Smithsonian.com

Hanny's Voorwerp and spiral galaxy IC 2497 (credit: NASA/ESA)

Did you watch IBM’s Watson supercomputer trounce two humans playing Jeopardy last week and do you now fear a future controlled by these jumbles of wires and circuitry with really boring voices? No? Me neither. And not just because I refuse to be intimidated by an invention that contains more information than I could possibly remember and has reflexes faster than any human. You see, computers just aren’t good at some things, including science, as I was reminded recently at a session at the recent American Association for the Advancement of Science meeting.

Discoveries in science often depend on finding some piece of data, like a weird green cloud in a picture of a galaxy, and saying, “that’s funny.” Computers aren’t that good at doing that, and humans are also much better at spotting patterns visually. That makes us much better prepared to look at, say, a picture of a galaxy and properly classify it. That’s how the first Galaxy Zoo was born–out of the need to identify all the galaxies imaged by the Sloan Digital Sky Survey.

That first project finished in 2007 with the classification of 10 million galaxies (and the identification of plenty of odd stuff, including Hanny’s Voorwerp, the weird green cloud I mentioned above), and now there is an entire Zooniverse, where you can help scientists to complete tasks like find planets, study the moon, or recover weather observations from World War I-era Royal Navy ships. But even people who aren’t actively participating in projects like these are being mined as human computers for grand projects–though you might not realize it.

You know when you fill out a form online and get to that box with the difficult-to-read jumble of letter or words? That’s called a CAPTCHA. You can figure out what the letters say or spell, but a computer can’t. It’s a block for spammers. The latest iteration is called a reCAPTCHA, and these boxes contain two words. What you might not have noticed is that when you decode those words, you’re helping Google to digitize books. Google puts one word it knows and a second one that its digitizing program has labeled as a word but can’t identify into that box and asks you what both are. By decoding 200 million of those words each day, we’ve helped Google to digitize millions of books.

Chris Lintott, an Oxford astronomer and one of the Zooniverse founders, noted that soon the tide of data will be so large that it will overwhelm what humans can handle. When the Large Synoptic Survey Telescope goes online in a few years, for example, it will scan the sky every three days, producing as much data as the Sloan Digital Sky Survey did in years. At that point, humans will still be needed, Lintott said, to train the machines.