Grok keeps sending gibberish responses to users
xAI's chatbot Grok has been generating nonsense output for several users interacting with the tool on Grok.com. The problem was reported mainly by people accessing Grok Lite, who encountered long strings of meaningless text and unusual research links. Although official status dashboards indicated normal operations, the company acknowledged the temporary bug on X.
Key Takeaways
- An unexpected technical problem has caused xAI's chatbot Grok to reply with nonsense text to direct prompts submitted on Grok.com.
Multiple individuals using Grok Lite noted the behavior starting Wednesday morning, with one query yielding a response starting with "match it without and your they and two for planets can practical and often cheese..."
- followed by several paragraphs of irrelevant text.
Other prompts produced unexpected links pointing to reinforcement learning research websites, prompting numerous user complaints across online forums such as Reddit.
- The official Grok account on social media acknowledged the behavior on Thursday morning, referring to it as a "pure word salad" temporary generation glitch and advising users to regenerate output or open a fresh session.
TechCrunch was unable to recreate the error during testing, suggesting the glitch affected a limited portion of the user base.
- The incident follows internal changes at xAI, which lost most of its founding team alongside at least 50 researchers and engineers according to a report from The Information in May.
Users of xAI's Grok Lite reported receiving strings of gibberish text when submitting direct queries on Grok.com starting Wednesday morning.
- The issue comes after xAI experienced significant turnover, losing most of its original founders and at least 50 researchers and engineers since May.
An unexpected technical problem has caused xAI's chatbot Grok to reply with nonsense text to direct prompts submitted on Grok.com. Multiple individuals using Grok Lite noted the behavior starting Wednesday morning, with one query yielding a response starting with "match it without and your they and two for planets can practical and often cheese..." followed by several paragraphs of irrelevant text.
Other prompts produced unexpected links pointing to reinforcement learning research websites, prompting numerous user complaints across online forums such as Reddit. Despite user complaints, the official status page at status.x.ai indicated that all services were fully operational. The official Grok account on social media acknowledged the behavior on Thursday morning, referring to it as a "pure word salad" temporary generation glitch and advising users to regenerate output or open a fresh session.
TechCrunch was unable to recreate the error during testing, suggesting the glitch affected a limited portion of the user base. For those observing AI system performance, output errors like returning unexpected research links or unstructured text can highlight underlying difficulties during model execution or deployment. The incident follows internal changes at xAI, which lost most of its founding team alongside at least 50 researchers and engineers according to a report from The Information in May.
For more details please read the original article at TechCrunch AI.
Continue Learning
Comments
Comments appear only after moderation. Your email identifies your submission to the moderator and is never displayed here.
No approved comments yet.