Last year, something remarkable happened on Oscar Night. Far from Hollywood, a group of casual moviegoers made Oscar-prediction history by out-forecasting the majority of industry experts, including the New York Times and FiveThirty Eight. They did this using a new and different form of A.I. called Swarm Intelligence. It works by combining the knowledge, wisdom, emotions, and intuitions of real people, in real-time.
But this is no poll or survey – it’s a dynamic system built by swarming algorithms and it empowers users to combine their intelligence in a way that far exceed the abilities of individual members. Simply put, Swarm Intelligence allows groups to put their heads together in an efficient way, turning average fans into super-experts. This is what it looked like when the swarm predicted the BEST ACTOR last year:
As you can see, swarms work fast. In fact, the group predicted the winners of the top 15 awards in only 15 minutes. As reported by Newsweek and Discovery, the average person in the study, when working alone, only gotof 15 correct (40%). They did slightly better when the researchers combined the data as a simple poll, taking the most popular answer for each category. This achieved 7 of 15 correct (47%).
But, when the group worked together as a swarm using a software platform called UNU, they predicted 11 of 15 correct (73%). That was far better than even the highest performing individual in the group, who got just 9 categories right. It was also better than the New York Times, which also got 9 of 15 correct. Clearly, many minds are better than one…
Encouraged by early results, researchers at Unanimous A.I. have tested swarms in a variety of fields, from financial markets to the NFL playoffs. And the swarms continue to impress, beating ESPN experts week after week, and outperforming 99% of gamblers when betting on the Super Bowl two weeks ago. The swarm placed 19 bets, predicting everything from yardage to turnovers and earned a +36% ROI.
Now, Unanimous A.I. is challenging the swarm to beat the Oscar experts one more time. As you may know, FiveThirtyEight famously models their Oscar predictions on their founder Nate Silver’s historical 2012 Presidential election predictions. You can read more about their time-intensive process HERE. In contrast, swarming uses the smarts of average people, not rooms full of mathematicians. Each magnet you see in the image below represents a single user, somewhere around the world, pushing and pulling toward their desired answer:
So, who is going to WIN this year? Here are UNU’s picks for the six major categories.
Want to learn more about Swarm Intelligence? Here is an ACADEMIC PAPER that describes the Oscar Predictions made last year, and reviews why the swarm beat the experts. It also explores how “swarming” could be a safer approach to A.I., by keeping humans in the loop.
Pingback: An AI Tries to Predict the Academy Awards Results - eTeknix