“Using new technology to offer new methods of expression for artists”

AlphaTheta Corporation which develops DJ equipment and software under the Pioneer DJ brand, has been a collaborator on a joint project since 2020. We spoke with Ryuichi Suganuma and Yiming Wu from the company’s Development Department about this ongoing collaboration.

Cooperation project 1: rekordbox ver.7.0.0 / AlphaTheta

Cooperation project 2: rekordbox ver.6.0.1 / AlphaTheta

Can you introduce yourself and explain your role in the company?

My name is Ryuichi Suganuma. I work in the Development Management Division, where I am responsible for product development and the development of key technologies. Our department focuses on creating new value for DJs and developing the necessary technologies in advance to realize that value.

My name is Yiming Wu. I am also part of the Development Management Division, where I focus on the development of advanced technologies. Specifically, my work involves exploring new features using machine learning technologies and developing prototypes.

Can you tell us how you came to know about Qosmo?

Suganuma: Before we started working together, Tokui-san’s AI DJ project had already been a topic of discussion within our company, and I’ve known about Tokui-san as an artist since the early 2000s when he was releasing records.

Wu: Since my student days, I have been engaged in research on music information processing, so I knew about Tokui-san as someone conducting research in a related field. I was also paying attention to Qosmo as a company founded by Tokui-san.

Project: AI DJ Project

Can you tell me about the issues and challenges you were aware of when starting the project?

Suganuma: As a company, Pioneer DJ first collaborated with Qosmo on the development of a deep-learning model to detect the presence and position of vocals in music feature which was released in 2020 in rekordbox. I officially joined the project as a team member around the beginning of 2021. Since then, we have mainly worked together on developing features such as the Cue Point Prediction Engine and the Recommendation Engine, which were included in rekordbox ver.7.0.0 released in May this year. The issues we were aware of included, for example, the burden on DJs of manually setting cue points as markers for the song’s intro, chorus, and other points to understand the song’s development and mixing points. We thought that by improving the efficiency of this task, DJs could focus more on the creative aspects. Regarding the recommendation feature, professional DJs often have vast libraries of thousands of tracks, and with the ability to play songs from streaming services, it becomes challenging to select the best track from an enormous number of options. Many tracks in their libraries remain buried and never get selected. Therefore, we started the project with the idea of supporting DJs by providing inspiration for selecting the next track to play.

I see. So, AI is seen as a tool to allow users to focus on creativity or as a source of inspiration.

Suganuma: When it comes to recommendation functions, deciding the order of tracks and the flow of the performance is precisely the creative part of being a DJ. We’re not looking to automate that process.

Why have you chosen Qosmo to collaborate on this project and not other AI vendors?

Suganuma: While general AI vendors would need to start by explaining concepts like “What is a DJ?” or “What are cue points?”, your company’s understanding of these concepts allows for seamless communication, which is incredibly helpful.

The “RADAR” feature, a music recommendation system, featured in rekordbox version 7.0.0

Wu-san, what were the issues or challenges you were aware of at the start of the project?

Wu: We were at a stage where we had already developer a prototype when we consulting with Qosmo. The prototype itself was already highly evaluated by internal team members. However, there was still a need for further improvement to translate it into an actual product, and we sought your advice to make it happen.

What were the difficult and enjoyable aspects of advancing the project?

Suganuma: This is a personal issue, but when I joined the project with your company, my knowledge of AI and machine learning was almost zero. At that time, there weren’t many experts within the company, so I struggled to understand the technical details. However, as we progressed, my understanding gradually deepened, and I believe I gained valuable experience.

Wu: Since the project is still ongoing, it’s hard to say definitively, but after sharing the concept and updates at the regular monthly meetings, we generally leave the details to Masuda-san (AI engineer at Qosmo). When necessary, we handle detailed checks and data exchanges without waiting for the next regular meeting. Overall, I believe things are progressing without any significant issues.

Can you share your thoughts on reflecting back on the collaboration/project with Qosmo?

Suganuma: Our mission is to provide value that DJs will appreciate, so we’ve had many opportunities to reconsider how to integrate AI technology into this. We don’t want to replace DJs with AI just because it’s convenient, nor do we want AI to automatically perform DJ sets. We need to carefully assess how to apply AI technology to enhance DJs’ ability to focus on their performance.

It seems that your company shares the same vision as ours.

Suganuma: I believe we have a similar vision. It's about how we can enhance the creative aspects for artists and what kind of tools we can create provide to achieve that.

The project with Wu-san is still ongoing, but if you have any expectations or thoughts for the future, please let us know.

Wu: I think the current project we’re working on is somewhat different from the direction that Suganuma was pursuing. Rather, it’s more about using new technology to offer new methods of expression for artists. Artists are always seeking new forms of expression, but it’s difficult to articulate exactly what this new expression is. As a engineer, it’s also challenging to determine what kind of technology can be provided, or even if it’s appropriate for engineer to be involved at all. It’s a challenging job in that regard, but I see it positively and believe that there is definitely work that engineers can contribute to. I hope to continue working on it with perseverance.