Artificial intelligence UNU seems to have conquered not just the Super Bowl and the Oscars, but also the Kentucky Derby: it just made betters $11,000 richer with the correctly predicted derby winner. How does the so-called swarm intelligence work and arrive at a successful forecast?
Swarm intelligence seeks to amplify, not replace, human intelligence, with the idea that large groups predict an event outcome better than just one individual can.
According to UNU inventor and Unanimous AI chief executive Louis Rosenberg, forcing polarized groups into a swarm lets them find that answer that will satisfy most people.
“Our vision is to enable the power of group intelligence for everybody,” he said, adding the concept could be applied to completely AI systems and, therefore, leaving the decision-making to machines in the future.
The UNU system lets users log on to its online question-answer and decision-making forum, aggregating a swarm of opinions or knowledge from users. Queries are posed to a group alongside multiple answers onscreen, where the individuals use magnet-shaped cursors for dragging a digital puck toward their answer of choice.
Take note that every swarm can only produce one response, leading to a push-and-pull within the group until a final, collaborative decision is made.
The UNU Derby panel outperformed authorities – much like what it did in predicting Super Bowls, Academy Awards, an even presidential primary victors – and netted Unanimous a staggering $10,822 profit from its mere $20 bet.
In a post-Derby poll of the 20 UNU users, decisions via voting instead of swarm intelligence would have only resulted in a correct guess of one horse (the eventual winner, Nyquist) and not the final lineup composed of Exaggerator, Gun Runner and Mohaymen. Not one of the users forecasted this lineup on their own.
A swarm, Rosenberg explained, is more accurate than a poll, which will provide the most popular answer, but not the one optimizing a group’s preference. In the UNU platform, pulling toward one’s owns suggestion alone will lead the puck to stalemate. To get it to move, compromise is necessary: an answer may not actually be Person A’s first choice, but could be the answer the group agrees on.
While still imperfect – getting 11 out of 15 winners correct in the 2016 Oscars – the UNU platform remains strongly accurate.
Swarm intelligence is also demonstrated in the animal kingdom, such as in the chemical connections of ants, as well as the gesture detection of birds and bees.
A biological technique for human groups to behave as a whole, however, is not yet developed; swarm technology allows collective thinking and attaining the same kind of “intelligence amplification” other species have already achieved.
Rosenberg highlighted the recent Kentucky Derby prediction as a milestone in AI capability and a potential game changer in people's view and practice of sports gambling.
Swarm intelligence has promising days ahead, with a post on the Unanimous website even hinting at the possibility of an elevated “super-mind” that could be produced by organizing hundreds or thousands of people in a swarm.
Photo: Bill Brine | Flickr