news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

Key Points

Tencent Technology News, July 18, according to foreign media reports, Elon Musk has high hopes for Grok, a chatbot launched by his artificial intelligence company xAI, hoping that people can get news information through it. However, when faced with complex events in reality, its performance is a mess. In particular, shortly after the assassination attempt on former US President Donald Trump, Grok published multiple misleading headlines based on the information flow on X, highlighting the limitations of artificial intelligence technology.

One particularly notable false alarm, which falsely claimed that Vice President Kamala Harris had been attacked, appeared to stem from a satirical interpretation by some X users of a previous slip of the tongue by President Biden, in which he accidentally confused Harris' name with Trump's.

Another summary incorrectly identified the suspect and linked him to the "extreme left-wing movement (antifa)." This statement seems particularly hasty when the police have announced different suspects and the motive is still unclear.

Although there is a disclaimer below these summaries, which clearly states: "This report is a summary of posts on platform X, and the content may change over time. As an artificial intelligence tool, Grok may be wrong, so users are requested to verify its output themselves." However, this has not completely eliminated the misleading influence of Grok in news dissemination.

Musk had high hopes for the X platform and its artificial intelligence robot Grok, claiming that it could automatically extract the essence of hundreds of millions of user posts and generate headlines and news summaries. He has publicly criticized traditional news media for their slow response and declining credibility, and encouraged users to get used to getting instant news updates from Grok.

Musk emphasized at an advertising industry event in June this year: "On the X platform, we are achieving the aggregation and refinement of information. We use the power of artificial intelligence to transform the massive input of millions of users into refined summaries. I firmly believe that this will be the new normal for news dissemination in the future."

X has not yet made an official response to Grok's false alarm.

Grok is a product of Musk's artificial intelligence company xAI. Since it started research and development last year, it has been gradually integrated into X, bringing subscribers a number of innovative features including intelligent chatbots. In terms of news summarization, Grok has shown a certain degree of accuracy, indicating its huge technical potential. Musk specifically pointed out that Grok's ability to instantly capture a large number of posts on X is one of its significant advantages. X calls it "an artificial intelligence search assistant with a sense of humor."

However, recent false alarm incidents have also exposed Grok's design deficiencies, especially when dealing with emergencies, where the algorithms and models it relies on may become weaknesses. Faced with real-time information flows from multiple channels, a computer model that tends to interpret humorously obviously faces considerable challenges in screening and integrating information.

In response, Katie Harbath, former Facebook public policy director, said: "We still have a long way to go. When it comes to emergencies like shootings, when the facts are not yet determined, human professional judgment and contextual knowledge are indispensable."

Even experienced journalists chasing breaking news at this sensitive time are bound to make mistakes, but some of Grok's mistakes went far beyond the realm of routine confusion.

One notable Grok headline falsely claimed: "Actor from the movie Home Alone 2 was shot at a Trump rally?" Trump did appear in the 1992 movie Home Alone 2: Lost in New York, which became a hot topic among some X users, but the Grok summary failed to clarify whether the "actor" was Trump himself. Instead, it confused the public by saying: "An actor from the movie Home Alone 2: Lost in New York was reportedly shot at a public event, presumably related to a Trump campaign rally."

Grok isn’t the first generative AI tool to stumble in summarizing news events. Previously, Google also announced a correction because its AI function occasionally made bizarre suggestions, such as suggesting that cheese be glued to pizza.

It is worth noting that some technology companies have adopted a different strategy from Musk, avoiding positioning their products directly as real-time news tools. For example, when OpenAI's ChatGPT received inquiries about the attack, it not only summarized the incident itself, but also provided a link to the fact-checking website PolitiFact, while clearly stating that it is "not a real-time news product."

Meta executives said they do not encourage the posting of political content on the short video platform Threads, which is designed to differentiate itself from X. Meta executive Adam Mosseri expressed this position last year, hoping that Threads can avoid the negative effects and other drawbacks that often accompany politics and hard news.

At the same time, over the weekend, some Threads users complained that the platform was not as fast as X in updating information about the shooting, reflecting that Musk's platform is still one of the important sources of news information.

In response to the assassination attempt, X CEO Linda Yaccarino emphasized in an internal email sent on Sunday that the company has treated this as a "major incident" and quickly mobilized resources to prioritize any content on the platform that violates the rules, especially those posts that praise the attack.

This is not the first time that Grok has encountered challenges in summarizing news events. After the US presidential debate in June, when user X mentioned California Governor Gavin Newsom, Grok mistakenly generated the headline "Newsom won a big victory in the recent debate", even though Newsom did not participate in the debate, although he attended the relevant event. In addition, although some political observers mentioned that Newsom might be a potential successor to Biden, this statement has been publicly denied by Newsom himself.

When Musk took over X (formerly Twitter), the platform had a dedicated team responsible for writing summaries of hot topics. However, with Musk's large-scale layoffs, the team was also disbanded. Former Twitter executive Evan Hansen revealed that he was a member of the team until the layoffs occurred.

Hansen recalled that before Musk took over, Twitter had tried to use artificial intelligence to assist in writing summaries, but manual review and correction of content was still required. He pointed out: "Although artificial intelligence can play a role, it must be handled very delicately." (Compiled by Jinlu)