Have you ever dreamed of seeing a new colour? Well, unless there’s a gaping hole in the electromagnetic spectrum, and every rainbow in the world is wrong, it will likely remain so: a dream. But luckily, there may be all sorts of new hues coming soon to the waking world, augmenting a different palette.
Because we cannot yet change our senses. But we are getting increasingly good at controlling what goes into them.
So it is natural to ask: What is the future of art? A century from now, the way in which humankind consumes media may be unrecognisable. In thirty years, we’ve gone from vinyl records to the iPod. My dad has ten thousand records in the attic; you could fit them all on a hard drive the size of small book. It’s clear that technology is having a huge impact on music in particular.
But we’re not talking about software that plays music, or software that stores music. Major change does not linger at the end of straight lines; you have to bend them. We’re talking about software that writes music.
That inexplicable gift of a great composer: the sublime matrimony of imagination’s infinity and infinitesimal detail. Everything we love about music, as a movement at once created and consumed by an entire race, albeit in different ways and places, seems to suggest that there is something inexplicable delivering it. Man seeks to decompose such art, just as a physicist describes the motion of the planets with his tools of analysis, and man gives names to musical devices that form the building blocks of composition. However, just as the physicist can describe accurately the elliptical orbits of the planets, but is at a loss to explain why an inverse square law applies in the first place, the musical scholar can never determine precisely why the particular composition of building blocks in any given piece of music evokes in him the feelings it does. As a result, our scholar cannot paint music by numbers; he may reassemble the building blocks in a novel shape, but there is no guarantee that this will sound good – because the execution of celebrated composition is not to be found in the crude architecture of such rigid blockwork.
And this is the challenge. How can computer-written music ever sound good if the essence, the truth of composition cannot be pinpointed?
And even worse: all computer software is made from building blocks. Is this the gulf between silicone and psyche – discrete memory versus continuous consciousness – or is it just the sense of scale deceiving us? For on the other hand, the fact that machines are now passing the Turing test with ease could lead a provocateur to argue that psyche is made of building blocks as well – just really really small ones.
Fortunately we can sidetstep the AI debate.
Still, imagine a Turing test for computer music! Developments in programming languages and the availability of free software may call for this, and sooner than we think. The salient quality of object-oriented programming is the power to create program structures that mirror the way things work in the physical world. More precisely: object-oriented languages allow us to create program structures that mirror the way we think about things in the physical world. So, perhaps the processes of thought we use in composing a piece of music can be faxed to binary.
At any rate, there is really only one way to illuminate the issue, and that’s where the fun starts. Leave theory to the theorists: let’s build! From this seat, first steps seem attainable. An attractive target slips into view: the 32-bar standard. The scaffolding of this historic jazz song format is already clearly delineated. At the heart of the idiom is a reasonably small pool of chordal notions directing a simple tune, and that’s it: two instruments. It’s doable. The core issue for the designer of computer composer is simple to set down: how to reconcile the notion of structure with the embellishment of probability? Or put another way: how to ensure an intelligible randomness? And the difficulty is that, whereas the question is asked in the language of music, the answer must be supplied in the language of source code.
Progress aplenty abounds in this orchard, and the fruits of recent labour taste promising. Beginning in 1981, David Cope’s project Experiments in Musical Intelligence has demonstrated our combined capabilities. In 1997 he released Virtual Mozart – machine-written music – and later exploited computers to imitate the styles of Beethoven and Bach. Established software like Band-in-a-Box uses stochastic grammar to improvise melody over chord sequences in real time – primarily as a rehearsal aid – but the potential there is enormous, and such methods are easily capable of writing original melodies for our 32-bar standard, once we’ve got the chords down.
From there, all we’ll need is a random-sentence generator to write words! The following is not at all far-fetched: get a dictionary, group words by rhyme, prescribe sentence structures, interpret syllabic meters gleaned from melodic phrasing, write full verses. This is the icing of the project, for it allows the results to be performed in context; by humans, for humans, and for fun. That is, the instrumental results are unlikely to be mind-blowing; they may be interesting. But if the computer is given a voice, speaks to us in our own language, and – most importantly of all – makes us laugh, these compositions can be reasonably aired in the public domain. And hence, no effort will be made to pass the lyrical Turing test!
Because after all, music is to be enjoyed. Can we not harness the creativity of programming for that purpose? Imagine a world where intelligent music responds to stimuli in real time, a world where the soundtrack to a party evolves in keeping with the mood, taking its lead from the temperature, noise levels and kinetic energy of the room. The old music was performed, or recorded; the new music participates. We have lights that take cues from sound, and soon we can have sound that takes its cue from lights. Feedback; a regulated loop that makes the whole show evolve like an interactive organism, like a sensory exhibition of colour and sound delivered with the rush of disbelief that only newness can bring.
All of this seeks to endow our lives with an audible new colour. Listening back to the era of repetition could be like watching a film in black-and-white. Here is potential for AI to prove itself, and stake its claim well beyond the grim debates of a species whose average outlook is the ultimate chauvinism.
Perhaps we shall no longer have to argue for or against Artificial Intelligence. Instead, we will just be enjoying Artistic Intelligence.