Recommendation Algorithms
Ongoing coverage on Recommendation Algorithms.
Latest coverage
Netflix pitted an LLM against its own feature-engineered recommender — and the LLM won
Netflix has developed a new recommendation system called GenRec, which utilizes a large language model (LLM) and has outperformed its previous feature-engineered system by 1.6 percent in offline ranking quality. This new system requires significantly less training data and is more adaptable to new content formats like games and live programming, though it still requires a separate component to handle hallucinations and ensure only existing catalog items are recommended.
Roblox shares fall 70% following lower-than-expected monetisation during Q2
Roblox shares experienced a significant 70% decline following the company's Q2 earnings report, which revealed that monetization fell 2% below guidance, impacting overall bookings. This shortfall is attributed to a shift in engagement from high-monetizing games to experiences with lower hourly monetization, compounded by changes in their recommendation algorithm that prioritizes long-term retention. The company forecasts a year-over-year decline in Q3 bookings and anticipates continued monetization weakness despite investments in AI and other areas.
Steam store redesign finally live for everyone
Valve has fully launched its redesigned Steam Store homepage, featuring a wider layout, larger game artwork, and improved experiences for Steam Deck and Big Picture Mode. Key additions include a Personalized Calendar for tailored game recommendations and expanded visibility for DLC and wishlist deals. The update also shifts focus in the 'Popular Upcoming' section towards major releases, potentially impacting indie game visibility.
Steam gets huge tags overhaul for tailored recommendations and 'helping players identify the games that best fit…
Valve has significantly overhauled Steam's tagging system to improve game discoverability and player recommendations. The update includes adding 17 new tags, removing 28, and merging others, with the goal of helping players find games that match their interests and allowing developers to better describe their titles.
Prime Video is adding a TikTok-like feed
Prime Video is introducing a new vertical recommendation carousel feature, similar to TikTok's feed. This update aims to enhance content discovery for users by providing a more dynamic and engaging way to find new shows and movies.
Spotify’s new Taste Profile feature lets users fine-tune their algorithm’s recommendations
Spotify announced its new "Taste Profile" feature, which allows users to customize their algorithm's recommendations for music, audiobooks, and podcasts. This AI-powered feature, currently in beta for Premium users in New Zealand, enables users to specify preferences and exclude unwanted genres. The company is also expanding its AI features, following the recent Prompted Playlist functionality.