Last year, Google launched a game that pairs you with someone, asks you tag random images and gives you points if one of your tags is identical to one of your partner's tags. Google Image Labeler was inspired by a system developed at Carnegie Mellon University whose purpose was to improve the results of an image search engine.
After 8 months since the application is live, the top contributor has 100,000 matched labels. Google Image Search has more than 200 million images in its index.
Until now, for each match you got 100 points. But many people realized that they could easily win points if they typed generic tags like "man", "people", "photo". To make the game more exciting and to improve the quality of tags, Google decided to change the way you get points. Now you can get anywhere from 50 to 150 depending on how specific your tag is.
"Labels that are more descriptive receive more points than generic labels. For example, for an image of a red corvette, matching on "red car" or "corvette" would receive more points than a match on the label "car". This is because more specific labels are even more helpful in our efforts to improve the quality of image search results, so we wanted to credit players accordingly."
Also the time for a round was increased from 90 seconds to 2 minutes and the thumbnails look bigger (they aren't bigger and your browser does the resizing).