SpringHill Company — for the opportunity to create

[zakiyatasnim]

VFacebook announced that it has created a computer vision program that can "see" what it looks at. The artificial intelligence was trained using 1 billion public images from Instagram, the company reported on its blog.



The program, called SEER, outperformed existing artificial intelligence models in an object recognition test, Facebook said.

SEER received an “identification accuracy score” of 84.2% when malta number data tested by ImageNet, which tests whether an AI program can identify what is in a photo.

A new approach
According to Facebook's chief artificial intelligence officer Yann LeCun, the idea is to create an artificial intelligence that can learn in the same way as a baby — by observing and building a mental model of the relationships between objects.

To do this, SEER studied entire photos, parts of them, and similar images. As a result, the artificial intelligence learned to recognize data based on its individual fragments. It is noted that this step is considered critical to achieving human-level intelligence in the future.

LeCun also noted that the next steps in the development of SEER will be video recognition and “multimodal learning,” where the artificial intelligence system simultaneously studies images and text.

According to Facebook, the program and the algorithm underlying it will be open and free for all users.

Regarding the use of Instagram users' photos, the social network said that its data policy allows it to do so. It includes a clause stating that "the company uses the information it has to support research and innovation."