Knowledge Micro-outsourcing

Amazon's Mechanical Turk is a new service to let companies treat people like machines, which is probably better than they treat them now.

The "Mechanical Turk" name comes from a famous 18th-century hoax where a robot dressed up in Ottoman regalia played chess and defeated most challengers. It was revealed a century later that there was a chess master hidden inside.

Amazon calls the project "artificial artificial intelligence". It has set up a programmable interface to let programs call people as though they were functions, and ask them to do small bits of work best done by humans. For instance: type this scanned handwritten page, or associate keywords with this picture. These microtasks (under a dollar) are proposed to thousands of people, and the person who does the work gets paid. For certain tasks, the work is done in duplicate and the correct ones get paid.

Steve Tibbett, one of my favourite programmers since he sold me my first home computer (an early Amiga) and pointed me to some utilities he had written, has written an interesting post on this. He would like to see a Mechanical Turk function in the Visual Studio IDE to turn routine software tasks into micro-outsourcing. There are plenty of things that I would let strangers do for a few cents, if only asking someone weren't as much work as doing it. For instance: here is a web page address; find me the e-mail address or the phone number of the person whose name or title is mentioned. It's probably on the web site somewhere.

Innocentive is a site that does this on a larger scale. It lets research companies post difficult scientific problems for people to solve. Typically these are problems in chemistry and biology for pharmaceutical firms. The first or best solution is awarded the price, typically in the tens of thousands of dollars.

This is a way for pharmaceutical companies to manage their risk and reduce the cost of managing a scientific staff. There could very well be a scientist somewhere in the world who has already solved this problem or a similar one, and can undertake the work for a low risk-reward ratio.

There are plenty of scientifically trained people out there, in Russia and China and retired scientists who only want to work on problems that interest them. It is an interesting approach to outsourcing where you are only paying for results and need not worry about how likely is your employee or contractor to get the results.

There are of course difficult ethical problems to be resolved. Are companies who do this washing their hands of a responsibility to have good labour standards, safety, and environmental practices? I can imagine third world chemistry labs set up by individuals to work on dangerous substances without the expensive equipment that the law would require here.

But as globalization goes, the ability to set your own hours and work on the tasks that you choose has certain potential advantages. It's the old empower-the-knowledge-worker lie actually coming true.

The other possibility I see here is ways to advance artificial intelligence. If there are repetitive intellectual tasks being doled out with dollars attached, good programmers may be able to create real automata that carry out these types of tasks for money, a sort of artificial artificial artificial intelligence program if you will.

More on that and on the relationship between information and knowledge in an organization in a later post.

http://www.mturk.com/
http://en.wikipedia.org/wiki/Mechanical_Turk
http://www.innocentive.com/
http://www.csfb.com/thoughtleaderforum/2003/bingham_sidecolumn.shtml

[] permanent link Comments: 0