Tapping the wisdom of crowds
Could the next revolution in application development involve the wisdom of crowds? At least one start-up thinks so. MobileWorks uses large collections of workers in developing countries to process tasks normally undertaken by computers.
App developers using this service can farm out user input to the crowd for processing. Workers are sent tasks including handwriting recognition, data classification, audio transcription, and article proofreading. They process the tasks more or less in real-time, returning accurate answers via the Internet to the developer's app. The app then delivers those results to the user.
Applications that use this crowd-sourced service include MobileWorks' Digitizer, which recognizes hand-written forms and transforms them into Excel spreadsheets. The forms are farmed out to workers who could be anywhere across the globe. Another, called Excavator, uses large teams of humans to collect email addresses and generate leads by reading websites.
Using the MobileWorks service is as simple as cutting and pasting a short code sequence into your own applications source code. It effectively exposes the crowd of individuals available to MobileWorks as an API.
This isn't the first time that an online service has used the power of crowds to make tasks easier for its users. Amazon's Mechanical Turk allows users to farm out basic tasks to a wide variety of individuals for micro-payments. Such tasks include tagging images, finding websites for particular businesses, and extracting email addresses from lists. However, this is more of a manual brokerage system, than an application that automatically farms out tasks.
microtasks on cellphones
Other systems, such as TxtEagle, are more aligned with MobileWorks' strategy. Developed by former MIT Research scientist Nathan Eagle, it was designed to marry crowd sourcing to cellphones. It divides work into ‘micro-tasks’ that can be sent via SMS to a mobile phone. Most of its workers are based in Africa and are paid via the mobile phone, either in airtime minutes, or in cash.
Another service, Microtask, was founded in 2009 by entrepreneurs based in Finland. It divides tasks into small pieces and farms them out to digital workers around the world. Its services include digitizing difficult-to-read text as found in handwritten forms, and archives of analogue material.
IQ Engines has a unique take on crowdsourcing; it uses computerized visual analysis to identify images submitted by users. It also sends those images to its crowd of workers, which labels the images based on manual analysis. Its computerized system then learns from the labels that the crowd has given the images, continually refining its results. The crowd effectively trains the system to analyze more images automatically. It could be argued that this effectively forces the crowd of workers to make itself redundant, but of course, new images appear all the time, and the computer vision API will always need refining.
Crowd sourced mobile and desktop apps of this kind are likely to make computerized analysis services increasingly accurate and effective. This will improve the user experience, while also providing workers in the developing world with much-needed revenue opportunities. Are there any downsides to this? As long as the companies employing these crowd-sourced mini-contractors treat them well, we don't think so.
What do you think?