Crowdsourcing - The rise of people powered digital products
It is a well-known fact that computers can handle complex computational tasks very well. Infact, there have been several advances made in recent years to make computers smart, and acquire intelligence as they do tasks. However, there is still at least one area in which they remain unreliable - at understanding human emotions, expressions and the various nuances and inflections used in expressing ourselves. There is really no one better than ourselves at understanding us! This realization has given rise to a phenomenon known in these circles as “crowdsourcing”.
Very simply crowdsourcing is about the realization that there is wisdom in crowds. But crowds are equally famous for causing havoc and anarchy. Can the crowd be tamed to get at their wisdom and help us interpret data where computers give up? Turns out the answer is yes.
Almost two weeks back I attended a talk given by Rob Miller, from CSAIL, MIT who has been researching crowd-computing. Crowd-computing is yet another word for crowdsourcing. Rob pointed out that the most obvious and well-known examples of these are wikis and wikipedia.
Amazon also latched onto this idea several years back when it started a service called Mechanical Turk, the name riffed-off a hoax perpetrated in the 1900s. This service provides a marketplace for buyers of on-demand workers who could be hired for a mere pittance. Not only economical, on-demand workers also have relatively quick turnaround times.
Infact Google caught on to this too, and used the search terms entered by people to power its auto-suggest feature. Auto-suggest as many would know are the terms that come up when users start typing in a word into the search box.
Rob and CSAIL took this a step further and tried out several experiments to see how far the process of human intervention could be applied. Could humans for example help read and transcribe illegible or blurred text? In experiment after experiment they recruited humans without any pre-requisite skills, and saw that over successive incremental passes, the entire passage of text could be transcribed successfully by a human chain. In doing so, they varied testing method by trying both parallel and sequential chains, and both results worked. Their toolkit can be seen at: http://code.google.com/p/turkit-online/
There are several other prototypes in experimental stages which share the same premise. One such is an iPhone app called VizWiz, which helps blind people recruit remote sighted workers to help them with visual problems in near real-time http://hci.cs.rochester.edu/currentprojects.php?proj=vw. Another called Soylent is offered as “an add-in to Microsoft Word that uses crowd contributions to perform interactive document shortening, proofreading, and human-language macros” http://projects.csail.mit.edu/soylent/.
Indeed, this phenonmenon has become so popular that many of the latest iPhone and iPad apps for social and location based services are making use of it. A new entrant unveiled at the recently completed SXSW conference, Localmind, makes interesting use of it. A location based app, it wants you to ask questions of people checked into a particular location, and on the flip side be the expert of a location for the duration you are checked-in and help people by answering questions. Can you imagine the possibilities that exist for such a service at venues like sporting events, concerts or even on crowded highways as the crowds help each other out?
With so many applications of crowd-power, how can journalism be far behind? Here’s an interesting infographic powered by the crowd: http://www.economist.com/blogs/dailychart/2011/03/big_mac_crowdsourcing?fsrc=scn%2Ftw%2Fte%2Fdc%2Fburgerbill
Computer algorithms are still no match for people power. What other products would you like to see powered by the crowd?
2 Notes/ Hide
- meetamathur-blog posted this