Home > Cool Tech > Will swarm intelligence predict America’s…

Will swarm intelligence predict America’s next president?

Swarm intelligence has generated quite a buzz this year by predicting the outcome of events more accurately than experts have. In February, a swarm of 7 laymen predicted 11 out of 15 Oscar winners, outperforming the New York Times’ film buffs by 18 percent. In May, a group of 20 Kentucky Derby fans correctly placed the race’s top four finishers, besting even Churchill Downs’ equestrian experts.

Today, a swarm session will be set up on Reddit to answer questions about an even more important contest: the 2016 presidential election. To participate in the swarm, head over to the IAmA subreddit at 1:00 p.m. ET.

RelatedDespite 540-to-1 odds, a swarm AI system correctly predicted the Kentucky Derby ‘superfecta’

These sessions are made possible by UNU — a swarm intelligence platform developed by California startup Unanimous A.I. — that allows groups of participants to consider questions in real-time and combine their knowledge for more accurate answers.

To understand swarm intelligence, look towards nature. Schools of mackerel and flocks of starlings swarm to overwhelm predators. Ants, bees, and termites swarm to find food and build structures. When you think of swarm intelligence, think of a “’super-organism’ that can outperform individual members,” Unanimous A.I.’s CEO, Louis Rosenberg, told Digital Trends.

“Swarm intelligence is a result of millions of years of evolution, which found that groups can amplify intellect by forming a closed-loop system,” Rosenberg said, “a brain of brains.”

While birds, bees, and other animals evolved the ability to swarm, humans did not. Instead, evolution equipped us with a network of cells that function as a unified system – i.e. our brain. But the capabilities of the internet now allow UNU participants to pool together individual intellect and intuition for collective insight.

If this seems like simple crowdsourcing, Rosenberg insists it isn’t. “A swarm is a dynamic closed-loop system that converges on solutions in real-time,” he said. In other words, UNU users collaborate to come to a conclusion; they don’t answer independently as one would in a poll or a survey. In fact, when the 20 Kentucky Derby fans were individually asked to predict the race’s winners, none of them correctly predicted more than one horse. “That is why swarms are so amazing,” he said. “They produce a whole that is far greater than the parts.”

“The swarm does not simply pick the most popular answers,” Rosenberg added. “It goes much deeper than that.” The real-time swarm session sees participants pull a shared “puck” towards their preferred decision, kind of like a 360-degree tug-of-war. Since users can — and often do — change opinion mid-session, the speed at which the puck moves fluctuates as users move their cursor to another answer. Meanwhile, a timer counts down to determine “brainpower” — or consensus — as the swarm deliberates. The quicker the swarm decides on an answer, the higher the percentage of brainpower and, in theory, the more accurate the decision.

Powerful as it may be, Rosenberg said swarm intelligence “is not psychic,” so events with pure or highly random outcomes are impossible to accurately predict. Considering the oft-unforeseen developments of this year’s presidential primaries, one may wonder what insights swarm intelligence can offer to the general election.