With Artificial intelligence being the current buzz, a lec-dem titled ‘Music appreciation and technology’ drew up in the mind visions of the performing arts in the near future.
Delivered by Kaustuv Kanti Ganguli recently at the Indian Music Experience museum in Bengaluru, the lec-dem’s topic was not just relevant but also crucial to understand.
While we have come a long way from when musicians used to be afraid that the microphones would suck the soul out of their voices, there is still a certain distance and disconnect between many musicians and technology.
What made the lecture-demonstration more interesting was that it was delivered by an accomplished classical music performer, who is simultaneously at the cutting edge of research in computational musicology and allied domains.
At the heart of this tangle is the juxtaposition of the subjectivity of aesthetics of an art form over the emotionless rational objectivity of a machine. Even Kaustav while demonstrating the capabilities of a computer system in performing tasks relating to fairly intricate musical perception, reminded the audience that this was not an advocacy of adopting such methods. It was simply a demonstration of the possibility. A neutral observer would probably wonder at the necessity for such disclaimers, but musicians and musicologists are occasionally at loggerheads about the value (or lack thereof) of the latter’s work, especially when there’s technology involved. Add AI-machine learning into the mix, and it just gets all the more convoluted.
Classical musicians spend years mastering the technical aspects of their craft and developing their aesthetic sense. AI can effortlessly tackle both. Technical precision is a breeze; AI systems can ‘sing’ perfectly in tune and in rhythm with just the right input parameters. Moreover, AI can analyse extensive archives of musical recordings to refine its understanding of complex concepts like rāga development. ChatGPT is already doing interesting things in the realm of language. How long will it take to see a ‘RagaGPT’ come up and disrupt the world of Indian classical music in a similar way? While concerns about such disruptions are reasonable and need to be factored in as the AI models evolve, it’s also not possible to rebel against them by turning a blind eye. What if one could actually lean into the strengths of a machine and find ways to assist and enhance the human experience of the domain? That’s probably the most pragmatic and responsible way of dealing with a complex situation.
While we rejoice in the convenience of home streaming of content, there’s an undeniable added joy in participating in the social transaction of going to a live event. The social factors aside, witnessing a human performer execute their magic in person is still an incomparable joy to behold, and we may not need to worry about artistes being replaced by AI for some time. Let’s take a look at some of the principles and phenomena involved in performing a musical task, like perception, cognition, (pattern) recognition, etc. While the human mind takes a lot of it for granted -- like differentiating the sound of the guitar from the voice in a popular song - a machine requires a lot of training to be able to develop some usable sense of cognition. The perception of the machine can however be refined to be as precise as one wants, even surpassing human perceptive ability. The scientist then requires a musician to define useful parameters of cognition and build usable models that can valuably exploit the computational capabilities of a machine.
The exchange goes the other way too, as was pointed out in the presentation. The artistic expression may not necessarily benefit from such a collaboration - at least as yet (specifically in the context of Indian classical music performance). The training and pedagogy, however, can be aided immensely through a collaborative approach with technology. Music archiving and documentation becomes a much easier task, and huge volumes of audio archives can be tagged to enable accessing them in far more useful ways than is commonly possible. We take text-based online searches for granted through the massive amount of work done in the domain of language processing. How about searching for a rāga by humming a characteristic phrase, and looking through terabytes of archival material to find maestros’ rendition of the same?
In addition to numerous applications in MIR (music information retrieval), a detailed enough technological solution could be of great aid in assisting the training and also regular self-evaluation process of a novice enthusiast. As one tries to go deeper into the subject, psychoacoustics becomes an important domain to engage with. Practitioners understand the nuances of the role of conditioning and memory in the process of musical training. The objectively measurable parameters of sound first need to be mapped to the subjective frameworks within the genre. Much like how optical illusions demonstrate the extensive capability of the mind to extract meaning beyond the actual information contained in the visual (to the point of even misleading the observer!), such phenomena abound in the aural realm and need to be accounted for. This conscious engagement with a subjective experience of sound, when supported by the neutral, rational, and objective evaluation of the audio signal can be invaluable in taking research in diverse domains such as music therapy, significantly ahead.
As Kaustav pointed out during the audience interaction during his session, the ‘right questions’ can actually apply this field to varied invaluable applications. The tools are already available at a usably developed stage. Constructive work on these lines has to be interdisciplinary. The best part of the session was that he ended it with a proper concert presentation. This seamless coexistence of the emotionless rationality of the technology with the aesthetic beauty of the art was the best takeaway of the evening. One that can help find the best way to harmonise music (or any art) with rapidly evolving technology
COMMents
SHARE