PhilOnAlgorithmicMusic

ThoughtStorms Wiki

Context : AlgorithmicMusic

Compare with : AIPainting

I was interviewed by a Brazilian journalist about it, and here's how it went down.

Let's look at the bigger picture.

What is an "algorithm" in art?

In a sense, an algorithm is a formalized representation of a "process".

And process or "technique" is something that has always been an essential part of art. The Greek word "techne" (which was used for "art" and "craft") is the root of both "technique" and "technology" etc.

And in history, an artist was a "craftsman" who learned the technique of making particular things : pots, paintings, music, rhetoric etc.

Then two things happened :

  • the Renaissance : which created our modern idea of the artist as someone with higher "sensibility" to the world, a "genius".

This strand of art history has led through the Romantic period to the 20th century to the idea of art as the outer expression of some "inner-being", inherent in the artist. Something personal, something beyond any technique that can be explained or learned.

  • the industrial revolution, the invention of photography and other ways of mechanically / chemically / electronically producing representations of things.

Photography is widely seen as having shocked painting out of its previous role of "representing" the world. Why try to capture a portrait or a landscape in paint when photography can do it more accurately? And so it forced art to find a new role.

Something very similar happened with recording and music. Why try to find the right singing technique to represent a love-sick young woman in an opera, when you can put a microphone in front of a real love-sick young woman and get a better "picture" of that emotion that you can sell on a disc? That's the origin of 20th century blues, jazz, rock and pop. An attempt to "photograph" real, authentic emotions rather than "fake" them using "technique".

But technology is also part of a historical move towards "abstraction", that is, extracting and writing down the "rules" for making a piece of art, distinct from actually making it. Historically we have been abstracting the rules for how we produce art out of the process since the Renaissance too. The rules of perspective in painting are an abstraction and formalization. Separating the musical score from the musical performance is an act of abstraction. It's this abstraction of the score that let classical composers reflect upon and develop more complex rules for harmony. Classical music is the result of the development of musical notation.

There's no need for a machine.

Everyone you read on this will talk about Mozart's "Musikalisches Würfelspiel" (https://www.youtube.com/watch?v=nFVmh9WNnDw), a game for creating music by throwing dice. Once you have rules which can be notated and expressed separately from the music itself, then you can use those rules to generate new examples of the music.

This is "algorithmic art"

Now the 20th century, particularly music, became obsessed with the abstract rules or compositional process for music. By the end of the 19th century, composers like Mahler and Strauss and Debussy had pushed and stretched tonal theory to its limits. The next generation of composers wanted to invent new rules.

Schoenberg invented "12-tone" or "serialism" which was the creation of a new set of rules to compose by.

Meanwhile, John Cage was experimenting with various new processes to generate music. His Imaginary Landscapes, starting in 1939, included record players, random objects, electronic circuits, radios and tapes of other people's music, to generate, often unpredictable, sounds. (https://en.wikipedia.org/wiki/Imaginary_Landscape) He used the I-Ching to help compose piano music (https://en.wikipedia.org/wiki/Music_of_Changes), he wrote the infamous 4 minutes 33 seconds, not to play "silence" but to make music from the background noise in the concert hall (https://en.wikipedia.org/wiki/4%E2%80%B233%E2%80%B3).

Pretty much all innovation in music composition since then has been about experimenting with process.

In the erudite / academic music it's very explicit that most new composers come with some kind of new theory. In popular music, each generation's new sound is often based on a new technology. And each new technology embodies a new theory of how music is made. Using close microphones to get "crooning". Or electrical amplification to allow a guitar to fill a stadium. Or a turntable to make hip-hop out of existing recordings. Or a computer with the latest software which makes pretty much everything in popular music today.

In the visual arts we can see Duchamp's invention of the "readymade" (ie. finding an existing object and putting it into the gallery), the Dadaist and Surrealist games with chance and explorations of the unconscious, Moholy-Nagy's "Telephone Pictures" (https://smarthistory.org/moholy-telephone/), Jackson Pollock's "Action Painting", everything by Fluxus, Richard Serra's sculptures based on particular actions. And basically most of the art of the 20th century that people felt was important and exciting, was largely based on play with new processes for creating art. That goes for the more explicitly technological art made with computers or cybernetics or living beings etc.

OK ... that was long. But I'm explaining this because I want to make two points : the first is that "algorithmic art" is really just a continuation of this long tradition of abstracting the rules for the process of art out of the execution / performance or "making" in art. It's not a separate thing. All art is, in some sense, "algorithmic" today.

The second. We have this dichotomy, between the idea of the algorithm and the idea of art as the pure expression of the inner genius of the artist.

In a sense, these are two sides of the same coin. It's like "beta decay" where the neutron splits into an electron and a proton. Art's "technique" split in two. Part became the "abstraction", the "ideal of the process", or algorithm. The other part became the personal expression, spontaneous "genius" of the artist.

But both of these are the result of the "breakdown" of "technique"

If art was still constrained by learnable technique, then you couldn't have either the idea of pure process. Nor the idea of pure expression. Technique would bind the two together.

Does this make sense?

So. Is there an opposition between people who want to use algorithms to make things look like they are the product of algorithms vs. people who want to use algorithms to make things look like they weren't algorithms

I don't believe that this difference can be very "real" or important. There may be some people who think like this. But consider, for example, computer graphics in games. The people behind computer graphics certainly aim for "photo-realism", something which looks as little like a computer, and as much like a film or even seeing the real world, as possible.

At the same time, I don't think they are ashamed of their algorithms. I think they celebrate the cleverness of inventing algorithms that can look so realistic.

I think there are some people who want to resist the collapse of technique. And want to return to an art that gives a high value to technique. These are the people who complain about "modern art". And complain that pop musicians "can't play their instruments". These people would certainly be against any sense that the art can be captured either by an abstract algorithm, or is mere self-expression.

I think there's a stronger quarrel there than between people who want to use algorithms visibly vs. those who want to use them secretly. Of course, an area which many people are excited about is using neural networks to produce realistic fake video. Eg. https://www.youtube.com/watch?v=S1OwOd-war8

After watching that, we should realize that you should never trust the evidence of video again in future.

We can now fake very realistic video where a movement is transfered to a different face or body. Where a scene in a city can be moved to the countryside or vice-versa.

In this kind of research you might find people who want to use an algorithm but pretend that the algorithm wasn't there. But their motive will be political, not artistic.

On to the next question : Paula Perissinotto says that the artist is the person setting the parameters and launching the algorithm, not, a computer.

I think she is right, but maybe not for the "obvious" reason.

I had a friend in the 1990s who was a jazz musician whose PhD research was to create a program that would improvise jazz. And at some point he was in a terrible crisis because he didn't know if he wanted to succeed or fail. He obviously didn't want to fail. But if he succeeded in creating a program that could really "improvise jazz" would he be invalidating everything he valued about himself as a jazz musician?

This is, again, the idea of art as personal expression, or outward sign of inner genius.

We don't believe that the computer has an inner personality or genius. And we don't want the computer to negate ours. So I think we won't let it. I think we'll continue to define art as the bit that the human does and not the technique that we've managed to teach to the computer.

In practice, the computer will be able to master most of the technique. I've been impressed recently by FlowMachines and the song Daddy's Car which is composed by an AI. . I think they have the right idea. They aren't trying to make Artificial Intelligence that takes over from the human. They are making AI as "compositional assistant". Something that conveniently leaves the human in control.

Even though the computer can do 99% of the technique. We'll continue to emphasize the story that human genius is the real source of art.

Daddy's Car by FlowMachines

Further Thoughts

Hi Joao.

I realize it was a lot. I hadn't really thought this stuff through before. But once I started writing, a lot of ideas started flooding out. >8-)

And I realize this might not be quite what you are expecting. Perhaps you are really more interested in something like "computer art" or "digital art".

The point I want to emphasize is that "algorithmic art" isn't a curious sub-genre of art distinct from all the others. As though it's a particular medium like "painting" or genre like anime or drum'n'bass.

Instead this shift towards making explicit "the processes" is part of the fundamental historical evolution of mainstream art. Computers are just one way of doing that. But many people are interested in it.

Now ... to your questions :

"- How did your jazzist friend's story end? Did he succede or fail (I don't even know if there's an objective way of saying someone succeeded or failed in music, but)?"

He made a system that generated jazzy sequences. But really it wasn't very different from the Mozart dice game. He just had a database of Charlie Parker riffs which it could pull out and fit into a chord-sequence. It was a success in that it sounded quite good, and he got his PhD. I don't think he ended up believing that it was "really" a musician though.

"- Let me see if I understand correclty: when you talk about an algorithm being the 'abstract' half of art's technique, I can see how that relates do the Mozart example, where the output is a variation on a pre-established pattern. Do you think this still apply to paintings/music/writing/video that are done trough more complex AI systems (with machine learning, deep learning, etc)?"

What I'm saying is that when we thought of art as "technique" we were talking about a process that needs the skill of the artist to execute. There's no way to separate "knowing how" to paint and "being able" to paint. Or you can't play the guitar in theory but not in practice.

But then we develop new tools and theories and technologies that DO allow us to separate the two. So now it is possible to talk about someone who knows enough about tonal theory and harmony to be able to compose music for the violin, even though they are very bad violin players.

Or someone who can program a drum machine and understands the groove that a particular pattern will give you, but couldn't play sufficiently well to keep the rhythm in time on a drum kit.

And yes, the more tools we use, the more opportunities we find to separate the knowledge of the process from the technique of executing the process.

Video. Or programming music on a computer. Or training an AI. These require you to think about the abstract processes and algorithms etc. in order to make the work, but allow you to leave much of the execution to the computer.

"- I find very interesting that you said some initiatives are more political than artistic. In the sense of this last questions, most of examples that I can think about are AIs made by Google or IBM, in which the resulting 'art' (be it what it may) is more newsworthy than, well, interesting. Is this what you meant?"

What I was saying was that the major MOTIVATIONS for pretending that something generated by algorithms was NOT generated by algorithms would be more likely to be political than artistic. I don't think there's much artistic reason to use algorithms but pretend that you aren't.

I mean, I can imagine some cases where a particular realistic effect is the product of algorithms, and someone might say "that's cheating, we thought it was because of your skilful technique". Or, to take a more concrete example, there's now a hypothesis that Renaissance painters used more optical devices to help them achieve their realistic perspective (https://en.wikipedia.org/wiki/Hockney%E2%80%93Falco_thesis) And for some people this might seem to devalue the works.

I don't personally find that a very interesting line of debate. It doesn't seem to me to devalue the "art" of the work. And I can't see any big philosophical points there. It's like complaining that an artist painted a portrait from a photograph rather than having the subject sitting in the studio for a week.

"- Lastly, what do you think of Brian Eno's work with generative music (the app and all that). Do you think if fits with all this you talked about?"

I'm an admirer of Eno as a thinker and I like a lot of what he did in the 70s and early 80s. And I think it was very naturally his thing to get into making those interactive / generative pieces from the 90s onwards.

However, personally, I think what I've seen and heard of them is a bit bland and "obvious".

When I started thinking about interactive art in the 1990s, I was very much reacting against what I saw people like Eno doing. (And Koan etc. https://intermorphic.com/sseyo/koan/ you might want to check them out as https://intermorphic.com/ )

I'm not dissing Intermorphic or Eno. I think what they did and continue to do is valuable.

But what I felt (and still feel) is that we need to further develop the "art of interactivity". What I mean by that is not some grand theory. But a set of qualities I look for in interactive art. These qualities include a kind of balance or harmony between what the user of the piece has to do, what the work is made of, and the resulting "movement" that comes back from the interaction. A similarity of "degrees of freedom" perhaps.

Another quality is that the connection between what the user does and the output is intelligible. The user can understand the result of their action and learn to change their input to gain some degree of control.

Another quality is that the user enters into some kind of engagement or "flow state" with the work. It's not something that just acts while the user ignores it. Similarly, a work should reward longer term engagement The more time you spend "practising" with it, the more you will see it will do.

In a sense, things like video-games or learning a musical instrument have these qualities. Video games are the leading "interactive art" form today. Because they really focus on the interactivity part.

"Again, thanks a lot. Just wanted yo know how can I credit you and if it's some of your work that I can link to?"

OK. I'll give you a couple of my works to try to illustrate some ideas.

Something I did in the 90s, partly as a response to the Eno type stuff is Gbloink! That's a kind of interactive automata that generates music from balls bouncing around a play area. You can see a short video on the site, and play with a prototype in the browser (tested on Firefox and Chrome). I'm also currently working on a new version as a VST plugin that can be used in FL Studio, Ableton etc. But this isn't released yet.

The point of Gbloink! And a lot of my other works, is that they occupy an intermediate position between something that you have to control, and an algorithm which works autonomously. I don't want to say to people "here's a passive thing. Now you go and push it around". I want my works to have their own inner "life" in some way. But at the same time, I don't want to set up an algorithm to generate things and say to people "just watch my beautiful algorithm at work"

What I want from my interactive works is something where the user participates in and learns some degree of control and collaboration to come to a result that neither the user nor the algorithm would have made independently.

Zewp! (http://synaesmedia.net/portfolio17/zewp.html) is a very similar idea based on a slightly different principle. It's never worked quite as well as Gbloink! and I'm still dabbling with different implementations.

Today I'm a musician with and write software for the Brasilia Laptop Orchestra. (BSBLOrk) Most of what we use has the principle of being somewhat out of our control. The founder of the orchestra, Eufrasio - who you may want to talk to for your article, I can put you in touch if you like - likes to have a lot of input from the cameras of our computer and then distort it with more chance influences.

He's also very influenced by Koellreutter (https://pt.wikipedia.org/wiki/Hans-Joachim_Koellreutter) who brings together the 12-tone serialist tradition in music with John Cage's chance-based ideas. So Koellreutter has scores which look like network diagrams. And musicians play them by randomly choosing routes through the network.

I've taken the Koellreutter diagram and adapted it as a program we use in the orchestra. I've created other interfaces to control the synthesizers we use in slow motion. And we have music based on various kinds of "games" or "role-play". One where we use a dice to chose which movement we play next. One where we pretend the cameras of our computers are security cameras we are trying to avoid being seen by. (And when they catch us, the sounds are based on samples of guard-dogs barking). We did a performance a couple of month ago where each musician played on branch of the government, and the improvisations / dialogues between us were based on internal political manoeuvrings.

Beyond the Orchestra, I'm also doing some work with Sonic Pi and other programming languages for music. I wrote the soundtrack for a dance piece two years ago ( https://www.facebook.com/Conflu%C3%AAncia-das-Coisas-In%C3%BAteis-1547281532242527/) of which, about 70% composed in Sonic Pi. Here's a sketch for a section : https://soundcloud.com/mentufacturer/useless-things-1-fiveminutes and here's the program that created it : https://gist.github.com/interstar/33098d0fa9cb843cd9b5300fe9bb9655

This is algorithmic but not interactive. There's a little bit of chance used in the program. For example the line :

play c.choose, amp: 1.2, attack: 0.5, decay: 2 if one_in 9

is choosing a random note from the chord to play. But the rest is largely deterministic as it has to fit the choreography.

I'm also doing some tutorial videos in Sonic Pi eg. https://www.youtube.com/watch?v=RfLAauDOvU0 and run various sessions on coding and live-coding music) (plus making music with electronics) at our local hackspace. (http://calango.club/projetos/musichacking/inicio)

For some physical works, I've a couple that follow what I call a "triptych format". They have something the user does as in input. An inner computational / algorithmic part. And then some kind of physical output. Another of my interests in art is the relation between art and craft. The role of the craftsman when art was both more social and more "useful". Before the Renaissance and Romantic era created these ideals of artist as self-expressive genius. And these works kind of talk about that.

So, for example DJ Oleiro (http://synaesmedia.net/portfolio17/pot.html) starts with a visual or kinetic "pun" between the contemporary DJ spinning records and the potter making pots on the wheel. Which have very contrasting ideas around them. I took a MIDI DJ controller and turned it into an input device for making pots. Basically you use the controller to form the shape of a cylindrical vase or cup. And then the program exports a stereolithography file which can be sent to a 3D printer to make it. Similar is Machine Gardens (http://alchemyislands.com/machines.html) which is a program where the user can design machines (with a very simple fractal algorithm) and export them to be made on a laser cutter. A final, somewhat elaborate example is Drawer, the Explorer (http://synaesmedia.net/portfolio17/drawer.html) which starts with the user drawing maps in pencil on paper. The movements are detected to be interpreted as the movements of an "explorer" looking for lost civilizations in the rainforest. When the right combination of movements occurs, the explorer finds an "artefact" in the form of a "mask". Actually a computer generated stylized face. Which gets exported as a Pdf to be printed and worn by the participant. It's kind of playing with everything from the heroically mythologized role of the European explorer, my love of masks, the role of maps and the blurring of the distinction between discovering / inventing etc.

Finally, recently I'm working on a library in the very high-level functional programming language Clojure to help me create complex patterns for textiles or decoration, using very short elegant programs : http://alchemyislands.com/patterning.html

OK ... that was all waaaaay too much about me. The point I kind of wanted you to grok though is that probably very little of this stuff is "algorithmic art" in the way you may have thought of it : something like a fractal or Eno-style "generative" music.

But pretty much everything I'm working on is in the form of a computer program that has algorithms in it. And the algorithms are not "neutral". They shape the artistic outcome. Deliberately. There's no pretense that you would get the same result without using this artistic piece / this "making system".

That's also true of everyone who is making pop music or hip-hop or EDM on a DAW like FL Studio or Ableton. They like to use samples of acoustic instruments like "pianos" and "orchestras". Because they like the sonic qualities of those instruments. But no-one is particularly interested in convincing you that somehow they brought a grand piano or an orchestra into the studio to bring a "human touch" to the music. No, they just like "organic sounds". They're happy to post videos of themselves on YouTube showing how they used that organic sample, transformed it electronically and then sequenced it on their computer.

So it's all "algorithmic art" in that sense.

Meanwhile contemporary art is about making us think about the process. The game-playing we do in the BSBLOrk is as much an "algorithm" as the programs we write. Or rather, "innovation" in all arts is increasingly tied to innovation in the processes / rules / algorithms that went into generating it.