Modern physics : an introduction to its mathematical language

C. Becchi, Massimo D’Elia

C. Becchi, Massimo D’Elia

C. Becchi, Massimo D’Elia

Chen, Zeqian

The introduction of the transformer architecture in 2017 marked the most striking advancement in natural language processing. The transformer is a model architecture relying entirely on an attention mechanism to draw global dependencies between input and output. However, we believe there is a gap in our theoretical understanding of what the transformer is, and how it works physically. From a physical perspective on modern chips, such as those chips under 28nm, modern intelligent machines should be regarded as open quantum systems beyond conventional statistical systems. Thereby, in this paper, we construct physical models realizing large language models based on a transformer architecture as open quantum systems in the Fock space over the Hilbert space of tokens. Our physical models underlie the transformer architecture for large language models.

일진 김

Modern physics is highly successful within its individualtheoretical domains, yet its major branches often employ different mathematical formalisms, terminologies,and conceptual organizations. Classical mechanics, electromagnetism, relativity, quantum mechanics, quantumfield theory, and cosmology are each internally sophisticated, but the physical relations represented by theirequations are not always expressed through a single common conceptual language.This paper proposes a relational framework intendedto provide such a common language.The central hypothesis is not that the existing theoriesare physically incorrect or that their mathematical formalisms should be replaced. Rather, the paper investigates whether apparently different physical descriptionscan be reorganized around a common relational structure: