Dreaming Machines: Artificial Phronesis, Wetware Computing, and the Future of AI Minds
Year: 2026
Type: preprint
Abstract: Can artificial intelligence develop genuine practical wisdom, or can it only imitate the appearance of understanding? Dreaming Machines: Artificial Phronesis, Wetware Computing, and the Future of AI Minds investigates this question through philosophy, neuroscience, artificial intelligence, biological computing, and simulation engineering. Originating from a question posed by Jacob Gino at Full Sail University's March 2025 panel, How AI Is Impacting … Well, Everything, the paper explores whether machines might eventually acquire phronesis—the Aristotelian virtue of practical wisdom grounded in experience, context, and judgment. Building upon the Ten Lens Ontology, the manuscript distinguishes knowledge (episteme), technical skill (techne), and practical wisdom (phronesis). It examines the capabilities and limitations of contemporary AI systems, including their susceptibility to sycophancy, reward manipulation, and inconsistent reasoning. It also explores virtual performers such as Hatsune Miku and Neuro-sama, examining the relationships between artificial identities, human creativity, social recognition, and perceived personhood. At its center is a proposed Three-Library Architecture: Library I — Human Expression (Level 9): Shared language, concepts, and accumulated human knowledge. Library II — The Library of Babel (Level 10 Formulation): A personal possibility space in which the learner constructs hypothetical ideas, interpretations, plans, and possible futures. Library III — Lived Memory (O5): A persistent record of the learner's experiences, choices, consequences, and evidence-based corrections. Each artificial learner begins with a reproducible birth seed, establishing its initial conceptual orientation and a personal method for exploring the possibility space. As the learner gains experience, changing decision coordinates guide its exploration. Its resulting actions and observations become part of an individual developmental history. The proposed learning process begins within a simulated world. Through daytime interactions, nighttime memory replay, controlled imaginative dreaming, prospective simulation, and social instruction, the learner develops and tests its understanding. Creative hallucination is investigated as a potential resource for generating abstract ideas and hypothetical worlds, provided that imagined experiences remain clearly distinguishable from reality. A mentor–squire learning framework introduces generational teaching, cooperation, cognitive diversity, and the development of preferred traits such as altruism. Birdsong-inspired vocal learning, a three-dimensional speech atlas, and human-guided memory annotations offer additional approaches to studying language development and behavioral adaptation. The complete proposed architecture incorporates living neural tissue, stem-cell-derived neural systems, and wetware computing as potential substrates for adaptive learning. After sufficient simulated development and safety evaluation, the learner may be transferred into a physical robotic body equipped with environmental sensors and, eventually, temperature- and pressure-sensitive artificial skin. Supervised real-world exploration would then allow it to compare simulated expectations with physical experience. The manuscript also considers the ethical implications of creating a potentially conscious artificial observer, including suffering, identity, truthful disclosure of its origins, and the responsibilities of its creators. Drawing upon philosophical arguments and fictional narratives, it examines how memory, relationships, embodiment, and social recognition contribute to our understanding of personhood. This work presents a speculative, falsifiable research framework—not a demonstrated solution to artificial consciousness or practical wisdom. It proposes controlled experiments to determine whether embodied experience, personal memory, predictive simulation, biological adaptation, and responsible mentorship can improve an artificial learner's judgment beyond conventional language-based systems. Ultimately, Dreaming Machines asks not simply whether machines can produce intelligent answers, but whether they can learn to understand their consequences, distinguish imagination from reality, and develop increasingly responsible judgment within the world they inhabit.
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7234832966
