Post #1953629

@xs4me2 @shadur @dynamite_ready @reading_recluse > I think you are wrong… You are not spelling out who is wrong, or wrong about what. LLMs can't be trusted. They do not "sometimes hallucinate". They _always_ hallucinate, 100% of the time. They cannot do anything else, and cannot be made to do anything else. Making up new text that resembles the input is how they work. It is all that they do or can do or will ever do. Yes, critical thinking is always needed... but the problem is that LLM use destroys critical thinking ability. They do not save time: they waste the users' time, and make them spend _more_ time and _more_ effort on checking... While destroying people's ability to check. https://www.media.mit.edu/publications/your-brain-on-chatgpt/