Which festive film should you watch next? Let Netflix's AI pick for you

Original

There’s only so many times you can watch Home Alone, Elf, Love Actually and Miracle on 34th Street before you have a Christmas movie overload. Gone are the days of highlighting your top picks in the Radio Times and setting an alarm so you don’t miss the best ones. Now, when you’re looking for the next film or series to queue up, Netflix is your best friend. Never mind trawling through reviews and websites to find ‘films like Die Hard’, Netflix knows your preferences and is able to suggest titles that you’re most likely to be interested in based on your watching history through their AI that ranks a large catalogue by determining the relevance of each of their titles to each user. This doesn’t only include title selection, but also takes into account which images you will respond best to, 'producing both personalised content and image selection.’ At the Deep Learning Summit in San Francisco this January 25 & 26, Yves Raimond, Director of Machine Learning, and Justin Basilico, Research Engineer Manager at Netflix will share their most recent work on DL for recommender systems and will survey the methods they’re currently applying to personalisation and recommendations. Netflix have recently published research at the intersection of deep learning and recommender systems and how they relate to traditional collaborative filtering techniques, which Yves and Justin will discuss.

Netflix has been available to stream in 190 countries worldwide since early 2016. This worldwide roll-out was met with some skepticism - what about bandwidth? With slower internet speeds in many countries, would streaming performance suffer as a result? Netflix however, had this covered. Their AI is able to use ‘algorithms to review each frame of a video and compress it only to the degree necessary without degrading the image quality’ so you’re not stuck with poor image quality. ‘This not only improves streaming quality over slower speeds, but also tailors content for customers that view Netflix on tablets and phones, as is the case in countries like India, South Korea, and Japan.’ Netflix also takes time into consideration - you’re not going to get recommended the summer holiday road trip movies you were watching back in the summer - the AI will find similar time relevant titles to suggest and the more you watch, the more the AI learns and your recommendations get better and better.

Netflix are so focused on providing personalised experiences for its users, and are succeeding to the point that it represents 35% of internet traffic in North America alone. So what’s the next step? Well, researchers have recently announced their experiments to create personalised trailers for each subscriber to highlight the reasoning behind the selection for each individual. For example, if a high portion of your viewed titles are action themed, they can show you an action scene within the comedy that the trailer is advertising, making the viewer more inclined to give it a watch. With over 100 million subscribers the amount of data collected is able to accurately train the machines, constantly improving suggestions and building its knowledge on each subscriber.

Can’t be bothered to decide what film to watch? Just pick the first one Netflix recommends - you’ll probably surprise yourself!!

If you’re interested to learn more about Netflix’s research join us at the Deep Learning Summit in San Francisco this January 25 & 26. Register now to guarantee your place.

Original

Deep Learning Pattern Recognition Deep Learning Summit AI


0 Comments

    As Featured In

    Original
    Original
    Original
    Original
    Original
    Original
    Original
    Original
    Original
    Original
    Original
    Original
    Original
    Original

    Partners & Attendees

    Intel.001
    Nvidia.001
    Acc1.001
    Ibm watson health 3.001
    Rbc research.001
    Mit tech review.001
    Facebook.001
    Graphcoreai.001
    Maluuba 2017.001
    Twentybn.001
    Forbes.001
    Kd nuggets.001
    This website uses cookies to ensure you get the best experience. Learn more