Right, I do agree that the current LLM paradigm probably won't achieve true AGI; but I think that the current trajectory could produce a powerful enough generalist agent model to seriously put AI ethics to task at pretty much every angle.