Guessing Ages

Recently, there has been a great deal of fun and action around a new offering from Microsoft.  It’s called How-Old and the premise is pretty simple.  You upload an image with people’s faces in it and the site tries to identify the gender and age of those in the photo.  The results come back looking like this:

The story behind the project is fascinating (albeit it a little technical) but would be inspirational for computer studies students.  You’ll see results, I’m sure, from friends on Facebook or you can follow the hashtag #HowOldRobot on Twitter.

I find the discussion interesting with people pleased that the guess of their age is way off, and sometimes, even the gender.  It’s one of those meme-y things that will probably come and go.  Interestingly, my friend Diane labelled it for what it’s doing with all of these incorrect guesses – #fail.  I tend to agree with her; I tried it with an image of my daughter and wife.  It came back correctly with gender but with only 10 years age difference.  You know what that makes me.

So, the bottom line is that it’s not ready for prime time yet.  I hope that the project continues and, like many things AI, gets smarter as more folks use it.  For example, there doesn’t come a time on the news when a camera recording of a crime isn’t played and “anyone who knows the people in the video are asked to contact police”.  Imagine how helpful it would be if it was done immediately.  Of course, the camera recording equipment would have to be upgraded – I don’t care how powerful your computer is – you’re not going to identify some of the images that are shared.

Image recognition has come a long way.  I’m sure that all readers are aware of Google’s search by image feature.  Or, who hasn’t used Google Goggles and had the service do its best to identify what’s happening in an image uploaded.  This application is a powerful extension of the concept and I think would be incredibly useful once perfected.

In the meantime, the best image I’ve seen and enjoyed so far….