such government interventions are inherently limiting, by intent and design, so they’re not consistently good or bad. For instance, starting in the 1400s they institutionalized monopolies and enriched royalty (they took a cut in the profits of the exclusive charters they granted), yet they can be credited for stimulating the entrepreneurial invention that gave us the 19th century’s Industrial Revolution.

The guys writing the US Constitution considered it important enough to the new country’s development that they listed it as a Power of Congress right after coining money and operating post offices.

And, at face value, why shouldn’t an inventor of a song, novel, or steam engine be able to derive proprietary benefit from their creation?

Musicians are the world’s original recyclers.

The limited number of notes in the Western music octave is not some new revelation, as musicians have been borrowing and reusing the one another’s and their own melodies since Day One. 

I’m a big Aaron Copland fan and he brilliantly used a melody called Simple Gifts in his Appalachian Spring, the former written in 1848 by two Shaker elders and the latter composed in 1945. Only the Shaker melody borrows from at least two songs written in the 16th century (one composer readily admitted he wanted to imitate what he imagined, and therefore possibly remembered, earlier folk music).

Mozart was notorious for reusing his “own” melodies, as do most of today’s leading pop stars. There’s a reason why their songs all sound familiar along with those of their wanna-be competitors, and sounding similar doesn’t have to be synonymous with being identical. The Rutles were band in the 1970s born of a comedy sketch that not only parodied The Beatles in their looks but performed songs that sounded sorta, kinda like Beatles hits. It was hilariously good.

There are many Reddit conversations dedicated to identifying riffs and songs that sorta sound like other ones. The emergence of sampling has made those activities and conversations (and lawsuits) even more commonplace.

But the vast majority of such recycling doesn’t end up in a lawsuit, and those that do are the result of multiple parties competing with one another (from composers and recording artists to publishers, record labels, and catalog owners). Few artists wholly rip off one another.

Copyright lawsuits have little to do with incentivizing or protecting creators.

AI to the rescue?

There are loads of problems with the All the Music project’s effort to copyright every conceivable melody.

For starters, it’s uncertain whether or not an AI can take credit for creating something because the applicable laws require a conscious human being play a role in it (it’s why animals don’t have rights). 

Second, that unconscious computing thing is central to asserting “accidental infringement” — the AI has no preconceived awareness of any songs — which is itself a dubious legal concept. One could imagine an AI working through every permutation for cola and “discovering” the formerly proprietary recipes for Coca-Cola or Pepsi, or writing code that overlapped with Apple’s MacOS. Or writing a million imperfect plays and one that’s identical to Shakespeare’s Macbeth.

Accidental infringement is in the eyes of the infringed?

Third, freeing the musical octave from onerous litigation would necessarily preclude anyone from claiming a part of it. If nobody can own a song, what’s to stop an AI from mimicking phrasing, intonation, orchestration, and other aspects of its performance? 

This is a far bigger issue for artists who’re already seeing their music (and literary works) imitated online. AI probably creates more of a litigation quandary than the Free the Music project hopes to resolve.

There’s no melody shortage.

Like so many problems today, copyright lawsuits over music are a people problem and therefore not solvable with AI or any other technology; in fact, tech might actually make it worse in this instance by multiplying many times over the opportunities for theft and mimicry. 

The Free the Music project deserves kudos for highlighting the weakness in the way the law understands creative content and recognizes infringement. If the solution they’re proposing is to blow up ownership entirely, they’re naive and wrong. 

There is no melody shortage. We’re just coming up short on the laws designed to incentivize and protect those who create them…however familiar those notes might be.