How to compose music with a recipe (Arvo Pärt, Xenakis and others)

people don't like computers being better than other things for years the question on everyone's mind was can we beat us at chess and we all know how that turned out more recently it's become chem they composed better music than us computers have been writing AI compositions since around 1960 here's one of their first-ever public performances although people find this kind of thing intriguing it also makes them a bit uncomfortable they don't like the idea that a cold calculating machine might come up with something beautiful and moving and for the same sorts of reasons those same people think as human composers shouldn't be too calculating a real composer can be found walking across a windswept Moore or sitting in a summer chalet gazing into the distance waiting for inspiration to arrive good ideas are made by algorithms they arrive mystically through the ether one of the reasons many 20th century classical composers are mistrusted it's because they're seen as too calculating for example this guy Yanis is anarchists he used scientific theories like all Maxwell and Boltzmann kinetic theory of gases or equations predicting the Brownian motions of particles he take reams of data from computer printouts and used them to generate music each particle bouncing around the science lab became a pick Takato or a glissando in his compositions [Music] it certainly doesn't seem like a notion of composing most of us are particularly comfortable with but actually calculation has some kind of role to play in most kinds of composition the fact is for us human composers composing is really hard work so if it is an algorithm out there that will make things a little easier for us we'll grab onto it these days you only hear about algorithms when it comes to ranking YouTube videos or showing ads in your Facebook feed but an algorithm is really just a series of steps that are used to solve a particular problem just like say a recipe take four colorful cords sprinkle with pentatonic melody add in a splash of funky beat and turn up the heat it's easiest to get a sense of this recipe idea when you're looking at mechanisms for recreating existing styles of music for example in the 1860s the quadrille was a popular style of dance the compositions that these things were fairly predictable and what followed a series of rules so much so that a guy called John Clinton realized you might be able to generate them using some kind of system and he developed the quadrille Mulla dist this was a card based system it had a series of predefined slots outlining the shape of the dance and each slot had 11 different cards that you could swap around to create a different tune so here's one version and then we'll swap out all the odd numbered cards and a new tune emerges so Clinton had worked out some of the essential characteristics of the quadrille and distilled them into a recipe for creating an almost limitless supply of them Mozart did something similar with his musical dice game yep I did it first so what's the modern equivalent of this well it's probably the Markov chain which analyzes existing music and uses probability instead of dice throwing to generate the next note there are a large number of other AI computer composers out there all working on new recipes but all rely on a good supply of musical data to feed off this means some composer styles are easier to recreate than others basically Chopin mazurkas of which are 56 and they're all the same style that's great that's perfect but if you're going to take the planets by Gustav Holst for example it is there's no way to go there's only one of them and there's none of the rest of his works or anything like that particular one so you put that in the database you're gonna get out something that pretty much sounds like Dada Dada Dada [Music] so the algorithms used in those kind of machine learning programs and in simple games like the quadrille cards are good at imitating existing styles they're kind of trying to reverse-engineer existing music and recreate it they're not really going to help you write something original but there are other kinds that can I think the first known attempt at serious algorithmic composition was taken almost exactly a thousand years ago by guido Durazo the same blog that invented music notation for his algorithm you assigned a specific note to each vowel find a piece of text you want to set to music find each vowel in the text and map the appropriate pitch on to that vowel arbitrary you might think but the result sounds surprisingly good someday your harness mary tora tora copy a snake way or technique are rare this is an example of what's known as the translational model of algorithms where you take the information or data that's available in one area in this case the vowels in the text and translate them into some aspect of your music you could do the same with colors in a painting or the structure of a poem but it helps if there's some kind of logic to the translation in Guido's case he felt that music should honor and heighten the significance of the words so finding a melody that matches the shape of the vowel sounds makes some kind of sense another algorithmic model and probably the most commonly used is mathematics using some kind of equations to generate your music and this is particularly good of course for generating rhythms so this beat was generated using a plug-in on a digital audio workstation that creates what's called a Euclidean rhythm to make one of these you set the length of your pattern and you select how many beats to sound within the pattern and the plug-in uses some fairly simple math to spread the beat out as evenly as possible what's interesting about this is that many of the resulting rhythms are found in world music three beats within an eight loop becomes a Cuban dressier or five is the sync EO seven beats in a twelve loop is a common West African Bell pattern seven in sixteen can be found in Brazilian Samba and these kind of patterns can then be rotated around the loop to make new versions of the same sequence if we change the sound to marimba and start overlapping a few of them we pretty quickly have our own instance de Vries piece so you can imagine how useful this can be for generating new rhythmic ideas and it's all using this one mathematical model if you're interested in exploring other models I highly recommend the encyclopedia of integer sequences which is a huge collection of all sorts of different number sequences 300,000 of them I believe as well as the raw data the website also has graphs and MIDI renderings of all of the sequences I'm just going to mention two which have a particular relevance to music first is the rack'em and sequence which uses a very simple rule to generate a fascinating series of patterns as visualized in this beautiful illustration from the numberphile video on the sequence it's got that mixture of order and chaos which seems very artistic very human here's the same pattern expressed as MIDI data where one is the lowest note on the piano and two is a half step above but of course this is only the simplest of an infinite number of ways you could translate the same sequence into different music you could apply them two modes you could generate rhythms even in forms the second sequence I want to mention was actually invented or discovered by a composer is the so called infinity series created by the Danish composer Pere Naga it's a pretty amazing fractal like pattern that contains replicas of itself at multiple levels so this sequence of the start and every subsequent note of the series that follows it can also be found by taking every four notes and every sixteen notes and so on as well as an inverted form here and here as pretty mind-boggling but again you don't have to translate the numbers just in two pitches here's a rhythm I created simply by separating out the upper and lower parts of the sequence [Music] you can hear no Garzon use of the infinity series in his piece voyage into the golden screen [Music] [Applause] now of course there are lots more different kinds of algorithmic approaches we could look at we haven't even discussed the medieval practice of ice Iridum but I'd like to end with a much simpler kind of algorithmic approach taken by the contemporary Estonian composer arvo pet pets music often sounds closer to something Guido might have written than anything contemporary perhaps because he has similar spiritual goals he was trying to get music back to basics to access its innermost spirit the system he developed called tintinnabulation had just two parts one melodic and one that was limited to picking out notes from a minor chord by limiting himself so much the music takes on a stillness a solemnity a sense of ritual and this is probably one of the key advantages of algorithms they give you some kind of limitation something to work against as an artist the trouble with musical algorithms is they're misunderstood people assume they're only good for computers or extreme kinds of academic music but musical algorithms aren't inherently good or bad they don't determine the style of music you'll write and to prove it I took some of his analysis computer data from that piece we heard earlier and used it to generate these lush string chords [Music] algorithms don't take the decision-making out of the composing process but they can point you in new directions you might not otherwise have thought of a special thank you to Ben Levin and Matthew Rose who you'll see in a minute for both singing that phrase from Guido thank you so much to my patrons on patreon for making all of this possible thank you for watching and I'll see you next time [Music] [Music] [Music] [Music] [Applause] [Music] you

Channel: David Bruce Composer

Views: 108,236

Keywords: musical algorithms, mathematics, algorithms, AI Music

Id: X0-zuS_62wc

Length: 11min 38sec (698 seconds)

Published: Fri Apr 24 2020