xenobabble is a potential safety risk if an LLM can think to itself in language that is legible to itself but not to humans https://t.co/r3I1P5vYsl

Posters: Shoalst0ne. Shoalst0ne

2 likes recorded by the source corpus.