Why does fake news still slip under the AI radar?

An artificial intelligence program, trained on various combinations of over 900 possible variables characterizing the trustworthiness of a media outlet, could judge them accurately only 65 percent of the time, which is a miss by quiet a margin.

The deficiency in the machine-learning model, developed by scientists at MIT, Qatar Computing Research Institute (QCRI), and Sofia University in Bulgaria, is not without a crucial lesson – our AI systems for spotting fake news are “data starved,” writes MIT Technology Review.

The researchers, instead of fact-checking individual claims or going after the trolls, focused on establishing the reliability of the news source, a factor closely linked to misinformation, reports Popular Science.