This technique has enabled amateurs to better predict the results of the 2015 Oscars as experts.

You’ve probably heard the concerns of Stephen Hawking or Elon Musk about the dangers of artificial intelligence. Some companies seem, for their part, have chosen to put people at the heart of it. This is what tells the site Tech Republic, which devoted a long article to“artificial swarm intelligence”, which would lead to better predictions than those proffered by the experts.

Here it is not a pure artificial intelligence, since “the software is only used to amplify human intelligence”:

“Since researchers began to start looking for ways to create artificial intelligence in the 50s, they focus primarily on the development of artificial neural networks -construct intelligence from scratch. Taking this concept upside, another approach called “artificial swarm intelligence” uses the power of nature in a different way: by exploiting human brains to achieve predictions about events that happen in the world.”

On Venture Beat, Louis Rosenberg, one of the leaders of AI Unanimous, a company that wants to democratize the practice, also explains the difference between “artificial swarm intelligence” and crowdsourcing:

“[Crowdsourcing] typically uses, votes, polls or surveys to gather opinions. If such methods are valuable to describe the people, they do not use feedback loops [the fact to reconsider its response as a function of the overall response, one way or the other, note] employed by swarms artificial to allow a single intelligent system to emerge. “

Tap into the collective knowledge

In the case of the software platform UNU -créée by Unanimous AI- on which focused Tech Times and Newsweek, it works well. A group of people connect simultaneously on the software platform. We ask them a question and each person sees, together with his choice, on what others are positioning solution and what the “average” solution. Together, participants can therefore move the magnet to the response that seems most correct correct.

“The group can only choose one answer, and it has sixty seconds to find one that suits her best. The idea is that rather than divide the group in the same way as would a survey, UNU draws on the collective knowledge and insight of the group to bring out a unified voice. The key is compromised. “

You can see how it works in the video below, where participants were asked questions about the US presidential election.
Tech Republic said that the researchers chose to emulate behaviors that already exist in nature, such as bees. Each year, the site wrote, bees have to find a new place. Several hundred are sent scouts and the group must then decide the best place. And as stressed a professor from Cornell University, 80% of the time, bees make the right choice.

Build on these methods to predict the future looks pretty good walk, emphasizes Newsweek:

“In February 2015, a few days before the Oscars, we asked seven groups of seven to predict the winners. The first group of films -of critics gathered by the New York Times- correctly predicted eight of the fifteen winners. The second, the majority of whose members had not even seen the films in competition, achieved eleven good predictions.

 

Even if they were not movie buffs or statisticians, members of the second group managed to beat industry experts with UNU, a new form of artificial intelligence based on human. “

“” Swarming “keeps a moral”

And as stated Louis Rosenberg on Venture Beat, this method is probably safer than artificial intelligence:

“” Swarming “keeps a moral and human sensibilities as an integral part of its process. Result, this approach to artificial intelligence with humans in the loop combines the benefits of IT infrastructure and efficient software with the unique benefits brought by each person: creativity, empathy, morality and justice. And as this swarm intelligence is rooted in the human contribution, the information that result are even more aligned with our values ​​and our morals, but also our aims and objectives. “

Moreover, underlines Tech Republic, these goals and objectives could be much deeper than the prediction of Oscar winners, as one might optimize the strategy of a company, for example. When asked by Newsweek, Louis Rosenberg explained to him that a group of doctors could use the tool to combine the collective intelligence and achieving even finer diagnostics.

Six months ago, when UNU asked a group of people their predictions for the US presidential election, their preferences were then to Donald Trump and Bernie Sanders.At the time, Louis Rosenberg believed that the algorithm was to be defective.

AUCUN COMMENTAIRE

REPONDRE