Putting Our Heads Together
A surprising resource is revolutionizing how we solve problems and get information. It is available in great abundance and can be harvested for almost nothing. This resource is the simple power of human thought.
We all know that computers can beat us at chess, find enormous prime numbers and remember our friends’ birthdays. If a task involves only calculations and large quantities of information, then a computer is perfect for the job. On the other hand, ask your computer to find the photos of a lovely peacock that you took on your last holiday, and it’s likely to throw in a picture of an elderly relative. Of course, any of your friends could do a much better job, but only if you can persuade them to bother. The emerging field of human computation is doing just that by using computer games to harness human brainpower.
A familiar example of a task we humans can do but computers cannot is decoding the fuzzy, distorted letters you are shown when you go to sign up for a Yahoo email account. You probably know that this simple test distinguishes flesh and blood from spambots. What you may not know is that reCAPTCHA, a brilliant variation on the original test, digitises books at the same time, presenting people with scanned images of words that a computer could not decipher. With humans typing 200 million reCAPTCHAs every day, in the past few years the project has digitised more than a century of The New York Times.
Evaluating the content of pictures, music and videos is also notoriously difficult for computers. To handle this problem, reCAPTCHA’s creator, Luis von Ahn, came up with a way to get people to do work tagging images with labels, for hours on end, for free.
His solution was to ‘gamify’ the task, thereby reclaiming some of the brainpower whittled away each day on games like Farmville and Angry Birds. He developed a series of (highly addictive) ‘Games With A Purpose’ (GWAP). In his ESP game, for example, you are teamed up with a partner in cyberspace. You each see the same image and type words to describe it. Matching words will earn you points. At the same time, your image tags are stored to help future Internet searches.
Both reCAPTCHA and GWAP show that people still have an edge when it comes to decoding text and understanding images, but we can excel at more complex problems too. A celebrated example is Foldit, which again uses the lure of a game. This time, the task is finding the best ways to fold proteins.
In 2005, the journal Science named protein folding one of the biggest unsolved scientific problems, and it remains a mystery. Proteins start out as chains of amino acids and many fold quickly into their natural shape. The question is, how can we predict the shape from the sequence? It has proved to be a very tough problem and computational cost is a big obstacle: testing all the ways a protein could theoretically fold would take much too long, even for a computer. A smarter approach is required.
Biochemists and programmers from the University of Washington decided to see if humans could help solve the protein-folding mystery. They created Foldit, an online puzzle game where players (who don’t need to know anything about biochemistry to get started) manipulate an on-screen protein, trying to earn points by finding more stable shapes for it. In a paper published last year in Nature (and co-authored by ‘Foldit players’), the game’s inventors showed that skilled players could outperform computer algorithms. They attributed this success to our ability to try different strategies and our willingness to explore avenues that may at first seem wrong.
Since Foldit, other games have begun to appear, in the hope that adding a little human intelligence might help with other questions. For instance, Foldit’s creators recently launched EteRNA, where players design new RNA molecules with the hope that some will have real-life biological function. The designers have even introduced a new dimension, pledging to synthesise and test the game’s most promising molecules each week.
As another example, in November 2010, researchers at Canada’s McGill University launched Phylo - another puzzle game where players try to align fragments of DNA from different species. The goal behind it is to find DNA sequences that are conserved across species and so may carry important genetic information.
Computers amaze us as they advance, growing constantly faster and more powerful, but with over one billion people online, the Internet offers access to a vast network of minds capable of tasks that computers cannot manage alone. Games, once seen as a waste of time, are gaining new respect as the key to tapping into this massive, living problem-solving resource.