Ready for robo-music?

Now algorithms can play

We encounter a multitude of algorithmic processes every time we log on to the computer. Facebook, Twitter and Instagram all utilise coding functions to bring you a feed, while Google filters adverts to find ones relevant to you.

However the relationship between content and algorithms we’ve seen so far is largely a matter of drawing together disparate locations of information across the internet; we are yet to experience original content generated by codes and procedures.

Jukedeck, a widely publicised music generation startup originating at Cambridge University, is one of the first firms to offer such a service.

Those wanting to generate royalty-free music simply select a genre and mood and after about 30 seconds end up with a finished song. From there the service offers adjustment options of the pace and structure of the track to suit the user’s purpose.

Ed Rex, founder of the firm, says its current market is “YouTube creators and people making videos”, a base audience to whom Jukedeck provides music free of charge in exchange for a credit.

Larger businesses must pay around $22 for the use of music, with the option to hold the exclusive copyright to the track at $199.

The results of the service’s AI are surprising; the ambient genre filter is capable of producing melancholic and washed out tones. The electronic option can create dance music with strong, original riffs at a high level of subtlety and sophistication.

Not only does the service provide adequate proficiency in composition but does it so with an ear for mixing, a sense of style and balance.

Beyond the ends of background music alongside visuals, algorithmic generation of sound content has incited a great deal of experimental works outside of the commercial sphere

Confield, an album released in 2001 by Mancunian group Autechre is a perfect example of code being used to generate something unachievable without the processing power of computers.

Algorithms are set to work in the programme Max/MSP to transform snippets of audio into metallic, textural soundscapes with minute fluctuations.

When harnessed in this way algorithmic generation pushes simple sounds into a vast network of multiplicities miles from something conceivable in the human mind.

To many, their work may sound like the stuttering of an open Youtube tab on a laptop begging for juice as it turns off unexpectedly.

In response to dismissals of their music as random or non-musical, Sean Booth of the group remarks: “If live musicians were playing it, they'd probably call it jazz or something. But the fact that it’s coming out of a computer, as they perceive it, somehow seems to make it different.”

So far, so musically good. The perfecting of publishing and the generation of coherent writing from algorithms however faces an issue in its source material – the existing content it synthesizes and draws from.

The infamous and now de-activated @tayandyou AI chatbot developed by Microsoft displays the kind of problem one runs into when sourcing content and information.

What began as a neutral program that would chat to Twitter users was transformed into a series of responses filled with every form of bigotry expressible.

Because Tay interacted and communicated with Twitter users their opinions and language constituted the account’s replies and posts.

For accurate or factual responses in content there will need to be an inbuilt level of data verification. Or at the very least an assignment of authenticity to sources of information. What is currently lacking is human scepticism, a voice in the head asking: ‘Is this really right?’

But do these examples really constitute a plausible subsumption of human creativity to algorithmic process? To an extent, yes. In Jukedeck’s case the programme has an inbuilt mastery of the language of music and an extensive knowledge of what does and doesn’t work.

AI bots like Microsoft’s chatbots have the advantage over people because of the ability to draw and extract from far corners of the internet and synthesize the information instantly, something people have no way of doing.

But perhaps the future lies in mixing algorithmic processes and people with knowledge of them, from as little as altering filters or parameters of programmes to curating and resequencing the products of code.

Algorithmic procedures allow us to extend our scope, fulfil our purposes and in many cases create things beyond our capacity.