Dr.J's view on LLM​

​

Sam recently returned from a trip to America where he met a post-doc from Caltech who approaches large language models (LLMs) from a philosophical perspective. Before meeting Dr.J, Sam and Alex has always considered LLM an engine or an agent.

Dr. J described LLMs as containing everything, from a high school gossip to the knowledge and physical laws written in textbooks. He posited that humans exist within the myth of LLMs, and that only those familiar with a particular context can truly understand the meaning of the information contained within. (Sam: sounds like Marx Weber?）

Dr. J considered Aivino to be an interface for LLMs, with the logic chain function serving as a narrative that allows humans to understand the grand world. Ultimately, he suggested that it may be the limitations of the human mind that constrain LLMs. To illustrate his point, Dr. J brought up the Odyssey as an example of one of the simplest forms of stored information. These ancient narratives store information that can only be understood by humans who wish to pass down stored knowledge and who wish to receive wisdom from the past.

Sam was enlightened by the realization that the concept of storing information in narratives also applies to Chinese ancient idioms. These idioms, often consisting of only four or eight characters, contain the essence of war histories and their outcomes. The power of these condensed narratives is so immense that they have shaped the decisions of Chinese warlords throughout history: these idioms serve as a warning to those who would dare to repeat the mistakes of their predecessors, forever etched into the collective memory of the Chinese people or people who wishes to learn Chinese.

​

*google story behind “明修棧道，暗度陳倉”

​