Combining human and computer intelligence could help solve the world’s most vexing problems, researchers envisioned in an article published Monday in the journal Science.
Researchers from Cornell University and the Human Computation Institute want more humans to help out in accelerating research and finding solutions to life’s most difficult problems, such as cancer, HIV, climate change, and drought.
Crowdsourcing analysis of research materials isn’t new. There are already “games with a purpose” that offload some of the work of analyzing data to humans, who receive the tasks packaged as games.
When the purpose of those games involves forwarding science, it’s “called citizen science,” said Pietro Michelucci, director of the Human Computation Institute.
New tools and infrastructure have made it easier to combine “various methods of crowdsourcing and create more complex and sophisticated systems,” he told TechNewsWorld.
“So this means, instead of creating just one-off human computation systems from scratch each time, we now have the ability to connect different methods of engagement and have real-time access to crowds,” Michelucci said.
The discovery of an HIV-related finding, which had eluded researchers for a decade and a half, took only ten days for citizen scientists to reach. The crowd experts were manipulating 3-D models in the protein folding game Foldit.
“If you could create an expert by combining people in the crowd, then you have access to a lot more crowd experts,” said Michelucci. “So we use the ‘wisdom of the crowd’ method to combine contributions from the general public.”
For a smartphone app called “Malaria Spot,” researchers found that every 23 diagnoses from members of the general public were as accurate as one diagnosis from a certified pathologist. But 23 might not be the right number for Michelucci and the company’s WeCureAlz initiative for researching Alzheimer’s disease, and it may be the wrong number of crowd experts for many other systems.
“We have to figure out how many members of the public it takes to analyze a certain amount of data before that analysis is as accurate as the one from the trained scientist working in the lab,”he said.
“When we can make that work, we have this force multiplier. If we have 30,000 people in the general public and it takes 30 people, then we have 1,000 crowd experts,” Michelucci added.
Crowd experts could fix problems of all sizes someday, according to Rob Enderle, principal analyst at the Enderle Group.
For example, human computation can impact “litigation, both case law and discovery, primary research on trends and causes,” he told TechNewsWorld. “Politics would be timely, for instance.”
Man and Machine
Human computation is far from humans doing the bidding of machines. It is about leveraging humans in areas where machines fall short, according to Michelucci.
“Another way of looking at it is, if machines could do everything humans could do, we wouldn’t need AI researchers anymore,” he said.
For example, in chess, humans still have the edge in being able to focus on only the logical moves, Michelucci pointed out. Computers are good at chess because they consider every possible move, including the ones that clearly make no sense.
“Even if machines can evaluate millions of moves per second, they waste most of their time on moves that won’t be successful,” Michelucci said. “Whereas human abilities — like abstraction, complex pattern recognition, imagination, and creativity — those kinds of things make it possible for people to find solutions quickly that machines can’t because there are too many possibilities to search through.”