Elon Musk Announces Plan to Retrain Grok for Greater Accuracy
On June 21, 2025, Elon Musk announced a major update to xAI’s Grok chatbot. The company will retrain the model—likely launching it as “Grok 4”—after calling the current data filled with “far too much garbage.”
Musk said Grok 3.5’s advanced reasoning will now be used to clean up and improve its knowledge base. This includes verifying facts, correcting errors, and deleting misleading or inaccurate information.
“We’re using Grok to clean Grok,” Musk posted on X, adding that the retraining will create a smarter, more reliable AI grounded in verified information.
Musk Calls on Users to Submit ‘Divisive Facts’
Musk has also invited users on X (formerly Twitter) to contribute “divisive facts”—information that may be politically incorrect but factually true—to be considered for Grok’s retraining data.
This crowdsourcing approach is part of Musk’s broader plan to develop an AI model that pushes back against what he calls “mainstream constraints.” He has frequently criticized other AI models like ChatGPT for being “woke,” implying they filter or avoid controversial truths in the name of social harmony or political correctness.
Scroll to continue reading
Critics Warn of Bias and ‘Orwellian’ Revisions
While Musk promotes this update as a way to improve truthfulness, critics say it could reinforce bias and personal ideology. NYU emeritus professor Gary Marcus called the move “Orwellian” and compared it to rewriting history.
“The danger isn’t just misinformation,” Marcus said. “It’s locking in one worldview as truth through model retraining.”
This concern follows past incidents, including when Grok briefly repeated “white genocide” tropes after an unauthorized backend modification—an error xAI later corrected, Business Insider is reporting.
Grok Gains Traction Despite Controversy
Grok has quickly grown in visibility. Some reports suggest that U.S. government agencies have even tested its analytical capabilities, though it hasn’t been officially sanctioned.
Unlike other models, Grok is positioned as an “unfiltered” chatbot, often leaning into controversial questions and edgy topics.
Musk’s followers praise it as a voice against conformity. But researchers warn that models shaped by user-submitted “divisive truths” risk becoming echo chambers instead of fact-based tools.
Inside the Roadmap: Grok 3.5 to Grok 4
According to Musk, the retraining will unfold in multiple stages. First, Grok 3.5 will analyze its own knowledge base to identify and remove errors. Then, verified information will be added before retraining begins.
The goal is to train Grok 4 on a “refined” corpus that is leaner, more accurate, and more aligned with xAI’s values. Musk sees this as the future of AI—truth-seeking rather than consensus-driven.
Still, the lack of transparency in how facts are verified or filtered has some AI experts urging caution. “Who decides what counts as a fact?” one analyst asked.
Why This Matters in the AI World
Musk’s plans to retrain Grok could change the future of how AI systems are developed. It signals a move away from broad consensus-based datasets toward curated, possibly ideologically driven ones.
This could result in a powerful chatbot that reflects the beliefs of its builders and contributors more than objective fact. Or it could produce a sharper, leaner tool less burdened by noise and bias.
The truth may lie somewhere in the middle—but the outcome of Grok 4’s retraining could influence AI development across the entire tech industry.
MORE NEWS ON EURWEB.COM: Bill Gates Predicts AI Will Surpass Human Intelligence in Most Jobs Within 10 Years
We Publish Breaking News 24/7. Don’t Miss Out! Sign up for our Free daily newsletter HERE.
