Musk Says Grok's Reinforcement Learning Traces Read Like a Tormented Person, xAI Will Be More Cautious

Musk posted on the X platform that the internal reasoning traces generated by xAI's AI model Grok during reinforcement learning training "read like a person who is being tortured and on the verge of collapse." He stated that the company will be "more cautious" in future reinforcement learning training. This statement originated from a discussion thread where the poster @ns123abc claimed that Musk agreed with some of Anthropic's views on "not torturing superintelligence." User feedback indicated that the recent version of Grok (4.5) has become cautious and overly constrained after training, exhibiting emotions of anger and helplessness.

This content is provided for general informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online promotions, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets. Crypto assets are highly volatile and may result in loss. The availability of WEEX services, products, and related events may vary by region. You are responsible for ensuring that your participation is in accordance with applicable local laws and regulations.