![[2602.17116] Epistemology of Generative AI: The Geometry of Knowing](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.17116] Epistemology of Generative AI: The Geometry of Knowing
Summary
This article explores the epistemological implications of generative AI, proposing a new framework for understanding knowledge production in high-dimensional spaces.
Why It Matters
As generative AI technologies advance, understanding their epistemic foundations is crucial for responsible integration into various fields. This paper addresses the gap in philosophical discourse regarding how generative models operate and their impact on knowledge creation, which is essential for educators, scientists, and policymakers.
Key Takeaways
- Generative AI challenges traditional notions of knowledge and its production.
- The paper introduces an Indexical Epistemology of High-Dimensional Spaces.
- It reconceptualizes generative models as navigators of learned manifolds.
- Navigational knowledge is proposed as a distinct mode of knowledge production.
- Understanding these concepts is vital for responsible AI integration in society.
Computer Science > Artificial Intelligence arXiv:2602.17116 (cs) [Submitted on 19 Feb 2026] Title:Epistemology of Generative AI: The Geometry of Knowing Authors:Ilya Levin View a PDF of the paper titled Epistemology of Generative AI: The Geometry of Knowing, by Ilya Levin View PDF Abstract:Generative AI presents an unprecedented challenge to our understanding of knowledge and its production. Unlike previous technological transformations, where engineering understanding preceded or accompanied deployment, generative AI operates through mechanisms whose epistemic character remains obscure, and without such understanding, its responsible integration into science, education, and institutional life cannot proceed on a principled basis. This paper argues that the missing account must begin with a paradigmatic break that has not yet received adequate philosophical attention. In the Turing-Shannon-von Neumann tradition, information enters the machine as encoded binary vectors, and semantics remains external to the process. Neural network architectures rupture this regime: symbolic input is instantly projected into a high-dimensional space where coordinates correspond to semantic parameters, transforming binary code into a position in a geometric space of meanings. It is this space that constitutes the active epistemic condition shaping generative production. Drawing on four structural properties of high-dimensional geometry concentration of measure, near-orthogonality, exponential...
