New Software Writes Convincing Fake News Articles

A new software program has been developed that can produce authentic-looking fake news with just a few pieces of information. Created by OpenAI, an artificial intelligence research group co-founded by billionaire Elon Musk, the software is able to near-instantly create fake news articles, including quotes from government officials, from a single sentence or small set of targeted keywords. An example was published by OpenAI in a research paper to show the public what the program could do.

The program created by OpenAI uses language modeling to predict the next word of a piece of text, kind of like auto-complete for text messages. While the program did write text with scenarios that didn’t quite make sense, some of the text created closely resembled the work of disinformation artists.  A basic version of the program has been released to let artists, writers, and journalists experiment with the algorithm to see what kind of text it can generate.

Due to concerns about what the program could be used for, OpenAI made the decision to not publish or release the most sophisticated versions of the software or the 8 million web pages it used to train the system. Jack Clark, the organization’s policy director, said, “One of the not so good purposes would be disinformation because it can produce things that sound coherent but which are not accurate.” However, according to Clark, the system is not trained well enough to pose an immediate threat.

The concerns around fake news and technology’s role in the spread of disinformation have been steadily growing over the past few years. Several tech companies have recently added language to their financial filings warning investors that their AI software could raise ethical concerns and harm the business. Last month, Google put constraints on research software it has shared due to fears of misuse.