Friday, October 5, 2007

The next big idea: human computing

Saw a fascinating lecture on "Human Computation" by a Carnegie Mellon professor who holds several patents on games like Peekaboom, Phetch, and the ESP Game. The idea is to harness free computing cycles from otherwise idle human minds to accomplish tasks, like image classification, that computers cannot do.

2 comments:

  1. The talk is from July 2006, and the very last question that Prof. von Ahn answered (which asked if these types of games could be used to have people solve 'captchas') must have sparked an idea: Prof. von Ahn recently launched another project at CMU called "reCAPTCHA" which uses the same concept to get folks to help process texts for Brewster Kahle's "Internet Archive" project by passing out unrecognizable words as captchas (see www.recaptcha.net and www.archive.org).

    Like Brian's earlier example of fast-food restaurant and airport shop managers using "purchase is free if you don't get a receipt" policies to get customers to help reduce employee theft, this is a pretty clever way to get people to do your drudge work...

    ReplyDelete
  2. Thought about this some more. Suppose the folks behind games like "Peekaboom" and the "ESP game" have a deal to share that keyword data so Google can use it in their AdWords system. What happens when people find out that their little coffeebreak diversion (or 15-hour-a-day obssession - good God!) is making money for Google?

    Something like this happened about 15 years ago - when CD-ROM drives in computers were a new and novel thing. Some folks wrote a CD player program called "xmcd" and thought it would be nice if the program could tell you the title and artist of the CD you were playing, and the names of the individual songs. (Many people are surprised to find out that CDs don't actually contain that information.) They came up with a little database for their program that would hold the title & track information, and developed a way to 'identify' the CD to link to that data. If you put in a CD it didn't "know", you would manually enter the information (title, artist, genre, track names) and it would be stored in the database. All the information was stored locally on your own computer, but people would sometimes e-mail their CD information to the program's developers so the data could be included in later releases of the CD player software.

    In the mid-90's they set up an internet server that everybody could access, and the information that everyone entered could be uploaded and centralized there. It worked quite well, and all kinds of multimedia software (including Microsoft's own CD player application) was using the system to retrieve and update CD information.

    Then a few interesting events transpired:

    a. The original programmers applied for and received a patent for their method of identifying CDs even though the source code for the program was freely available (in the industry, this is known as the "'680 patent")

    b. The original programmers sold their project to a company (and also joined that company).

    c. In 2001, the company (CDDB Inc, d/b/a Gracenote, Inc.) changed the terms under which people could access the data - only companies who licensed the company's patent and software would be allowed to use the information. Gracenote also began suing companies who had sold software in the past that accessed the database.

    As you might imagine, many of the people who had freely and voluntarily provided information to the database in the preceding 8 years were quite incensed.

    You can read more here:

    http://www.news.com/2009-1023-258109.html

    http://en.wikipedia.org/wiki/CDDB

    http://www.wired.com/entertainment/music/commentary/listeningpost/2006/11/72105

    ReplyDelete