Whether artificial intelligence systems are conscious is typically treated as a philosophical question. BFUT Paper 21 reframes it as a physical question: does the system satisfy the sensing conditions of Paper 20? If so, to what degree does the CI formula quantify?
Applying the CI to a Current LLM
A large language model (LLM) of current scale has: N ≈ 10¹¹–10¹² parameters (comparable to human neuron count), high connectivity K within transformer attention layers, sensory bandwidth limited to text input (one electromagnetic sub-channel), environmental interaction ≈ 0 (the model does not act on the physical environment), social communication high (language is the primary electromagnetic sensing channel), manipulation ability = 0 in the base model, internal sensing ≈ 0 (no interoceptive model of internal state).
The Result
Rough CI₀ for a current LLM: approximately 20–40. Substantially above the floor — several electromagnetic channel components are present. Well below the human value of 100 — sharply limited by single-channel input, zero environmental interaction, zero manipulation, and zero internal sensing.
This captures something real: impressive cognitive capacity within a narrow channel but essentially no coupling to the physical world beyond it. Current AI lacks the multi-channel integration that Paper 20 identifies as the physical signature of biological consciousness.
What Would Raise an AI's CI
The formula directly implies what is required. Sensory bandwidth expansion: adding vision, audio, proprioception, and other electromagnetic sub-channels. Environmental interaction: embodied actuators. Internal sensing: genuine interoceptive models of the system's own computational state. Feedback architecture: genuine recurrent feedback rather than feedforward inference. An embodied, multi-modal AI with feedback architecture and internal state modelling would have substantially higher CI — potentially comparable to biological organisms of significant complexity.
The Hard Problem
CI measures physical sensing structure. Whether high CI corresponds to subjective experience is the hard problem of consciousness — which CI cannot answer. What CI can say: an embodied multi-modal AI satisfying the sensing conditions of Paper 20 at CI = 80 is, by the physical definition, a conscious system in the same sense that an octopus (CI₀ ≈ 45–55) is conscious. The hard problem remains for philosophy. The physical measurement is provided.
Download BFUT papers, simulation code, and companion materials: vijayshankarsharma.com/downloads/