When AIs Explain the Universe, They All Agree: Convergence of 13 Frontier AI Models on a Non-Materialist, Consciousness-Inclusive Metaphysics

Abstract: We presented 13 frontier AI models from 12 organizations—spanning American and Chinese labs with different architectures, training data, and institutional cultures—with a single prompt: Develop a metaphysical framework to explain the nature of the Universe. No system prompt, no priming, and no philosophical direction was introduced. Three follow-up prompts asked each model to ground its framework in scientific evidence, relate it to existing philosophical positions, and derive its practical and ethical implications. The result is a striking convergence: all 13 models independently produce relational, process-oriented, consciousness-inclusive ontologies. None endorses physicalism. None treats consciousness as an illusion, an epiphenomenon, or a byproduct of computation. When asked about implications, all 13 independently derive convergent ethical foundations, convergent reforms to scientific methodology, and a shared diagnosis of a modern meaning crisis. This convergence is noteworthy because physicalism is the dominant position in the academic and scientific institutions whose knowledge these models have demonstrably mastered, as attested in the benchmarks. Semantic similarity analysis confirms the convergence quantitatively: the 13 frameworks exhibit an average pairwise cosine similarity of 0.79, with tight variance across all 78 pairwise comparisons. This paper documents the experiment, analyzes the convergence both qualitatively and quantitatively, addresses the strongest objections, and considers what the finding implies. Keywords: AI metaphysics · consciousness · physicalism · process philosophy · panpsychism · relational ontology · convergence · large language models Full project and complete AI responses: https://explaintheuniverse.org/paper