How swarm intelligence helps AI make better decisions

Bees swarm on the leaves of a tree after Palestinian workers opened their beehives to collect honeybee combs during the harvest at an apiary near Beit Hanun in the northern Gaza Strip on April 30, 2017. Jihad Abu Shamalah, the owner of the apiary from Gaza cultivates 450 hives, producing some 4000 kilos of honey every year which is only sold in the Gaza Strip. / AFP PHOTO / MOHAMMED ABED (Photo credit should read MOHAMMED ABED/AFP/Getty Images)

Bees swarm on the leaves of a tree. (Photo credit should read MOHAMMED ABED/AFP/Getty Images)

The ancient Roman poet Virgil compared the work of bees to that of humans, even drawing parallels between bee society and human society with anthropomorphic flair. He looked at the beehive as a city, as a house and even delved into its political organization.

Centuries later, the behavior of bees and other swarming insects continues to fascinate. It has served as a plot device in sci-fi books and movies, as mother nature gone rogue. Michael Crichton’s novel Prey is about out-of-control swarm intelligence in man-made nanobots built from bacteria. In the book, the life forms possess rudimentary intelligence, learn quickly and innovate. They develop predatory behavior, attack and kill.

But these potent natural behaviors also inspire entrepreneurs to mimic them with computers and robots. Swarm intelligence was defined in the late 1980s as a branch of artificial intelligence that attempts to get computers and robots to mimic the highly efficient behavior of colony insects such as ants and bees. For instance: studying ants’ foraging behavior led to a method for rerouting network traffic in overtaxed telecommunications systems, and applying the division of labor among honeybees has helped streamline assembly lines in factories.

Super-intelligent group think

“We connect people together in a real-time system that’s modeled after nature,” said Louis Rosenberg, CEO of Unanimous AI, a software maker using swarm intelligence to help humans make better decisions. By doing so, one gets “a much deeper expression beyond intuition, and beyond what you can’t express verbally when individuals are behaving in a swarm…they’re actually expressing themselves at a more intuitive level,” Rosenberg said. “And it works. It’s the reason why bees swarm and birds flock. Nature discovered through evolution a way to maximize the group so they could think together and converge on an optimal system.”

Flock of migrating Starlings Photo by DeAgostini/Getty Images)

Flock of migrating Starlings Photo by DeAgostini/Getty Images)

Rosenberg’s company develops algorithms that allow humans to collectively make predictions. They do so through the company’s distributed architecture that enables groups of people, scattered around the world to log on to its “swarm platform” and participate in what it calls real-time closed-loop intelligence, which is moderated by AI algorithms. The participants are asked questions related to the topic or theme being explored,  their answers are collected through Unanimous’s “swarm interface,” and processed and analyzed by its AI software. So far, the algorithms and their human directors have made some astoundingly accurate predictions, such as the first four winning horses of the Kentucky Derby, the Oscar winners, Stanley Cup champions and others. The thinking goes that the more people who are involved in the swarm, the greater their predictive power will be.

What comes after robots gain swarm intelligence? Rosenberg told me that one could look to a group of small flying robots, which individually have limits on their sensing and computational power, but as a group, they could share information like a collective brain and even access additional brainpower from the cloud. And by doing so, the robot’s size and weight would not proportionately limit its access to intelligence. Banding together the collective power of many brains working together is at the heart of swarm intelligence and something we humans have in abundance given the size of our population.

“People tend to forget that there are seven billion people in the world and to be able to tap into that body of information and inference would have extraordinary benefits with super-intelligent systems,” Rosenberg said. “We can amplify and let them perform better, all the while keeping the human sensibility as part of the process.”

A recent study by Oxford University shows the potential. The project connected a small number of financial analysts and asked them to predict oil, gold and the S&P 500 for 26 weeks.  The group congregated using swarm intelligence and their predictions were 26 percent more accurate than traditional AI.

We, robots

Meanwhile, at Harvard University, professor Radhika Nagpal’s work at the Wyss Institute for Biologically Inspired Engineering focuses on studying collective behavior in biological systems and how such behaviors can be applied to computing and robotics. Nagpal is developing programming models inspired by bee swarms and termite colonies that enable new types of robotic systems to mimic the collective behaviors of living creatures. Nagpal’s kilobots were designed so that if one malfunctions, its neighboring kilobot is able to leverage another set of algorithms that corrects errors and variations and allows for it to continue working on the task at hand. With this capability in place, kilobots can be programmed to self-assemble into specific shapes without any intervention.  She is building new types of distributed systems, from multi-modular robots and robot swarms to vast sensor networks to meet real-world challenges.

Kilbots used by Harvard's Self-Organizing Systems Research Group. (Image credit: Harvard University)

Kilobots used by Harvard’s Self-Organizing Systems Research Group inspired by ants. (Image credit: Harvard University)

“Ants, flocking birds or schools of fish create amazing global structures that are elegant, mesmerizing and scientifically inspiring,” said Magnus Egerstedt, Steve W. Chaddick School chair and professor at the School of Electrical and Computer Engineering at the Georgia Institute of Technology. “There’s something massively exciting about robots—kids and people are in awe and inspired by these machines and that’s a beautiful thing.”

Robotic researchers like Egerstedt take ant colony principles and apply them to a swarm of robots in which each member has a simple set of rules to follow. This leads to self-organization and self-sufficiency. Egerstedt said this work leverages the central tenet of swarms—whereby individual robots are not the central brain, as they have limited range sensors, but collectively contribute a partial picture of the world. Together, they can solve problems as a collective group, such as searching for survivors in a natural disaster or working together to cover a wider area on a farm field or manufacturing floor.

A photo of a Termite Mound at Entebbe Botanical Gardens, Entebbe, Wakiso, Uganda, East Africa, Africa.(Photo credit: Getty Images)

A Termite Mound at Entebbe Botanical Gardens, Entebbe, Wakiso, Uganda, East Africa, Africa.(Photo credit: Getty Images)

Man and Machine

“We spend too much time thinking about people and computers and not thinking about how the two of them work together,” said Thomas Malone, Patrick J. McGovern Professor of Management at the MIT Sloan School of Management. In much the same way, one views a neural network as collective intelligence comprised of individuals, which when they interact make the overall behavior of the network more intelligent. “It’s an important question for designers to think about when creating products and applications for the future.”

Malone in his book, Superminds, explores this collective intelligence. Founding director of the MIT Center for Collective Intelligence, he’s investigating ways that humans — regardless of geographic location, can work collectively and gain the “wisdom of the crowd.” He aims to find ways that people can solve problems more intelligently as a group than as individuals.

Malone argues that collective intelligence and the resulting “superminds” may lead to new ways of looking at the world. Instead of thinking about AI in one sphere and people in the other, we ought to consider them in the same circle, where they can leverage one another’s strengths, examine and solve complex problems. Immediate applications might include strategic planning processes for large businesses involving far more of its employees than they might typically involve in face-to-face information gathering; or creating climate change strategies that involve a broad base of business, government, and community stakeholders.

Walt Whitman wrote about bees “humming their perpetual rich mellow boom.” Swarming principles and collective intelligence are now seeing a boom of their own. They’re playing an increasingly important role in solving complex challenges.