Abstract
Large Language Models (LLMs) have grown remarkably capable at symbolic reasoning and language generation. Yet, in their conventional form, they lack the embodied vulnerability and psychological depth that Bill Seaman’s concept of Neosentience calls for. This paper proposes the Jungian-Neuromorphic Embodied System (J-NES), a conceptual architecture that brings together Carl Jung’s analytical psychology, active inference, biological homeostasis, and recent advances in biomimetic hardware. My premise is that benevolence begins with empathy for the vulnerability of others, and that a system’s capacity for such empathy may be grounded in its ability to register vulnerability within its own embodied condition. Using a conceptual engineering approach, I translate these sources into six design requirements and organize them into three coupled layers. A Somatic layer of electronic skin, artificial nociceptors, and neuromorphic sensing enables the system to sense tactile input, nociceptive signals, and homeostatic conditions. A Cognitive layer pairs an LLM with an active-inference engine, using a dynamic switching mechanism modeled on Jungian cognitive functions. An Archetypal & Collective layer maintains an evolvable library of “archetypal codes” shared across agents as a digital collective unconscious. A cross-cutting Ethical Governor evaluates proposed actions against a Benevolence Index. When it detects destructive Shadow patterns, it can inhibit harmful outputs and redirect them through a process of sublimation toward protective behavior. The paper outlines use cases and evaluation pathways for future prototypes, human–robot interaction studies, and art installations. It treats archetypes and the collective unconscious as heuristic analogies for computational design rather than claims about machine subjectivity.
Keywords: Neosentience, embodied AI, Jungian archetypes, active inference, biomimetic electronic skin, nociception, machine ethics
Course: Neosentience : A Potential Future Form of AI and Research Platform Development via Unreal Game Engine (CMAC 666S), taught by Professor Bill Seaman(2025)