4 ms·

There are few examples of this as well: https://www.reddit.com/r/singularity/comments/1fqjaxy/contextual_training_and_overreliance_on_llms/ https://www.reddit.

• andix 11mo ago
  It really shows how LLMs work. It's all about probabilities, and not about understanding. If something looks very similar to a well known problem, the llm is having a hard time to "see" contradictions. Even if it's really easy to notice for humans.