AI-driven tech startup Beatoven.ai is a platform for music composers and content creators to create royalty-free, affordable, easy to licence, exclusive music. Beatoven.ai’s web-based application allows content creators to create and customise tunes from a vast library of mood-based tunes across genres. The platform is user-friendly, and even individuals with little or no knowledge of music can create original soundtracks with the help of AI.
In an exclusive interview with Analytics India Magazine, Siddharth Bhardwaj spoke about how Beatoven.ai embeds ethics in their processes.
AIM: Tell us how Beatoven.ai leverages AI.
Siddharth Bhardwaj: Firstly, we use ML and AI systems to accurately tag music data that our partner artists submit to us. Then, we are building our custom music composition and production algorithms for regional (currently limited to Indian) and Western genres. The mix of music theory rules and signal processing is tied together with our continuously evolving AI algorithms. These composition algorithms use our music sample library to create a unique track based on user inputs like duration, genre, moods for different sections of the content, etc.
AIM: Elaborate on the AI governance methods, techniques, and frameworks Beatoven.ai use to ensure the best possible experience for users.
Siddharth Bhardwaj: The data we use to train our models or create compositions have been sourced through our artist partnerships. Music licensing and royalties are complicated, and hence we work with many independent artists to source our own data, and we own the copyrights to all the tracks we use to create our compositions. Being musicians ourselves, we wanted to create monetisation opportunities while creating a human-AI collaborative tool that solves a real problem.
We use the Indian classical music theory established over centuries like ragas, thaats and their associations with moods and incorporate this with AI to help compose music; the same is applicable for western music theory, where we use associations of scales and moods and their equivalent visual representations. This accurately represents each geography, and the plan is to expand this globally. We take pride in studying these intricacies and hopefully accurately representing them in our tool. We regularly evaluate our compositions on several parameters by expert musicians in each genre/region and improve our algorithms accordingly.
AIM: What explains the growing conversation around AI ethics, responsibility, and fairness?
Siddharth Bhardwaj: Being at the forefront of AI in music, we have a responsibility to be fair to the artists contributing to the platform, and we take this very seriously at Beatoven.ai. Our vision is to become an AI-powered platform for all the musical cultures of the world, and we work with many amazing local artists to try and represent these cultures as accurately as possible.
In AI music, the most important question is about the ownership of the tracks being created – does the ownership lie with the artists, developers, users of the tool or the company? We own all the data that we use to develop our composition algorithms and then licence the final tracks to our users. We also incorporate the music theory rules wherever possible to enhance the algorithms and accurately represent the genre/styles of music being composed. These rules have been in existence for centuries, and no single entity has ownership over them. This approach also helps us normalise any biases creeping into our algorithms.
AIM: How do you ensure compliance of AI governance policies and best practices at Beatoven.ai?
Siddharth Bhardwaj: We take great care in ensuring that the musical data we collect does not infringe on any past copyrights by manually and algorithmically checking them. We work with our in-house music producers and industry experts to help us develop our AI algorithms and evaluate the quality of music being produced. We are a monetisation channel for our artists who put in the effort of producing and tagging our music data. This enables our algorithms to produce diverse music while accurately capturing the artist’s intent.
AIM: Do you have a due diligence process in place to make sure the data Beatoven.ai uses is collected ethically?
Siddharth Bhardwaj: Currently, we have in-house music producers who work closely with the artists to source the specific types of musical samples and tracks that we require. We pay the artists up-front for their efforts in producing and tagging these samples. While sourcing our music, we ask the artists to tag the data. Then, our in-house producers use proprietary algorithms to double-check this data and ensure correctness. With every release, we try to talk to as many users as possible to find any ethical concerns or inaccuracies in our tool. We also regularly work with expert musicians in each genre/style/region to evaluate and appropriately represent the music.
AIM: How does Beatoven.ai protect user data?
Siddharth Bhardwaj: We ask only tool-related questions to our users, like the channels where they intend to use the music created on Beatoven.ai. Some other questions are related to improving the product. We store these in a secure database managed by AWS. Also, we take great precautions in ensuring the video or podcasts that our users upload on our platform are stored securely on our servers using the best industry practices. For user authentication and login, we use Auth0, which is the industry standard and provides the best possible security.