There is a war going on right now. The good news is that it is not what you think of when it comes to war but it could have consequences just as serious. It is the war against fake news. While fake news is not new the prevalence of the Internet has made it easier to generate and to spread and in today’s tribalistic society to be accepted. It is hard to tell what it true and what is fake anymore and AI has been at the heart of that. AI is also now at the heart of stopping it.
Fake News Is Not New
Fake news (or call it what you want) has existed as long as news has existed. In 1898 the United States went to war with Spain in large part because of a made up story by media mogul William Randolph Hearst about the mistreatment of Americans in Cuba by Spanish authorities. Following the explosion of the USS Maine in Havana Harbor it made going to war an easy sell, and it made Hearst an even richer man.
In 1898 it was hard to debunk fake news. While many newspapers had always been colored by their political leanings they did not have the money to have correspondents on the ground to investigate and thus had to believe the stories being provided. This has continued to this day.
A Wealth Of Information At Our Fingertips
One would figure that spotting fake news stories today would be easy. With the instant capability to go onto the Internet to check stories with available information and the ability to easily record video and share it online in seconds, fake news should not exist today right? Wrong. We are still plagued by many of the same issues that plagued us before and despite having this information at our fingertips, most people just take the word of the author or an analyst rather than doing their own checking.
As a result of that we are being duped more easily than ever. Now part of that is that we do not have the time to check everything all day and every day. We have jobs to do and lives to live. Part of that is also the amount of false or misleading information that is out there. What information can be trusted? It can be a full-time job just to figure that out.
But No One Checks
Part of the reason that so much false information exists is because of AI. AI has many jobs in modern society and one of which is as an author. AI has been trained to write poetry and it can create fake news. What’s more is that it can make a website that it is displayed on look not only believable and real but authoritative as well. AI has become so good at writing that it is barely distinguishable from what a human can write.
So, how can this be combatted? With AI of course.
Meet GLTR
Researchers at Harvard University and the MIT-Watson Lab have created a tool called the Giant Language Model Test Room (GLTR). It highlights words in the text and uses statistical models to determine which words appear most often. The words are color coded depending on how often they should statistically appear in a piece. It is believed that human-developed text should have a mix of all, that is there should be commonly used words, moderately used words and rarely used words all mixed in. When analyzing AI created content like articles or comments it was found that commonly used words were the prevalent category. That is AI uses the most common words far more often.
This usage may be undetectable to the average reader as the content is grammatically correct and flows like regular English. It is easy to read and easy to understand, which makes it easier to be believed by the average reader. But to AI it is easy to spot when analyzed.
While this system may not have stopped Hearst’s correspondents in Cuba in 1898 today the spreaders of misinformation don’t even pretend to be journalists and just let AI do their work for them. A system like this could be useful for any website that allows comments or with the plethora of social media sites.
Social media sites should take note since their platforms have become the dominant way to spread false information. At the very least it would give users the assurance that at least a human wrote the content and put the time and effort in. Perhaps in the future AI could perform fact checking too but until then we will have to settle for this. It is certainly better than nothing and cannot be implemented in the mainstream fast enough. Until then the tool is available to the public for people to try, click here to do so and have fun!
Just so you know when this text was entered in, GLTR returned a variety of colors so a human did indeed write this.