NASA 'Collective Intelligence' Can Send Space Messages Faster
NASA scientists have shown that future fleets of spacecraft using 'collective intelligence' can send more data to Earth faster and more efficiently than ever before.
Using insights from economics and physics, NASA researchers have learned how to mix the separate, 'selfish' goals of distinct computer programs, robots and human beings to achieve larger, common goals - a process known as collective intelligence. Scientists use algorithms - recipe-like procedures - especially designed to knit computer programs together so they can achieve global goals.
"We have run computer simulations that verify that our recipes for collective intelligence work," said David Wolpert, a scientist at NASA Ames Research Center in California's Silicon Valley. He said that his group might well be the first team to write algorithms based on collective intelligence that successfully integrate assorted computer programs.
Image left: Collective intelligence software team discusses application at NASA Ames Research Center. Image courtesy: NASA Ames' Dominic Hart.
"The Internet is a huge network of computers relaying messages to one another," Wolpert explained. "We figured out how to change the goals of those computers so messages arrived at their ultimate destinations faster, with improvements of up to five times in certain Internet-based experiments," Wolpert said. The same type of collective intelligence will enable spacecraft to send messages faster to Earth and return more data.
"Just as human beings can self-organize into companies and achieve overall goals, so can these selfish computers, or even robots, self-organize to achieve overall goals," he explained. Scientists envision using collective intelligence to control any large group of robots or people. "If you have a whole bunch of rovers that you want to create a human habitat on Mars, the global goal is constructing that habitat," Wolpert said.
"We don't have to specify details of the environment, or how the individual computer programs go about achieving their goals. The individual robots can run various artificial intelligence programs to achieve their goals," Wolpert stated. The algorithm that knits the computer programs together to achieve a goal does not depend on details of those programs, according to Wolpert.
Just as economic policy makers can use many different recipes, such as lower interest rates, changing regulations and reducing deficits to improve the economy, NASA can use different algorithms to solve bigger problems, according to Wolpert.
Image right: NASA Ames scientist David Wolpert. Image courtesy: NASA Ames' Dominic Hart.
Concerning the future of collective intelligence, Wolpert predicted that "this is a method that someday could be used to program very powerful artificial brains made of nanoscale (miniature) components." Scientists measure nanoscale objects in nanometers, each of which is one billionth of a meter (one billionth of 3.3 feet).
These procedures also can help carry out other tasks such as programming nano-computers, controlling unpiloted aerial vehicles (UAVs) and running the national airspace where airliners fly, Wolpert ventured.
Scientists think that NASA-developed collective intelligence could even work successfully in business to motivate a company's staff so that employees would achieve corporate goals. An example of a business goal that collective intelligence could attain is maximizing a company's stock value.
"In the last year we stumbled upon a relationship between physics and economics that greatly advanced our understanding of collective intelligence," Wolpert said. "In economics, you have many, many individuals with separate goals. In a physics system like a gas you have many, many molecules, which can be viewed as if they all have separate goals. The commonalities of how you control an economy and how you control a gas -- that commonality is the field of collective intelligence," he explained.
The collective intelligence development effort includes NASA Ames, Stanford University, Stanford, Calif.; and Oxford University, Oxford, United Kingdom.
For more information, please send an e-mail to email@example.com
John Bluck, NASA Ames Research Center