Microsoft generated its own meme earlier this year when it tried to guess everyone's age. That produced some pretty amazing results, from 30-year-old kids to zombies aged 87. Microsoft is back at it again, but this time its trying to guess facial expressions in images. Emotion recognition is the name of the game, and Microsoft warns "recognition is experimental, and not always accurate."
Microsoft thinks Sad Keanu is only 0.01831 sad, and much more likely to be neutral and apathetic in his classic bench pose. Mystery solved, perhaps? Other memes generate some equally puzzling results. Success kid is apparently mostly angry, and no face is detected for first world problems. Thankfully, Microsoft gets it right with bad luck brian and rates him as happy. Xzibit's "yo dawg" face is also happy, apparently.
Aside from the fun and games of feeding this algorithm random meme images, there's actually some important work going on here. Microsoft is trying to improve its machine learning for images, and the plan is to allow developers to dig into various APIs to use that learning inside apps. That could allow apps in the future to detect when someone is moving in a video, or identify exactly who is speaking. Developers could create apps to filter out sad images (bye Keanu) or sort photos by how happy people are. Microsoft obviously has a long way to go, but it's this type of machine learning that search engines are scrambling to build. You can try your own sad or happy face over at Microsoft's Project Oxford page, and just make sure you hover over your face to see the full result.