The boundaries of human language are the boundaries of cognition.
Artificial intelligence based on language naturally has boundaries brought by language, and its cognition cannot transcend the representational system, thus it is not truly self-supervised in the genuine sense.
Humans are constrained by language, and large language models are likewise constrained by language. Humans have experiential knowledge based on body and sensation, and artificial intelligence will also have this in the future, but perhaps not based on LLMs.
From a philosophical perspective, AI is currently only in an infa
原文表示