07 January 2009 ~ 4 Comments                                       

Part 7: Many hands make light work: The atomization of work

Some tasks cannot be accomplished by automated, centralized means but can be broken up into smaller, manageable chunks that anyone can do. Shockingly, the internet comes in handy for these types of projects. Tasks that would be simple for a human (identifying an object or poorly scanned word in an image) can be difficult if not impossible to efficiently do on a computer.

The big kahuna in this space is Amazon’s Mechanical Turk. Break up a project into distinct HITs (human intelligence tasks), set your price for each task, and let the results roll in. Pricing ranges from as little as 1 cent to a few bucks per task depending on the complexity. The work itself ranges from the rewarding (helping search for a missing Steve Fossett and his plane using satellite images) to the unsavory (creating links and comments for spammers).

I see this as yet another example of the decreasing importance of the firm and a further lowering of the transaction costs of going outside your company for work. Have a huge project that you can break down? Without vetting and hiring a freelancer or outsourcing firm you can immediately post a project and start getting results while fine-tuning your request and pricing to get the best results. Pay can be dirt cheap (sometimes you can spend an hour doing something and only get $1 or $2), but in vast swathes of the world, that is a decent wage.

(On a side note and on the flip side of this: Some computing projects require unfathomable amounts of supercomputer time, so these projects, as well, are broken up into data crunching chunks that can be handled in a reasonable amount of time by home computers – I’ll call these CITs (computer intelligence tasks). Again the work spans a range of reputability: from helping Seti@Home search for alien signals to, unwittingly, sending emails or attacking sites for a botnet.)

In a twist on being paid to complete individual tasks, people either do it because they enjoy it, are getting an ancillary benefit from it, or, in a sort of slight of hand, are accomplishing more than they understand. GalaxyZoo was a project allowing anyone to participate in scientific research by identifying cosmological elements in telescope photos (Enjoyment). Gwap develops online games that accomplish image and music tagging while tagging sites on del.icio.us allows you to find them easily once you bookmark them (ancillary benefit). Finally, a stunningly brilliant application: reCAPTCHA. CAPTCHAs are those annoying tests given to you when you register on a site in an attempt to verify that you are a human and not a bot. But with reCAPTCHA, you are actually identifying words that computers could not identify when scanning books (Slight of hand). CAPTCHAs are a sad necessity given the current state of the world, but to put it to good use at the same time is strikingly elegant and simply awesome. (Further, by definition, you are using images that a computer couldn’t identify in the first place. Damn!)

(In a strange irony, it is possible to thwart CAPTCHAs by employing a system similar to Mechanical Turk in which you pay people a penny or two to identify CAPTCHAs.)

Any other examples out there?

For pay:
Mechanical Turk

Ancillary benefits:
Photo tagging on Flickr
Site tagging on deli.cio.us

Fun:
Gwap games
Write transcripts of Parliament sessions
GalaxyZoo

Unwitting participant:
CAPTCHA and reCAPTCHA

Let the computers do it:
Seti@home
Botnets (evil).
Hundreds of other distributed computing projects

Related Articles:

4 Responses to “Part 7: Many hands make light work: The atomization of work”


Leave a Reply