In November 2013, I was invited to the Festudy conference in Lille as the Dour Festival's Director of Programming and Communication. At one point, conversation strayed to the surveys that were sent to festival-goers intended to find out what they wanted to see at the next edition of the Dour Festival. Some of the audience were outraged and remarked, rightly, that cultural programming should not be based on figures but rather on a desire to share emotions and moments.
The majority of the audience thought that we were simply collecting the top performers that festival-goers wanted to see and that the line-up was foolishly based on this information. Our method was more subtle than this, however: we used this data to detect upcoming artists with high demand. First and foremost, Dour Festival is based on the discovery of new genres and new artists. This data was therefore primarily used to pick up on new emerging musical niches. I also responded by informing the audience that reading music columns in specialist magazines, listening to journalists on the radio, or asking a friend if they have seen any good bands recently is also a means of collecting data.
At that time, we weren't talking about recommendation algorithms, and YouTube and Spotify had not yet come to control our daily music suggestions.
As a programmer for major European musical events (Dour Festival, Marsatac and BSF), and having studied computer science and advanced algorithms at the University of Liège – and therefore not trusting any algorithms that I don't understand (*) – I decided to write my own algorithms to help us find new artists and new sub-genres. Don't get me wrong, I don't think AI is going to replace music programmers any time soon: this technology is sorely lacking in emotion. On the other hand, however, it is impossible for me to keep up with all the latest music news and listen to everything. This is where algorithms can help us spot things, as they are much better at analyzing millions of data points than I am. This is what Frédéric Martel has labelled Smart Curation, in an article published on Slate.fr (*):
The Internet is decentralised, decentred and plural: it is characterised by disintermediation. And it is unlikely that we will return to an elitist model in which judgment is left in the hands of a small number of critics, lambasted even in the days of Balzac through his character Lucien de Rubempré.
However, the second option, that of strictly mathematical "machines", which involves delegating this prescription to automatic algorithms, does not appear to be any more effective. It is too imperfect to be efficient.
"Smart curation" offers an alternative solution: it is a combination of two models, the algorithm on the one hand, and curation on the other. It is a "double filter" that makes it possible to combine the power of "big data" with human intervention – the association of machines and humans, engineers and "entertainers". This algorithmic curation will be carried out both by those who use words and by those who use numbers.
It motivated me to create my own database by collecting millions of public data entries relating to artists. I add private data collected from target audiences and generate new data based on my specialized know-how. From there, I can analyse all this data and play around with it: programming automatic playlists for bars or restaurants, suggesting artist recommendations for festival stages, and much more.
Through these articles, I attempt to offer an understanding based on fun examples, by explaining the methodology used. Algorithms are neither good nor bad: you have to give them the right amount of importance, understand how they work and program them according to your needs.
Today, agents use data to justify artist fees, while record companies are finding ploys to artificially inflate streaming numbers. In the world of tomorrow, or perhaps even today, a lack of reliable data makes programmers easy prey: they must equip themselves with tools and stop dismissing data as something incomprehensible that they cannot master. But they can rest assured: this data will never stop them from developing new obsessions!
(*) Read also :
Wired.com : How to break out of your Spotify feedback loop and find new music
Slate.fr : La «smart curation» est à inventer (French)
BONUS
New “Smart Curation” Playlist :
Eurosonic 2021 : weekly playlist by SmartCuration.io