This metamorphosis isn’t happening in a vacuum. No company can merely snap its fingers and decide to become one of the world’s foremost players in AI. And to develop capabilities in this area requires more than computing power and human expertise. It also requires massive amounts of data about patterns in human language, in addition to intelligent algorithms that can interpret those patterns.
How did Microsoft come to acquire all this data about language patterns? In a word, Bing. Yes, the search engine many dismiss as an also-ran behind Google has had an outsized influence behind the scenes, serving as the foundation for much of the company’s AI expertise. Since 2009 (when Bing was launched), Microsoft has accumulated billions of natural-language questions about the kinds of things people ask and the ways they ask them. For example:
- “What’s the weather for next Tuesday in Florence?”
- “health benefits kale”
- “best burger joints Madison”
Bing has similarly has an impressive history of images submitted for searches saved in its vast database. The huge data store of natural language and images searches has given Microsoft an opportunity to develop AI algorithms, such as ones that can finish your search term for you, interpret your search term in the way you most likely intended it, or recognize objects in images. Machines learn by looking for patterns in data, and Bing has acquired far more than the critical mass of data needed to make useful insights possible.
The data provided by Bing, plus the desire to improve Bing search results, is one of the key factors that spurred Microsoft to accelerate its efforts in AI. As an example of this new focus on improving AI capabilities, Microsoft has adopted a type of reprogrammable computer chip—one that uses field programmable gate array (FPGA)-integrated circuits—to power the servers behind Bing and its AI-driven features. Using these chips, Microsoft engineers can write algorithms as needed right into the hardware instead of waiting years for better AI-optimized processors to appear on the market. And the new expertise and feature capabilities that Microsoft is developing in its AI efforts are being channeled to important products such as Microsoft Azure and Microsoft Office 365.
Just as Bing has galvanized Microsoft in its AI efforts, the fruits of these new AI initiatives are also being funneled back into Bing. Recently, for example, Bing introduced the ability to respond with multiple perspectives to questions that don’t have one clear answer, such as “Is coffee good for you?” And back in December, Microsoft announced a number of other new AI-driven features in Bing. For example, Bing can now formulate a single answer to a question based on information drawn from multiple sources, which itself builds on improved reading comprehension announced earlier last year, in addition to FPGA algorithms. Other new AI-powered features in Bing include interactive search capabilities, in which Bing asks follow-up questions to enable more precise answers, and improved image-search capabilities.
Bing has been an unsung hero in Microsoft’s transition toward AI. Massive amounts of search data collected using Bing over the years has provided an opportunity for Microsoft to develop cutting-edge AI algorithms, in addition to an incentive for developing AI and improving Bing through these same algorithms. This motor that has spurred Microsoft to improve Bing has had positive spillover effects into other services the company offers. In many ways, Bing is a key reason Microsoft is positioned at the front lines of the AI revolution that is now fully underway in technology.