Microsoft's new emotion-detecting AI can be hilariously bad at detecting emotions
Productivity killer of the week: A free Microsoft tool that can guess what you're feeling, but not very well
Now, Microsoft has launched an algorithm designed to recognize how humans feel, simply by reading our faces – and anyone who's not freaked out by it can try the beta version for themselves right now.
Announcing new <a href="https://twitter.com/hashtag/ProjectOxford?src=hash">#ProjectOxford</a> APIs, including emotion, video and spell check: <a href="https://t.co/ZWcshCRhNq">https://t.co/ZWcshCRhNq</a> <a href="https://twitter.com/hashtag/FutureDecoded?src=hash">#FutureDecoded</a> <a href="https://t.co/NXkLYcyFfL">pic.twitter.com/NXkLYcyFfL</a>—@msdevUK
The team behind Microsoft's Project Oxford, a set of advanced machine learning technologies for developers (or "intelligent services"), announced the public beta release of its new Emotion API Wednesday morning.
"Humans have traditionally been very good at recognizing emotions on people's faces, but computers? Not so much," reads a blog post published to Microsoft's website today. "That is, until now."
The post's author goes on to explain that this emotion tool can currently be used to help computers identify eight core emotional states in photos of human faces: Anger, contempt, fear, disgust, happiness, neutral, sadness and surprise.
While Microsoft does not reveal which pictures (or how many of them) were used to "train" the system, the emotions it can recognize are said to be based on facial expressions that are "cross-culturally and universally communicated."
Like most artificial intelligence systems, the Emotion API was designed to become smarter as it receives more data.
That's where you come in.
Anyone with an internet connection can currently access the beta version of the emotion tool on Microsoft's website and upload a photo to test it out.
The computer will analyze a picture within seconds to reveal how much of each specific emotion the subject is expressing on a scale from zero to one.
A score of, say, 0.900261164 in the "anger" category – as shown on the above image of NHL star Sidney Crosby – would mean that the person photographed is roughly 90 per cent angry.
Some of the images we tested pulled a pretty accurate reading, but others were way off.
That photo of Crosby, for example, was taken directly after he scored the gold medal-winning goal in overtime at the 2010 Vancouver Olympics.
Was he angry? Possibly, but it stands to reason that he was probably more joyful than anything when that picture was taken.
Many of those who were playing with the tool today shared similarly wonky results on Twitter.
i think you're a bit off <a href="https://t.co/b3U8q17nP7">https://t.co/b3U8q17nP7</a> <a href="https://t.co/cdG1wqMEPc">pic.twitter.com/cdG1wqMEPc</a>—@Brassrhino
Microsoft’s AI thinks I look happy here. I was not. <a href="https://t.co/Kh1NBuLrwy">https://t.co/Kh1NBuLrwy</a> <a href="https://t.co/zvkwSFJ7xL">pic.twitter.com/zvkwSFJ7xL</a>—@mcwm
Microsoft says Sad Keanu is only 2% sad <a href="https://t.co/Mimtj2wuis">https://t.co/Mimtj2wuis</a> Emotion Recognition <a href="https://t.co/RfNOSNxCbt">https://t.co/RfNOSNxCbt</a> <a href="https://twitter.com/Azure">@Azure</a> <a href="https://t.co/BbbtgT63ib">pic.twitter.com/BbbtgT63ib</a>—@jeffwids
This Microsoft Emotion Recognition is NOT optimized for Japanese subtleties in expression. I'm clearly smiling here. <a href="https://t.co/PR7X1v6hiS">pic.twitter.com/PR7X1v6hiS</a>—@kahodesu
<a href="https://twitter.com/suppadoopa">@suppadoopa</a> Emotion detection is pretty meh. <a href="https://t.co/p9l1PIBOBR">https://t.co/p9l1PIBOBR</a> <a href="https://t.co/p2Fpp5Kj04">pic.twitter.com/p2Fpp5Kj04</a>—@cyborg008
Conversely, others seemed pleased with the results they got.
<a href="https://twitter.com/hashtag/ProjectOxford?src=hash">#ProjectOxford</a> thinks (correctly) that I am neutral towards happy. Well at least they got this one right. <a href="https://twitter.com/hashtag/Microsoft?src=hash">#Microsoft</a> <a href="https://t.co/iujzs4fJ2K">pic.twitter.com/iujzs4fJ2K</a>—@_crazyredhead
Yep, that Microsoft emotion-detecting AI is super good at its job <a href="https://t.co/GjcVmBhZ1o">https://t.co/GjcVmBhZ1o</a> <a href="https://t.co/KroahVSd0g">pic.twitter.com/KroahVSd0g</a>—@TheNextWeb
Microsoft created a tool they say can guess how you were feeling in a picture. Mine wins :) <a href="https://t.co/su6U0i1gP7">https://t.co/su6U0i1gP7</a> <a href="https://t.co/VjpzNZWixa">pic.twitter.com/VjpzNZWixa</a>—@PhillyD
One clever sleuth even used the tool to find out what Mona Lisa was hiding behind her famous smile.
Monalisa Smile decoded by Microsoft's Project Oxford, interesting this is! <a href="https://t.co/sMC5BqwDCE">https://t.co/sMC5BqwDCE</a> <a href="https://twitter.com/hashtag/ProjectOxford?src=hash">#ProjectOxford</a> <a href="https://t.co/6vsbDOwgBs">pic.twitter.com/6vsbDOwgBs</a>—@GangulyAmrita
Neutral happiness. Thanks, computers.