Post #1656309

IMO, one of the greatest limitations of LLMs is that they don't have any idea about what they *don't* know. A gigantic "context" without any kind of introspection. I asked ChatGPT for an image of a "sgruf on the moon eating bergles". Generated image is nice. Now tell it "It is not a sgruf, it is a pitoc, please use a sgruf, even the green one is ok". Almost identical image, but green (obviously). And you can go on and on... It will never say "what is a sgruf?" *unless* you hint at that...