Monday, February 20, 2012

Online gaming helping science

The concept of using computers as tools to solve our problems today feels like a bit of second nature. But what can we do when we’re stuck with datasets our computer clusters simply can not compute or can’t compute efficiently enough for it to be feasible? An interesting approach is doing the reverse - having users solve your problems for you online, but then again you have the problem of how to get sufficient interest for people to spend their time doing this.

Zooniverse, which is a citizen science webportal, together with Yale University started up doing something really cool as they setup a game called PlanetHunter using public data from the public data archive. The amount of data generated by Kepler was too much for the NASA team to handle so they thought, what if we make a brief tutorial and let hundred of thousands of users look at our data and flag interesting spottings?
The outcome of this is clearly impressive as two possible new planets were identified outside our solar systems.

Another, perhaps even more significant finding, made by users of online games is the finding of an HIV DNA protein. Previously a normal distributed computing approach had been used where users let the researchers run tests on their individual machines, trying to generate random DNA structures and see if they worked out. However a group of Washington University students thought why not have the users help us with their brainpower as well? The result was the game Foldit, where users get highscores for how stable a protein they are able to make and solo efforts as well as team rankings are public. Along with this in order to gain even higher popularity the social features of chat and forum were added along with a wiki page. After only three weeks a solution close to perfect had been obtained and the scientists could then use this and only do slight modifications to find the actual protein.

I believe this is a very interesting approach to problem solving which seems to be fairly new in implementation. It combines probability theory of having enough people working on finding a solution eventually somebody will get it with creating something entertaining to attract these people. Of course this is only suitable for a specific set of problems but it’s truly mind boggling how efficient it can be with a year long mystery being solved in a mere few weeks. Go gamers!


1 comment: