I witnessed the future of AI on Netflix. It bypasses the hype and identifies a clear purpose.

      When you think of AI, names like Google, Microsoft, and OpenAI likely come to mind. However, Netflix, the largest streaming platform globally, doesn’t seem like the typical place where you would expect a generative AI chatbot—capable of processing vast amounts of information—to appear.

      After all, you subscribe to Netflix for films and TV shows. Maybe you catch a few short clips or even play some games. Nonetheless, Netflix has historically invested significantly in tools like machine learning in various ways, particularly to refine its recommendation algorithm.

      This year, Netflix is venturing into the “usual” AI territory, specifically generative AI, the technology behind products such as ChatGPT, Gemini, and Copilot. However, unlike other big tech companies, Netflix is adopting a more measured approach. Rather than overwhelming users with numerous AI tools that may not be useful, Netflix is taking a thoughtful route.

      Let AI align with your mood

      Netflix

      I find myself spending far too much time aimlessly scrolling through the Netflix catalog to select what to watch. It’s an exhausting process. I’m not the only one; a 2016 study by Reelgood and Learndipity Data Insights revealed that the average Netflix user spends 18 minutes deciding what film or TV show to watch.

      By 2019, that time had decreased to 7.4 minutes per session, which may seem minor, but accumulates to about 45 hours annually. According to a report by Talker Research and UserTesting last year, Americans dedicate around 110 hours (nearly five days) a year merely browsing streaming service catalogs.

      The struggle is genuine. To combat this, Netflix introduced a feature called Play Something to encourage users to watch actual videos rather than just scrolling through titles. What if there were a tool that could assess users’ moods and suggest a selection of perfectly suited titles?

      Netflix

      This concept forms the basis for an experimental search feature on Netflix, powered by an AI system developed by OpenAI. Instead of inputting vague keywords like genre or actor names, users can casually type out conversational phrases.

      During a press briefing, Netflix indicated that users could enter something simple like "something scary, but not overly scary" to find the exact type of movie that fits their mood in the moment.

      “We want you to discover shows and movies using natural conversational phrases,” a senior Netflix executive stated during a virtual press event. Currently, this is an experimental opt-in feature, as the company works to address "niche" risk scenarios, including explicit keyword searches.

      Find me something light-hearted. With a bit of romance. And maybe some suspense?

      Once the search system processes the natural language request, it will provide a carousel of movies and TV shows that match that specific theme. The entire system operates on OpenAI's tech stack and will be initially available on the iOS platform starting this week.

      Has Netflix trained an AI tool on its catalog? Did it conduct human labeling and use it as a "mood" classifier for the AI-driven search tool? Will there be age restrictions? Those details remain undisclosed, but the concept could profoundly alter our interaction with Netflix.

      A relatable instance could be looking at ChatGPT and mirroring that situation since the underlying technology is also what Netflix is implementing:

      Nadeem Sarwar / Digital Trends

      Allowing users to articulate their content preferences in detailed terms and assisting them in locating exactly what they want serves two purposes. First, it minimizes the time spent aimlessly scrolling through the content catalog.

      Second, it aids in content discovery and establishes a feedback loop for improved recommendations. This leads us to the next point…

      Accurate and dynamic recommendations

      Up to now, Netflix has utilized various “signals” to recommend content, including what viewers watched, the ratings they provided, preferred genres, actors, time spent watching, and language, among others. However, the experience hasn’t been fully personalized for viewers.

      The recommendation system also considers what “other members with similar tastes and preferences on our service” are viewing. Basically, if a TV show is receiving enthusiastic reviews and generating significant streaming time, it might be featured in your recommendations on the home page.

      Netflix

      This method is an effective way to inform users about the latest and most popular content, but it may not always align with individual preferences. Netflix acknowledges that there is room for enhancing its renowned recommendation system, and it is adopting a more dynamic approach to achieve this.

      When searching for content, the search terms will serve as signals. The system will detect details such as genre, actor names, or overarching themes. Based on this information, it will adjust the content feed in real time.

      What you search for and the trailers you watch will help determine what you wish to watch at that moment. Consequently, the home feed will suggest content worthy of your attention. The entire system “adapts to you as you browse,” the company explains.

      Netflix states that the home page will subtly adjust to each user, with everything happening