Comment by saubeidl

LLMs are not the way to AGI and it's becoming clearer to even the most fanatic evangelists. It's not without reason GPT-5 was only a minor incremental update. I am convinced we have reached peak LLM.

There's no way a system of statistical predictions by itself can ever develop anything close to reasoning or intelligence. I think maybe there might be some potential there if we combined LLMs with formal reasoning systems - make the LLM nothing more than a fancy human language <-> formal logic translator, but even then, that translation layer will be inherently unreliable due to the nature of LLMs.