Why AI Can't Spot Fake News

Fake News

AI technology has become deeply embedded in nearly every aspect of our world. In fact, chances are you are exposed to artificial intelligence if you use social media applications, voice assistants, or even chatbots. 

Download Amazon HQ2 Whitepaper "Using  AI-Processed News to Perform Predictive Analytics"


Though some may say AI is still in its infancy, there's no questioning the vast capabilities of the technology. AI algorithms can write realistic text nearly convincing enough to pass for human. The good news is AI technology can also be used to differentiate real and fake text. 


The problem is that AI cannot readily determine the validity of the information, which has caused a widespread flood of "fake news" across the internet. The impact of misinformation has the potential to affect the way people receive news adversely. 


How AI Writes Text

AI systems are becoming increasingly proficient at generating written dialogue. Applications are simple and accessible to nearly anyone with a computer. People begin by entering a text prompt, and the AI software automatically generates cohesive language (at least from a structural standpoint). 


With simple algorithms, the technology generates text by studying large datasets from the internet and other various resources. It then looks for pattern information and creates a surface-level narrative. The problem is much of the text may be falsified information.


A global survey found that 86% of internet users have been misled by fake news, mostly through social media outlets such as Facebook. Widespread misinterpretation of news has created a sense of distrust and skepticism amongst users. 


The Potential Ramifications of Fake News

Though fake news can appear to be satirical if not harmless on a surface level, truth is the spread of misinformation can have extreme consequences. 


In 2016, a 28-year-old man walked into a Washington D.C. bar and pizzeria and fired into the establishment after being told Hilary Clinton led a child-trafficking ring from the restaurant. Better known as "Pizzagate" this is just one extreme example of fake news gone wrong.


The problem is the line between satire and realism has become blurred, leading people to question the validity of all news outlets. AI text generators are consistently becoming smarter, faster, and more precise, creating an ever-growing potential of malicious attacks and slanderous claims. Many even claim "fake text" to be the next global political threat. New technologies have emerged to spot and determine AI text. 


AI Can Detect AI Text

Not only can AI write text, but the same technology can also detect AI text. Researchers from Harvard University and MIT recently collaborated on a new technology designed to spot AI-generated text. Known as the Giant Language Model Test Room (GLTR), it uses statistical patterns to determine if language model algorithms versus humans have written the content. 


The technology uses algorithms to identify inconsistencies over sixty-word sequences to spot predictable words and phrases. The limitations of the technology mean that sentence structure and grammar selection may be intact. Still, the software is unable to determine if the text is lacking any significant depth or realism.


Why AI Can't Spot Fake News

Despite advancements, recent MIT research indicates detecting fake, computer-generated writing is still a work in progress. Experiments showed fake news detectors could not differentiate fake from real text if both were machine-generated. 


Research has indicated that identifying machine-generated text is not enough. Programs also must determine valid, truthful text from intentionally malicious and misleading information. The difficulty lies in the fact that AI text recognition mainly uses statistical contextual clues. 


Teams hope to improve existing models through the development of new algorithms and the construction of datasets that cover more types of misinformation. The issue is AI text writing software continues to improve at a rapidly increasing rate. Creating new technologies that spot AI text beyond simple pattern recognition may prove to be complicated.


The Future of Fake News

Thanks to AI text generators, a massive influx of fake news has swept the internet. AI can now be used to recognize AI text, however limitations still exist beyond the scope of current technological capabilities. Although AI cannot yet readily determine the validity of information, strides are continually being made. 


Artificial intelligence continues to shape the world of today. The future of AI precision intelligence is here with Bitvore AI. As a corporate research platform that uses unstructured data sources and advanced AI-techniques, Bitvore uses comparative and predictive analytics to better drive corporate decision making.


For additional information on Bitvore AI and how it's changing the world, be sure to:Check Out Our Whitepaper: Using AI-Processed News Datasets to Perform  Predictive Analytics


Check Out Our Whitepaper: Using AI-Processed News Datasets to Perform Predictive Analytics

Subscribe to Bitvore News Blog Weekly Email

Recent Posts