Facebook has launched new AI-powered systems to detect misinformation on its platform.
These systems rely on a variety of technologies, including ObjectDNA. Details of this technology were published in a study titled “An Analysis of Object Embeddings to Image Retrieval” earlier in the year.
ObjectDNA focuses only on the key objects in the image, rather than other computer vision tools that look at the whole image to determine the content. Facebook explained in a blog post that this allows them to identify reproductions of claims that are based on pieces from images they have already flagged.
California-based technology company, LASER, use cross-language sentence level embedding to evaluate the semantic similarity between sentences. It can be used for text, images, or both.
Facebook warned users about more than 180,000,000 pieces of content they viewed between March 1 and U.S Election Day 2020. These warnings were disproved by third-party fact-checkers. Facebook acknowledged its AI tools and stated that they helped to flag potential problems for review, and automatically identify new instances of misinformation.
SimSearchNet++ was used to apply warning labels. It is an improved image matching system that uses self-supervised learning and matches variations of images.
SimSearchNet++ can withstand a greater variety of image manipulations such as cropping, blurring, and screenshots. This makes it critical for visual-first platforms like Instagram. Facebook has also developed systems that can predict whether two pieces of content will convey the same meaning, even though they may look different.
John is working as a professional news editor at TechieGuard. He is accustomed to finding daily reports. Therefore, this keen working and addiction towards his work, it helps him to find good news.