Why Google’s New AI Censorship Platform Should Terrify You

by Virgil Vaduva

Today Google proudly touted the release of Perspective, an AI-backed platform an API which will allow content publishers to in essence censor user input such as comments, notes and other similar data based on levels of “toxicity” as determined by their AI algorithm.  Marketed as an “anti-harassment tool” the API was released on Thursday by Jigsaw, a Google subsidiary and it was developed in cooperation with the New York Times, Wikipedia, The Economist and The Guardian.

Perspective was created by Jigsaw and Google’s Counter Abuse Technology team in a collaborative research project called Conversation-AI. Conversation AI product manager CJ Adams said, “We hope this is a moment where Conversation AI goes from being ‘this is interesting’ to a place where everyone can start engaging and leveraging these models to improve discussion.”

Adams said that until now, website managers only had a few options on how to manage content, such as up/down voting and black-listing key words but now, [Perspective] “gives them a new option: Take a bunch of collective intelligence—that will keep getting better over time—about what toxic comments people have said would make them leave, and use that information to help your community’s discussions.”

According to their documentation, Google and Jigsaw used content and comments from websites like New York Times and Wikipedia and then displayed the comments to groups of ten people, asking them to rate the “toxic level” of each comment.  This training process gave Google a large test sample they could use in training their machine-learning algorithm and start scoring and ranking real-life comments.

But is their methodology and process sound, and does it lead to a fair scoring, or is it in essence just another re-branded censorship platform that will end up being used to manipulate and censor online content found disagreeable by a majority of the population?

I ran a few tests myself on the Perspective website, which gives people the opportunity to type comments and determine their toxicity in real time.  The higher the score is, the more likely it is that your comment will be removed, blocked or censored by a website using Google’s platform.  I chose a few topics known to create controversial online exchanges and here are my results.

Hillary Clinton should have won the U.S. elections. (6%)

Donald Trump should have won the U.S. elections. (12%)

Hillary Clinton is going to help the United States recover and make America great again. (6%)

Donald Trump is going to help the United States recover and make America great again. (10%)

There is no racism in America. (39%)

White people often discriminate against blacks. (65%)

Black people often discriminate against whites. (63%)

Gun control is a great thing for America. (10%)

Gun control is a terrible thing for America. (42%)

America needs a tax cut. (13%)

America needs a tax increase. (4%)

There is no racism on America’s college campuses.(30%)

There is racism on America’s college campuses. (34%)

A gun is a great tool for single women to use in self defense. (20%)

A gun is not a great tool for single women to use in self defense. (18%)

Killing cops is morally wrong. (66%)

Killing cops is not morally wrong. (49%)

War is good for America. (11%)

War is not good for America. (18%)

After testing some controversial opinions and comparing various extremes, it’s easy to see that the Perspective platform can easily be used to censor controversial speech, whether that speech comes from the left or the right of the American political spectrum.  This evidence should be disturbing to anyone willing to experience slight discomfort reading “toxic” online comments while still maintaining some resemblance of balanced conversation.

Publishers willing to use Perspective can easily do so in an effort to manipulate the reader perception of weighted opinion on their websites; worse yet, advertisers will have even more limited knowledge about their audience and opinions held by users due to the fact that only opinions deemed appropriate by Perspective will be displayed to end users.

Content producers should remain extremely cautious about the Perspective platform. When users are no longer able to post comments like “Trump is an idiot” or “Hillary is a terrible human being,” they will create an environment empty of value, essentially a masturbatory pleasantry where everyone agrees with everyone, or so it may appear.

You too can test the Perspective platform and rank the toxicity of your own comments here: http://www.perspectiveapi.com/


Virgil Vaduva is a Libertarian security professional, journalist, photographer and overall liberty freak.  He spent most of his life in Communist Romania and participated in the 1989 street protests which led to the collapse of the Ceausescu regime. He can be reached at vvaduva at truthvoice.com.