The AI system has been used to update the company’s assistant application for visually impaired people & quot; seeing AI & quot; and will soon be incorporated into other Microsoft products such as word, outlook and PowerPoint. There, it will be used for tasks such as creating alt text for images, which is particularly important for improving accessibility. < / P > < p > these applications include Microsoft’s own seeing AI, which was first released in 2017. Seeing AI uses computer vision to describe the world to the visually impaired through a smartphone camera. It can identify family items, read and scan text, describe scenes, and even identify friends. It can also be used to describe images in other applications, including email clients, social media applications and messaging applications such as WhatsApp. < p > < p > Microsoft did not disclose the number of users of seeing AI, but Eric Boyd, vice president of enterprise at azure AI, told the verge that the software is one of the leading applications for blind or low vision people;. Seeing AI has been selected as the best application or best auxiliary application by applevis, a community of blind and low vision IOS users for three consecutive years. < p > < p > Microsoft’s new image captioning algorithm will significantly improve the performance of seeing AI, because it can not only recognize objects, but also describe their relationships more accurately. Therefore, when looking at an image, the algorithm can not only tell what objects and objects are included in the picture (e.g., a person, a chair, an accordion), but also the interaction between them (e.g., a person sitting on a chair playing the accordion). Microsoft says the algorithm is twice as big as its previous image captioning system, which has been in use since 2015. < / P > < p > the algorithm was described in a preprint paper published in September and scored the highest ever score on an image subtitle benchmark called & quot; nocaps & quot. This is an industry-leading image subtitle scoreboard, but it has its own limitations. The nocaps benchmark consists of more than 166000 human generated captions, describing about 15100 images extracted from the open images dataset. The images range from sports to holiday shots to gourmet photography. It is said that “gta5” will be launched on Google cloud game platform stadia