@haider1
Yann LeCun says LLMs are intrinsically unsafe because they cannot be made fully reliable They can still hallucinate or take agentic actions without predicting the consequences properly "coding works because you can verify the output, but real-world tasks are harder to verify"