Developers at Microsoft's Build conference severely underestimated the popularity of their experimental app, which guesses how old you are after you upload a photo of yourself.
Microsoft's How-Old.net will analyze any photo you upload, detect any faces, and guess the person's gender and age.
It's powered by Face APIs, a new tool that launched at this week's Build Developer Conference. The program's page describes it as "a cloud-based API that provides the most advanced algorithms for face detection and recognition." Proposed uses include verifying whether two faces in separate photos belong to the same person, or using one person's photos to find him or her in multiple other photos.
Microsoft's Machine Learning Blog team said that they emailed "several hundred people" asking them to try out the API, and expected about 50 people to use it. Instead, they logged more than 35,000 tests in only a few hours.
Clearly, they underestimated the general public's interest in age and appearance.
Nearly 30,000 of their hits came from Turkey. "Apparently there were a bunch of tweets from Turkey mentioning this page," they wrote.
The page also includes a series of stock photos to test the API. The Microsoft team said they assumed more people would use these instead of uploading their photos. Instead, they found that fewer than half of the tests were done with the stock pictures.
It's still very much a work in progress, though: users online described getting a wide range of ages with multiple photos, thanks to differences in resolution, lighting or framing.
That's probably part of the fun, though. "Forget Minesweeper or Solitaire — this is the best game Microsoft has bestowed upon humanity," writes The Verge's Josh Lowensohn.
Of course, no one said you had to upload a photo of yourself. Cue the deluge of pictures of ad billboards, celebrities and movie stills.
Have you tried the How Old Do I Look program? How accurate (or inaccurate) was it?