Online public forums and social media platforms are havens for trolls and bullies--some of them agents of repressive governments--who launch campaigns of mass harassment that in many cases lead to the self-censorship of the people they target.
With the looming launch of its Conversation AI toolkit, Google's Jigsaw technology incubator hopes to "level the playing field," says Jigsaw president Jared Cohen. He notes the Conversation AI software uses machine-learning artificial intelligence to automatically detect abusive and harassing language with unmatched accuracy.
The software was trained on a vast corpus of comments supplied by The New York Times and Wikipedia, and it can assess the tone of a string of text with more than 92% certainty and a 10% false-positive rate, according to Google.
Jigsaw also will open source Conversation AI for use by any Web forum or social media platform.
Cohen says Jigsaw's projects are linked by a common focus on what he calls "vulnerable populations" that researchers are advised to become intimately familiar with so they are personally invested in the projects' success. By striving to tackle such problems as organized crime, terrorism, and human trafficking, Jigsaw signals Google's intent to become a positive force for empowerment.
View Full Article
Abstracts Copyright © 2016 Information Inc., Bethesda, Maryland, USA