Grover Exhibits Sense of Humor in Its Fake News Stories

With all of the AI driven shenanigans that have come to light in the past few years its maybe shouldn’t be surprising that AI can now help you write fake news. And its so good,you probabaly wont be able to tell it was fake or machine written. Like the last sentence. Which is it? Well, it’s really a human writing this story.

Now meet Grover.

A new AI developed by the Allen Institute for Artificial Intelligence and the University of Washington, Grover allows you to enter a headline, and it will generate hundreds of words of convincing, fake text that looks like it belongs in The New York Times or on CNN.

Mark Wilson’s story on fastcompany.com explains how Grover may be able to help in a time when parents need to teach their children to be more discerning about the information they encounter.

Trained on 120 gigabytes of real news articles, gathered from the 5,000 publications on Google News, it can tell a human-written story from a computer-written story with 73% accuracy. The researchers put Grover’s fake news generator online for anyone to try.

Here’s a fake one generated by Grover:

chicagotribune.com
‘The Last Sharknado: It’s About Time’ is every bit as silly as the original
July 5, 2017 – Kenneth Turan
Some heroes take a vacation every now and then, even temporarily, a couple of beachfront condos in a movie world they hope, a la Rick James in “The Beautiful Daze,” won’t have to live through. Unfortunately, that has never happened to one of the most exceptional directors in the business, Anthony & Joe Russo, so the moment is altogether too good to spoil, at least for us Southern Californians. In “The Last Sharknado: It’s About Time,” which had its Los Angeles premiere at the Tower Theatre on Tuesday, we can expect the San Diego-born Anthony Russo to somehow make it through seven years without being inside a sharknado. Oh, he has a couple of other movies during that time, mostly that stupid’d-up New York “Avengers” picture, but they have nothing to do with sharks or serious action-film plotting. Don’t worry, he lives to fight another day.

This description on the official Grover page run by the Allen Institute For AI, or AI2, explains Grover’s genius.

Our model, Grover, is a generator that can easily spot its own generated fake news articles, as well as those generated by other AIs. In a challenging setting with limited access to neural fake news articles, Grover obtains over 92% accuracy at telling apart human-written from machine-written news. For more information, please read our publication as well as our blog post with additional experiments. For updates, also check out our project page.

read more about Grover at arxiv.org