What is Swarm Intelligence? Can humans swarm? Do we make smarter decisions and more correct predictions as a collective? And how these questions led to the development of a gamified platform which is at once a fun chat room and a potentially powerful think-tank – find out more about UNU, a project brought to us by researchers at Unanimous AI.
What is Swarm Intelligence?
Swarm intelligence (SI) is a subfield of artificial intelligence. The concept originated in the field of robotics and refers to the amplified intelligence of flocks of birds, colonies of ants or swarms of bees, vs the performance of their individual members. SI is concerned with the design of multiagent systems modelled after the collective behavior of these self-organized animal populations.
The brainpower, self-organization and problem-solving abilities of such collectives in nature have always exercised fascination on researchers and scientists. More recently, they’ve inspired advances and applications in robotics and optimization. Now artificial swarm intelligence is brought within the reach of human groups as well.
Of Bees and Men
Every spring honeybee swarms engage in a complex collective decision-making process to find a good nesting site for a new colony. This task is actually delegated to only a few hundred members of a colony of up to 10,000. The scout bees find up to a dozen alternative locations in trees and use body vibrations – the so-called “waggle dance” – to influence the collective decision of the group. This decision appears to be reached by quorum – when a sufficient number of scout bees ‘vote’ for a particular site (by spending time at it). And whether or not the choice is made by consensus or quorum, the fact remains that the bees usually settle at the site that best meets their needs.
Human groups and societies have developed a wide range of voting tools and procedures of varying sophistication to tackle the problem of aligning individual and group interests, preferences, and choices.
Modelled after honeybee swarms and neurological brains, the UNU platform means to put another such tool at our disposal – one that taps into individual insights and knowledge to make “a brain of brains” out of any group of networked users, “enabling them to make complex and nuanced decisions as a unified intelligence”.
Swarms vs. Polls
Says Dr. Louis Rosenberg, founder of Unanimous AI,
“The birds and the bees don’t vote, or take surveys, or use sequential prediction markets. They form real-time dynamic systems that explore a decision space together, in synchrony, and converge on optimal solutions in unison. Now, with UNU, any human group can login and do that too, and the results are pretty amazing.”
While humans do not possess this innate ability, Unanimous AI, the company behind UNU, has developed software that simulates swarm behavior and decision-making within a real-time system. It allows groups of individuals in an online network to negotiate optimal decisions and arrive at collective predictions of sometimes astounding accuracy. And it doesn’t hurt that they can have some fun in the process 🙂
When asked a question in a UNU swarm, each user holds a magnet and is pulling a puck on the screen towards their preferred choice of answer. Swarms are given only 60 seconds to answer each question, and time pressure is another factor driving collaborative decision-making.
“This results in a real-time physical negotiation among the members of the swarm. With everyone pushing and pulling at the same time, the group collectively explores the decision-space and converges upon the most agreeable answer.” (Artificial Swarm Intelligence vs Human Experts)
The key terms to understanding the difference between essentially divisive, polarizing polls/votes and human swarms are ‘real-time dynamics’ and ‘compromise’. UNU swarming is a process and rather than a statistical average, the outcome is a mutually acceptable solution that maximizes group satisfaction.
“Taking a vote or poll is a great simple way to take a decision but it doesn’t help a group find consensus,” Rosenberg says. “It actually polarizes people and highlights the differences between them. People end up getting entrenched in their views.” (Swarm Intelligence: AI Algorithm Predicts the Future – Newsweek)
“While such methods [votes, polls, and surveys] are valuable for characterizing populations, they don’t employ the real-time feedback loops used by artificial swarms to enable a unique intelligent system to emerge. It’s the difference between measuring what the average member of a group thinks versus allowing that group to think together and draw conclusions based upon their combined knowledge and intuition.” (How swarm intelligence could save us from the dangers of AI)
How is UNU a gamified experience?
Asking and answering questions about any topic of human interest, solving problems, making choices in a fast-paced environment, all while interacting with others – these make for an intrinsically engaging and intellectually challenging experience.
To attract and retain the large swarms of users that the platform needs to function at its best, UNU features a user-friendly, intuitive interface, personalization options, social sharing functionalities (Twitter and reddit) and a vast range of discussion groups and themes. It incorporates a healthy amount of competition, as well as its own PBL system.
The UNU points and badges are called credits and kudos. To ask questions in a group, you have to pay with credits, and you earn them by answering questions posted by other members.
How many credits you are awarded for each answer depends on the brainpower (BP) indicator that measures group activity in each session. Groups that converge quickly achieve high BP scores while groups that converge slowly, achieve a lower BP score. In the event of a Brain Freeze – when a group fails to converge upon an answer at all – the BP score is 0%. This encourages all participants in the group to seek common ground. You may also lose credits is if you ask a question that the swarm deems is a “Bad Question”.
Mini-games add to the fun during group discussions. Whenever a question calls for participants to come up with their own suggested answers, they compete for swarm approval. The chosen option earns its author several hundred credits and 1 point in the contest. The first to win 3 points wins more credits and the scoreboard is reset.
As for Kudos, they are given to users who achieve particularly high levels of performance when contributing to answers. In addition, Kudos are given to users who give “suggestions” that are selected by the swarm as the top answer. Unlike credits, Kudos are never spent and thus indicate total user performance over time. They will serve as the basis for forthcoming features such as rank and high-score boards.
As in any other anonymous online community, a certain amount of profanity and aggressive behavior is only to be expected in the chat room but if you are lucky to join a good group and stick around for a while you’ll find the swarm dynamics can be really interesting. The intellectual competition and the challenge to quickly come up with winning suggestions are exciting and fun and by the time you win your first three-point contest, you might be totally addicted.
Implications and applications
The best thing is, the fun aspect is just a side effect and the technology has fantastic potential in many areas where “many minds are better than one”. Not to mention that it is ‘safer’ than AI insofar as it is human-driven and better aligned with our moral values and our goals.
In 2015, a UNU swarm famously predicted correctly 11 out of 15 Oscar awards. More recently, UNU made headlines with the Kentucky Derby, when a swarm of 20 participants predicted the top four horses in the correct order, at odds of 540 to 1!
“Personally, I was speechless,” said Louis Rosenberg. “We’ve been blown away by how smart UNU has been in prior predictions, but when the horses crossed the line I almost didn’t believe it, especially since we put ourselves out there by publishing the picks. And here’s the amazing thing — while the Swarm A.I. got the picks perfect, not a single individual who participated in the swarm got the picks right on their own — not one.” (Swarm A.I. Correctly Predicts the Kentucky Derby, Accurately Picking all Four Horses of the Superfecta at 540 to 1 Odds)
The latter point is quite important as it demonstrates how intelligence is amplified in the swarm. All participants were asked to fill out a survey with their picks. The swarm was 100% correct in picking horses to finish in the correct order, while the individuals who comprised the swarm were only 23% correct. This means the AIQ (Amplification Intelligence Quotient) of the swarm showed a 435% amplification of intelligence. If the group had taken a vote instead and bet on the horses that were the most popular picks made by the participants on their own, the group would have collectively gotten only one correct prediction — the favorite to win.
While it may be some time before UNU actually revolutionizes human democracy and government, one possible application envisioned by Dr. Rosenberg is in the field of medical diagnosis. UNU can also be great for research purposes, brainstorming and planning sessions, online community contests and awards, and of course, predictions.
UNU is still in Beta, eager for feedback and new members. So next time your office is planning a new project (or party) maybe you can invite your coworkers to a swarming session on http://go.unu.ai/ for a change?..
Feature image source: http://whenonearth.net/bee-attracting-competition/