After correctly guessing 11 out of 15 Oscar winners last year, Silicon Valley startup Unanimous A.I.’s UNU software will now answer questions related to the U.S. presidential election. UNU uses the collective knowledge of a group of people to develop its opinions and often accurate predictions, also seen when the software determined the winning results of the Kentucky Derby. The presidential predictions surfaced while UNU engaged in an AMA (“ask me anything”) on Reddit.
UNU, which combines groups’ “knowledge, opinions, and instincts into a Swarm Intelligence,” works with this existing human intelligence rather than copying it, similar to robotic “AIs.” Those interested in participating may sign up for UNU or even try it out as a guest before converging on an answer in real-time. A swarm forms in different ways depending on the prediction taking place – before the Kentucky Derby prediction, Reddit and Amazon ads asked for volunteers who knew about horse racing. However, a swarm made of amateurs also has the potential to succeed in these kinds of scenarios, as the group’s decision will always outperform individual ones. The size of groups answering questions varies, with predictions for the Oscars and the Derby coming from 20 people or less. However, about 100 to 200 people determined results at June 1’s livestreaming political debate.
As with the previous experiments, a group of decision-makers privately answered questions in unison. When a question appears, everyone has 60 seconds to drag a puck towards an answer with a graphical magnet. Answerers can only see their own choice, but a collective decision that best describes the swarm’s intelligence will ultimately appear.
The software is inspired by studies that explore how honey bees in swarms find solutions. Bees live together in large groups and arrive at decisions such as new nesting site locations through a group method called the “waggle dance.” Each year, when bee scouts split from their hive to find a new home and come across a potential spot, they return to the hive and break into this dance to display what they prefer. Based on the bees’ movements, the swarm negotiates toward a decision through body vibrations. “Biologists have found that bees will pick the optimal site almost 80% of the time,” said Louis Rosenberg, CEO of Unanimous A.I., in an interview with Engadget . “No individual bee can understand the question or find and evaluate the site. But when a group of bees forms swarm intelligence, it functions as a brain of brains.”
By copying the nature of a swarm of bees, UNU’s AI presents the concept that people are smarter when grouped together and even begin to resemble a super-organism. Such collective wisdom provides a crowdsourcing opportunity to gather opinions. However, Rosenberg is quick to say that UNU isn’t like a poll, but rather amplifies a group’s intelligence to create its final view. “The individuals who participate in the swarm don’t have the answers the swarm does,” he said. “But even though no single individual in that swarm came close, the swarm was right.”
In its shining moment on Reddit, UNU answered a variety of questions, ranging from “Does Bernie have a snowball’s chance?” (“No,” according to the swarm) to “Do you see conservative or liberal trends in politics?” (The swarm leaned towards liberal). The UNU responses also clarified which answers had been selected after some struggle, as seen with its final pick of Hillary Clinton becoming the next President. The UNU commentary stated, “This was a difficult decision…with the swarm highly divided.” A link to see a replay of the swarm arriving at the decision also appears.
Of course, the prediction’s validity won’t be determined until November 2016, but the UNU has yet to make a wide impact beyond this predictive format. However, if the idea attracts numerous global users, UNU could become an insightful tool for problem-solving. “What we’ll see is the ability to democratize expertise,” said Rosenberg.
Source: Newsweek, Engadget, Reddit, UNU
Advertisement
Learn more about Electronic Products Magazine
