Posts Tagged ‘swarm intelligence’
[LINK] “Watch How Bees Teach Each Other to Solve Problems”
National Geographic‘s Brian Clark Howard describes a new study that demonstrates how bees, that epitome of a swarm intelligence, learn.
Bee see, bee do. At least that’s the conclusion of research published earlier this month, showing that bumblebees learn to solve problems by watching each other.
In the first study of its kind in insects, scientists constructed experiments that challenged bees to pull strings in order to access rewards of nectar. It’s a technique that has long been used to test cognition in various vertebrates, but hadn’t yet been tried with insects.
[. . .]
The first step was proving that bees could learn to solve a simple problem. But what’s more interesting is that other bees that hadn’t encountered the problem before picked up the ability to solve it more quickly when they had a chance to watch a trainer bee that had already figured out the puzzle.
Further, that knowledge was shown to spread from bee to bee throughout a colony, even if the first bee that figured out the trick died.
The scientists hoped their study would shed light on a bigger picture: how social learning spreads through a population. That might even have implications for the evolutionary roots of culture in human beings, they noted.
The study in question is available here.
[LINK] “Can Social Insects Have a Civilization?”
In the past, I’ve blogged about the idea of swarm intelligence, the sort that might emerge from social insects or other highly social if individually simple-minded creatures. Centauri Dreams features a guest post from one Michael Chorost in which he imagines a trajectory for a social insect species to evolve to high civilization, even sentience.
[H]ere’s the idea I want to test on you all. I asked myself, “Would it be possible for social insect colonies on some other planet to evolve to have language and technology – in other words, a civilization?”
Of course, the idea of swarm intelligence, or hive-mind intelligence, has been around forever in science fiction. To give but one example, Frank Schatzing’s The Swarm posits an undersea alien made of single-celled, physically unconnected organisms that collectively have considerable intelligence. But I need to examine the idea with much more rigor than can be done in fiction.
I refined the question by deciding that, as on earth, the individual insects would have brains too small for serious cognition. The unit of analysis would not be individual bugs but colonies of bugs. The intelligence would have to emerge from their interaction.
After much thought, my answer to the question is “No – but…”
Let me explain both the No and the but. It is these explanations on which I want your feedback.
[LINK] On termites as models for construction robots
Wired’s Nadia Blake pointed towards the latest development in applied swarm intelligence.
Termites normally inspire thoughts of insect invasions, obnoxiously colored house-tents, or the Orkin man. But for a group of scientists at Harvard University, the cooperative insects inspired a whole new army of robots, described in Science.
These robots, each about the size of your head, follow a minimal set of rules when building a structure. Instead of detailed plans, they rely on environmental cues to accomplish their task — a minimalist strategy that’s based on a principle called “stigmergy.” Conceived in 1959 by a French biologist, the term describes how termite colonies cooperate and self-organize to build massive, intricate mounds. In some places, these mounds can be 8 feet tall; they’re also air-conditioned by a network of internal tunnels, and are often oriented along the Earth’s north-south magnetic axis.
But unlike humans, termites don’t follow blueprints or plans while constructing their homes. All they know is what the finished product is supposed to be, and what to pay attention to along the way. As each termite scoops up mudballs and deposits them in various places, it leaves a trail of chemical cues for other builders. Based on these cues, the other termites modify their building behaviors and deposit their own mudballs where the stuff is needed. Boom. Termite mound.
The fleet of Harvard robots follows this strategy, too. Their minimal programming includes the ability to move forward, backward, and rotate; they can also climb, sense, pick up and deposit bricks. Where they lay those bricks depends on what the other bots are doing and what the final structure is supposed to be.
Meeri Kim’s Washington Post article goes into much more detail.
“Around here, you hear about termites destroying buildings,” said Justin Werfel, a Harvard University computer scientist and author of a study in the current issue of the journal Science. “But in Africa and Australia, they are known for building enormous, complicated mounds of soil.”
[. . .]
Each termite is an organism of fairly low complexity, but, using stigmergy, a colony can build a highly complex structure. So the team started with this simple framework: Each robot must have its own basic brain and sensors, and be programmed with certain “traffic rules” it must obey.
The sensors enable them to see bricks and robots next to them, and the traffic rules depend on the final structure. They prevent robots from placing bricks in places where they might easily collapse, or constructing a scenario in which a brick would have to be squished in between two others.
Each robot, about eight inches long, consists of internal metal gears and hardware as well as 3-D printed parts. The bricks themselves are also made in a way that helps the robots climb and align them better.
“In our system, each robot doesn’t know what others are doing or how many others there are — and it doesn’t matter,” Werfel said.
The main difference from the real-life insect is that termites don’t have a desired end product. Rather, there is a random component involved; given the same starting place, a colony will build a slightly different structure every time. But for constructing a house, for instance, the robots would need to follow a specific blueprint. So Werfel created the option for a user to input a picture of a predefined structure, and the robots will go to town on building it.
The paper in question is here.