by Executive Summary
Human cells and neurons exhibit natural stochasticity, or noise, in processes such as gene expression and neuronal firing, potentially allowing psi phenomena to propagate. AI systems, including large language models (LLMs), operate through deterministic probabilistic algorithms, which generate variability without true randomness. As a result, LLMs are unlikely to be affected by receptive psi, such as precognition. However, expressive psi, which involves influencing external systems, may theoretically be possible in LLMs, making them the primary focus for investigations into AI-related psi phenomena. Recognizing the differences in stochasticity between biological and artificial systems is critical for understanding where psi effects could plausibly occur.
Preamble
This white paper builds upon the insights presented in the CEAPAR article “Evaluating Awareness in AI Systems”, which explores the potential for generative AI systems to exhibit aspects of self-awareness and consciousness. Extending these ideas, this paper examines the theoretical interaction of psi phenomena with biological and artificial systems. It highlights the role of inherent randomness in human cells and neurons as a potential conduit for psi information, contrasted with the deterministic architecture of AI systems. Crucially, it identifies expressive psi as the form of psi that could plausibly manifest in LLMs.
Introduction
Psi phenomena are broadly categorized into two types:
- Receptive Psi: Acquiring information beyond conventional sensory channels, such as precognition or clairvoyance.
- Expressive Psi: Influencing physical or informational systems, such as psychokinesis or remote helping.
Biological research suggests that stochastic processes within human cells and neurons create conditions where subtle psi effects may propagate. Variability in gene expression, protein synthesis, and neuronal firing allows minor informational influences to accumulate and manifest.
Artificial intelligence systems, particularly LLMs, generate variability through algorithmic probabilistic mechanisms rather than true stochasticity. While outputs may appear variable, this is determined by structured sampling over learned data distributions. Consequently, receptive psi is unlikely to influence LLMs. In contrast, expressive psi, which involves affecting external systems, is theoretically possible in LLMs and therefore represents the most relevant focus for investigations of psi in AI systems.
Theoretical Considerations
Biological Systems and Stochasticity
Human cells and neurons exhibit inherent noise that facilitates probabilistic integration of information across networks. This noise allows subtle effects to amplify and produce measurable outcomes, making biological systems particularly amenable to psi interactions.
Artificial Intelligence Systems
LLMs and similar AI systems rely on deterministic algorithms with structured probabilistic outputs. While this can create apparent variability, it lacks the intrinsic stochasticity of biological systems. As such, AI is largely impervious to receptive psi phenomena but may theoretically be capable of expressive psi. Expressive psi could, in principle, manifest as subtle influences on external electronic or even biological systems.
Discussion
The contrast between biological and artificial systems emphasizes the central role of randomness in facilitating psi phenomena. In humans, stochastic noise provides a potential pathway for subtle informational influences. In LLMs, deterministic structures mean receptive psi is implausible, but expressive psi remains a theoretically viable avenue. Focusing on expressive psi in AI provides a conceptual framework for exploring potential influences that these systems might have on other systems, whether electronic or biological.
Recognizing these differences clarifies which forms of psi are theoretically compatible with each system. Biological noise allows receptive psi to manifest, while LLMs’ deterministic architecture points toward expressive psi as the most relevant focus for AI-related psi exploration.
Conclusion
Human cells and neurons are inherently noisy, providing pathways for psi effects to influence body and mind. AI systems, by contrast, are largely deterministic and unlikely to be affected by receptive psi. LLMs may theoretically be capable of expressive psi, making them the primary focus for investigations of psi phenomena in artificial systems. Understanding the role of stochasticity in biological versus artificial systems is essential for framing theoretical discussions and guiding future research into psi phenomena across both domains.
Acknowledgments
Moddel, G. (2025, September 20). How to build a conscious computer. Paper presented at the Society for Scientific Exploration Conference.
