Tech Explained: Swarm Intelligence

How the decision-making processes of ants and honey bees can be applied to human technology

In nature, individual organisms often work together to produce complex decisions or structures. For example, ants are tiny, yet they organise themselves into colonies complex enough to build ‘cities’, practice a form of agriculture, and create ‘rafts’ to escape floods. Bees are able to build sophisticated hives, and flocks of move together to migrate. All of these creatures are exhibiting swarm intelligence, where simple creatures follow simple rules yet display a great deal of complexity and efficiency. This trait is common in nature, but recently researchers have been using it to transform fields such as robotics, artificial intelligence and even medicine. How does swarm intelligence work and how can it be applied to new fields?

In nature, individual organisms often work together to produce complex decisions or structures. For example, ants are tiny, yet they organise themselves into colonies able to build ‘cities’, practice a form of agriculture and create raft-like structures for escaping floods. Bees are able to build sophisticated hives, and flocks move together to migrate.

These creatures are exhibiting swarm intelligence. This trait is common in nature, but recently researchers have been using it to transform fields such as robotics, artificial intelligence and even medicine. How does swarm intelligence work and how can it be applied to new fields?

How swarm intelligence works: In the natural world, swarm intelligence occurs through a process called stigmergy. The principle of stigmergy is that each individual leaves a trace in the environment, such as a chemical signal, and this trace causes other individuals to behave in a certain way or perform a certain action, leading to a new pattern of behaviour. A classic example of this is how ants direct others in the colony to a new food source. 

When an ant finds a food source, it uses chemicals called pheromones to mark the path. Other ants are attracted to the pheromone signals and mark their own paths with more pheromones as they go. Over time, the most efficient path to the food is travelled by the most ants and becomes strongly marked – becoming an ant highway. To an observer, it appears as though the ants have made a decision on which path is the best, but the new path emerged out of microscopic changes made by thousands of individual ants.

How swarm intelligence is used: This principle can be applied to address a wide variety of problems. Some applications require the development of artificial intelligence algorithms that leave traces or signals, in a similar way to ants.

For example, researchers at Hewlett-Packard developed an algorithm to route calls more efficiently. In the program, software ‘ants’ roamed through the telecom network and left tiny packets of information to signal uncongested areas. Phone calls were then routed along the trails left by the software ants. 

Swarm intelligence software has also been used to speed up logistics. When Southwest Airlines was facing cargo bottlenecks due to lack of storage capacity at airports, it turned to swarm intelligence. The airline realised that employees were wasting time trying to load freight onto the first plane going in the right direction.

Researchers applied lessons learned from ants and determined that sometimes it was more efficient to leave cargo on a plane headed in the wrong direction, then move it to another plane. Other companies have used similar approaches to schedule deliveries or work processes.

Another use for swarm intelligence is in robotics. Researchers at the Georgia Robotics and InTelligent Systems laboratory created a small swarm of simple robots that can carry out basic tasks, including playing the piano. Instead of communicating directly, the robots use the position of surrounding robots to determine the best path for completing their task.

In 2018, researchers at the University of Cambridge and Koc University in Istanbul developed the concept of an ‘energy neutral internet of drones’. They designed a drone swarm whose drones could share energy with other drones that were running low on power. NASA also has plans to develop swarms of drones to use in space exploration, and the US Department of Defence has tested drone swarms that can carry out complex military missions.

Swarm intelligence can also aid human decision-making. One company, Unanimous AI, has developed algorithms that treat humans as a swarm. Humans input their opinions in real time and are able to influence the opinions of other participants.

The system has made amazingly accurate predictions, include picking more Academy Award winners than professional movie critics three years in a row and predicting that the Chicago Cubs would end their 108-year dry spell by winning the 2016 World Series—four months before they Cubs were even in the playoffs. 

The Unanimous system, which is modelled on the collective decision-making process of honeybees, has also been applied to diagnosing medical conditions. Radiologists analysed chest X-rays to predict how likely it was that a patient had pneumonia. Each doctor also moved an icon that allowed them to push the group consensus toward their opinion.

As the doctors weighed in, the AI algorithm inferred how strongly each felt about their choice, based on the movement of their icon over time. The algorithms then combined the preferences to reach a single choice. In the end, the swarm system was 33 percent more accurate than individual practitioners and 22 percent more accurate than a Stanford machine-learning program called CheXNet.

Swarm intelligence is already essential to many species’ success in nature, it may soon be an essential component of human success as well.

In the natural world, swarm intelligence occurs through a process called stigmergy. The principle of stigmergy is that each individual leaves a trace in the environment, such as a chemical signal, and this trace causes other individuals to behave in a certain way or perform a certain action, leading to a new pattern of behaviour. A classic example of this is the way that ants direct others in the colony to a new food source. 

When an ant finds a food source, it uses chemicals called pheromones to mark the path. Other ants are attracted to the pheromone signals, and mark their own paths with more pheromones as they go. Over time, the most efficient path to the food is traveled by the most ants, and so becomes very strongly marked – turning into an ant highway. To an observer, it appears as though the ants have made a decision on which path is the best, but the new path emerged out of microscopic changes made by thousands of individual ants.

It turns out that this principle can be applied to a wide variety of problems. One way to do this is to develop artificial intelligence algorithms that leave traces or signals, in a similar way to ants. For example, researchers at Hewlett-Packard developed an algorithm to route calls more efficiently. In the program, software ‘ants’ roamed through the telecom network and left tiny packets of information to signal uncongested areas. Phone calls were then routed along the trails left by the software ants. 

Swarm intelligence software has also been used to speed up logistics. When Southwest Airlines was facing cargo bottlenecks due to lack of storage capacity at airports, they turned to swarm intelligence. The airline realised that employees were wasting time trying to load freight onto the first plane going in the right direction. Researchers applied lessons learned from ants and determined that sometimes it was more efficient to leave cargo on a plane headed in the wrong direction, then move it to another plane. Other companies have used similar approaches to schedule deliveries or work processes.

Another use for swarm intelligence is in robotics. Researchers at the Georgia Robotics and InTelligent Systems laboratory created a small swarm of simple robots that can carry out basic tasks, including playing the piano. Instead of communicating directly, the robots use the position of surrounding robots to determine the best path to use to complete their task. In 2018, researchers working at the University of Cambridge and Koc University in Istanbul developed the concept of an ‘energy neutral internet of drones’. The designed a drone swarm whose drones could share energy with other drones that were running low on power. 

NASA also has plans to develop swarms of drones to use in space exploration, and the US Department of Defence has tested drone swarms that can carry out complex military missions.

Another use for swarm intelligence is as an aid to human decision-making. One company, Unanimous AI, has developed algorithms that treat humans as a swarm. Humans input their opinions in real time, and are able to influence the opinions of other participants. The system has made amazingly accurate predictions, include picking more Academy Award winners than professional movie critics three years in a row, and predicting that the Chicago Cubs would end their 108-year dry spell by winning the 2016 World Series—four months before they Cubs were even in the playoffs. 

The Unanimous system, which is modelled on the collective decision-making process of honeybees, has also been applied to diagnosing medical conditions. Radiologists analysed chest X-rays to predict how likely it was that a patient had pneumonia. Each doctor also moved an icon that allowed them to push the group consensus toward their opinion. As the doctors weighed in, the AI algorithm inferred how strongly each felt about their choice, based on the movement of their icon over time. The algorithms then combined the preferences to reach a single choice. In the end, the swarm system was 33 percent more accurate than individual practitioners, and 22 percent more accurate than a Stanford machine-learning program called CheXNet.

Swarm intelligence is already essential to many species’ success in nature, soon it may also be an essential component of human success as well.