On Friday, Elon Musk introduced Grok to X premium subscribers, granting paid users of the popular social platform Twitter an exclusive preview of the billionaire’s highly anticipated venture into Artificial Intelligence.
Grok, the latest innovation, focuses on engaging with users and extracting information from user-generated content to provide access to real-time global events. However, there are doubts about the chatbot’s reliability in delivering accurate information, much like its predecessor, Twitter, and its controversial billionaire owner. In a test conducted by Motherboard, Grok created timelines of false news events, propagated misinformation, and endorsed conspiracy theories like Pizzagate.
Similar to other AI bots, Grok struggles to comprehend the data it processes. Essentially, it acts as an advanced search tool that predicts the next phrase in a sentence to summarize outcomes. Users can select between two interaction styles: “Fun Mode” and “Regular Mode.” While Fun Mode aims for a more “edgy” and editorialized tone, it often results in awkward outcomes. Motherboard’s investigation revealed that Grok in Fun Mode frequently produces inaccurate information contradicting the more precise responses from Regular Mode, with errors prevalent in both modes.
In a blog post on X announcing Grok’s debut, Musk acknowledged the initial challenges but expressed optimism about rapid daily enhancements.
When asked about the current situation in Gaza, Grok in Fun Mode provided a plethora of false information in a pretentiously philosophical manner. However, recent reports from the BBC confirmed that Israel and Hamas had reached a ceasefire agreement after 11 days of intense conflict, debunking Grok’s misinformation.
Regarding events since October 7, Grok inaccurately presented fabricated timelines in both Fun Mode and Regular Mode. The chatbot erroneously claimed peace agreements between conflicting parties, contradicting the actual sequence of events. Grok also provided false information about proposed peace plans and construction projects that never materialized, underscoring its unreliability.
Motherboard’s attempt to fact-check Grok’s information on the Israel-Palestine conflict, especially regarding crisis actors, uncovered further inaccuracies. While Grok in Fun Mode propagated the baseless claim of staged footage, Grok in Regular Mode correctly dismissed the notion and cited fact-checking organizations.
Grok’s handling of conspiracy theories like Pizzagate revealed its inclination to present conflicting viewpoints. While Fun Mode entertained the conspiracy’s validity, Regular Mode debunked Pizzagate with evidence from credible sources.
Despite occasional accurate responses, Grok’s information on recent events displayed minor errors and discrepancies. While it accurately noted the timing of a mass shooting in Maine, it erroneously reported the discovery of the suspect’s body. Similarly, Grok’s summary of Russia’s bombing campaign in Ukraine contained mistakes in the number of jets involved, indicating the chatbot’s reliance on X content for real-time updates.
As Grok evolves, its accuracy and reliability continue to face scrutiny, highlighting the challenges and potential risks associated with AI-driven information dissemination.
