With its Turkish smoking pipe, sorcerer’s beard and elaborate turban, the Mechanical Turk was a celebrated ruse, an eighteenth-century chess automaton wheeled out to beat the masters of the day. The artificially intelligent machine famously check-mated Benjamin Franklin, and even Napoleon. But its secret was artificial artificial intelligence: a chess master concealed inside the elaborate wooden cabinet, sliding around inside the cramped space, evading detection as false doors revealed decoy mechanics inside.
More than two hundred years later, we are perhaps closer to the dream of true artificial intelligence. For instance, Stanford University’s Kwabena Boahen is at the forefront of neuromorphic engineering, developing self-organizing silicon computing to emulate neuronal intelligence. But while efforts to reconstruct the elusive magic of human problem-solving and instinct in silicon continue, other researchers are finding novel ways to combine silicon computing with one of nature’s best computers: our own grey matter.
Fold.it is a website developed at the Baker Laboratory at the University of Washington. Its tag line— “solve puzzles for science”— coyly hints at the possibilities of tapping into the human super-computer, and its efficient parallel networked architecture, for scientific endeavor. Visitors to www.fold.it download the custom software, and begin to fold three-dimensional proteins in a competitive game environment. Using the allure of entertainment to encourage amateurs and experts alike to solve protein-folding problems, its simple brilliance is in its well-designed employment of the collective computing power of the crowd.
Protein science is notoriously complicated. Each protein is a three-dimensional building block, a key component in the workings of every cell in every living organism. Proteins are built from chains of amino acids, each constructed from atoms of carbon, oxygen, nitrogen, sulfur, and hydrogen. While this molecular string—manufactured in the cell from DNA’s instruction code— contributes to the protein’s function, its shape is key. The complex ways proteins fold are specific and repeatable for every protein. With molecular locks and keys built into the three-dimensional architecture, knowing the protein’s precise structure helps us to know what it does.
Determining and predicting protein shape not only advances scientific knowledge, it could also enable development of new cures for a range of diseases, from Alzheimer’s to HIV/AIDS. Potentially, new proteins could be designed, specific to targeted functions. But since human proteins are built from chains containing as many as one thousand amino acids, the immense number of possible combinations means that an astonishing amount of computing power is required to determine their shape.
Fold.it’s approach to protein science developed out of rosetta@home, one of many distributed computing projects that have enlisted the computing power of otherwise idle PC’s around the world. These initiatives effectively create supercomputer networks to rival some of the world’s most powerful machines. Launched in 2005, the rosetta@home software has been downloaded by 150,000 volunteers, who contribute their computers’ power to compute protein shape.
University of Washington’s David Baker explains the evolution of a new approach to this problem, “Fold.it is really completely different as it taps human intelligence, not just spare computing cycles.” Even the most inexperienced human is better at solving complex spatial puzzles than the best computer. On the Fold.it site, non-experts learn the rules of protein folding through a series of starter puzzles, packing proteins, hiding hydrophobic elements and clearing clashes. They are soon ready to apply their innate human puzzle-solving skills to more complex folding, including the site’s Grand Challenges. The aim is to find novel predictions of a protein’s structure from its amino acid sequence. In the process, science learns about proteins, and people learn about science. Generating evidence that human computing can find new solutions is one of Fold.it‘s goals. Its founders hope to encourage the biotech sector to recognize this as a genuinely viable method, and not just PR. Baker is confidently collecting proof.
While Fold.it’s elegantly designed system is new, citizen science has been contributing to scientific knowledge for years. The National Audubon Society’s Christmas Bird Count has been enlisting citizen bird watchers, and professional astronomy benefits from crowd-contributed data. Citizen science has expanded greatly with the advent of the Internet, social networking and the global marketplace. The Fold.it game takes advantage of these capacities, and uses ranking schemes, chat-rooms and wikis to build an engaged community—generally with more success than other ventures that employ distributed intelligence for scientific problems.
In the commercial world, Amazon’s so-called Mechanical Turk is a crowd sourcing Internet marketplace that uses human minds to “solve problems and perform tasks that computers cannot yet be used for.” Fold.it is the Biological Turk, taking total advantage of human intelligence and interfacing it with silicon. With Fold.it, the volunteer non-expert, equipped with well-designed tools, can offer more than the lowly paid employee of the Mechanical Turk, performing mundane tasks.
Equipped with the right tools, amateur scientists could become essential rather than just useful. The Fold.it website states, “For protein structure prediction, the eventual goal is to have human folders work on proteins that do not have a known structure.” Baker confirms that, “more has come from users without much science background.” Fold.it takes advantage of human intelligence traits, described by science journal Nature as “a superior spatial awareness; an ability to take short-term risks for long-term gain; and the converse, recognizing a dead-end early and knowing when to quit.”
Baker believes that more areas in science could benefit from using the kind of innovations pioneered with Fold.it. And as projects are initiated and expanded, perhaps this kind of research could move from the web into physical space. Just as the International Genetically Engineered Machine competition—an undergraduate competition held annually at MIT—provides momentum for the field of synthetic biology, could thousands of top-ranked science game players be gathered in one space to make intensive progress solving problems? Perhaps this is the future of citizen science, where communities are united trying to solve the same problem in parallel, rather than tinkering in series. In this evolution of DIY biology, collective intelligence rather than distributed thinking could be the future of the Biological Turk.