Microsoft is bringing the machine-learning technology behind the company’s surprisingly successful How-Old.net Website to Bing Image Search.
Branded as the “#HowOldRobot” by Microsoft, the age- and gender-identifying site first appeared as a tech demo during last month’s Build developer conference. Three hours after Microsoft staffers sent an internal email, thousands of users began sharing screenshots on social media and the Internet depicting their supposed age.
“Within hours, over 210,000 images had been submitted and we had 35,000 users from all over the world (about 29K of them from Turkey, as it turned out—apparently there were a bunch of tweets from Turkey mentioning this page),” Corom Thompson and Santosh Balasubramanian, engineers in Information Management and Machine Learning at Microsoft, said in a May 4 statement. As with any work in progress, the technology’s guesses ranged from spot-on to amusingly inaccurate.
Microsoft is now putting HowOldRobot to work on its search engine, the company said this week.
“Seeing how much fun people had with #HowOldRobot, a small group of Bing engineers spent the last few days making the robot available in the new Bing Image Search,” stated Microsoft in a brief announcement on its Website. “We’re rolling the feature out over the next week, so when you see the Robot, you can have it analyze any image search result with one click.”
An accompanying screenshot shows an image of U.S. President Barack Obama, whose face is highlighted in a box and is crowned by the now-familiar How Old graphic. While it correctly identified him as male, the age—54—was off by a few months. (Obama turns 54 on Aug. 4.)
At the heart of HowOldRobot is the face-recognition component of Project Oxford, a collection of machine-learning APIs and software development kits (SDKs). The initiative’s other offerings include betas of speech and computer vision APIs, along with an invitation-only natural language technology called Language Understanding Intelligent Service (LUIS).
In the Bing implementation, Microsoft hopes to enhance the search engine’s image-based capabilities as well as provide users with novel new experiences. “This is one more example of putting computer vision and image understanding to work to increase the richness of Bing’s Image Graph and help our Bing users be inspired, learn more and do more with image search,” stated the software giant.
As with How-Old.net, users should expect some hiccups, said Microsoft. “This is cutting-edge image-understanding tech that improves every day, so please understand if it’s not always correct.”
Microsoft isn’t the only technology company banking on machine learning to build better search technology. Highspot, a startup launched by former Microsoft execs and engineers, is employing machine learning to advance enterprise search.
The company’s aim is to wean businesses off using tags, which are prone to losing relevance over time, to index and search for information. Instead, Highspot’s cloud-based platform creates a knowledge graph based on various types of content (spreadsheets, documents, videos, audio, etc.) stored on-premises or in cloud services like Dropbox, Box, Google Apps and Microsoft Office 365. By applying observed and learned facts, the platform automatically presents users with the most relevant information.