Facebook launches new AI project to learn from videos

On March 30, according to foreign media reports, Facebook developers are building an artificial intelligence project called "Learning from Video" that can learn from public videos.

[[390603]]

The project uses available audio, text, and visual data and adds content from users around the world to its dictionary content. The focus of the AI ​​project is to provide content recommendations to users, implementation of content policies, and enhancing the ability of artificial intelligence to learn like humans. The above project is believed to be able to recognize any video content.

In the past decade, AI systems have made significant progress in speech and visual recognition, making them less dependent on acquired datasets.

Facebook's developers are focused on using "learning from videos" to enable the platform to absorb large amounts of data from different regions and cultures around the world. With this technology, the team hopes to create an artificial intelligence project that can predict user habits while browsing the Facebook platform. The development team has adopted a transparent privacy policy for user content to ensure the ethical deployment of learning from videos. Facebook developers confirmed that by implementing user trust and safety guidelines, content creators feel safe when using this new feature.

It is reported that this semi-supervised and self-supervised artificial intelligence product has brought great success to Facebook. Self-supervised learning products are believed to reduce speech recognition errors by 20%. Automatic captioning can be improved by reducing the above errors.