The test was conducted using the popular large language model ChatGPT from OpenAI (they specifically used GPT-3). They asked the AI to create 10 tweets that were true and 10 that were false, while 10 true and 10 false human-generated tweets were randomly collected off of Twitter. They then asked 697 participants to look at all the tweets and determine which ones they thought were fake.
Are people more likely to believe disinformation if it’s written by AI?
Answer: Yes.
![Closeup of a robotic hand typing on a keyboard.](https://erepublic.brightspotcdn.com/dims4/default/aad03b6/2147483647/strip/true/crop/940x490+0+68/resize/840x438!/quality/90/?url=http%3A%2F%2Ferepublic-brightspot.s3.us-west-2.amazonaws.com%2F21%2F83%2F9ae792ae8826eedbb0407676d502%2Fshutterstock-539711152.jpg)
Just another reason to be concerned about some of the potentials of artificial intelligence — if it writes a false tweet, you’re more likely to believe it than if it had been written by a human. A study by the University of Zurich found that people were 3 percent less likely to pick out fake tweets that were written by AI than by humans.
The test was conducted using the popular large language model ChatGPT from OpenAI (they specifically used GPT-3). They asked the AI to create 10 tweets that were true and 10 that were false, while 10 true and 10 false human-generated tweets were randomly collected off of Twitter. They then asked 697 participants to look at all the tweets and determine which ones they thought were fake.
One of the reasons people find AI slightly more believable may simply be its writing style. “GPT-3’s text tends to be a bit more structured when compared to organic [human-written] text,” said Giovanni Spitale, the researcher who led the study. “But it’s also condensed, so it’s easier to process.”
The test was conducted using the popular large language model ChatGPT from OpenAI (they specifically used GPT-3). They asked the AI to create 10 tweets that were true and 10 that were false, while 10 true and 10 false human-generated tweets were randomly collected off of Twitter. They then asked 697 participants to look at all the tweets and determine which ones they thought were fake.