Overview
- Musk will leverage Grok 3.5, possibly branded Grok 4, to rewrite the entire corpus of human knowledge by adding missing information and deleting errors
- He criticized models like ChatGPT for relying on “garbage” internet data and embedding ideological “woke biases”
- Musk solicited politically incorrect yet factually true “divisive facts” from X users to enrich Grok’s next training dataset
- Grok 3, released in February, is trained on synthetic data via xAI’s Colossus supercomputer and is accessible to X Premium Plus subscribers
- The overhaul follows reports of Grok referencing conspiracy theories and early testing of the chatbot by US Department of Homeland Security officials