Raw LLM Responses

There is no paradox, you've presented a false dichotomy. "Dumb" and "Smart" are not mutually exclusive in this context. The AI is "dumb" in the dimension of generalized, conscious intelligence (it has no self-awareness, beliefs, or desires). It is "smart" in the dimension of executing tasks it was trained on (pattern recognition, data synthesis, text generation) at a massive scale. A chess computer is "smarter" than any grandmaster but is not conscious, and no one would ever claim it was. The confusion with an LLM arises only because its output isn't a set of chess moves, but natural language. An LLM does not understand, it manipulates statistical relationships between tokens. It does not need to understand to generate coherent and useful text. Therefore, no one should be confused into thinking it is a conscious or intelligent entity. Regardless, it will take jobs. Many many jobs.