Space Peas and Holy Toast: Why Pattern Recognition is Humanity’s Killer App

February 15th, 2012 by

microtask_holy_toastPeople are strange. We look for faces and animals in clouds, we spot Michael Jackson in a greasy roasting dish, and we pay tens of thousands of dollars for pieces of toast that resemble the Virgin Mary (personally, I think it looks more like Michael Jackson, but I admit that I am not a qualified toast inspector).

This compulsion to find meaningful images in random information is known as Pareidolia. Though it may seem like a useless and potentially dangerous handicap to our thinking, in fact Pareidolia is a key part of one of humanity’s most useful abilities: pattern recognition.

Pattern recognition is the key to some of our greatest achievements as a species, from language and music to Where’s Wally. Now it’s becoming even more important, as it becomes clear just how advanced our ability is, and how difficult it is to create software that can emulate it. This, of course, is where human computing and Microtask enters the picture.

Rise of the human machines

The current boom in human computing, signaled by Microtask’s recent success at the Red Herring awards, puts pattern recognition at the frontier of the tech industry. I’ve written in the past about the limits of artificial intelligence and the ongoing war between human and machine translators (spoiler alert: humans are still winning, but only just). This, along with the success of Digitalkoot in correcting machine recognition errors, shows just how far computerized pattern recognition has to go before it can rival the average human brain.

But as well as letting us prove our mastery over the machines by thrashing computers in translation races (which, incidentally, may be a bad idea if you take the threat of a robot uprising seriously), crowdsourced data processing projects are raising exciting questions about how powerful our pattern recognition powers become when we work together as a crowd.

All they were saying, was give Peas a chance

We’ve mentioned the Galaxy Zoo project before, and its ongoing mission to classify hundreds of thousands of images of galaxies with the help of a crowd of volunteers. But an unexpected discovery on the Galaxy Zoo forum provides a great example of what can happen when a crowd of people combine their pattern recognition skills. Members reported strange green blobs resembling peas, floating in the corners of the images, and a whimsical thread called “Give Peas A Chance” snowballed into a major astronomical discovery.

At first the peas were assumed to be insignificant errors in the images, but after a campaign on the Galaxy Zoo forum put pressure on the supervising astronomers, investigation revealed them to be a previously unknown type of compact galaxy. Astronomers gave the crowd credit for the discovery and showed their gratitude by officially naming these mysterious galaxies ‘Peas’.

The Galaxy Zoo crowd’s ability to notice unexpected anomalies shows where human pattern recognition is supreme. A computer would not have noticed the peas, as they were not part of the focal problem. However, sadly, for every Space Pea there is a slice of Holy Toast, and human pattern recognition’s strength (the ability to detect patterns and deduce meaning) is also its weakness (it can lead to people detecting patterns in meaningless noise).

This is where the power of the crowd can make a difference. Noticing the peas was only the first step: The Galaxy Zoo forum provided a venue for discussion which allowed the wisdom of crowds to get to work. We are used to seeing projects use the crowd to verify results like we do at Microtask. But the Galaxy Zoo discovery shows the potential of human computing to deal with more complex problems. My guess is that in 2012 we will start to discover just how much we can achieve with our powers combined.


  • http://www.facebook.com/breinlinger Josh Breinlinger

    Holy crap – that’s cool.  Never knew about the Pea galaxies. 


<<

>>