Last week, I attended ISMIR, my favourite conference, which I have only missed a couple of times since I presented my MSc. work there back in 2007 💛
It is where, year after year, I keep up with the state of the art in my field and reconnect with my people 🤝
For the 5th time in a row, I chose to participate remotely 🍃
Here are my takeaways:
👏 Running a hybrid conference is not easy. It feels like the virtual participants and those on site have two very different experiences. To me, the virtual conference is best to deeply experience the content (papers, tutorials, keynotes, panels, Late-Breaking Demos) with very little distraction, while attending it in person provides a lot of networking opportunities. That being said, the organisers this year did an amazing job including the virtual participants: authors could present their work remotely (including Late-Breaking Demos), replays of the entire conference to accommodate several time zones, special online events dedicated to virtual participants (including social/networking events). I, and other seasoned ISMIR participants, were also invited to be backchannel hosts: commenting on the papers live during one of the sessions in the conference Slack, providing fun facts and additional context information. This last initiative was engaging even the people on site who were following along. Kudos to the organisers, especially Blair Kaneshiro, Katherine Kinnaird, Vinoo Alluri and the virtual volunteers for taking such good care of the virtual participants.
🤖 As already seen in the previous years: AI is a massive part of MIR now. Most papers used AI approaches, and the three best papers, although very diverse, were all making use or focusing on AI. Interestingly though, the community seems to be taking a balanced approach to it. Indeed, two of the keynotes (+ an online invited talk) were discussing the ethics and/or (il-)legality of using copyrighted material to train AI models. One of the nominees for Best Paper studied the environmental cost of AI-driven MIR research, and several papers were looking into whether LLMs can actually understand musical content or not.
🎨 Diversity is still at the core of ISMIR: the DEI keynote, part of the main track of the conference (as always at ISMIR), was illustrating how to "Listen for Diversity". It was announced that the 10-year initiative of Women in MIR (WiMIR) becomes Widening Inclusion in Music Information Retrieval (WIMIR), to better reflect its broader vision. The music studied in the papers was very diverse as well, including 15th-Century Korean Court Music, Carnatic Music, Japanese Idol-Group Songs, Classical Guitar Duets, Latin American Music, Choral Music, Larynx Microphone Recordings, Wagner Operas, and more. Finally, there is still a good balance between the number of papers focusing on audio and those focusing on symbolic music (whether for analysis or for generative purposes), and other modalities, such as lyrics are gaining attention too.
Business developer / Uitgever bij Boom hoger onderwijs en Boom Docentacademie
1wThe floppies are awesome! 🤓👌