@nprofile1q... I am very AI and tech positive but very anti-hype. Current LLMs are fanstastic statistical predictors of likely best next word over a huge corpus. They do not "reason". They can simulate reasoning because they were trained on a lot of actual reasoning. But they do zero conceptual thinking or even know, as far as using them, what a concept is. Trying to stuff all human work possible into them is insane.

That said they are really really good for somethings like churning out reasonably accurate code or websites from descriptions in languages they are trained on. They are really good a pulling together a lot of relevant related things in a reasonably coherent manner. They are good because of approximation push for best next word stuff in breaking logjams that would halt a more deterministic knowledge systems. But that makes them useful adjoint technology rather than being THE TECHNOLOGY.

I have been in tech since the Homebrew techno-hippie days in SF. I have seen so very many silver bullets and hype cycles. This is the most over-hyped HUGE BUBBLE I have ever seen. I don't think it will end well.

I am a big believe in real AI generally of which LLM is just one variety. I believe Artificial GENERAL Intelligence would be a very good thing to have in the world. We could sure use more effective intelligence active in this world than what we humans seem capable of putting together. But that GENERAL in AGI is for general learning which requires concept formation and one shot learning like humans do. LLM can never ever get there by the nature of its architecture.