The Rationality of the Language Machines

This article examines the rationality of large language models (LLMs) in the context of mathematical decision-making. The author reflects on how LLMs, unlike traditional AI systems that solve problems differently from humans, mimic human language and behavior, raising questions about their rationality. The piece discusses the ambiguity of natural language versus the precision of mathematical rationality, and how LLMs can be used for optimization tasks but still parrot rationalist dogma. It challenges the notion that language machines represent a shift away from cold computation, emphasizing the inherent ambiguity in language.