Facebook has been training its artificial intelligence systems for a long time so that they can be able to analyze and know, with increasing precision, the content that is available after each photograph uploaded by its users.
The company directed by Mark Zuckerberg has the enormous advantage of having the contents that are shared publicly daily by its millions of users in its different services, following an existing point in its terms and conditions that users must accept when creating their accounts. , in which it leaves the doors open to the use of user content for research and product development, with which their participation is essential to enable ever more precise and effective training of neural networks.
An ambitious project with which to have public content from users
In fact, a week ago it presented SEER (SElf-supERvised), its new artificial intelligence system that tries to be more effective in all areas of computer vision, for which millions of Instagram images have been used, and as its acronym indicates, there has been no human supervision in their training.
Well, now Facebook wants to do something equivalent but with the content in video format. For this, the company has just presented the project Learning from Videos For which to try to train Artificial Intelligence models so that they can be able to recognize with precision and accuracy everything that happens within the videos.
In this project I will make use of the millions of user videos that have been shared publicly.
The company points out that the resulting machine learning models will allow the creation of new content recommendation and moderation systems, although there is a possibility that it could be applied in other functions, with which there are already those who fear that it could be used in advertising systems. to better know the users themselves to offer them more targeted ads.
There are also fears for its use at a future time when they have already launched their smart glasses, of which not much is known about it yet. Facebook has been vague in the data provided, where at the moment it has not wanted to pronounce on the need to know whether to allow users to oppose the use of its content for training, among other aspects.
Facebook is not the only technology company that wants to generate advanced machine learning models in video analytics, as Google also does the equivalent with a huge selection of videos available on YouTube.
In the case of Facebook, they will also choose to use a training system in which the human factor does not intervene in the entire process. It will already be a matter of seeing the possible uses once the new systems that are being developed along the way are put into production based on the new automatic learning models.
More information: Facebook