Von Ahn’s Gwap

God, I love saying that headline.

Von Ahn’s Gwap.

Will a computer ever experience the kind of pleasure I derive from saying “Von Ahn’s Gwap,” or will that be reserved for humans?

As The Register notes, a new site was launched this week, by Carnegie Mellon’s School of Computer Science, that aims to entice humans into playing simple games that will help computers get smarter. The site, called Gwap (an acronym for “games with a purpose”), is the brainchild of computer scientist Luis von Ahn (who also cofathered the Captcha). “We have games that can help improve Internet image and audio searches, enhance artificial intelligence and teach computers to see,” he explains. “But that shouldn’t matter to the players because it turns out these games are super fun.”

The site includes the already familiar ESP Game, an image-tagging competition that Google previously launched as Google Image Labeler. It’s intended to create computer-readable metadata about pictures to facilitate image searches. And it offers four new multiplayer games:

Matchin, a game in which players judge which of two images is more appealing, is designed to eventually enable image searches to rank images based on which ones look the best

Tag a Tune, in which players describe songs so that computers can search for music other than by title – such as happy songs or love songs

Verbosity, a test of common sense knowledge that will amass facts for use by artificial intelligence programs

Squigl, a game in which players trace the outlines of objects in photographs to help teach computers to more readily recognize objects

One thing the Internet enables, which wasn’t possible before, at least not on anywhere near the same scale, is the transfer of human intelligence into machine intelligence. (Google’s search engine, which aggregates the human intelligence embedded in links, is a great example.) That capability can also, in theory, help train computers to do things that they haven’t been able to do before, such as identify the contents of pictures or make subjective or qualitative distinctions between similar things. If you can get enough people to tag enough photos of mountains as “mountains” in a machine-readable way, then eventually the machine will start to “see” the mountains in images without needing people’s help.

The challenge, of course, is to figure out a way to get people to do these kinds of routine chores – to work for the machine. (Tagging pictures gets old fast.) Amazon’s Mechanical Turk uses small payments to get people to contribute their time to extending computer intelligence. Von Ahn’s Gwap uses the pleasure of gaming as a lure. As Von Ahn says:

Unlike computer processors, humans require some incentive to become part of a collective computation. Online games are a seductive method for encouraging people to participate in the process. Such games constitute a general mechanism for using brain power to solve open problems.

In fact, designing such a game is much like designing an algorithm – it must be proven correct, its efficiency can be analyzed, a more efficient version can supersede a less efficient one, and so on. Instead of using a silicon processor, these “algorithms” run on a processor consisting of ordinary humans interacting with computers over the Internet.

In other words, we become part of the processor, part of the machine. In Gwap and similar web-based tools, we see, in admittedly rudimentary form, the next stage in cybernetics, in which it becomes ever more difficult to discern who’s in charge in complex man-machine systems – who’s the controller and who’s the controllee.

“Human computation doesn’t work unless you have people,” says von Ahn. “That’s why we’ve made the games on gwap.com as fun as possible. We need people.” For the time being, anyway.

12 thoughts on “Von Ahn’s Gwap

  1. Kaoticfen

    If i am understanding this correctly, GWAP is assuming that people are honest. Imagine if I went in and decided to choose the wrong answers (I understand that these games are entirely subjective so wrong may not be the best word but you get the idea). For example for Matchin I could always choose the less appealing one or for squigil instead of tracing outlines I always would draw a square. I know this sounds absurd but If thousands of people were to do this we wouldn’t have overall increased intelligence. Realistically im sure the people at Carnegie Mellon will get the user input they need to give their computers the ability to discern.

    I for one would hate to see the day that when I search for “pretty” it would return image results with “Paris Hilton” because of her appearance of being aesthetically pleasing to the eyes.

  2. Nick Carr

    One of the reasons they’re designed as multiplayer games is to filter out seditionaries like you. And, of course, once our overlords really wise up, they’ll use these kinds of games not only to filter out seditionaries but to identify them, at which point an electromagnetic pulse will be automatically sent through your Google Brain Plug-in and – well, suffice it to say that what happens next will not be Super Fun.

  3. Sergey Schetinin

    So Sci-Fi authors and other futurists thought that machines will be used to enhance humans. Wrong! It’s the other way around.

  4. Luis

    It isn’t the first time we become part of a wider system that we contributed to creating. After all, we created the financial markets, and you could arguably say that we have become “servomechanisms” of these markets.

    A positive thought : these new interactions with the computers should push us further as creativity, for instance, is concerned, and probably make new skills appear. But these evolutions do demand a new level of leadership in our society, not just within our corporations.

    Not there yet if you look for instance at problems within most national education system …

  5. Jarred

    I wrote about Von Ahn and human computation a little while back, tying it to the idea of “spare cycles.” Check it out here.

    To see Von Ahn explain the concept itself (something he’s quite good at), check out his presentation to Google a few years ago.

  6. Seth Finkelstein

    There’s a subtext of these sorts of posts which bothers me. It’s the flip side of what might be called the machine-worshipping posts of net-evangelists. You’re taking the same sort of analytical framework, but inverting it. So while they’re writing AI-good, you’re writing AI-bad

    (more nuanced, AI-liberating versus AI-subjugating).

    But it’s not about AI, but rather more at how humans use other humans.

    “In other words, we become part of the processor, part of the machine” – just like factory workers became part of the factory in a way. But the factory was not an autonomous entity except in a metaphorical way, and neither is machine “intelligence”. I wouldn’t say there’s no value in that way of looking at things, but there’s also a lot of potential for reactionary fogeyness and distraction.

    When I was a system administrator, I used to say I was “Servant Of The Machines”. But that was a joke, not a sociological analysis.

  7. Tom Lord

    Seth,

    I agree that it’s the human element that’s most critical but it isn’t so crazy to say “AI bad” in this context (not to claim that that is what Mr. Carr is saying, exactly).

    It’s a cybernetic thing. You build controls and they control — signals and systems. “We become what we build.” C.f. Weitzenbaum.

    It’s trivial to program computers to really mess with people’s head. That’s why we mostly don’t do it. Except that, of late, some do — at industrial scale.

    -t

  8. Charles

    When you speak of “machine intelligence” I immediately wondered if this was some new interpretation of the word “intelligence.” Coordinating human intelligence is not in and of itself intelligent.

  9. Nick Carr

    Machine intelligence is certainly a very different thing than human intelligence, though it’s probably easier for the former to mimic elements of the latter than vice versa. Of course, if you define “intelligence” to mean human intelligence (which is a reasonable definition) then “machine intelligence” would be a contradiction in terms. So you’d have to come up with a different phrase.

  10. Linuxguru1968

    Nick:

    Doug Lenat at Cyc Corp has been doing similar work with a game playing program to help develop the Cyc Common Sense Database: FacTory. AI is been pretty much a discredited field these days; aside form show ponies like IBM’s DeepBlue most major technology companies don’t spend a lot of money on it.

    General intelligence is based on the self organization properties of biological neurological systems which computer programs running on single processors are incapable of even simulating. Carver Mead at Caltech was creating silicon neural systems that replicate animal senses but creating large scale system of this nature is still decades away – none of the major technology companies see profits from it in the short term.

    Consider this, suppose that to create an intelligent machine capable of passing the Turing Test we just re-invent the human? What have we really accomplished?

  11. Tomek

    A few months ago, inspired by work of Luis Van Ahn, I decided to build a game where as people play and have fun they help to solve one of the current challenges in the area of computer vision and content based information retrieval. Snipetag http://www.snipetag.com is one of my free-time experiments and is now in BETA mode. Please forgive me bugs. If you have any suggestions, ideas of improvements please don’t hesitate to drop me a line.

Comments are closed.