The launch of Grok – Artificial Intelligence risks proving premature

X’s chatbot is supposed to be an artificial intelligence model that leverages posts on the social networking platform to show summaries of the latest news, but last week Grok’s weaknesses were exposed for the umpteenth time, when the chatbot got confused and unfairly accused NBA star of vandalism.

“Klay Thompson Charged in Series of Brick Attacks,” said the Grok-generated headline in a post that sat normally on X for days. Under the headline, Grok went into detail to analyze the fake news:

“In a strange development, NBA star Klay Thompson is accused of vandalizing several homes with bricks in Sacramento. Authorities are investigating the case as several people claim their homes were damaged, with windows smashed by bricks. Klay Thompson has yet to make a statement regarding the allegations. The incidents have shaken the local community, but no injuries have been reported. The motive behind the unconfirmed vandalism remains unclear.”

Grok is apparently confusing a common term in basketball where players are accused of unleashing “bricks” when they miss shots completely. According to SF Gate, one of the first to spot Grok’s error, Thompson had an overall rough night as he missed none of his shots in his emotionally-charged final game against the Golden State Warriors before left free to negotiate the next stage of his career.

In fine print below Grok’s post, there’s also a disclaimer, stating that “Grok is in its early stages and may make mistakes. Please verify the accuracy of its posts.”

But instead of confirming the accuracy of this particular post, some users chose to amplify the misinformation. Below the chatbot’s post, statements from alleged victims appeared. Some of these humorous messages recorded millions of impressions.

In the past, both Microsoft and OpenAI have faced defamation lawsuits, as ChatGPT falsely accused a politician and a radio producer of completely bogus criminal acts. Microsoft was also sued by an aerospace engineering professor who Bing Chat falsely labeled a terrorist.

It remains unclear whether a disclaimer like the one appearing under Grok would allow the companies promoting these products to avoid legal consequences should someone decide to take action against them. Possibly, whether there has been defamation will be determined by whether it can be shown that the platforms “knowingly” publish false information.

The fake news story with Thomson’s alleged vandalism may be the first time a Grok failure has become widely known, but it wasn’t the first time Grok has peddled nonsense by clicking on posts by X users who were joking on the platform. During the recent solar eclipse, for example, as reported by Gizmodo, Grok posted a headline that read “The Sun’s Weird Behavior: Experts Bewildered.”

While it’s fun to watch users poke fun at Grok, the whole situation reveals that Grok may prove vulnerable to manipulation by malicious actors to push more serious fake or propaganda news.

The capabilities of AI models remain impressive, but there is growing concern about the quality of the data they are trained with, and the incident with Grok and Clay Thompson’s “bricks” adds to a string of similar situations where chatbots reproduce completely incorrectly Results.