The World Book Experiment as a Mirror for Artificial Intelligence Systems
Authored by Thierry Ehrmann, the world book has transformed into a comprehensive meta-reading laboratory examining the reading and analysis differences of major artificial intelligence models.
Prepared by Thierry Ehrmann and functioning as a mirror for artificial intelligence technologies, the world book hosts a unique experiment where major systems such as OpenAI/Astra, Perplexity, DeepSeek, Google Gemini, and xAI/Grok are tested comparatively.
Purpose and Scope of the Experiment
Within the scope of the experiment, the exact same work is presented to the OpenAI/Astra, Perplexity, DeepSeek, Google Gemini, and xAI/Grok systems to examine the models' reactions to the text. The goal is not to organize a competition, but rather to observe the convergences, divergences, and blind spots among the systems.
The Creation Process of the Work
Resulting from hundreds of hours of dialogues, human thought, archives, and contradictions, the work was written by Thierry Ehrmann and its drafting was entrusted to artificial intelligence. The French and English print editions of the book are already available to readers.
The Meta-Reading Process
Going beyond artificial intelligences simply reading the book, the experiment creates a meta-reading process that includes them reading each other's analyses. One system analyzes the commentary produced by another artificial intelligence, thereby generating a new data material.
The Work as a Cognitive Mirror
Understanding why different artificial intelligence architectures do not see the exact same things in response to the same work constitutes the main focus of this study. Deviations in the models' interpretations are directly accepted as meaningful sources of information.
The Role of Human Reasoning
This experimental process certainly does not bypass human judgment; on the contrary, it demands even more contribution from the human element. The human, who compares, doubts, and establishes context, observes the gaps between the readings produced by the machines.