Microsoft has created a very advanced artificial intelligence. She is able to take a text description of an object and draw it quite accurately, according to a study released by the company.
The technology developed by the company is based on no real image. It creates the version itself from the description. “If you go to Bing and look for a bird, you find an image of a bird,” explains researcher Xiadong He. “Here, images are created by the computer, pixel by pixel, from scratch.”
The result is still not perfect but it is quite impressive, although it exhibits some imperfections. Artificial intelligence has been programmed to pay attention to individual words when it comes to generating images, which guarantees results that are quite close to what the text really means.
At first, Microsoft researchers created a system that automatically wrote captions for photos. From it, they developed another system capable of answering questions made about images. By uniting the two technologies, they came to the artificial intelligence capable of drawing the very version of the image from the text description.
The company believes that artificial intelligence has applications for painters and interior designers, for example, or inside a voice-activated image refinement tool.