Yes. I thought this was rather obvious to anyone that ever tried to use llms for complex tasks. It’s not there yet, LLMs by themselves most likely will never be.

“Anybody who thinks LLMs are a direct route to the sort [of] AGI that could fundamentally transform society for the good is kidding themselves.”

A quote in an article that still uses the generic “AI” to refer to LLM models, thus losing any credibility. Probably was written by an LLM - sorry, AI, since that’s what it means now. AI is popular jargon now to mean anything that seems like it’s thinking, only serious people use AGI/ASI and even they often slip up and say AI sometimes. Tainted word.

I do think LLMs are/will be part of the tools needed for AGI, but alone, no, they aren’t processing what they’re being asked, so of course on anything more complex than their training they can go astray.