jump to navigation

Solving intractable problems through gameplay April 30, 2007

Posted by jeremyliew in Consumer internet, gaming, Internet, social media, social networks, user generated content.

I’ve posted in the past on applying game mechanics to social media. Robb Web’s blog pointed me to a fantastic lecture by Luis von Ahn about how to design games to take advantage of human computation. In effect, he gets people to solve problems that computers can’t by making them into a fun game that people want to play. His research interests also include Captchas, automated tests that humans can pass but that current computer programs cannot. The most common version of this is the distorted set of words and numbers that you sometimes have to type when getting a new account or entering a comment on a blog to prove that you’re human.

I believe this lecture was part of the Google Tech Talk series. It was given in July 2006 and was widely covered at that time, but I missed it.

Examples of the games that von Ahn has developed include the ESP game that gets people to tag images, the Peekaboom game that gets people to identify objects within an image, and the Phetch Game that gets people to find a particular kind of image on the web. I’ve played them all, and they are all pretty fun!

Put these pieces together and you have the ingredients for an image search engine…

(In the lecture he also mentions the Verbosity game that helps collect well known facts to help fill the corpus of knowledge for AI research, but this seems to not be working at the moment.)

This guy is a genius. The a video is about 40 minutes (with another 10 minutes of questions) but it really is time well spent. I don’t make a 40 minute recommendation lightly.


1. Charles - April 30, 2007

In case you haven’t seen it, Google licensed von Ahn’s IP for use in Google Image Labeler – http://images.google.com/imagelabeler/

Recent AI advances have largely been in pattern matching with simple algorithms on large existing data sets, so naturally this research is quite helpful. Peter Norvig has a great talk on this for Google Langage Translation.

Recent computer vision work has shown that simpler classification algorithms (NNMF) with larger data sets to yield excellent results.

2. Rob Webb - May 1, 2007

Thanks for the shout out!

Robb Web?

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: