The Architecture of Thought: Reflective Structures in Mental Constructs

Abstract

Natural language, the fundamental medium of human communication, has long been a subject of fascination for linguists, philosophers, and computer scientists. Its fluidity, complexity, and depth offer rich areas for exploration and analogy. One intriguing way to view natural language is through the lens of programming paradigms, in particular, the homoiconic concatenative paradigm. This paper couples existing work on the concept of synesthetic language of thought and expands upon it by suggesting that in addition to being synesthetic, it is also homoiconic, and largely concatenative in nature. I use this model of human thought to resolve a contention raised by van Rooij, et al. 2023.

A Scholar's Year in Review: Navigating the Convergence of AI, Economics, and Physics in 2023

“…Yet another paper on the topic of machine learning and language, "The Architecture of Thought: Reflective Structures in Mental Constructs" considers the limits to what machine learning can do in terms of mimicking human language and thought, in this paper, I argue against the narrative that transformer models are merely poor simulations [13]. The argument centers on the nature of both our communication languages and our internal language of thought.…”

“…Linguistic Diversity in the Age of AI As Guo et al [15] contemplate the decline of linguistic diversity amidst the recursive training of LLMs on synthetic data, they touch on a core concern of my research: the consequence of AI's language capabilities on the richness of human thought and expression [10,7,13]. Their insights not only serve as a critique to be heeded but also as a warning that the development of AI should not come at the cost of diminishing the human linguistic palette.…”