Grok, the chatbot built by Elon Musk’s AI startup xAI, is promoting false news reports on X, a report by Ars Technica stated.
The chatbot, which is meant to go through all the posts on the microblogging platform and summarise breaking news reports, falsely accused an NBA player of criminal vandalism last week. In a trending post that was written by the AI chatbot, Klay Thompson was shown to be on a “brick-vandalism spree.” The story then went on to make up details to support the false news.
SF Gate, one of the first news outlets to spot the post, said Grok seems to have misunderstood a common basketball phrase with the crime mentioned. When basketball players take an airball shot that doesn’t hit the rim of the basket, it is referred to as “throwing bricks.” Thompson reportedly had a tough night in his previous match, which led to the mix-up. After the incorrect news report was posted, X users took to fuelling the misinformation by commenting with fake victim reports in the same format.
X hasn’t taken down the post yet. It does include a disclaimer, however, stating, “Grok is an early feature and can make mistakes. Verify its outputs.”
(For top technology news of the day, subscribe to our tech newsletter Today’s Cache)
Grok has been caught spreading fake news multiple times in the past. A report by PC Mag published on 17 April last month stated that it published a news report claiming that Indian Prime Minister Narendra Modi lost the general elections, even though they haven’t happened yet.
The headline stating, “PM Modi Ejected from Indian Government,” appeared on the feed of an X user who called it out.
A few days prior to that, another AI-generated story appearing on the main feed of the platform announced that Iran had attacked Tel Aviv with missiles.
Musk reportedly laid off Twitter’s news curation team in November 2022, shortly after acquiring the platform. This summarising and contextualising of news content is now done by Grok.
Meanwhile, Grok was rolled out to Premium X subscribers a couple of weeks back with Musk saying that his AI engineers are working on the second version of the large language model behind the chatbot.
Musk revealed that they hope to complete training on the AI model by May. The training required as many as 20,000 Nvidia H100 GPUs, which are the most advanced AI chips available in the market.
Musk also said he anticipated that the Grok 3 model could need around 100,000 Nvidia H100 GPUs for training.
