GILG!
September 14, 2006 7:35 PM Subscribe
Google Image Labler game You and a random partner try to pick tags for random images. If any tags match up, you both get points.
um, is this the nineteenth or the twentieth time I've seen this?
I think you mistook metafilter for del.icio.us
posted by unSane at 7:39 PM on September 14, 2006
I think you mistook metafilter for del.icio.us
posted by unSane at 7:39 PM on September 14, 2006
What shysters. People actually get paid to tag photos in the real world. Anyone want to play the "wash my dishes" game? Come over and wash my dishes, and you get points!
posted by lovejones at 7:46 PM on September 14, 2006
posted by lovejones at 7:46 PM on September 14, 2006
Yeah, I "played" three times before feeling like a total sucker.
posted by mrnutty at 7:49 PM on September 14, 2006
posted by mrnutty at 7:49 PM on September 14, 2006
Well, it's the first time I saw this. It's fun! My highest score is like 800.
posted by daninnj at 7:57 PM on September 14, 2006
posted by daninnj at 7:57 PM on September 14, 2006
This might work better if for every successful tag team tag you win .001 cents or something.
posted by TwelveTwo at 7:58 PM on September 14, 2006
posted by TwelveTwo at 7:58 PM on September 14, 2006
HEY EVERYBODY. If you are going to play, make your first guess "blimp". If you are matched with a MeFite then the matching will go quickly.
Here you go: blimp
Just mouse and copy that word. When you get to the game, all you have to do is type Cntrl-V in the box while you are starting to think of what to type.
Let's see how many pics can be tagged with "blimp".
posted by MonkeySaltedNuts at 8:05 PM on September 14, 2006 [1 favorite]
Here you go: blimp
Just mouse and copy that word. When you get to the game, all you have to do is type Cntrl-V in the box while you are starting to think of what to type.
Let's see how many pics can be tagged with "blimp".
posted by MonkeySaltedNuts at 8:05 PM on September 14, 2006 [1 favorite]
Anyone know if Google ripped it off 'The ESP Game' (wiki), or did they pay the guy?
posted by MetaMonkey at 8:20 PM on September 14, 2006
posted by MetaMonkey at 8:20 PM on September 14, 2006
MonkeySaltedNuts: Let's see how many pics can be tagged with "blimp".
I just did ten sets and got two 'blimp' matches. One was a piece of paper with a x b dimensions, the other was a photo of a guy dressed up in an Indian outfit.
posted by tellurian at 8:29 PM on September 14, 2006
I just did ten sets and got two 'blimp' matches. One was a piece of paper with a x b dimensions, the other was a photo of a guy dressed up in an Indian outfit.
posted by tellurian at 8:29 PM on September 14, 2006
Tellurian, I'm pretty sure that was me. Did you put blimp for eveypic (there was a fish and a green poster thing) cos I did and it only matched on those two. Seems like they have an inbuilt quality control or something.
posted by scodger at 8:36 PM on September 14, 2006
posted by scodger at 8:36 PM on September 14, 2006
I believe the guy in question, Luis von Ahn, works for and/or consults to Google. Here is his excellent lecture on human computation. The implications of this are fascinating and profound. Certain problems are particularly hard for computers, but trivial for humans. Provide a vast quantity of examples of human solutions to the problems, and computers may be able to derive rules.
Curiously in the video he didn't bring up, and wasn't asked about, the issue of emotive reaction to images. Porn is the most obvious example of this, but not the only one: there are images that will provoke disgust, laughter, sadness, empathetic pain, anger, familial comfort, etc etc. Some of these, notably porn, will be considered inappropriate for a game. (That said, I think a method of consistently describing and rating porn images would be of immense benefit, incidentally could produce an actually-working "decency filter", and will have fascinating implications for psychology.)
(On preview: sorry to puncture your blimps, but they have in fact given a lot of consideration to moderating the griefer/idiot factor.)
posted by aeschenkarnos at 8:44 PM on September 14, 2006 [2 favorites]
Curiously in the video he didn't bring up, and wasn't asked about, the issue of emotive reaction to images. Porn is the most obvious example of this, but not the only one: there are images that will provoke disgust, laughter, sadness, empathetic pain, anger, familial comfort, etc etc. Some of these, notably porn, will be considered inappropriate for a game. (That said, I think a method of consistently describing and rating porn images would be of immense benefit, incidentally could produce an actually-working "decency filter", and will have fascinating implications for psychology.)
(On preview: sorry to puncture your blimps, but they have in fact given a lot of consideration to moderating the griefer/idiot factor.)
posted by aeschenkarnos at 8:44 PM on September 14, 2006 [2 favorites]
I've tried several rounds, and many of them seem like I'm not playing agains a person.
The other "person" makes 3 guesses and then stops. Meanwhile I'm generating 7...8...9 guesses and the "other" is not going beyond 3.
posted by MonkeySaltedNuts at 8:56 PM on September 14, 2006
The other "person" makes 3 guesses and then stops. Meanwhile I'm generating 7...8...9 guesses and the "other" is not going beyond 3.
posted by MonkeySaltedNuts at 8:56 PM on September 14, 2006
MSN: Same for me. I wondered if my partners just got bored after 3 guesses.
posted by chiababe at 9:26 PM on September 14, 2006
posted by chiababe at 9:26 PM on September 14, 2006
Not just only 3 guesses, but 3 tepid and lame guesses.
Such as, round, red and planet for a picture that is clearly of Mars and even has a caption calling it mars.
And I try to make specific and useful captions and my partners always just want to do 'man' or 'woman' or the predominant color of a picture.
That would, when coupled with someone else with the same approach, serve to max out your points though. I just don't see how it could provide useful tags.
posted by ursus_comiter at 9:41 PM on September 14, 2006
Such as, round, red and planet for a picture that is clearly of Mars and even has a caption calling it mars.
And I try to make specific and useful captions and my partners always just want to do 'man' or 'woman' or the predominant color of a picture.
That would, when coupled with someone else with the same approach, serve to max out your points though. I just don't see how it could provide useful tags.
posted by ursus_comiter at 9:41 PM on September 14, 2006
collaborative indexing - a fascinating field of enquiry. In theory it works best within homogenous segments (think of the lexicon teenagers use to describe and categorise music), rather than amongst the population at large.
The main game here is coming up with a search engine that can deal with the proliferation of non-text content - video, audio. If you can get to this before Google, you'll be super-ultra-narco-petro-rich.
posted by achilles_hunt at 9:42 PM on September 14, 2006
The main game here is coming up with a search engine that can deal with the proliferation of non-text content - video, audio. If you can get to this before Google, you'll be super-ultra-narco-petro-rich.
posted by achilles_hunt at 9:42 PM on September 14, 2006
Get paid instead (by Amazon). Not much mind you, but it's candywork for the obsessive-compulsive.
posted by stavrosthewonderchicken at 9:50 PM on September 14, 2006
posted by stavrosthewonderchicken at 9:50 PM on September 14, 2006
It seems more about guessing the most obvious tags rather than the best tags. I guess those could be one in the same though.
posted by bob sarabia at 9:58 PM on September 14, 2006
posted by bob sarabia at 9:58 PM on September 14, 2006
Nope. It's one and the same, but good guess, bob.
posted by weapons-grade pandemonium at 10:03 PM on September 14, 2006
posted by weapons-grade pandemonium at 10:03 PM on September 14, 2006
Man, I hate doing Google's work for them, but this is fucking addictive.
Especially when one has important, but boring, work to do....
posted by mr_roboto at 12:48 AM on September 15, 2006
Especially when one has important, but boring, work to do....
posted by mr_roboto at 12:48 AM on September 15, 2006
aeschenkarnos: Won't somebody think of the porn?!? Think how easy it would be to find MMF lefthanded albino midget porn if we had reliable metadata! I'm just saying...
posted by Harald74 at 1:36 AM on September 15, 2006 [1 favorite]
posted by Harald74 at 1:36 AM on September 15, 2006 [1 favorite]
I played this when I first heard about it (I guess it was last week?) and it was fun for a bit, but then I got really frustrated with my stupid "partners" who seemed to not be paying attention at all. Either that, or they were typing with a single finger, and stopping to look up at the screen and make sure that each character showed up. I'd get on average 5-6 guesses in before their FIRST guess, and it seems like they NEVER pass! Or even worse - they pass before even trying a single guess! Grrrr.
posted by antifuse at 1:40 AM on September 15, 2006
posted by antifuse at 1:40 AM on September 15, 2006
sorry, antifuse. I'm new at this keyboard thing.
posted by TwelveTwo at 2:08 AM on September 15, 2006
posted by TwelveTwo at 2:08 AM on September 15, 2006
Won't somebody think of the porn?!?
I read the post title as 'The Google Image Labia game', so I *was* thinking of the porn -- only to be sadly disappointed by the actual content.
posted by PeterMcDermott at 2:16 AM on September 15, 2006
I read the post title as 'The Google Image Labia game', so I *was* thinking of the porn -- only to be sadly disappointed by the actual content.
posted by PeterMcDermott at 2:16 AM on September 15, 2006
Could it be that sometimes....sometimes you get paired with automated software? it suggests what it thinks is there, and if you say the same it can learn and improve.
Not a bad way to "teach" image recognition software.
posted by lemonfridge at 4:04 AM on September 15, 2006
Not a bad way to "teach" image recognition software.
posted by lemonfridge at 4:04 AM on September 15, 2006
Wow, what a testament to combined ignorance. :-)
My proudest moment would have to be labelling a nebula as a cell. :-P
posted by Deathalicious at 4:15 AM on September 15, 2006
My proudest moment would have to be labelling a nebula as a cell. :-P
posted by Deathalicious at 4:15 AM on September 15, 2006
antifuse,
I might be one of those slow partners. I'm slow because I don't want to just punch in labels that I know are wrong, so I limit it only to ones that I'm sure are right.
See my comment above about the galaxy=cell. I *knew* it wasn't a cell, but also I knew that none of my more conservative guesses would work. Finally I just typed in cell so we could move on to the next image, but honestly I think too many guesses is worse than too few.
posted by Deathalicious at 4:17 AM on September 15, 2006
I might be one of those slow partners. I'm slow because I don't want to just punch in labels that I know are wrong, so I limit it only to ones that I'm sure are right.
See my comment above about the galaxy=cell. I *knew* it wasn't a cell, but also I knew that none of my more conservative guesses would work. Finally I just typed in cell so we could move on to the next image, but honestly I think too many guesses is worse than too few.
posted by Deathalicious at 4:17 AM on September 15, 2006
lemonfridge, as somebody who's taken a machine learning class, it's definitely possible.
posted by onalark at 4:59 AM on September 15, 2006
posted by onalark at 4:59 AM on September 15, 2006
Is it just me, or do I find this rather insulting? Why should I do work on behalf of one the largest IT companies in America?
Can't they at least follow their own "Do No Evil" ethos and get some temps in to help catalogue their image cache like they did with the book scanning initiative?
posted by Funmonkey1 at 4:59 AM on September 15, 2006
Can't they at least follow their own "Do No Evil" ethos and get some temps in to help catalogue their image cache like they did with the book scanning initiative?
posted by Funmonkey1 at 4:59 AM on September 15, 2006
aeschenkarnos, good link. Really interesting.
posted by verisimilitude at 5:00 AM on September 15, 2006
posted by verisimilitude at 5:00 AM on September 15, 2006
This is a rip off of Fastr, only it isn't fun at all.
posted by fire&wings at 5:06 AM on September 15, 2006
posted by fire&wings at 5:06 AM on September 15, 2006
No returned "blimp"s at all. That game just lost all its aspect of fun for me.
posted by NinjaTadpole at 5:49 AM on September 15, 2006
posted by NinjaTadpole at 5:49 AM on September 15, 2006
In some Neal Stephenson or William Gibson novel there's a reference to a (fictitious) picture book, a visual dictionary called What Things Are, which exists because, if I'm remembering this correctly, people are so widely illiterate and inured to their lives being mediated by a sea of graphical user interfaces that no one readily learns the names of anything. I wonder if we're really headed that way.
posted by pax digita at 5:55 AM on September 15, 2006
posted by pax digita at 5:55 AM on September 15, 2006
This is so painful. If they really want to direct the mile a minute wham bazzle energy and mindless dedication of today's modern world it really shouldn't give four minutes to label some itty bitty picture.
posted by shownomercy at 6:14 AM on September 15, 2006
posted by shownomercy at 6:14 AM on September 15, 2006
It's ridiculous to expect people to identify and give useful tags to most of those images when they're only showing you the thumbnail versions. The resolution is shitty.
posted by needs more cowbell at 6:49 AM on September 15, 2006
posted by needs more cowbell at 6:49 AM on September 15, 2006
I can't believe my partner didn't match my tags of "shitty music" and "sucks" when a picture of Matisyahu came up.
posted by sbrollins at 8:46 AM on September 15, 2006
posted by sbrollins at 8:46 AM on September 15, 2006
I might be one of those slow partners. I'm slow because I don't want to just punch in labels that I know are wrong, so I limit it only to ones that I'm sure are right.
See, maybe i'm just doing it wrong then - I'm just trying for the highest score. Accurate image labelling be damned! If you want me to do it accurately, you'd best not make it time limited :)
posted by antifuse at 9:11 AM on September 15, 2006
See, maybe i'm just doing it wrong then - I'm just trying for the highest score. Accurate image labelling be damned! If you want me to do it accurately, you'd best not make it time limited :)
posted by antifuse at 9:11 AM on September 15, 2006
Why did this take a week to post to MeFi?
Watch the video that was linked up above. It's by the guy who came up with the ESP game, and Google worked with him to implement this.
posted by smackfu at 10:55 AM on September 15, 2006
Watch the video that was linked up above. It's by the guy who came up with the ESP game, and Google worked with him to implement this.
posted by smackfu at 10:55 AM on September 15, 2006
Peekboom is made by the ESP Game guy, I found it a lot more fun than the labelling game. The idea is one player tries to tell another player a word by revealing parts of an image. In the process you are teaching the system what words correspond to what objects in an image.
From the interesting video, I liked the dude's idea of distributed game playing as a general approach to problem solving, though he gave out frustratingly few clues of how the idea might be extended. I was however distracted by his awesome accent, finding his cadence mystifyingly compelling. Anyone know what the accent is?
posted by MetaMonkey at 5:09 PM on September 15, 2006
From the interesting video, I liked the dude's idea of distributed game playing as a general approach to problem solving, though he gave out frustratingly few clues of how the idea might be extended. I was however distracted by his awesome accent, finding his cadence mystifyingly compelling. Anyone know what the accent is?
posted by MetaMonkey at 5:09 PM on September 15, 2006
In some Neal Stephenson or William Gibson novel there's a reference to a (fictitious) picture book, a visual dictionary called What Things Are, which exists because, if I'm remembering this correctly, people are so widely illiterate and inured to their lives being mediated by a sea of graphical user interfaces that no one readily learns the names of anything.
What would be very cool (and self-referentially satisfying, not to mention necessary) would be if the first page of that book was a picture of a book (in fact that very book), illustrating what one uses a book for.
posted by stavrosthewonderchicken at 10:00 PM on September 15, 2006
What would be very cool (and self-referentially satisfying, not to mention necessary) would be if the first page of that book was a picture of a book (in fact that very book), illustrating what one uses a book for.
posted by stavrosthewonderchicken at 10:00 PM on September 15, 2006
« Older A quarter pounder with cheese - in a cup | Stay out of trouble. Newer »
This thread has been archived and is closed to new comments
posted by justkevin at 7:39 PM on September 14, 2006