AIChE Journal, Год журнала: 2024, Номер 71(3)

Опубликована: Ноя. 30, 2024

Abstract Large language models (LLMs) are often criticized for lacking true “understanding” and the ability to “reason” with their knowledge, being seen merely as autocomplete engines. I suggest that this assessment might be missing a nuanced insight. LLMs do develop kind of empirical is “geometry”‐like, which adequate many applications. However, “geometric” understanding, built from incomplete noisy data, makes them unreliable, difficult generalize, in inference capabilities explanations. To overcome these limitations, should integrated an “algebraic” representation knowledge includes symbolic AI elements used expert systems. This integration aims create large (LKMs) grounded first principles can reason explain, mimicking human capabilities. Furthermore, we need conceptual breakthrough, such transformation Newtonian mechanics statistical mechanics, new science LLMs.

Язык: Английский