The Amen Break and AI art
The Amen Break and AI art
If sampling can be art, so can Large Sampling Machines
In 1969, The Winstons, an American R&B band, released their debut single ‘Color Him Father.’ It hit #7 on the Billboard Hot 100, #2 on the R&B charts, and won its composer Richard Spencer a Grammy. Alas for them, it was a one-hit wonder, and The Winstons would have been consigned to the dustier footnotes of musical history… if something truly extraordinary had not happened to that single’s B-side.
“Amen, Brother” is not a great song. It did not chart. But eighty-six seconds into it, drummer Gregory Coleman unleashed a drum break that will sound very very familiar, because those six seconds, known today as “the Amen break,” ultimately made that B-side the most-sampled song in musical history by far.
Sampling erupted into the public consciousness in the 1980s, with the simultaneous rise of hip-hop and new hip-hop-enabling technology, such as the E-mu SP-1200 and Akai MPC workstations / samplers. The backlash was immediate and furious:
A few choice quotes from Virginia’s Daily Press Newport News when they asked readers their opinions on rap, sampling, and art in 1990:
“I think sampling in rap music is disgusting. It’s definite copyright infringement and the absolute opposite of creativity.”
“I think sampling in rap music stinks. It’s out and out plagiarism.”
“I personally feel that what makes a musician is talent and originality. Rappers who use other people’s hard work have neither.”
“Rap is trash. It’s all trash. It has nothing to do with music and nothing to do with art.”
Conversely, as art critic Andrew Ross writes in his essay ‘Princes Among Thieves’:
“For every layperson’s casual dismissal—“it’s not real music”—there was a musician who saw rap as a threat to his or her livelihood as a performer … Most working musicians, their labor power directly threatened by the growing popularity of the new DJ culture, had little patience for [sampling] …
The courts ultimately agreed sampling was a copyright violation; since a 1991 decision that invoked the Seventh Commandment (“Thou shalt not steal”), US musicians have been required to get, and usually pay for, approval from the work’s rightsholders. I’m sympathetic, though I think a compulsory mechanical license would have been better. I see the fair-use argument, but samples are outright copies … and Gregory Coleman, the drummer whose Amen Break features in more than 6,500 songs, died homeless and penniless.
The question “can great art be built on sampling?” has a clear answer, though, and that answer is: hell yes. Pre-1991 records that were wall-to-wall samples, such as It Takes A Nation Of Millions To Hold Us Back and 3 Feet High And Rising, are today widely acclaimed as among the greatest albums of all time.
1980s hip-hop was of course only one incarnation of a longstanding artistic tradition. Visual collage is very old indeed. Avant-garde French musicians used existing music in their “musique concrète” experiments in the 1940s. “Is sampling art?” is actually an ancient debate, which rears its head anew almost every generation…
…which takes us, of course, to AI art.
It is an article of vociferous faith among many people that ‘AI art’ is an oxymoron; that it is simply not possible for generative AI to contribute to meaningful creativity; that, paraphrasing, it too “has nothing to do with art,” just like sampling in the 1980s. No less an authority than Ted Chiang, one of our greatest writers, advances this thesis in a recent New Yorker piece:
How good could [LLMs] get? Could they get better than humans at writing fiction—or making paintings or movies […] Whether you are creating a novel or a painting or a film, you are engaged in an act of communication between you and your audience […] the fact that you’re the one who is saying it, the fact that it derives from your unique life experience and arrives at a particular moment in the life of whoever is seeing your work, is what makes it new. We are all products of what has come before us, but it’s by living our lives in interaction with others that we bring meaning into the world. That is something that an auto-complete algorithm can never do, and don’t let anyone tell you otherwise.
This is so weird — that an imagination as great as Chiang’s comes to this conclusion (and articulates it, by his lofty standards, not especially eloquently.) He actually cites “film” as an example, and previously in the same essay writes
When photography was first developed, I suspect it didn’t seem like an artistic medium because it wasn’t apparent that there were a lot of choices to be made; you just set up the camera and start the exposure. But over time people realized that there were a vast number of things you could do with cameras, and the artistry lies in the many choices that a photographer makes. It might not always be easy to articulate what the choices are, but when you compare an amateur’s photos to a professional’s, you can see the difference. So then the question becomes: Is there a similar opportunity to make a vast number of choices using a text-to-image generator? I think the answer is no.
Again, this is so weird. It is true that art comprised of generative AI outputs will be, definitionally, remix art, like sampling for 1980s hip-hop … except stochastic sampling, because LLMs are, at heart, large sampling machines. (Which would be a better name for them.) But that doesn’t mean it won’t be art!
Of course there will be “similar opportunity to make a vast number of choices using a text-to-image generator.” There are today. Chiang in fact cites an example of someone doing just that! … and dismisses it on the grounds that it’s only art if it “enable[s] extremely fine-grained control over the image you’re producing” and, extra weirdly, “[OpenAI] probably isn’t trying to build a product to serve users like that” — in a piece otherwise about whether AI art is even theoretically imaginable.
But sampling has never offered ‘extremely fine-grained control.’ It’s so bizarre to see someone like Chiang on the side of, basically, ‘sampling is trash and has nothing to do with art.’ In fairness, though, in his métier and mine, fiction, sampling / collage has always been limited to extremely-out-there avant-gardeness. I too think authors should not use generative AI to write fiction; but then my literary advice has been “Write only what only you can write” since well before LLMs appeared.
Chiang’s essay was widely cited, many loudly agreeing “Yes, just so!” But it is not so. Modern AI won’t generate out-of-distribution work, i.e. that beyond the boundaries of its training data … but it will let us generate and remix an infinite number of in-distribution samples — and these remixes can, and will, be genuinely creative art.
It is especially not so for film, which, again, Chiang specifically cited as an example. Modern movies have an average shot length of ~2.5 seconds. A feature-length film is composed of literally thousands of individual shots. Imagine a person using a text-to-video generator — Runway, Kling, OpenAI’s Sora (contradicting yet another Chiang argument), Google’s Veo, etc. — to generate (and then re-generate with a new prompt, and again, and edit, and again) every shot of a full-length movie.
Yes, that’s a lot of work, but less work than writing a novel. Yes, each shot will be a stochastic sample from all the videos the model has been trained on, and (to a greater or lesser degree, depending on the prompt) somewhat generic. But this is also true of almost every individual shot in every movie today!
It is true that today’s AI models cannot yet generate a coherent cinematic story. Persistent characters, settings, and objects are not yet available. But they’re getting there at speed. Full-length generative-AI animated movies are closer than most expect … and while it seems many people don’t want to believe this, the inescapable reality is that when they arrive, their 3 Feet High And Rising won’t be far behind. In less time than you might expect, the argument won’t be “can we make art with generative AI?” but rather, as with 1980s sampling, “how could people ever have believed otherwise?”