Crowd sourcing replaces journalists
Friday, 11 February 2011

Crowd sourcing journalism is just a step on the way to completely deskilling any complex task. Can Amazon's Mechanical Turk do the job well enough to prove the point?

Can you break any skill down into small easy to specify chunks that almost any even slightly intelligent agent can perform?

Writing a news item - like this one, say - clearly involves sophisticated intelligence that cannot be reduced an algorithmic process. However, there have been attempts to do exactly that and the latest experiment is novel because it uses crowd-sourced intelligence to complete separate tasks. Is an experiment that uses crowd-sourced intelligence a test of AI? It is as long as the activities are sufficiently well specified.


The latest experiment is being conducted by a group of journalists and programmers using Amazon’s Mechanical Turk.

In case you haven't heard of the Mechanical Turk before it is a web site where simple tasks can be allocated to any of a huge number of people who sign up to get paid for completing said simple tasks.  In general the tasks are simple and repetitive - but even with an ideal task the Mechanical Turk has been plagued by problems of spam pollution.

The experiment is hosted on the website. The test is to make use of the Mechanical Turk to take a short scientific paper and construct a, hopefully readable, article for the general public. Work on the article is to be split into different tasks - what is the most interesting aspect of the story, locate other researchers who might provide a quote, get a quote and so on. Many of the tasks are designed to be completed more than once and a consensus average is used to construct the final article.

This isn't the first time that AI has been applied to the task of creating robot, but it probably is the first time that crowd sourced intelligence has been applied to the problem. The experimenters don't rate their chances of success very highly - but they feel that the experiment is worth trying before some publisher or other tries it for real.


The important point is that if this skill can be crowd sourced so can others of a similar nature. This would reduce the skilled to the unskilled and change the economics of the modern world.

More Information



Azul To Tackle Java's Warm Up Issues

Azul's ReadyNow is yet another attempt to tackle Java's warm up issues, on the cloud and large fleets of JVMs or JVMs in containerized environments.

SQLite 3.43 Released

SQLite 3.43 has been released with new support for Contentless-Delete FTS5 Indexes, as well as better JSON processing.

More News

Last Updated ( Friday, 29 July 2011 )