I just don't think this is it, although I freely admit that I'm amazing at the technology.

As it stands, Chat-GPT is probably better at giving the appearance of reasoning that some percentage of humans. That reflects poorly on those humans, but I've certainly known people in my life who seemed like they rarely knew the end of a sentence when they began speaking.

However, the focus of LLMs on text-prediction is not a path that leads toward human-like reasoning, just the appearance thereof. If and when we have AIs that actually do reasoning, it will be some other branch of AI, possibly also incorporating an LLM for communication, but not based solely on LLMs.