The journey of the term “fake news” has been an interesting one to watch. What originally gained prominence as a phrase used to describe stories about presidential candidate Hillary Clinton in 2016 later became a rallying cry for election winner Donald Trump, who uses “fake news” as a means to discredit news media.
But fake news — real fake news, if such a thing can exist — is a huge problem. It reared its head throughout last year’s presidential campaign and continues to operate all over the Web. And though people are behind the unverifiable claims and flat-out falsehoods at the moment, it might not stay that way for long, as researchers from the University of Chicago have shown.
The team of researchers, working in the university’s computer science department, have successfully built artificial intelligence (AI) capable of writing fake reviews online — think the types of reviews you’d read about products on Amazon or restaurants on Yelp.
This may not seem all that impressive if you consider what a simple review could look like: a bot simply saying “bad food” wouldn’t be all that impressive, right? But these researchers, led by Ben Y. Zhao, were able to develop AI that can leave highly believable reviews that are both unable to be detected as fake, and have been considered reliable by those who’ve read them.
According to Business Insider, one review looked like this:
“I love this place. I went with my brother and we had the vegetarian pasta and it was delicious. The beer was good and the service was amazing. I would definitely recommend this place to anyone looking for a great place to go for a great breakfast and a small spot with a great deal.”
I don’t know about you, but that looks like a pretty normal, human-written review to me. But it’s not. It was written by a machine, powered by artificial intelligence, that’s never been outdoors much less sipped a beer.
There are immediate implications here for companies who use reviews pretty heavily on their websites. It’s understandable the Amazons and Yelps of the world would be concerned about fake reviews like this becoming widespread.
But it’s tough not to think about the wider problems this type of artificial intelligence could present when it comes to the news. As the recent election showed, opinions can be shaped by fake news. Markets go up and down depending on the news, and fake news can send a market into a fall, just as it did when the Syrian Electronic Army hacked the AP’s Twitter account in 2013 and posted about a fictitious bombing at the White House.
If we reach a point where humans can step away from the keyboards and let artificial intelligence write fake news that’s just as believable, we could soon have a real crisis on our hands — one where legitimate news outlets have to spend more time debunking fake news stories than reporting on real ones, and one where people are less trusting of news media in general.
It’s something to think about.