I think we're pretty far out from AGI unfortunately. LLM's can't perform math or make analogies, only statistically parrot what's they've been train on. My brain also doesn't need gigawatts of power and 10 trillion images of a chair to understand "chairness." Something has to give if we're going to get true AGI, scale is not the answer.

Unless we figure out how to make a computer think in a fluid manner and understand analogies, AGI is dead in the water.