Elon Musk says his AI model Grok will be retrained using improved data without mistakes or bias.
- Musk wants Grok AI to rewrite human knowledge and delete false info.
- Critics warn the project may change facts to match personal beliefs.
- The update aims to fight bias, but may spread misinformation instead.
Grok 3.5 Will Use “Corrected” Knowledge
Elon Musk’s company, xAI, is preparing a new version of its chatbot, Grok. The updated version, called Grok 3.5, will be trained on what Musk calls a new “corrected” dataset. He said the current data used to train AI has “too much garbage.” His goal is to rewrite existing content to improve its accuracy before using it to retrain Grok.
AI Will Help Rewrite Human History
Musk posted on X (formerly Twitter) that he wants Grok to correct and add missing information to the entire body of human knowledge. He believes existing AI models include many errors. Musk said this rewrite will help Grok reason better and become less biased than other AI systems like ChatGPT from OpenAI—a company he once helped create.
Critics Say Musk’s Plan Is Risky
Not everyone agrees with Musk’s plan. Gary Marcus, a professor and AI expert, said the idea sounds like something from a dystopian novel. He compared it to rewriting history to serve a personal agenda.
Another professor, Bernardino Sassoli de’ Bianchi, said powerful people shouldn’t edit facts just because they disagree with them. He warned that controlling training data could lead to dangerous outcomes.
User Input Could Fuel Misinformation
Musk invited X users to send him “politically incorrect but true” facts to help train Grok. Some users replied with misinformation, conspiracy theories, and offensive claims. These included vaccine lies and racist ideas, which raised concerns that the AI might spread false or harmful content.
Balancing Free Speech and Truth
Musk has often criticized content rules he sees as too strict. He wants platforms like X and tools like Grok AI to highlight all sides of a topic. But critics worry that this approach might encourage lies and hateful views instead of real facts. As Grok 3.5 moves forward, many are watching how it handles truth and bias.
Source: cointelegraph.com