Detecting fake news by using machine learning

Meet Grover – a program that is capable of not only detecting fake news but producing it as well. In order to produce it, the researchers tuned the GPT-2 neural network.

It seems the age of digital transformation and innovative technologies is the present reality we live in. The researchers at the Allen Institute of Machine Learning have created a program that is capable of detecting and producing fake news. It’s based on a natural language processing system that can generate articles perfectly capable of fooling a human reader.

Imagine a scenario when one machine is faced against another, a battle of good and evil. A war consisting of light and darkness that the bots carry out. Well, this just might be it.

For those who are wondering how it’s possible to detect fake news, a simple way of explaining it would be to say that fake news spewing algorithms leave a trace or a signature behind. This is what the researcher team refers to as “artifacts”. The very same patterns are used when constructing and spotting these word combinations.

According to the authors, combating the spread of fake news is a pressing matter, both in the technical as well as the political sense of the word. They see the power of GPT-2 as an excellent opportunity to make a stand against it.

If you’re interested in the full scope of the research paper, head on over to the arXiv pre-print server and search for the paper titled “Defending Against Neural Fake News”.