In an experiment designed to use intelligent systems to write news stories an coalition of journalists and computer scientists aim to test current computing technology and human power to see if they can’t generate automated news stories. The experiment doesn’t expect to produce anything that could replace actual human reporters; but it will demonstrate the current capability of data processing to produce human-like reporting out of already-existing work.

The experiment is called “My Boss is a Robot” and it will use the distributed crowdsourced power of Amazon’s small-task outscourcing engine, Mechanical Turk, by writing an algorithm to break up the task of developing and writing a news article into a series of separate, human tasks and then coalesce them into a final product.

Amazon’s crowdsourcing engine is named after the Mechanical Turk, or an automated chess-playing automaton from the 18^th century. The Turk, in fact, was not a mechanical automaton as it seemed to be for excited crowds and dignitaries, but instead was a complex puppet for a small-person hiding inside the machine who directed its movements. Amazon’s process is very similar: it breaks up large tasks into small tasks, hands them out to tens or hundreds of people, and then welds them together afterwards. This Turk similarly seems robotic, but really relies on the human power of tens or thousands of people and the robotic part happens to be deciding how to break up the task and how to merge the results into the final product.

It’s essentially automated project management with a twist.

As for the “My Boss is a Robot” experiment, Courtney Boyd Myres at The Next Web gives us a run-down of the history of other similar endeavors and how they relate,

We’ve seen automated “robotic journalism” before. The NYTimes uses the semantic web to automate wedding announcements. Business Week used software to replace sports journalists. Infonic’s Sentiment software analyzes thousands of news stories to determine how a particular company is faring. And TV journalism could be replaced by a prototype of software that uses voice-over narration of a rapid-fire series of pre-taped stories, vidclips and images to present television news.

The Mechanical Turk has even written a simple encyclopedia entry about New York City. Based on this success, the scientists behind My Boss is a Robot believe Turkers (as the distributed workers on Mechanical Turk are called) could transform a research paper into a 500 word piece of original science journalism.

Crowdsourcing is an extremely powerful engine of creation when a product (especially data products) can be broken up into a multitude of discrete human-capable tasks. In fact, it’s been used for years even before the advent of the Internet by fan communities to sub animé films swiftly and then collate them together. However, with the emergence of the Internet, rapid communication permits even large, complex tasks to be completed rapidly by compelling only minutes a day from thousands of people and then aggregating them together.

We’ve seen crowdsourcing dabbled at by different industries already from GiffGaff trying to do it with customer service, the Gap (and others) running contests amid customers to generate logos and marketing paraphernalia, and we’ve seen it suggested in order to organize and aid in environmental crisis situations—like in the wake of the BP Gulf oil spill.

Here, with the “My Boss is a Robot” experiment, we’re probably not going to see any actual journalists replaced by crowdsourcing. What we will see, however, is simplistic articles which are simple aggregations of information produced in short periods of time through amalgamating the work of dozens of individuals—it allows for the generation of abstracts, summaries, and other rewriters of otherwise complex or confusing news stories into different contexts to allow varied audiences to connect with the news.

I think of this more along the lines of having a long document in a foreign language, then breaking it up into single paragraphs, and over the course of a few hours shop it out amid thousands of native speakers with good English to translate a paragraph or three each and finally putting it back together again.

Except in this case, the researchers want to do this with general reporting on a single subject.

As a writer myself, I’m curious of how this will end up rendering itself as it seems that the product may still need a final edit (or perhaps a series of crowdsourced editors) in order to give the final article a coherent voice. With multiple people all writing segments of the final article, it will introduce a sort of shifting-tone or even changing grammar problem that might not look weird to the average English speaker, but would still end up being very poor writing or at worst lead to a totally incoherent conclusion.

When the experiment produces a result, of course, be sure to expect that us at SiliconANGLE will give our own crowded creative opinion—another thing that services like the Mechanical Turk cannot yet replicate.

When they can, I guess I’ll be in trouble.