Thanks. If I'm reading this right, the limiting factor on the intelligence of current LLMs is not the network size, nor training data (size/quality) but rather

Thanks. If I'm reading this right, the limiting factor on the intelligence of current LLMs is not the network size, nor training data (size/quality) but rather the architecture? Do we know of a better one for complex computations / "reasoning"?