One of the most complex processes when working with Artificial Intelligence is related to the learning process, with how we do so that a software is capable of making decisions from data.
Large companies have been presenting different ways of training these systems, and now it is Facebook that presents an artificial intelligence project based on video learning, since it can learn from the publicly available videos uploaded to the website.
The objective of this AI is to use audio, textual and visual data so that it can gradually provide better content recommendations for users and enforce content policies. In the future it will be able to recognize any type of video content, as opposed to the specific content selected by expert developers.
With the name Learning from Videos, they will be able to absorb data from different cultures and regions of the world and anticipate the habits of users while browsing Facebook, which can even help to avoid malicious content before it is published.
To ensure the ethical implementation of Learning from Videos, the AI team has adopted transparent privacy policies regarding user content, so that all information absorbed will be anonymous.
They have not commented whether it will work the same way in all languages, but we suppose not, although in this case it is also about analysis of image sequences, so errors should not be as common as in other machine learning platforms. .
The first application is already available in the Instagram Reels recommendation system.