Giving Artists a Fair Chance to Get Heard<\/h3>\n

\u201cFirst Fans is an attempt to give new music a fair chance,\u201d says Aron. Imagine the untapped potential in those overlooked tracks \u2013 it\u2019s a proverbial gold mine waiting to be discovered.<\/p>\n

Addressing this \u201czero plays\u201d problem required a new type of recommendation system \u2013 one that didn\u2019t merely focus on what was already popular, but gave equal attention to new uploads. This meant moving away from the traditional \u2018collaborative filtering\u2019 approach, which recommends content based on what similar users enjoy, to a \u2018content-based\u2019 recommendation system.<\/p>\n

Technical Challenges<\/h3>\n

To bring this idea to life, there were 3 key technical challenges to overcome:<\/p>\n

\u2192 Knowing what music you have<\/h4>\n

Aron, along with lead engineers Vignesh Badrinath Krishna and Nikolai Tatarinov, and a wider team of over 20, used AI to construct a unique system to catalogue and tag new music as it\u2019s being uploaded. The Musiio AI tech they built analyses the characteristics of every song, considering factors like recording quality, mood, genre, energy and more.<\/p>\n

\u2192 Optimising for the user experience<\/h4>\n

\u201cNext, recommendations are performed by an AI that is figuring out what\u2019s a good match for a user and what\u2019s a bad match, and optimising for a good user experience,\u201d Aron explains. This process uses user information like listening history and followed artists, and inherently biases against recommending music to bots and fraudulent accounts. However, powering this process is no walk in the park. The AI needs to process thousands of recommendations every second \u2013 one of several challenges that took years of integration to overcome.<\/p>\n

\u2192 Speed and Scalability<\/h4>\n

\u201cThere\u2019s a lot of moving parts in it,\u201d says Aron, alluding to the big-data challenges that SoundCloud had to contend with. One part of the solution was building cloud computing systems that balanced cost efficiency with speed and reliability, ensuring every track uploaded to SoundCloud is analysed in near real time. The system also includes a data pipeline designed for real-time recommendations, which was another non-trivial challenge.<\/p>\n

Real-world Impact for Artists<\/h3>\n

Aron anticipates that the \u2018First Fans\u2019 initiative could exponentially increase the number of artists being discovered and making a living out of their music.<\/p>\n

With new tracks being recommended to around 100 real people through the system, SoundCloud\u2019s curators gain valuable insights into the song\u2019s performance soon after uploading. Therefore, if a track resonates with listeners immediately, it can rapidly catapult the artist into a much larger audience.<\/p>\n