When stories of Artificial Intelligence make it onto the news, they either have a way of exciting and fascinating us, or manage to downright outrage us or completely terrify us. With applications ranging from the mundane such as text auto-complete, to the more futuristic-seeming creation of driverless cars, to scarier scenarios of machines taking over jobs and causing mass unemployment, improved data collection systems for advertising and propaganda purposes, and even research into the development of self-running military weapons… well it makes for a mixed bag… Yet it is more and more apparent that AI is here to stay and is already a part of our lives in many ways, even if don’t quite realise it.

Which is why it is fascinating to listen to Professor Bob L. Strum’s ongoing story of ‘folkRNN’: the AI program he developed with his collaborators which generates melodies in the style of Irish Folk tunes. Compared to some of the scenarios described above, on the surface this would seem to be an innocuous experiment. Yet in Professor Strum’s talk we discover how it manages to bring out very mixed and often very intense reactions in the people that engaged with it. It seems his AI managed to touch on all the soft spots we have toward the romanticised human aspects we usually look for in art, culture. He unwittingly revealed some of the biases and deeply felt ideas audiences feel about music, and folk music in particular. ‘folkRNN’ wonders into this sensitive spot, in an experiment that very quickly seeks to be out of the research lab and in the real world. The experiment soon becomes about people’s reactions to the AI’s creations more that developing the AI itself, and in his engaging and entertaining talk we hear the various experiment situations set up to explore this question and some their results.

Bob L. Sturm is an Associate Professor in the ‘Speech, Music and Hearing Division of the School of Electronic Engineering and Computer Science at the Royal Institute of Technology KTH’ in Sweden. He has published numerous articles on topics such as digital signal processing for sound and music signals, machine listening, and algorithmic composition. On his YouTube channel you can see him playing ‘folkRNN’ generated melodies arranged for his accordion, and it seems he does a pretty good job with them too!

Text: James Alexandropoulos - McEwan