Microsoft has announced an update to Intelligent Answers using Intel’s field programmable gate array (FPGA) semiconductor chips, allowing Bing to read and quickly analyze billions of documents across the web and provide the best answer in a fraction of a second.
Because people are asking more diverse types of questions, Microsoft built in a technology called machine reading comprehension. Some other improvements, which roll out this week, allow those who are searching for answers to hover over definitions for uncommon words to gain additional information, aggregate facts across multiple sources, receive multiple answers for how-to queries, and get support for object detection on all common top fashion categories.
The update also expands many of the answers into the United Kingdom, improving quality and coverage, adding new scenarios.
In addition to using FPGAs to speed the return of answers, the feature -- which now aggregates facts for given topic across several sites -- also relies on artificial intelligence to provide those searching for answers with more comprehensive information.
These features require massive amounts of computational power, supported by the Intel chips. The technology was built on a deep learning acceleration platform called Project Brainwave, which runs deep neural networks on Intel’s Arria and Stratix FPGAs.