Score Meaningless Points With Google Image Labeler!
- By: kidmercury [privmsg - website] On 2nd Sep 2006 In
Danny Sullivan reports on Google launching Google Images Labeler, a game intended at helping Google classify images better.
The game pits you against someone else. If you see a picture of a car, and you both label it car, you can proceed to the next image. You continue until your time has run out. Here is an image of Horcrux and Barry Schwartz from our blog (rustybrick) scoring 300 points for matching tags on three images.
Won't this just get overrun by spammers? Even if not, can Google do anything social successfully? Or is it all irrelevant as social software is just a passing fancy, as Nick Carr suggests?
- Y! MyWeb
Remains to be seen
I certainly see it as a spammer risk and a GEEK-o-meter. What segment of the online population wants to play an image game? A good cross section? Probably not so the data may be skewed.
As for social networking, I am a very lonely man in the view that it’s days are numbered. I don’t know how much more it can take as the PRIVACY of such models is waning and the BIG BOYS are moving in.
The video ads are sure to follow and it will be a commercial enterprise (it already is really) which to me goes against the whole idea of a community in the first place. Over commercialization will kill it.
Want to make some money? Offer hosted communities that are LOCKED DOWN from the commercial entities and people will gladly move in away from ‘the powers that be’.
Blogging, podcasting and RSS haven’t really taken off as the pundits predicted. I shall wait before in consider it a STAPLE of the web…
Outta here….
Will randomness protect against spammers?
Players are randomly matched. For a spammer to have a impact, the spammer must be matched up with a like-minded spammer.
Well I feel safer
Since I have never heard of spammers working together... or programming a sys to do so...
Whew.. (lol)
The biggest problem with this idea...
is most people are idiots.
I saw a picture of al franken giving a speech at harvard. It was obviously al franken, he jacket said harvard...
None of that matched, nothing did until i took a guess on the state of my fellow player. "Fag" matched. In a few minutes I found either really simple and useless concepts matched like sky, building etc or... no idea because we had no matches.
yeah I'm an idiot
I have no idea who Al Franken is either (yeah I can wikipedia him but that doesn't help when I'm looking at his picture)
Captcha challenges
Anybody else remember when the captcha people used this to research the potential for image-based captchas? That was more fun, because you knew the purpose. Banking has also researched the use of images as "private keys".
This "game" is a waste of time and IMHO poorly executed. Whoever had the need that this thing hopes to address should have hired a game developer and stated that need outright. I bet the game would have been much different, and way better.
I still don't believe it has anything to do with image search...smells like a 20% deal. If they actually think this might help search, THAT's scary. Who would ever submit a specific label as a guess in such a game? What are the odds of matching "Al Franken speaks at Harvard" exactly? Zilch. "Fag" or "dork" is so much easier, and more likely to match. Now for what query do you want to find that image? Exactly.
Interesting
Interesting
Better thanTexas Holdem'!
I messed around with it for a half hour. I thinks its a nice spin on things, but yes, very general...man, people, boay, lake...)
A game for 5 year olds
I like it....
dumb idea, same M.O.
google has always been great at getting people to do their work for them, and for free. many times they dress it up in "contribute to the greater good" horseshit - e.g., from their marketing copy:
they may be running low on that brand of goodwill these days, so they did something different. make it a game!! make it fun!!! earn points!!!!
recognize this for what it is. they see a business opportunity (image search being approx 10% of total search volume, from what i hear), and they've recognized that trying to understand everything with the neverending math equation that is their algo won't work in this case.
they can't understand images unless there are words around them, which means that even contributions from jackasses help - "douchebag" may not be an accurate descriptor of a given picture, but at least it's something for their algo to chew on, and it's probably 100% better than what they have to work with now.
Can we choose the images?
Can we choose the images? *eg*
Google Tech Talk Video
http://video.google.com/videoplay?docid=-8246463980976635143&q=esp+game
This video of the lecture given by Luis von Ahn at Google Tech Talk will explain how they combat spamming and other issues. He also presents two(?) other games that are equally interesting. One of which allows mapping of specific objects in an image.
The statistics given are pretty amazing and funny.