And how it portends future forensic failings when AI begins to compose at scale.
I’m not originally from the Bay Area. The phone number at Musicologize starts with a 212 area code so I’d prefer to quote Casey Stengel or similar, but today we’re going to borrow from Al Davis and say “a win is a win,” even if it wasn’t quite perfect.
Obviously, the verdict in Vince Vance’s “All I Want For Christmas Is You” case against Mariah Carey was the correct verdict. As I said in my first article on the matter, it only takes ten seconds to explain:
Truthfully, a great deal of musicological analysis can be effectively summarized as that second one, but forensic musicologists have to avoid reliance upon an “appeal to authority,” so it’s important to explain and I’m fond of saying, “teach.” And this is why I want the arguments to be sound. In fact, truth be told, I want them to be unassailable.
This case got tossed at summary judgment. What if it had not? It would’ve been because specious arguments and illusory ideas raised the spectre of uncertainty beyond some threshold. In other words there’s too much at stake for there to be so much in play. Consider the observation attributed to the plaintiff’s expert:
Many images and phrases are carried over from Stone 1989 to Carey
1994, including “mistletoe” and “Santa,” “sleigh bells” and “snow,” “asking” for something, the
phrase “I don’t want” or “I don’t need,” a “dream” or “wish” that might “come true,” the exact phrase
“underneath the Christmas tree,” and, of course, the exact title phrase “All I want for Christmas is
you.” Many of these are traditional seasonal images, of course, but the piling up of substantially
similar turns of phrase is striking. — Taken from the defense expert’s rebuttal report of the plaintiff’s expert’s report.
Presumptive phrasing like “carried over;” the fact that there’s hardly a “turn of phrase” listed there, and the casual and at best, again, presumptuous and I’d certainly argue inaccurate use of terms critical in copyright discussions — “substantially similar” and “striking” — these all offend my sensibilities.
And in it I see the ghost of the inverse ratio rule. The now largely discredited inverse ratio rule had in it the presumption of copying, and it’s worth examining it’s fallacy, because that’s the job. The job is to help the court arrive at its verdict with a clear understanding of what’s true. If you arrive at what’s true by faulty logic, you got lucky, and ultimately leave yourself open to review or reversal. When common words, short phrases, and holiday cliches “pile up,” as I’ll concede they do here, is that something? Possibly. Just as the inverse ratio rule suggested that when we question “access” — in this case, “do we imagine Mariah Carey heard Vince Vance’s song?” — and find, “yes, it’s on the radio, she probably heard it,” then we require less similarity to infer that she copied it, which is no better than plausibly not inconsequential, we might call all those phrases smoke enough of which is tantamount to fire. But that makes little sense. You’re allowed to be at the bank. Of course, you might be there to rob it, but are you? It’s a lawyers job to argue yes. It’s an experts job to be far more measured and accurate.
Equal opposite fallacy is “filtering.” Filtering in this context is a form of prior art argument where often it’s posed that a plaintiff’s preliminary report is inherently faulty because an economy of prior art research is met with an insisted assumption that if prior art research had been performed, and then the plaintiff’s music essentially redacted to not include any fragmentary musical elements that could be found elsewhere, there’d be little to nothing left of the plaintiff’s work to consider. To say the baby gets purposely thrown out with the bath water hardly tells the story. I’ve covered why this sort of strategy would be terribly wrong, especially if indeed the appeal is strategy.
Speaking of which I also do not want to stand idly by as the plaintiffs here besmirch the idea of soundalikes and imply, certainly as I inferred it, that it’s a form of infringing copying. A soundalike is a song that’s meant to sound like another song. You could do it for nefarious reasons or you might just be imitating something unprotectable by copyright and unassailable morally, as in when Robin Thicke says to Pharrel Williams — and I’m not going to look this up and get it word for word, “We should do something like Got To Give It Up!” which was of course what led to “Blurred Lines,” and that admission, I think from his deposition, surely led to, I think, the only major case I’ve ever predicted incorrectly. And I’m soothed in that it’s the dumbest and wrongest copyright infringement verdict in 30 years.
Let’s say this: copying is a necessary component of infringement. But copying is not inherently wrong. Pharell and Thicke were not wrong to copy aspects of an old song they admired. Indeed, it was noble and awesome of them to bring those aspects back in a new form. More than sufficient creativity should’ve been an absolute defense.
And here in Carey v Vance we have another inverse-ratio rule type of failing. The idea that in the signature phrases, “All I Want For Christ-” the notes from Vance, scale degrees 6, 7, 1, 2, 3, are a manipulation of Carey’s 3, 2, 1, 6, b6. If I sing fa, ti, do, re, mi, is that interestingly similar to mi, re, do, fa, le? Of course not.
(I think that’s the solfege for b6, but really who cares?)
But does it matter how those notes in a sorta kinda reverse order, but not really because we’re talking 1-2-3’s and 3-2-1’s mostly? Again I’m going to take from the defense’s presentation of what the plaintiff’s expert put forth, that this different series of notes…
“…betrays the sorts of manipulation that creators of “soundalike” tracks (tracks which are designed to mimic the sound and feel of famous songs without actually infringing on them) tend to impose on the tracks they copy. A classic trick, used here, is to preserve the basic contour, rhythm, and patterns of tension and release, but flip one or more sections of the tune upside down. Where the original goes up, you go down, and you’re safe. But this is a risky strategy…” — Taken from the defense expert’s rebuttal report of the plaintiff’s expert’s report.
Warren Buffett likes to say, and I think he got it from Maslow, “To a guy with a hammer, everything looks like a nail.” I’m thinking, to a forensic musicologist who has concluded infringement, do-re-mi may look like a manipulation, an imposition, a trick, and a strategy, all to disguise the unoriginality of mi-re-do. But really, mi-re-do is just three blind mice.
These are understandable failings though; just failings to consider and overcome.
Logic that too soon emphasizes its conclusion, and finds that it’s looking for the keys under the streetlamp because the light is good. Or confuses motive with method, like arresting someone for wearing gloves at the scene of a crime when they might be there to shovel snow.
This is going to explode when AI really gets cooking. These tools require prompts. And those prompts are going to be akin to, “Give me something like ‘Got to Give It Up.’”
So what happens when the output does sound like ‘Got to Give It Up’?
The hard fact is that the prompt containing the phrase isn’t damning. That’s backward. It’s the ghost of the inverse ratio rule. Infringement happens at the output, even if the input is a necessary ingredient. It’s directionally logical. And so should be the forensic musicology. That will take ever greater mindfulness.
