POLLS FAIL ONCE AGAIN TO PREDICT TRUMP’S VICTORY, BUT THE SWARM GOT IT RIGHT
Once again, a small “human swarm” has out-predicted a massive crowd, this time by correctly forecasting that President-elect Donald Trump would be TIME Magazine’s Person of the Year. This is yet another demonstration that a small number of people, when connected as a real-time closed-loop swarm (i.e. a Swarm Intelligence), provides faster, deeper, more accurate insights than polling thousands upon thousands of disconnected individuals.
.
.
In this case the competition was between UNU, the world’s first Artificial Swarm Intelligence from Unanimous A.I., and a massive online poll coordinated and intelligently processed by IBM Watson. The swarm, which had only 75 participants connected by A.I. algorithms, forecasted Trump, hands-down, would be the winner. And it made that prediction in only 10 minutes.
.
In contrast, the IBM Watson Poll, driven by its own A.I. algorithms, took weeks to conduct, required hundreds of thousands of participants, and picked Indian Prime Minister Narendra Modi by a wide margin. (UPDATE: A similar poll run by the TODAY show, had Hillary Clinton tabbed as the likely winner.) The Swarm, however, narrowed the field down to Donald Trump and Hillary Clinton as the two most likely candidates as seen above, and put Modi’s chances at winning no higher than sixth.
.
.
This is not the first time that a swarm has blown away the polls. In a 2016 study performed as a collaboration between researchers at Unanimous A.I. and Oxford University, a crowd of 469 football fans were asked to predict the outcome to a set of twenty official “proposition bets” for the 2016 Super Bowl.
.
This large group was tested against the predictive ability of an Artificial Swarm Intelligence comprised of only 29 football fans, connected online in real-time as a “hive mind”. Results revealed that the swarm outperformed 98% of the individual participants, achieving 68% accuracy while the large crowd achieved only 47% accuracy. Even more impressive, bets on the Swarm’s picks would have produced a 35% ROI, while the large crowd, would have lost money.
.
In the case of the 2016 TIME Person of the Year award, the magazine worked in collaboration with IBM’s Watson Opentopic to track each candidate’s impact on the public. According to the magazine, Opentopic evaluated more than 62 million documents from 3.5 million online sources, which Watson’s deep-learning technology classified and sorted.”
.
.
The TIME Person of the Year is awarded annually to the person who “most influenced the news – for good or for ill – and embodied what was important about the year.” Trump’s 2016 victory comes on the heels of his being named the runner-up in 2015 in the midst of a tumultuous primary campaign.
.
Naturally, Swarm Intelligence has demonstrated a wide-range of capabilities beyond predictions. Enter your email below if you’d like to stay in loop on insights from the Swarm.
.
