Science & Technology

By Using What He Learned About Swarm Intelligence on the Curiosity Podcast, a Listener Won $1,000

What if you had a piece of technology that made such accurate predictions that, with almost no background knowledge, you could win more than a thousand dollars on a $1 bet? That's exactly what Unanimous AI did for one lucky listener of the Curiosity podcast. Here's how.

Wise Guys Finish First

Imagine a global hive mind that could tap into the knowledge, wisdom, and insights of millions of people to produce a superintelligence way smarter than any individual. Now imagine that this hive mind can reach decisions that are more socially moral than if you took a vote, poll, or survey. A far cry from the dystopian hive minds of science fiction, this "swarm intelligence" might be the secret to superhuman intelligence — and it already exists.

You're probably familiar with the wisdom of the crowd, which refers to the way the average judgment of a random group of people can roughly match that of a single expert. The problem with this "wisdom" is that results vary wildly — sometimes it results in great wisdom, sure, but it can also lead to poor policy decisions and other mistakes. The wisdom of the crowd is usually tapped via polls — and that's a problem.

Polls "drive people to extreme positions and they force people to entrench. They actually reinforce our differences," Dr. Louis Rosenberg, CEO and founder of Unanimous AI, told us on the Curiosity Podcast. "When you take a poll, you're not trying to find where groups agree. You're trying to find where groups disagree, and then you're highlighting those disagreements and causing those disagreements to entrench." Enter: swarm intelligence.

"Swarms really do the opposite. Swarms highlight where groups agree," Rosenberg explained, and "encourage a group to find their commonality and converge on the solutions that really optimize the groups' collective satisfaction."

Birds of a Feather Swarm Together

Rosenberg got the inspiration for swarm intelligence from nature, where birds form flocks, fish form schools, and bees form swarms. Honeybees have fewer than 1 million neurons in their brains, but they have evolved the ability to put those brains together to form a swarm intelligence that solves problems with more ease than that a single human brain (which has about 85 billion neurons). For example, research shows that when scouting for a hive location, honeybees converge on the optimal decision 80 percent of the time.

"A swarm of bees does not have gridlock like human political systems ... if a swarm of bees entrenched the way we humans do when they're trying to pick a new home, that swarm of bees would have died out millions of years ago," Rosenberg told us. To harness the power of swarm intelligence in humans, his startup, Unanimous AI, has built a tool that crowdsources answers online. It lets hundreds of participants respond to a question all at once, pooling their collective insight, biases, and varying expertise into a single answer. But instead of weighing each answer equally the way a poll would, the system uses artificial intelligence to weigh the opinion of every person in the swarm appropriately.

"The AI algorithms that are watching their behaviors are actually figuring out their various levels of confidence even if they themselves don't know how confident they are," Rosenberg explained. "If you give somebody a survey and you say, 'How confident are you in this answer?' and somebody puts down 7 out of 10, well, how do you know that my 7 out of 10 is the same as your 7 out of 10 in confidence? How do you know that I even really know how confident I am? But in a swarm, it's this system where people are all interacting together." In this way, the swarm is able to converge on the most accurate representation of the group's collective wisdom.

The Big Payoff

So how does this perform in the real world? Well, in a recent study by researchers at Unanimous AI and Oxford University, a Swarm AI system was used to predict the winners of fifty English Premier League soccer games over five consecutive weeks. Results showed that the humans in the swarm, who averaged 55 percent accuracy (or roughly as well as a coin flip) when working alone, were able to boost their accuracy to 72 percent in the swarm. Other research studies have shown similarly impressive results.

Unanimous AI regularly invites people to converge in swarms online to predict the outcomes of sports matches, awards ceremonies, and other events, with tremendous results. Their horse-racing swarm famously predicted the 2016 Kentucky Derby Superfecta, where the 540-to-1 odds netted nearly $11,000 from each $20 bet — an extraordinary feat that they outdid in 2018.

Today, though, that 540-to-1 payout has been eclipsed by perfectly predicting a 10-game weekend for the English Premier League. Unanimous AI published the swarm's picks on its blog, social channels, and in its newsletter, and members of the community reported placing bets that paid 1,244 to 1 and 1,089 to 1, respectively. That means a $20 bet would have produced $25,000 win.

One Curiosity Podcast listener emailed Unanimous AI to say, "I know nothing about sports, but i [sic] like AI and technology, and this is a neat idea you guys are working with." This faith in technology turned his $1 bet into a $1,089 payday. Of course, Unanimous AI's predictions aren't 100 percent accurate, so please gamble responsibly. But $1,089 isn't a bad return for listening to a 38-minute podcast. It pays to listen!

To learn more about swarm intelligence, listen to our full conversation with Dr. Louis Rosenberg, stream or download the podcast using the player below, or find the episode everywhere podcasts are found, including iTunes, Stitcher, and Gretta.

Written by Cody Gough March 29, 2018