DJALCHEMI

Why online radio is the model for listening to music in the future

At the RSA Music and Technology Event last month, Paul Sanders of State 51 described a scenario ten years from now where more music than you could listen to in a lifetime will be available on demand wherever you want it (at home, on the street, in your car). As he elaborated, the question then becomes, How do you facilitate listener choice in this world of ubiquitous music?

Paul rightly pointed out that the collaborative filtering systems used by Amazon et al to make recommendations to consumers are tiresomely predictable ("Customers who bought music by Bryan Ferry also bought music by Roxy Music and David Bowie" — you don't say!). My instinct in the face of the limitations of artificial intelligence is to replace it with human intelligence (see my justification for this). So, hey presto, in place of filtering technology we've just invented the disc jockey.

And then there's the question of how you pay for your access to this ubiquitous music. Perhaps a far-sighted government would set up a corporation to manage and develop this incredibly rich resource as an asset for the public good. Citizens with the equipment to access the resource might pay a license fee. So, hey presto, we've just invented the BBC.

...Which isn't to say that we're forever consigned to the same centralised playlist-oriented and format-driven programmes. There are already some interesting developments in the space between filtering technology and career DJs. There's an emerging continuum of 'formality', investment and control that starts with giving you cheap and cheerful tools to 'roll your own' radio station and ends with the niche stations that digital radio is now making possible for mainstream and other broadcasters.

The Last.FM site uses technology probably no more sophisticated than Amazon's filtering, but what you hear is tied to the profiles of other listeners: you choose one of them, and they become your virtual DJ. Only their selection and 'personality' is itself partly a composite of other listeners' choices, which they have heard and said they liked — and so on in an endless post-structuralist chain of reference. The other part is based on more directed selections and favourites that the listener has registered on the site. Here's my Last.FM profile for you to listen to.

Alistair Fitchett complements the superb Tangents online fanzine with the Tangents radio stream from Live365.com. Though, despite the gorgeous mix of Galaxie 500 and Felt (and a great set of Phil Ochs and other protest songs during the Iraq war), this has a number of limitations: I think Alistair has to pay to run his own station, and listeners have to pay to listen without adverts. Most significantly, the playlist is limited to a fixed amount of data storage, so if you listen for hours on end it sounds like a loop — until, I guess, Alistair intervenes and uploads some new tunes.

It's become de rigeur for a certain type of music artiste to issue mix albums of some of their favourite tunes, fulfilling Brian Eno's prophesy that the roles of curator and artist would start to converge. Laura Cantrell has gone a step further, though she was, I think, presenting her Radio Thrift Shop programme before she earned recognition as an artist in her own right. Meanwhile Tom Robinson has moved in the opposite direction, more or less retiring as a writer and performer to focus more on his evening show on BBC 6music.

These latter examples also demonstrate the relatively new feature of being able to 'time shift' when listening to your favourite shows, bringing to radio what VCRs bought to TV viewing 20 years ago. Thanks to the BBC's Radio Player, I can now listen to any of the programmes I missed because I was out or working (all I need now is about six extra hours in each day). With ever better means to play Internet audio through decent sound systems this becomes a seriously attractive way of listening to music.

One of the great joys of the 'radio' style of listening is the discovery or re-discovery of a track you would never have consciously and actively chosen if you'd been compiling your own playlist. I have no Dr Feelgood records in my collection, but when Andrew Collins played their Milk and Alcohol the week before last, it came so completely out of the blue that it cut across my mood, instead of going with the grain, and made me, well, feel good! This can happen without human intervention as well, as many people who set their iPods to random play will testify, and as I found this afternoon when Last.FM threw a Ventures track I'd never heard at me.

That's what Paul Sanders was talking about when he complained about the predictability of most automated filtering systems. To make the kind of leaps that technology systems would never get away with — as when John Peel jumps from Skimmer's Girl From The Black Country to Bob Dylan's Girl From The North Country — you need to build trust with the listener. People and personalities are better at building trust than technology. Or, more precisely, people are infinitely better at spotting when trust is under threat and then repairing it.