Swarm A.I. Predicted Trump as Time's Person of the Year
The collective intelligence of people governed by A.I. algorithms demonstrates that when it comes to the wisdom of the crowd, swarms win where polls fail.
You know who might be most surprised that Time magazine picked U.S. President-elect Donald Trump as their Person of the Year? Time magazine's readers. They participated in an online poll last month based on a list of names compiled by IBM's supercomputer Watson. Trump was on the list, but the readers voted for India's president Narendra Modi.
Swarm intelligence nailed it, though. Today, the San Francisco-based tech company Unanimous A.I. confirmed what it had predicted yesterday: that Donald Trump would come out on top. It arrived at that conclusion after a mere 75 people randomly selected to work together in real-time used the company's online platform, called UNU, to narrow down the choices.
The collective intelligence governed by A.I. algorithms demonstrates that when it comes to the wisdom of the crowd, swarms win where polls fail.
"A poll just tells you the divisions in a group. That's why polls are polarizing," Unanimous A.I.'s CEO Louis Rosenberg told Seeker. "Whereas a swarm is unifying. It will actually have the group find the answer it can agree upon."
Anyone can become part of the intelligent swarm, which opened to the general public this past June. Signing up is free and once in, users can anonymously suggest questions or enter a "room" to help generate answers. Because questions gets answered in real-time, people can sign up for alerts to topics that interest them the most and then join the swarm at the right time. Who will win the Superbowl? Should you buy a certain stock? Who will win the Nobel Peace Prize? What will the new president tweet about next?
Using the platform is easy. Once inside a room to work in a swarm, a screen appears with the question at the top. For example, during the election, one group was asked, "What is the most important issue in the presidential campaign?" It was encircled with several possible answers, including healthcare, immigration, taxes, jobs, gun control and military.
At the center of every screen with a question is a clear "puck." Each member of the swarm has a golden magnet icon that she uses as a cursor to pull the puck toward her answer. She can see everyone else's magnet and react in real-time if the crowd's decision isn't going her way.
For instance, if the most important issue in the presidential campaign was the military, but few people appeared to be dragging the puck that way, she could switch to her second most important issue, healthcare. And if the crowd doesn't seem to be leaning that way, she can chose her third most important issue, jobs.
In this way, each member of the swarm is inspired to find common ground. Eventually, a larger number of participants begin to coalesce around one answer, which becomes the winner.
The algorithm is inspired by animals that swarm in nature, like fish that school, birds that flock and bees that swarm. Alone, any one of these animals may not be smart enough to evade a predator or fine a new location for a hive, but together they maximize their collective intelligence.
"In nature if a flock of birds can't agree on a new nesting ground, that's a life or death decision," said Rosenberg.
Although the Time magazine question wasn't a life or death decision, UNU showed that its approach was beneficial because it's efficient. Instead of gathering opinions from thousands of people and taking days to answer, UNU asked just 75 and the answer was generated in a few minutes.
They approached the question by showing the swarm members the names of Person of the Year candidates and then successively asking, "Who is the least likely to be named Time's Person of the Year?" In this way, UNU was able to narrow down the choices until there were just two: Hillary Clinton and Donald Trump. At that stage, UNU asked, "Who is most likely to become Time's Person of the Year?"
The gif shows the swarm tugging it out until Trump wins.
This is not the first time that UNU's swarm A.I. platform has correctly predicted the outcome of an event. Back in May, a swarm correctly predicted which four horses would cross the finish line - and in which order - at the Kentucky Derby. The so-called superfecta had 540-to-1 odds. UNU also predicted, mid-season, which eight baseball teams would make it to the playoffs, which two teams would end up in the World Series and who would win.
Perhaps the true test of UNU's algorithm will be whether it can predict what the new president is going to do next.
WATCH VIDEO: What Would Happen If all of the Bees Died?