The Turing team is excited to go into detail on how Turing Bletchley v3, a new multilingual vision-language foundation model, is used across Microsoft products. This model is the latest version of Turing Bletchley -- our series of multilingual multimodal foundation models which understand more than 90 languages. You might already be using Turing Bletchley v3 when searching for images via Bing, and in this post, we will share more details...
Read More

As we announced earlier this week, with Bing Image Creator you can now turn words into pictures to truly express your imagination, giving you access to infinite image possibilities right from within Bing. A vivid dream, a birthday invitation to fit a unique theme, the new concept you’re trying to land at work. These are personal and specific. These are things that require a unique image to express your unique idea. Even better is that you...
Read More

A year and a half ago, I wrote an article suggesting that search was far from a solved problem, and that in fact it was poised for a revolution. I suggested that by applying deep learning models and making search more modern and intuitive, we could significantly improve the experience for our users. I’m thrilled to see that we are making a large step forward towards this vision with the recent announcement of the new Bing. We’ve...
Read More

The advantages of large language models are undeniable due to the capabilities achieved and opportunities ahead. We must also consider the benefits and potential harms coming via uses of these models. In 2021, we founded the Microsoft Turing Academic Program (MS-TAP) that provides academic teams with access to some of the world's largest language models. The program reflects our belief in the importance of having diverse and talented teams...
Read More

Turing Universal Language Representation Model family (T-ULRv6) has achieved the 1st position on both the Google XTREME and GLUE leaderboards, demonstrating that a single multilingual model can achieve state-of-the-art capabilities in both English and Multilingual understanding tasks. Multilingual technology is not only a technical challenge but also a social responsibility. Therefore, we are committed to democratizing AI by...
Read More

The Bing search stack is built for speed at all layers. Fast performance delights users and drives loyalty and engagement. This blog presents the key performance techniques used by Bing’s front-end to deliver and render search results pages with world-class speed.
Read More

The Microsoft Turing team has a new addition to its family of image and language models. The Turing Image Super Resolution model (T-ISR) uses deep learning to deliver the magical “enhance” to any image in the world. It is already being used today to improve the quality of the aerial view in Bing Maps around the globe and is also rolling out in Microsoft Edge to bring its powers to the entire web. See examples of the model in action...
Read More

We are proud to introduce WebQA, a dataset for multi-hop, multi-modal open-domain question answering challenge, to be hosted at NeurIPS 2021 Competition Track. Designed to simulate the heterogeneous information landscape one might expect when performing web search, WebQA contains 46K knowledge-seeking queries whose answers are to be found in either images or text snippets, where a system must determine relevant sources first before reasoning to...
Read More

Rank Multimodal (RankMM) The RankMM model effectively combines the search paradigms of a text query, page context, and images to aid image and video retrieval. RankMM models are Visual Language (VL) models which take page context into account to improve image and video retrieval performance in a web-scale search engine.
Read More

Enhancing image quality is a critical but challenging computer vision task. We've released a new AI-based model for improving the quality of images on Microsoft Bing. Not only are the Bing image search results relevant, but they are also beautiful and high-resolution. Our new V3 model outperformed the V2 model by 36% in terms of click-through rate, demonstrating that users find visually appealing images in search results more interesting and...
Read More