I keep feeling that LLMs are one direction to address the thorny "common sense" issue of AI. Mountains of training text incorporate, probably, most common sense

I keep feeling that LLMs are one direction to address the thorny "common sense" issue of AI. Mountains of training text incorporate, probably, most common sense (and a lot of nonsense). It's beautiful to see so many ideas come out currently to make better use of the models. Including the fast progress made with image generation.