Qosmo Artificial Intelligence DJ and Human DJ -Back to Back-

Qosmo AI DJ and HUMAN DJ Abstract

“Qosmo A.I DJ and HUMAN DJ – Back to Back – ” hypothesizes the future of music and technology through experimental performances.
In the year 2045 it is said that civilization will reach the “singularity”, an event in which computer intelligence will surpass that of humans. In the present, human behavior is visualized by enormous amounts of data, and contemporary automation by computers deprives a person’s ability to work. In these ways, arti cial intelligence is eroding the necessity of a human’s “thinking” and “useful- ness”. Will technological advancements alter creative activities and spur human obsolescence?
How will this progression a ect music? A half century has passed since the IBM 7094 computer sang “Daisy Bell” in 1962. Nowadays, it is common and expected that a song is aided or even performed by a computer. Often, the technology is what makes a song a hit. Simultaneously, a huge amount of music is created on a daily basis, and simply lives on as data.
How then, can the expression of music, and its relationship to technology change in the future? If popular songs and playlist data from around the world can be extracted from iTunes, Spotify, or Soundcloud, then perhaps DJs and composers may be unnecessary. The expression of a DJ or composer will change even more as music becomes freely accessible via computer networks. It is important then to consider how we can co-create with arti cial intelligence systems rather than simply removing the human element of a DJ, VJ, musician, or composer.
Arti cial intelligence should not be considered as an imitation or emulation of humans. Rather, it is an intelligence with a logic di erent from humans. We believe in creating interesting and unique relationships between machines and humans. Through our performances, we want to show that A.I. systems can think with and alongside humans.

Deep Q-Network Training for Turntable Manipulation

Technical Features

As much as possible, the AI and human DJs perform under similar conditions. The AI system not only performs digitally, but it daringly uses real vinyl records and turntables. By using vinyl records and turntables, the AI system does not use a database to search for music instantly. Therefore, it does not have data “on-hand” such as a genre of song, artist name, tempo, release year, etc.
Rather, the AI listens to the songs played by the human DJ, detects the tempo, judges the genre, and processes other information on the spot. Following this, the AI chooses the next record to be played. Of course, since the AI does not have a physical body it cannot set the new record on the turntable, so we help it a little bit. After loading, the AI begins to adjust the tempo of next song to the tempo of the song currently played by controlling the pitch of the turntable.

In Back to Back, the AI occasionally chooses songs we do not expect. It is selected by an intelligence and logic di erent from humans. Sometimes the selection suits the atmosphere very well, but at times it feels random or unexpected. Thus the purpose of our performance is to have the human audience rethink what the AI system will do next. This is the type of awareness or experience we wish to co-create with the AI system.

Photo by Mao Yamamoto

いつか、QosmoのAI-DJと 別のロジックを持ったAI-DJのBack to Backが見られる日を夢見ています。


Concept / Programming
Nao Tokui
Shoya Dozono
Assistant Programmer
Yuma Kajihara
Project Management
Miyu Hosoi
Special Thanks
Chris Romero