Its useful to view LLM revolution as the amazing 21st century refresh of the "Expert Systems." Like the expert systems, they too depend on "our" collective declarative and procedural knowledge--except that it can be written in Natural Language. #NeurIPS2025 #AIAphorisms
To elaborate, LLMs don't learn about the world, they learn how we view our world. While this limits LLMs to humanity's knowledge closure, it also makes them infinitely easier to align to us: Even their "cheating" behaviors imitate us at some level! �x.com/rao2z/status/1…C