AI is a big focus for companies like Facebook. Every time you load your news feed, some kind of algorithm is working hard to curate the best content for you. Have your closest friends shared a video of their holiday that you would like, does your favourite brand have a promotion that would be interesting to you, is there a viral video clip that’s suited to your interests. One way Facebook is looking to improve this system is with the use of a neural net AI, that can scan not only images, but also video content and text, detect what kind of content is featured and categorize it to serve it up in the news feed, on other websites and even respond to natural questions and searches based around the information it has learnt.
To demonstrate this, Facebook’s team taught the AI how to detect hundreds of different sports just by showing it a short clip. It was able to pick them out and categorise them, which could help improve upon the current and rather limited tag based system we use for videos.
A further demonstration involved showing the system the books of Lord of the Rings. The AI generated a short interpretation of the events of the story and in the real-time demo, they were able to ask questions about what happened, such as “where was the ring before it was destroyed?” and “where if Frodo now?” and the system responded correctly. This could have a huge impact on the way we navigate content and conduct searches on websites; I’m looking forward to seeing how this platform develops.