Participants 2022
The Synthonauts
AI Song Contest 2022 / participantsTEAM / The Synthonauts
SONG / Tracing Our Roots
TEAM MEMBERS / Devashish Gupta, Nithya Shikarpur, Pronoma Banerjee, Debapriya Kar, Sneha Shah
About the TEAM
We are all a group of engineering students who graduated/will graduate from BITS Pilani University, Goa, India. During our time at university we were all active members of the Music Society where we performed and created music together. Inspired to invest our training in engineering and music into a creative product, we decided to take part in the AI song contest 2022.
About the SONG
Our song is a tapestry of sounds generated by several modular synths at different stages of evolution (guided by a human-AI collaboration). In our entry, we arranged the audios from their most evolved form to their least and found interesting parallels between their evolution and the evolution of life on Earth. Beginning with a utopian soundscape in a hopeful and distant future, we traverse rhythmic movements from the industrial revolution, sounds of discovery in ancient civilizations, the elementary chaos we imagine in Earth’s first organisms, and finally, a symbolic big bang. We see this song as a souvenir of gratitude towards nature and the millions of years of evolution that made us human and brought us to this very day.
Note: please check out our annotated soundcloud file which highlights the parallels we draw between our song and evolution of life on Earth (https://soundcloud.com/snnithya/tracing-our-roots).
About the HUMAN-AI PROCESS
Our idea revolves around capturing the evolution of sounds and their synthesis using AI. Inspired by the flexibility and the sheer extent of possibilities with modular synthesis, we wanted to use AI to select modules and discover the best way to connect them. To do this, we employ a genetic algorithm (a type of AI) and track the evolution of several modular synths. In our experiments, each modular synth represents an organism with a characteristic ‘genome’. The genome of our organisms is defined by the modules involved (instruments, note-effects, synths, audio-effects, modulators etc.), the connections/signal flow among the selected modules and the individual parameter values for each module. Along with human input that scores each of the organisms from a generation, the algorithm, through mechanisms of selection, crossover (combining genomes of 2 organisms) and mutations, helps organisms evolve over several generations.
The creation of a song is never a linear process, there is always a back and forth between ideas and new inspiration coming from unexpected places. Our goal with this project was to create this flexibility for the genetic algorithm to create sounds. By keeping a human in the loop to guide the AI based on their sense of musicality, this becomes a very interesting human-AI collaboration.