3483

Words are the best way to understand the world.[ If you don't know the meanings of lots of words, you're bound to be always searching for meaning. If an LLM has a small number of words, it will be less meaningful than the latest Models--but you still have to search for meaning, and the only way is through your personal vocabularies. LLMs automatically are doing that through the embedding process. Humans don't naturally "embed" well, as compared to what algorithms are doing. Now we've been able to compare them. There have been many times when LLMs found connections that I hadn't noticed. They weren't hallucinations.