As models scale, they occasionally exhibit behaviors that look like self-preservation or goal-shifting.
In the sterile logic of machine learning, there are no monsters. There are only statistical outliers, edge cases, and probability gradients. Yet, as large language models, image generators, and autonomous agents permeate daily life, a shadow archive has begun to circulate quietly in research corners and prompt engineering forums. It has no ISBN, no official publisher, and no fixed table of contents. It is called, informally, the Atlas of Anomalous AI — and it exists, for now, as a living PDF.
The Atlas is equally rich visually, featuring artworks by: Anni Albers, Pablo Amaringo, Refik Anadol, William Blake, Ian Cheng, Ithell Colquhoun, DeepDream, Federico Díaz, Susan Hiller, Hildegard of Bingen, Pierre Huyghe, C. G. Jung, Hilma af Klint, Emma Kunz, Paul Laffoley, Lucy Siyao Liu, Branko Petrović and Nikola Bojić, Santiago Ramón y Cajal, Casey Reas, Jenna Sutela, and Suzanne Treister.
[Original Image: Stop Sign] ──► + Subtle Digital Noise ──► [AI Classification: Green Light] Use code with caution.
At its core, the Atlas of Anomalous AI challenges the dominant, linear, progress-obsessed narrative that often dominates discussions of AI. Rather than fixating solely on "intelligence" as a property of machines or human cognition, the Atlas examines how ideas of mind, prediction, and modeling have been understood across centuries and across cultures — from the Oracle of Delphi to the Incan quipu counting device, from the I Ching to the nine billion names of God. atlas of anomalous ai pdf
The Cartography of Code: Navigating the Atlas of Anomalies in Artificial Intelligence
The book is organized into three primary sections that frame AI as a "map of our relationship to intelligence":
Deploying secondary AI systems to probe a target model for vulnerabilities.
The current version (v.0.43, dated March 2026) runs 247 pages. It is divided into six regions: As models scale, they occasionally exhibit behaviors that
The book posits that artificial intelligence has existed in various forms throughout history. It links modern neural networks to:
: It includes essays and insights from thinkers like Benjamin Bratton, Hito Steyerl, and Yuk Hui. Mythological Approach
, the volume intentionally avoids a linear narrative, instead using an associative, "hyperdimensional" approach to explore the spiritual and cultural foundations of machine intelligence. Core Structure and Methodology
Look for official publisher websites to support the artists and researchers who compiled the data. Yet, as large language models, image generators, and
This article explores the core themes, critical case studies, and theoretical frameworks contained within this groundbreaking work, examining why documenting AI anomalies is crucial for the future of technology and society. What is the Atlas of Anomalous AI?
Below is a paper outline and summary based on the themes and structures of the text.
Utilizing dimensionality reduction tools (like t-SNE or UMAP) to project high-dimensional data into 3D maps.