Spotify (SPOT.N) expanded its partnership with Google Cloud to use large language models (LLMs) to help identify a user’s listening patterns across podcasts and audiobooks to suggest tailor-made recommendations. Artificial intelligence-powered LLMs, such as OpenAI’s ChatGPT and Google Bard, are trained on a massive amount of data to generate text and other content. Alphabet’s (GOOGL.O) Google Cloud, which powers Spotify, has several models, including PaLM 2, Codey, Imagen, and Chirp.
Spotify will use the new tools to help users discover and explore more spoken-word content on its platform, such as audiobooks, which are growing faster than music. It will also be looking at how to use them for personalized recommendation systems, ensuring that users are listening to the best content for them. In addition, Spotify will be using the technology to improve safety measures by filtering potentially harmful content.
While the technology is still early, it could eventually reach all Spotify products. This would include the mobile apps, desktop software, and its ad-supported free tier. Some DSPs have been worried about audiobooks cannibalizing music streaming hours, but the latest data shows that it’s more of a complementary listen than a direct threat. Audiobooks account for about 15 hours of the average Spotify weekly active users (WAU), up from around ten years ago.
It’s an excellent example of how powerful AI is becoming, even in non-music areas, and a testament to the traction that Google has been able to build with its voice recognition technology. For many people, the most effective listening experiences are when they’re absorbed in an engaging story or narrative. And that can be more difficult to achieve with music.
The move is also likely to help bolster Spotify’s relationship with rightsholders. The company seeks to shift its focus from passive licensing to active consumption. It will be able to prove to them that more of its users are engaged with audiobooks and not simply stealing music streaming hours.
The partnership will see Spotify migrate its technical infrastructure to the Google Cloud Platform. It had been handling hardware technical issues by leasing or buying space in various leading network providers, including Amazon Web Services (AWS). Streaming services typically run on a cluster of servers and need fast networking to transfer petabytes of data. Spotify will use Google’s networking services like Direct Peering, Cloud VPN, and Cloud Router to speed up its transfer times. It will also use Google’s cloud storage services for backup and to deliver its music worldwide. The companies said they will work together on the migration project over the coming months. The two have a long history of engineering collaboration that goes back to Spotify’s open-source Backstage developer portal. It simplifies actions that slow developers down and has been used by over 2,000 companies. They’ll continue to work with Google on this to reduce further the steps developers take to get started with the platform.
