12 hours ago
Elon Musk Vows to 'Rewrite Human Knowledge' Using Grok AI, Slams Existing AI Data as ‘Garbage'
Billionaire entrepreneur Elon Musk is setting his sights on an ambitious new goal for his AI company, xAI: rebuilding the entire corpus of human knowledge using the latest version of its AI chatbot, Grok.
In a series of posts on X (formerly Twitter), Musk criticized current AI models for being trained on what he called 'garbage' data and unveiled his plan to retrain Grok using a revised dataset. 'We will use Grok 3.5 (maybe we should call it 4), which has advanced reasoning, to rewrite the entire corpus of human knowledge, adding missing information and deleting errors,' Musk shared. 'Then retrain on that. Far too much garbage in any foundation model trained on uncorrected data.'
Musk's goal is not just about refining Grok's capabilities—he wants to reshape how AI models are built, trained, and aligned with truth. Launched earlier this year, Grok 3 was introduced as the 'smartest AI on Earth,' boasting performance ten times stronger than its predecessor. The model is accessible via xAI's platforms, the Grok app, and to X Premium Plus subscribers.
One of the more controversial elements of Musk's announcement involves his call for user input. In an appeal to the X community, he invited followers to contribute 'divisive facts' to help train Grok—facts that may be politically incorrect but, as Musk emphasized, are 'nonetheless factually true.'
Musk founded xAI in 2023 to challenge established AI giants like OpenAI. He has often accused leading models, including ChatGPT, of harboring 'woke biases' and distorting facts to fit certain ideological perspectives. With Grok, Musk wants to break away from that mold and create an AI assistant grounded in what he considers cleaner, more accurate information.
At the core of Grok's development is xAI's Colossus supercomputer, a powerful system built in less than nine months using more than 100,000 hours of Nvidia GPU processing. Grok 3 uses synthetic data, reinforcement learning, and logic-driven techniques to minimize hallucinations—a common flaw where AI chatbots fabricate responses.
Now, as Musk and his team prepare to roll out Grok 3.5—or Grok 4—by the end of 2025, the focus is shifting toward using advanced reasoning and curated content to create a more reliable foundation for machine learning.
With this bold move, Musk is not just tweaking another chatbot. He's trying to challenge the entire approach the tech industry has taken toward artificial intelligence—and possibly redefine what AI knows as 'truth.'
