This morning there was a neat new toy in my web arsenal.  First let me set the stage for you, you see, I use a lot of Google services; from Gmail, Google Calendar and the Google Documents suite, all the way to Picasa and even Goog411.  From that last sentence you can already guess that I know a lot about Google, and that Google knows even more about me.  But, I guess I appreciate innovation when I see it.

This morning Picasa shot me an email, letting me know that my little sister had just posted her vacation pictures for all of her family and friends.  She went to New Mexico, and now she wanted to rub it in a little.  I was halfway through her photos when something caught my eyes.  Off o the right of the view area there was a new message… Picasa wanted me to add name tags.  You know, at first this didn’t shock me.  Flickr does it, MySpace does it, Facebook does it… but this is different.  Picasa first indexed my photos, all 5000+ of them.  That took a few minutes.

Then Picasa offered me a screen that showed me a view of faces it thought to be of the same person, but in different images.  And it was very, very accurate.  The image I scraped of my screen shows you some random faces.  (By this time I had spent too much time playing and already had the bulk finished.)  The face selected is that of Dana White, the guy that runs the UFC.  I selected him because he is already a celebrity, and because he’s a nice guy too.

Once I had selected his image Picasa prompted me to associate his face with a contact out of my Gmail account, smart.  (I will tell you why that is so smart in a moment.)  I typed in his name, as he was not a contact already in my Gmail system, and saved the tag to his image.   So now when Picasa finds his face in my other images, it has to decide if it is him or not.  Picasa is a machine, so it doesn’t exactly make arbitrary guesses, rather it uses mathematical equations to decide probabilities: is this image close enough to the tagged image to be considered a possible match or not.
Once the image is tagged, you see a cool little box around the person’s face when you are viewing a picture that contains them, like the one to the right here. (Click on it for a better view.)  If I had linked you directly to my Picasa album you could have clicked on his name to the right of the photo to see more images tagged for him.  On each of the photos there can be faces that are present in many more images on my Picasa site.  By tagging their faces you can allow the viewer to quickly move around the diverse albums but stay with an individual.  My wife thought this was so cool, she looked at pictures of the kids all night.  This is where the viewer gets the value, and where Google learns.

Google is using that mathematical equation that I mentioned a bit earlier, but it isn’t perfect, yet.  As I move through my Picasa pictures and tag all of these faces, the machine is learning, from me.  I find a picture with my 5 year old in it, tag it to her name, find her in another picture from three years ago, and tag that one; Google’s Picasa engine sees that, registers it and ‘learns’ that both images are of the same face.  This happens over and over, and soon enough, the Picasa engine will become very accurate at more than just facial recognition.  Not just accurate with pictures of my daughter, but with pictures of any face, or any two like objects.  You see, Picasa doesn’t really care that they are faces, in fact, they are just similar objects according to the math.  A set of measurements that do, or do not, coincide to produce a similar image.

But earlier I mentioned that Picasa was allowing me to associate these facial images with my Gmail contacts.  This is brilliant.  I get to associate more information with the person in my contacts section, and Google is again learning.  So what are they learning from this?  Well, they are getting a ton of useful data again.  Stuff like this:

• Am I in pictures with some of my contacts?
• What is the geographical location where the image was taken?
• In what other pictures is this person showing up in, and are those picture outside of my albums?

For instance, if I have you in my contacts, and you show up in some of my images, then there is a greater chance that we are actually more closely associated than just the many emails we exchange.  (Don’t forget that Google reads those emails too, only to provide ads, but still…) So now Google knows that you and I hang out and watch movies together, but hey, you show up in other Picasa albums across the ‘net too.  You know people I don’t know yet.  Perhaps Google should recommend that I meet your other associates?  Maybe Google will market similar products to you, me and the other person that has your images too?  If we attend social gatherings together, we might have similar taste in commercial products, right?

The first time I saw some technology like this from Google was the Google Image Labeler.  Image Labeler pairs you with an anonymous partner and then you are both shown an image, and a set of words that you cannot associate with that image.  Then you have a set amount of time to offer applicable suggestions as to what the image contains.  The more matches you get with your partner, the more points you both get.  This was, as far as I know, Google’s first real attempt to get the crowd to teach the machine.  Humans spend time looking at images, tagging them and the machine looks at the data to learn what the images mean to the humans.

Perhaps this technology will find its way into the enterprise in a way that will revolutionize our workforce.  Comparing your pictures of your products to every image Google can find, looking for the copyright infringement.  Or perhaps a service will spring up that can search images from public events,pictures taken by you and me, but tagged with the date and name of the event.  Once those images are known, the service could be searching for known people in the crowd so that you can market to them. Hey Bob, glad you enjoyed the event, here is a coupon for dinner in a deli close to the event.

All of this to show you that we are speeding head-first into a new era of computing.  One where the cloud is smarter than the individuals that make up the cloud, and Google is striving to be a massive current within this cloud.  The possibilities are endless…

Matt
matt (at) technologystory.com