Microsoft believes 2018 is the ''year of AI''. To that effect, the company has been investing in multiple projects pertaining to the field of artificial intelligence. In fact, only a couple of days ago, Microsoft announced an AI which reads and answers questions about a document with human-level accuracy. Similarly, the tech giant is aiming to decode the immune system using AI as well.
Today, Microsoft has revealed yet another new AI tech under development - a bot that draws what you tell it to.
This new tech, dubbed the ''drawing bot'', can generate highly detailed images based on caption-like text descriptions. According to Microsoft, a recent research indicates that the new techniques for text-to-image generation employed behind the creation of the bot produce an image quality with a three-fold improvement, in comparison to previously considered top techniques. Interestingly, each image also contains plenty of minute details which are not specifically mentioned in the text, which indicates that the AI contains an ''artificial imagination'', according to the company.
Xiaodong He, Principal Researcher and Research Manager at Microsoft Research AI, explained the working of the bot in the following way:
"If you go to Bing and you search for a bird, you get a bird picture. But here, the pictures are created by the computer, pixel by pixel, from scratch. These birds may not exist in the real world — they are just an aspect of our computer’s imagination of birds."
Creation of the drawing bot has involved considerable research in the fields of computer vision and natural-language processing (NLP). Exploring these fields, the researchers have advanced from image captioning to image generation. The latter is a much more difficult task as it requires the machine learning algorithms running the AI to imagine details which are not included in the caption.
The technology which is a major component of Microsoft's new AI bot is known as a Generative Adversarial Network (GAN). However, the tech giant believes that the technology is imperfect at the moment. Several minor flaws can be observed in the generated images, but the company believes it still a major step on the road to human-like artificial intelligence. In the future, the company believes that improved versions of this bot could be used as a sketch assistant, or even to generate animated films based on screenplays.
This new announcement makes it seem as though the tech giant is further cementing its footing in the field of Artificial Intelligence. Recent collaborations in AI with companies like Fujitsu and Nordcloud seem to indicate as much.