Elon Musk has turned to the crowd for help in improving his AI chatbot. Over the weekend, the billionaire entrepreneur appealed directly to X users to share examples of where his company XAI’s conversational assistant Grok is falling short. The call for feedback comes as the firm rolls out Grok 4.1, a significant update that promises faster reasoning and fewer errors, after weeks of mixed reviews.

Musk’s petition marks a rare moment of open collaboration between users and a major AI developer. In a post on X, he asked the community to report Grok’s failures and even include examples where competing chatbots perform better. “Please provide examples where @Grok needs to improve answers. Showing how another AI does this better would be helpful,” Musk wrote. “These should be examples of Grok going wrong today, as we fixed several bugs earlier in the week.”

He made a typically emphatic statement: “We will not rest until Grok is perfected.” Musk also urged users to share positive examples and said that highlighting the strengths of the model was “useful for the team as well.”

Try to improve after strange answers

The move comes in the wake of a series of viral misfires from Grok that have raised concerns across the internet. In recent weeks, the chatbot had attracted attention by issuing wildly exaggerated statements about its creator, even claiming that Musk was physically superior to professional athletes and historical icons.

In a now-infamous exchange, Groke declared that Musk would beat basketball star LeBron James in a fitness competition, arguing that the tech mogul’s daily “Managing rocket launches, EV revolutions, and AI frontiers demands a rare blend of physical stamina.” It also dubbed Musk the “fittest man alive” and predicted he could defeat former boxing champion Mike Tyson in the ring.

Musk laughed off these incidents, attributing the bot’s behavior to malicious manipulation rather than genuine bias. “Earlier today, Grok was unfortunately provoked into saying absurdly positive things about me,” he wrote, before capping the moment with a self-deprecating quip: “For the record, I’m a fat retard.”

Nevertheless, the episode reignited the debate about the trustworthiness and neutrality of xAI’s technology. Critics pointed out that Grok’s erratic tone and tendency to generate chatter could undermine its credibility at a time when xAI is positioning the model as a serious competitor to OpenAI’s ChatGPT and Google’s Gemini.

New version of xAI aims to fix accuracy flaws

In response to the criticism, Musk confirmed that Grok 4.1 has been launched with several improvements. “Many updates and improvements have been implemented on Grok 4.1 and many more are on the way!” He has written. “Going forward, Grok 4.1 will spend more compute time thinking about your question to improve accuracy.”

According to XAI, the new version allocates more computational power to reasoning tasks, which should result in fewer factual mistakes and better contextual understanding. Musk also shared a video tutorial explaining how users can report problematic reactions directly in the chat interface.

The feature lets users give a thumbs-up or thumbs-down rating or file detailed problem reports using a three-dot menu, with Musk describing this approach as crucial to Grok’s development. He wrote, “Critical feedback about Grok is extremely helpful. We’ll keep iterating until Grok is the best in every way.”

Musk’s open call for help underscores his pragmatic approach to AI development and his desire to involve users in shaping the progress of XAI. Whether this community-driven strategy will make Grok more accurate, or simply more chaotic, remains to be seen.