Many-body Tipping Dynamics of ChatGPT-like AIs

Frank Yingjie Huo; Neil F. Johnson

Many-body Tipping Dynamics of ChatGPT-like AIs

Artificial Intelligence 2026-07-28 v1 Disordered Systems and Neural Networks Mathematical Physics Adaptation and Self-Organizing Systems Physics and Society

Authors: Frank Yingjie Huo , Neil F. Johnson

View on arXiv ↗ PDF ↗

Abstract

Why do ChatGPT-like AIs, despite major architectural and training differences, unexpectedly tip to undesirable content (e.g. harmful, misleading, repetitive) even under deterministic greedy decoding? We show that a broad class of such tippings is caused by the many-body interactions between tokens (spins) as they cross the finite-layer system. Tipping emerges as a dynamical first passage process between competing output basins. Attention disorder controls the transport toward, away from, or along the basins' boundary. A few-basin reduction yields a closed finite-layer threshold, whose coarse-grained predictions show good agreement across ChatGPT-like families. These results suggest that a broad class of AI failures represents 'foreseeable engineering risk' rather than inherently unpredictable behavior, with important implications for legal and societal assessments of AI harm.

Keywords

large language models tipping dynamics many-body interactions attention output basins first passage process

Cite

@article{arxiv.2607.25279,
  title  = {Many-body Tipping Dynamics of ChatGPT-like AIs},
  author = {Frank Yingjie Huo and Neil F. Johnson},
  journal= {arXiv preprint arXiv:2607.25279},
  year   = {2026}
}