Here comes Google with yet another Beta version of a product. Only this one is like a game with a fun twist and could get quite addictive. The object of the program is to use human intelligence to label images. As powerful and ubiquitous as computers have become, there are still many tasks that us humans are still more skilled at. In this case, identifying a photo or an image (especially a blurry or a vague one) is a task best left to the human brain.
Amazon has capitalized on the same concept with its Mechanical Turk site. In that site people create tasks, called HITs (Human Intelligence Tasks) and invite others to respond. They could be research quizzes, surveys, categorizing web sites, or writing articles. Responders are paid for successfully finishing the tasks and Amazon keeps a commission.
In Google Image Labeler, Google harnesses its vast visitor pool to assign labels to images. Two people are paired at random for each round and for 2 minutes are shown random images in sequence. Participants are tasked with coming up with as many labels as they can for each image. One side doesn't see the other side's suggestions. If one of the labels match, the participants are given a score and they move on to the next image. Or they can skip the image.
What's in it for the participants? A journey into the psyches of 2 randomly connected people for 2 minutes at a time, and accumulating scores, perhaps for bragging rights. And for Google? A cost-free experiment to more accurately identify the images in its vast database. Since participants don't know each other and time is short, they are motivated to quickly suggest the most appropriate labels based on their visceral reactions.
If you get a chance, give Google Image Labeler a shot. Just be warned that it could get addictive. I had to stop myself after a few rounds, lest I waste hours in oblivion.
Speaking of Google, the stock climbed another 19 points or 3.25% percent today to $593. Since its low of $413 on March 10 (barely 7 weeks ago), it has risen nearly 37%. Could've, should've, would've.