we are the robots

I was recently checking twitter when I found out about this thread by Cherie Hu, who runs water & music, talking about AI & Music through the specific angle of “where's midjourney/DALL-E/GPT3 for music????”.

https://twitter.com/cheriehu42/status/1590006092779905024?s=20&t=zmZmAvEAlv4JLv0m-_Y9Ug

Quite an interesting question. Since the recent breakthroughs in AI powered creative tools, there’s been 2 things on my mind. The first one is basically, how do these tools fit into the legal framework we’ve been using for centuries? Can an AI own copyrights over a work? Can a human being own copyrights over a work created 99% by an AI? These questions have found early responses in the US, and we’ll get back to these later. The second thing that’s been bugging me is less practical and more philosophical: what is art? What is it to “create”? There’s been plenty of books written on the matter, and I will definitely not revolutionize this field. I do however think we’re in an interesting moment where it’s worth to take a fresh look at these questions, and try to bring them answers that are still relevant in the era of AI powered creative tools.

The idea that “we’re standing on the shoulders of giants” has been expressed since the middle age, and popularized by Isaac Newton, to who people often mistakenly attribute its origin. Philosophers like Hobbes or Rousseau have used the concept of “state of nature” to describe an initial state pre-existing any form of civilization, to try to describe the genealogy of civilization and social constructs. Of course, these states are purely theoretical and serve as basis for thought experiments about social contracts and the birth of state as a structure. It’s also interesting when it comes to the history of creation and art. We could indeed theoretically pre-suppose the existence of a similar “state of nature” for art creation where there’s no previous “human created art form” to draw inspiration from, and where the first humans would create things from scratch.

This hypothesis is actually Plato’s working hypothesis about art. Plato pre-supposed that ideas are the pure form of things, and that ideas solely exist in their true form in the realm of ideas. The nature that surrounds us is just an imperfect attempt at materializing these ideas into things. Plato declares that there’s no perfect tree on earth. Every tree on earth is just an imperfect representation of the idea of what a tree actually is. For Plato, nature imitates the realm of ideas.

How does this relate to art? Well, Plato develops the idea that art is “an imitation of an imitation”. Art imitates nature, which itself imitates ideas. If we accept this pre-supposition, then even in a so-called “state of nature” for art, no one creates in a vacuum. Even hypothetical “first artists” based their work on models present in nature itself, and the verb used to qualify the work of artists, “to create”, is merely an abuse in language. Cambridge dictionary defines “to create” as “to make something new, to invent something, or to cause something to exist“. But no artist does that. Based on Plato’s definition of art, artists don’t create, they imitate, they recycle, they mimic. But they don’t create. The misconception that artists create, the ambiguity that lies in the word “create” is at the root of almost every hot discussion that has happened around the definition of artists and art. This is also one of the reason why we’re struggling so much to label AI generated art nowadays.

No art happens without tools. Musicians use instruments, painters use brushes, sculptors use chisels. Even the most conceptual artists use objects they transform into tools: human bodies, a space, a light. Tools are the medium that are used by artists to turn an idea into a material representation of that idea. Historically, tools were very rudimentary. They did not contain strong artistic proposals in themselves. A brush doesn’t offer you a strong artistic direction, it leaves you with a choice. Although an argument could be made that choosing a brush over a pencil for example is already an artistic choice, with the same brush you could paint renaissance, cubism, or impressionism type of paintings. Same goes for music instrument: different variations of the same instrument, piano for instance, or even different types of music instrument, don’t hold a strong artistic proposal in themselves. With a piano you can choose to play classical music, pop music, or jazz music. So for a long time, the artistic proposal had to come from the artist. Nothing happened without the artist mastering their tool and materializing their ideas, which means the main differentiator between artists resided in:

1/The faculty to have great ideas

2/The faculty to make these ideas a reality

For a very long time, the execution, the faculty to translate an idea into art, was the crucial element that defined artists. The mastery of a craft was what differentiated amateur artists from revered ones, because elementary tools were unforgiving and they required a high level of mastery. Regular people could feel that it was out of their reach, and respected it for the time and talent it took to reach that level of mastery. Having ideas was good, but it felt more accessible to anyone to have an idea. Painting the Sistine chapel, Sculpting Roman statues or playing a violin solo, this required craft. This required mastery. Art would be valued through this prism mixing effort, work, mastery, craft and of course subjective beauty of the final result. This has forged for a very long time the conception we have of art and artists in general, and how we assign value to a work of art.

Photography was a major milestone for visual arts. Art had been historically trying to represent nature & reality, and there came a technology that basically signified the end of the game. No matter how good you were, you could never be truer to reality than a picture. That’s basically what it meant and consciously or unconsciously, new art forms, more conceptual, started to appear as a consequence: impressionism, fauvism, cubism, surrealism… The quest for concepts and ideas was on. It was not anymore about depicting reality as you can visually see it, but reality as you feel it. Reality as an idea. As a symbol. In the 20th century, it was not about who could paint a tree in the most faithful way, but who could represent a tree in the most clever, the most unexpected, the most thought provocative, the most conceptual way possible. We started to value & reward the idea over the mastery of the craft. Rewarding the artistic proposal. This created conflict between people arguing that contemporary art wasn’t really art, and others who thought this was the purest form of art. But this was one of the first time we started to question our own perception of what art is, since for such a long time we thought it was about mastery of a craft & technique. But it wasn’t.

Ok, ok. You got me. So how does this all apply to music and creativity in music? Well basically, if you look at it retrospectively, music instruments haven’t evolved a lot over time. Not saying they haven’t evolved at all, but music didn’t really have its “photography” moment. However, music had its “computer” moment, and it changed things a lot.

Historically, tools in music (instruments) were similar to tools in visual art, in the way that they didn’t hold strong artistic proposal in themselves. They required strong mastery, but the artistic proposal had to come from the artist using the instrument. With the birth of computer aided music, especially through synthesizers, drum machines or electronic keyboards, this changed drastically. Have you heard about the Sleng Teng Riddim before? That’s a fascinating story really, and whether you’ve heard of it or not, I strongly recommend watching this video to understand the story better.

https://www.youtube.com/watch?v=R_Hko8LRdyo

It’s the story of young Jamaican musicians Noel Davey & Wayne Smith playing around with a Casio MT-40 electronic keyboard in the early 80s, and unexpectedly reinventing reggae. Here’s how it’s described in wikipedia:

The MT-40 has a built-in bassline (internally referred to as the "rock" preset) that, along with the keyboard's suggested 1/16 note fill for that preset, formed the basis of a seminal reggae track, 1985's "Under My Sleng Teng". The track's riddim went on to spawn nearly 500 cover versions. This song's success is widely credited with single-handedly transitioning reggae from analog to computerized production. This transition to a music production that depended on digital instruments and sequencers is also seen as catalyzing the computerization of hiphop music, which was, like reggae, an analog musical tradition until the mid-1980s. Given the lasting consequences of the Sleng Teng riddim, the MT-40's "rock" preset has been the subject of considerable speculation.

Presets in electronic instruments were some of the first forms of “artistic suggestion” embedded in tools. This was the first time tools used by artists had a strong artistic proposal, and were not just “medium”, or mere tools for the artist to produce their art. Seen through that prism, some people would think that this was almost like “cheating”. Imagine if Da Vinci worked from a pre-drawn Mona Lisa that wasn’t his? Wouldn’t that be controversial? Wouldn’t people think that he doesn’t deserve all the praises he’s getting for it?

Another interesting example of technology chipping away at the “master of its craft” type of artist. Although quite recent in history, DJs started with vinyls, mixing records and syncing their tempo manually. This is a craft that gave birth to scratch & turntablism. But over time, and with technology, vinyl started to get replaced by CDs. Then CD turntables added an auto-sync functionality. Then laptops & Serato took over, and everything became more or less automatized. Since the beginning of this shift, there’s been voices complaining that non-vinyl DJs are not real DJs, because they’re not “doing the work”. But aren’t they though?

Whether they are or not is not really the point. The interesting thing here is that this controversy shows how much we value art as “the mastery of a craft”. We want to admire artists, not only for their ideas, but also for how brillant they are at their craft. How out of reach they seem to us. If we feel like “if I had the same machines, I could do the same thing”, then the magic is gone. It’s obviously grossly underestimating how complex some of these machines are, but again, we’re only interested in the expression of this discomfort that is a symptom of how we value art, and that is extremely important to understand why there’s a debate around AI generated art.

Even though it creates controversy to this day (see critics of contemporary art), it is widely accepted nowadays that art is not only about “the mastery of a craft”. Remember that quote, “we’re standing on the shoulders of giants”? Well, this is basically how we see art curriculum nowadays. You study and digest the different art currents that have existed in the past, you train, and then you produce your own work of art. Wait, didn’t I just describe machine learning? Nobody creates in a vacuum, and as we’ve seen earlier in this post, even with the hypothesis of a theoretical state of nature, art would be merely imitating the representation of ideas we find in nature. So how come we consider it’s ok when a human digest the knowledge of the world and recycle it through art, but it’s not ok when a machine does it? How come we’re ready to assign copyrights to someone for their art, but we’re not when it comes to assigning royalties to work created by an AI?

First, and so that we’re clear, there are 2 levels of analysis here:

-The first one asks the following question: “Can we assign copyrights to an AI?”. As linked above, the US copyrights office thinks not. They state: “Current copyright law only provides protections to the fruits of intellectual labor that are founded in the creative powers of the [human] mind“. This answer is just perfect for this blogpost. We have just showed that this “creative power of the human mind” is at best a highly controversial way to put things, and strictly speaking a myth. It’s also interesting to deconstruct the logic behind it: AIs don’t exist without the “creative power of the human mind” since they’re trained with tons of data issued from past human minds. So if anything, the result of an AI produced work can only be the fruit of intellectual labor founded in the creative powers of the human mind. Except in this case, the intellectual labor is simulated through code, used as a tool. Also produced by a human. Other countries like Australia seem to take a different path though, as they’ve ruled that “A.I.-created inventions can qualify for patent protection“.

-The second one asks the following question: “Can we assign copyrights to a human for a work partially produced through AI?“. Very interesting question again. There’s been several events recently that have brought a partial response to that question. First, we’ve seen platforms like Getty image deciding to ban AI generated artwork from their platform over concerns regarding legal and ethical matters. But on the other hand, we’ve also seen the first US registered copyright for an AI assisted comics. The fundamental contention point seems to reside in this part: “AI-assisted”. How far is it assisted? How much human input was there? Take the case of one of Cosmopolitan magazine’s cover this year, created using OpenAI. There’s been a question around it regarding copyright protection. Whether the copyright should go to its human author, Karen X. Cheng, to the AI, or if there should be any copyright protection at all. Mike Wolfe, a copyright lawyer, had this to say on the matter:

"Where AI has played an essential role in the creation of a work, there are still pathways to some copyright protection. Even with a very capable AI, there will probably be a lot of room for human creativity. If AI helps generate a song and makes the bass line, but the creative professional makes it more complete by filling in gaps to make a cohesive piece of music, that act itself would likely give right to copyright on the basis of human authorship."

There’s still heated debates on the matter, but there seems to be a school of thoughts that consider that as long as there’s a human input, then there’s a trace of human creativity, and copyrights are in order. This basically follows the logic behind the copyrighting of the “Sleng Teng Riddim” cited above, and more broadly, any music generated using computers. To a lesser degree than current AI tool, computer softwares, especially in music, do contain multiple artistic proposals. Same goes for drum machines, synthesizers and other electronic keyboards that have contained presets for decades. These presets have been used countless times in now copyrighted electronic music production. In the “Sleng Teng Riddim” story video shared above, there’s a theory that the composition was inspired by Bowie’s “Hang on to yourself” track. It’s true that when you compare the two, it’s absolutely undeniable. But then imagine this: If it was not Casio’s Hiroko Okuda who had designed this preset, consciously or unconsciously inspired by her listening of David Bowie, but an AI that had “machine learned” the work of David Bowie. Would that make it wrong? The funny story here is, according to the Sleng Teng Riddim wikipedia’s page, Casio actually considered defending their intellectual property against the authors of the “Sleng Teng Riddim”. Given that this preset was highly inspired by Bowie’s composition, it would have been quite the irony!

The history of copyrights is really enlightening for our topic today. What really started the wheel for the recognition of intellectual property and the birth of modern copyrights was the invention of printing by Gutenberg. Until that point, copying somebody else’s work was no easy task, and the impact was de facto limited. To copy a book, or sheet music for that matter, you’d had to hire an educated person that would hand write the whole thing. Doesn’t really scale, right? But then printing came along and allowed exactly that: copying at scale. Almost effortlessly. That was the starting point of a technological movement, enabling humans to copy what they create always more effortlessly, fast, and at low cost (anyone heard of sampling?). This brought the societies of that time to conduct a reflection that would lead to the birth of the concept of intellectual property, which would define copyrights for centuries. As shown in Peter Drahos book “A Philosophy of Intellectual Property“, the modern concept of intellectual property as we know it today originates from legal battles that took place in 18th century England. Here are a few lines of justification used by judges ruling one of the emblematic cases, the Millar Vs Taylor case:

“it is just, that an author should reap the pecuniary profits of his own ingenuity and labour. It is just, that another should not use his name, without his consent.”

“It is wise in any state, to encourage letters, and the painful researches of learned men. The easiest and most equal way of doing it, is, by securing to them the property of their own works.”

“A third line of justification is presented by Aston J. He bases the existence of the right upon the fact that the author owns the produce of his mental labour”

These arguments are foundational to how we understand intellectual property. It is, in that regard, very interesting to see that the judge’s 1st statement above separates “ingenuity” and “labour” in what should be rewarded. Which loops back to what we have mentioned in our earlier part (The faculty to have great ideas & The faculty to make that idea a reality).

Modern intellectual property and copyrights were established not only to preserve authors financial stability in a society that consumes art and culture, but also to incentivize people to produce art, to produce knowledge, to produce culture, providing the environment that would assure them fair returns. The judge’s statement sustaining this argument is very insightful, as it cites the “painful researches”.

We see that judges ruling on this case put a strong emphasis on the effort, the labour, the work, that is required to bring the “original” to life, and this conception has infused the world of art, to the point where people started to question the validity of some artistic movements because there was not enough work, craft or labour required. But then in a world where AI is going to be more and more prominent, we can see, as evoked previously, that the current concepts of copyrights & intellectual properties won’t fit with modern artistic productions. There’s going to be, and there already is, a mismatch between the way art is being produced, and the way art is being protected. Where do we go from here?

Let’s try to gather our thoughts before we draw a conclusion here. We believe the following statements to be true:

-No one creates in a vacuum / We’re standing on the shoulders of giants / Creating = copy + transform + combine (I borrowed this last one from the short documentary called “Everything is a remix”, that I highly recommend you watch since it’s completely relevant to what we’re discussing here)

https://youtu.be/nJPERZDfyWc

-There’s an evolution of artistic tools from tools with little to no artistic proposal but requiring a high degree of craft mastery, to tools with increasingly stronger embedded artistic proposal requiring little to no craft mastery.

-Modern Copyrights & Intellectual property frameworks are rooted in the notion that art creation requires work and labor, and therefore protects artist’s work in a way which conflicts with our first statement (but as long as said artwork required effort/labor to be produced, it indeed works)

-Machine learning based AIs mimic human creative patterns: copy + transform + combine.

I’m quite pessimistic here, but this could very well be the end of the “artist lifestyle” as we know it. AI will soon have more ideas (if we see ideas as a mix of copy / transform / combine) than any human could have, and bring them into existence in a better quality than any human could do. So soon, humans won’t be able to compete with AI on either quantity or quality, and given what we went over in this blog post, I don’t see our legal systems resisting granting copyrights to AI for a very long time. Philosophically speaking, it doesn’t hold. There’s an argument to be made that without the sum of human knowledge, AI wouldn’t be able to create. That’s true. But it’s also true for every invention & discovery in this world, to which we still assign intellectual property and copyrights.

AI laid bare our creative process, and some people hate it for that reason. We probably loved to think that we were so special and inventive. AI, after humiliating humans in Chess & Go, has now decided to humiliate humans in art. But just like AI didn’t kill Chess & Go as games, I don’t think AI will kill art. Fundamentally, I think people connect with art for another reason: the human connection. We like to see the human emotion behind the art, whether it’s a drawing, a picture or a song. We like to imagine the emotions, or feel the emotions, the human artist behind it went through, and sympathize with them. Something we can’t do (yet?) with a machine: sympathize. My bet for the near future is that there will be a class of artistic asset that will be “guaranteed human generated“, and people will be ready to pay a premium to get that guarantee, in order to make sure the connection is genuine. Just like nowadays people pay experts to authentify artworks and make sure of the provenance, before they let themselves connect with the work of art (you absolutely need to watch the netflix documentary “made you look” on that matter).

What’s funny to me, and I’ll conclude with this thought, is that current AI tools use text prompt, which are basically ideas, and turn them into art. That’s basically taking a shortcut in Plato’s definition of art, where he sees nature as imitating the world of ideas, and art imitating nature. So in a way, by creating this direct path between ideas and art, AI created the purest form of art production.