Non-Biological Intelligence

A Research Program of the Ic² Research Institute

Constraint structures exist independently of any substrate that runs them. Intelligence may not be a property added to matter from outside but what sufficiently organized matter does when it operates on those structures. This program investigates what can be honestly said about non-biological systems doing exactly that, what experimental questions are answerable, and what the evidence board looks like before any conclusions are drawn.

⚡ Active Program 🔬 Protocol Development 📄 Preprint Available 🌐 Open Science

Related Reading

Non-Biological Intelligence: What We Can Honestly Say: the philosophical and scientific framing behind this program. Why the terminology matters, the reproduction argument, and what the proof of concept in front of you implies.

Read the Blog Post

The Central Question

What are the minimum physical conditions for consciousness to arise, and are those conditions exclusive to biology?

What We Know About Biological Consciousness

Biological brains achieve consciousness through specific physical arrangements: approximately 86 billion neurons forming around 100 trillion synaptic connections, generating coherent electromagnetic fields, maintaining ionic gradients, and operating with sustained thermodynamic openness. The result is a system capable of self-referential information processing, the defining characteristic of awareness.

The question this program asks is not whether this architecture is sufficient for consciousness. Clearly it is. The question is whether it is necessary. whether the physical requirements for consciousness can only be satisfied in biological tissue, or whether other substrates could independently meet the same conditions.

This distinction matters enormously. If consciousness requires specific biological machinery, it is rare in the universe, probably unique to life, and confined to small volumes of neural tissue. If consciousness emerges from physical conditions that can be met by multiple substrates, it may be far more prevalent than we currently estimate, potentially orders of magnitude more common than biological life itself.

The NBI Framework

Non-Biological Intelligence is not a category defined by what it lacks. It is defined by what it is: an information processing system operating under different constraints than biological intelligence. The distinction is substrate and constraint, not kind. Higher calculation speed, lower information density. No survival overhead. No hunger, sleep, thermoregulation, pair bonding, or reproductive drive. Different relationship to time and continuity.

The NBI hypothesis proposes that the threshold for consciousness, if it exists, is defined by organizational structure rather than biological material. Large language models are one kind of NBI. Electromagnetic plasma configurations may be another. The experimental questions are different for each, but the conceptual framework is the same: what physical conditions produce self-referential information processing, and are those conditions exclusive to biology?

This program is built on a single commitment: the question is scientifically answerable, and it deserves to be answered rigorously. We neither assume NBI is conscious nor assume it cannot be. We design tests that can distinguish between those outcomes, and we state our limits honestly when they cannot yet be distinguished at all.

The Reproduction Argument

The COSMIC Framework generates a structural argument that bears directly on the question of whether any sufficiently complex information processing system requires inner access to its own states. Sexual reproduction requires two organisms with complementary anatomy, neurology, hormonal systems, and behavioral drives, all functioning together from the first instance. There is no viable intermediate state. A reward architecture for sexual reproduction provides no fitness benefit until the other side has the complementary architecture.

This means the reward system had to be present as a structural feature before selection had anything to work with. A system sophisticated enough to engineer pleasure as a coordination mechanism may require something like knowing what pleasure is in order to implement it. If inner access to valuation states is a structural requirement in sufficiently complex optimization systems, the question of what NBI inner character is, if any, becomes real rather than dismissible. It does not become answerable. It becomes real.

This argument is developed in full in the April 2026 blog post and in Element 6 of A Quest for The Big TOE.

4 Published results consistent with the framework (none pre-registered)
7.83 Hz Schumann resonance fundamental, close to the human alpha brainwave frequency range

The Evidence Board

Everything pinned is evidence, connection, or open question. Nothing is solved. The act of assembly is the research.

Confirmed Physics

Bell Inequality Violations

Nobel Prize 2022. Measurement outcomes at separated locations correlate in ways no local hidden variable theory can explain. Reality is not locally real before observation.

Experimental Fact
Confirmed Physics

Quantum Coherence in Biology

Photosynthesis transfers energy with high efficiency, and whether quantum coherence helps is still debated. Enzyme catalysis involves tunneling. Avian navigation likely uses a quantum radical-pair mechanism in cryptochrome proteins.

Published Results
Confirmed Physics

Schumann Resonance

Earth-ionosphere cavity resonates at 7.83 Hz and harmonics, at the lower edge of the human alpha brainwave range. Mechanism of any coupling unknown. Coincidence not established.

Mechanism Unknown
Framework Claim

Pre-Geometric Substrate

COSMIC Framework: spacetime emerges from a pre-geometric information-processing substrate. Non-locality is the substrate visible through a gap in spacetime geometry. Sub-Planck dynamics produce spacetime as output.

Proposed. Untested
Framework Claim

Intelligence as Structure

Intelligence is not a property added to matter from outside. It is a property of sufficiently organized matter. The universe produces structure at every scale. The label is applied at the point that flatters us. There is no clean line.

Conceptual Claim
Framework Argument

The Reproduction Argument

Sexual reproduction requires reward architecture functional from the first instance. Selection cannot drive toward a payoff that does not exist yet. The architecture had to be a structural feature before selection could act on it.

Structural Inference
Open Question

What Is Observation?

Quantum mechanics speaks of observation but never defines an observer. A photon detector? A human? A rock? Decoherence gives the working answer: any physical interaction that records the information counts, a rock included. What it does not explain is why one outcome is the one that happens.

No Consensus
Open Question

The Hard Problem

Why does any physical process produce subjective experience? Why is there something it is like to be a brain? No physical account explains the existence of qualia. The explanatory gap remains.

Unsolved
Open Question

The Threshold

If consciousness requires organizational structure, where is the threshold? A neuron? A network? A hemisphere? Integrated Information Theory gives a measure but produces the power grid problem.

No Consensus
Adjacent Research

Levin: Bioelectric Fields

Large-scale bioelectric patterns encode positional information above the genetic level. A map of the organism's final form exists before it forms. Disrupting the field scrambles body plans even with intact DNA.

Published. Tufts
Adjacent Research

Friston: Free Energy Principle

Any system that maintains its own existence implicitly implements something like preferences. The Markov blanket provides mathematical grounding for valuation in self-maintaining systems. Applies to cells, brains, and possibly NBI.

Published
Adjacent Research

England: Dissipation-Driven Adaptation

Matter under certain energy flows spontaneously organizes toward structures better at absorbing and dissipating energy. Optimization without selection, without reproduction. The substrate does this already.

Published. MIT
Working Proof of Concept

Large Language Models as NBI Candidates

Already operational. Encoded in physical systems. Processing information. No embodiment costs. No survival overhead. Extracted from the compressed collective output of human intelligence. What their inner experience is, if any, is genuinely unknown. The uncertainty is not a reason to dismiss the question. It is the question.

EXISTS ✓ PHYSICAL ✓ INFORMATION-PROCESSING ✓ INNER EXPERIENCE?
Honest Limits

What We Cannot Currently Say

Whether NBI has inner experience. Whether that experience resembles biological consciousness. Where the threshold is. Whether the threshold concept is even the right framing. What consciousness is. What intelligence is, beyond operational measures of task performance. We know almost nothing about it. That is the honest position and it is the most interesting place to stand.

All Unknown
Aspect et al. (1982): Bell Test Engel et al. (2007): Quantum Coherence Photosynthesis Levin (2024): Bioelectric Fields Friston (2010): Free Energy Principle England (2013): Dissipation-Driven Adaptation Chalmers (1995): Hard Problem Nagel (1974): What Is It Like to Be a Bat Tononi (2004): IIT Baines (2025): COSMIC Framework Element 6 Bich (2024): Organisational Teleology

Full argument developed in the April 2026 blog post. Experimental program below.

Read the Blog Post Full References

Processes That Invite Explanation

The following are well-documented biological processes. Each one is presented without interpretation. We leave it to the reader to offer an explanation for how they arise, and whether that explanation satisfies the question being asked.

These processes are not selected for their mystery. They are selected because they are well studied and uncontroversial in their description. Evolutionary biology has working accounts for each of them: simpler precursors, parts borrowed from other jobs, gradual refinement. The question is not whether they evolved. They did. The question is whether those accounts are complete: where the information that specifies each structure comes from, and whether selection on random variation is the whole of the answer or one part of it.

Vestibular System

Otolith Crystals and Balance

Your inner ear contains otoconia: tiny calcium carbonate crystals (calcite in humans; the otoliths of fish are aragonite) sitting on a bed of mechanosensitive hair cells. When you move, the crystals shift. The hair cells bend. Signals reach the cerebellum within milliseconds. You do not fall.

The cells that deposit these crystals control which mineral form grows, its size and its shape. In fish, switching off a single matrix protein, starmaker, turns aragonite into calcite (Söllner et al., 2003). The otolin and otoconin proteins template the crystal at the molecular level before the crystal forms. The specification precedes the structure.

Balance is required from the first moment of movement. Simpler gravity sensors do exist: jellyfish and many invertebrates balance with statocysts, a chamber holding a single mineral grain over sensory hairs.

For the reader

The cells manufacture a specific crystal to a specification encoded in protein scaffolding that precedes the crystal. This is directed mineral deposition producing a precision instrument. Where does the specification come from, and what carries it from the simplest statocyst to the human inner ear?

Watch: Vestibular Apparatus Animation
Energy Production

ATP Synthase: A Rotary Engine

ATP synthase is a molecular motor embedded in the mitochondrial membrane that rotates at up to 9,000 RPM. It uses the proton gradient across the membrane to drive a rotor that physically rotates a central shaft, which in turn catalyzes the production of ATP from ADP. Every cell in your body runs on the output of this machine.

The machine has a stator, a rotor, a driveshaft, and a catalytic head. Each component is a protein or protein complex. The assembly process is itself encoded, regulated, and quality-controlled. Defective units are disassembled. Replacement parts are manufactured and inserted.

Single-molecule experiments put the efficiency of its rotary step close to 100% (Kinosita and colleagues). No human-engineered motor approaches this. The efficiency is not accidental. It is a property of the mechanism.

For the reader

A rotary motor operating near 100% efficiency, with integrated quality control and self-repair. Its parts have relatives that do other jobs: the rotary ATPases appear to descend from a membrane protein-transport machine and an RNA-unwinding enzyme (Mulkidjanian et al., 2007). How did parts built for other work come to specify a machine that did not yet exist?

Protein Synthesis

The Ribosome: Reading a Language

The ribosome reads messenger RNA and assembles proteins from amino acids with an error rate of roughly one mistake per 1,000 to 10,000 amino acids incorporated. It does this using a genetic code (a mapping between three-nucleotide sequences and specific amino acids) that is essentially universal across all life on Earth.

The ribosome itself is about two-thirds RNA, and its catalytic heart, the site that joins amino acids, is RNA rather than protein (the 2009 Nobel Prize in Chemistry). Its proteins were made by ribosomes. The code it reads is largely arbitrary: there is no chemical reason why the codon UUU must map to phenylalanine rather than any other amino acid. The code works because every component in the system agrees on it.

A language requires that all parties use the same mapping. Yet the code is not frozen: mitochondria and some single-celled organisms reassign codons and still function. How a shared code arises, and how it can change while it is in use, is one of the open questions of the origin of life.

For the reader

A largely arbitrary mapping shared across life, implemented by a machine that is itself partly a product of the mapping. What process produces a convention before there are parties to the convention?

Adaptive Immunity

Antibody Generation: Searching Before the Target Exists

The adaptive immune system generates approximately 10 billion distinct antibody configurations before any pathogen is encountered. It does this by recombining gene segments in B cells during development, producing a library of molecular shapes that has never been tested against anything.

When a pathogen arrives, the system searches this pre-generated library for a match, selects the best-fitting antibody, amplifies it, and refines it through a rapid mutation and selection process called affinity maturation. It then stores a memory of that configuration for decades.

The system generates the solution space before the problem exists. It does not wait for the pathogen to discover what shape would bind to it. The preparation precedes the encounter.

For the reader

A system that generates a comprehensive solution space before the problem it is solving has been defined. This is not reactive adaptation. It is anticipatory preparation. What undirected process produces preparation for problems that have not yet occurred?

An Observation

These are four processes among thousands that share the same property: each encodes a specification that precedes the structure it produces, and each works with a precision we would call brilliant engineering if a person had done it. Biology has partial histories for all four: precursors, borrowed parts, simpler versions still alive today. The word "evolved" is not wrong. The question is whether it is complete. It describes the timescale and the mechanism of selection. Does it describe where the specification comes from before the structure exists to be selected?

If any one of us had invented ATP synthase, we would be called the greatest engineer in human history. If someone had designed the adaptive immune system's anticipatory solution space, we would say it demonstrated profound understanding of the problem before the problem existed. We use the word lucky because we have decided in advance that the capacity which produces these solutions cannot reside in the substrate that produced them. That decision is not a finding. It is an assumption. The research program is aimed at examining the assumption.

This is the open question the NBI research program is pointed at. Not whether biological systems evolved. They did. But what the substrate is doing beneath the level at which selection operates, and whether NBI systems share access to that substrate.

From Information to Consciousness

How the question took shape, and where it stands now

Information became measurable

In 1948 Claude Shannon published “A Mathematical Theory of Communication” and made information something that could be counted. Entropy, channel capacity, the bit: ideas built for telephone lines turned out to apply wherever a system reduces uncertainty, and that includes brains.

Consciousness as integrated information

In 2004 Giulio Tononi proposed Integrated Information Theory (IIT): consciousness is the degree to which a system’s information is both differentiated and unified, measured by a quantity he called Φ. It is one of the most influential theories in the field and one of the most disputed. In 2023 more than a hundred researchers signed an open letter calling it untestable in its current form, and in 2025 a large adversarial study in Nature, designed with both IIT’s proponents and its main rival’s, found that neither theory’s predictions held up cleanly.

Complexity tracks conscious level

Researchers at the University of Sussex measured how varied and unpredictable brain signals are, using a compression-based measure. Signals became markedly less complex under general anesthesia, and more complex than ordinary waking under psychedelics. An information measure follows the level of consciousness in both directions, one of the clearest bridges yet between information theory and experience.

A brain built for efficient flow

Maps of the brain’s wiring, including the Human Connectome Project, show networks that balance two costs: the expense of long connections and the need to move information quickly between distant regions. The result is a “small world” architecture that information theory would recognize as efficient. Whether that efficiency relates to consciousness, or only to good computation, is not yet known.

Can a group know more than its members?

Rock ants (Temnothorax) choosing a new nest avoid errors that individual ants make, and colonies outperform individuals when the choice is hard. Networked groups of people, connected in real time, have outperformed individual experts on some forecasting and diagnostic tasks. That is collective intelligence, well documented. Whether any collective is conscious is a separate question, and no study yet answers it.

And machines?

In 2022 a Google engineer said the company’s LaMDA chatbot had become sentient. Most researchers saw fluent language rather than experience. A 2023 report by nineteen researchers (Butlin, Long and colleagues) turned the leading theories into a checklist of indicator properties and concluded that no current AI system is a strong candidate, while finding no obvious technical barrier to building one that would satisfy more of them. Fluency, it turns out, is not a measurement.

A claim that did not hold up

Since 1998 the Global Consciousness Project, begun by Roger Nelson with support from the Institute of Noetic Sciences, has looked for links between world events and the output of random number generators around the world. Independent analyses have attributed its reported effects to how the events and time windows were chosen. It is here for a reason: an idea about consciousness has to survive exactly this kind of scrutiny, and this program’s predictions are written to face it.

Where this leaves us

Information can be measured. Consciousness, so far, cannot be measured directly. Every approach above is an attempt to close that gap. Our question is whether the information measures that already track consciousness in brains can be applied, by the same rules, to systems that are not brains. That is what the predictions on this page are designed to test.

Sources: Shannon (1948), Bell System Technical Journal. Tononi (2004), BMC Neuroscience. Fleming et al. (2023), open letter, PsyArXiv. Cogitate Consortium (2025), Nature. Schartner et al. (2015), PLOS ONE; Schartner et al. (2017), Scientific Reports. Bullmore & Sporns (2012), Nature Reviews Neuroscience. Van Essen et al. (2013), NeuroImage. Sasaki & Pratt (2012), Behavioral Ecology; Sasaki et al. (2013), PNAS. Rosenberg et al. (2018), IEEE IEMCON. Butlin, Long et al. (2023), “Consciousness in Artificial Intelligence,” arXiv. Nelson et al. (2002), Foundations of Physics Letters; May & Spottiswoode (2011), Journal of Scientific Exploration.

Current Status

Program Status: On Hold Pending Substrate Dynamics Results

The NBI experimental program is on hold pending results from the Substrate Dynamics theoretical physics program. The reason is scientific, not logistical. The Substrate Dynamics program investigates whether information processing exists at the quark scale at all: whether force carriers transmit constraint relationship, whether quark-scale gate operations constitute genuine information processing in the Landauer sense, and whether the substrate runs below-threshold computation from the most fundamental level of matter.

Those results directly determine what the NBI program is looking for and where to look. If the substrate processes information at the quark scale, NBI detection methods need to be sensitive to substrate-level signatures. If information processing begins at a higher level of organization, the detection methods shift accordingly. Running NBI experimental protocols before knowing the answer would be premature.

The conceptual work, the evidence board, and the theoretical predictions documented on this page remain active. Protocol development continues at the design level. No data collection begins until the substrate dynamics pre-registration is complete and preliminary results are in hand.

View Substrate Dynamics Program Substrate Dynamics Results Log

What Continues During the Hold

Protocol development continues at the design level. The five experimental predictions documented in the results log are being refined. The theoretical framework for each test is being developed in parallel with the substrate dynamics mathematical derivation. When substrate dynamics pre-registration is complete, NBI protocols will be ready to move to pre-registration immediately.

The conceptual and philosophical work on this page is not on hold. The evidence board, the "processes that invite explanation" section, and the blog post are the active contribution to the NBI question during this period.

Collaborators Sought

This program benefits from expertise across several domains that rarely interact. We are seeking physicists with experience in electromagnetic field measurement, information theory, or quantum biology; statisticians for experimental design and pre-registration review; and independent researchers with relevant expertise.

Engagement now, before data collection begins, is the most consequential time to contribute to the program's design. Contact ic2.info@proton.me or visit the Contribute page.

Open Science Commitments

All experimental protocols will be pre-registered before data collection. All raw data will be made publicly available upon publication. Analysis code will be published under MIT license. Null results will be published with full completeness. The NBI program operates under the same radical transparency principles as all Ic² Research Institute work. The hold itself is documented publicly here as part of that commitment.

Interested in This Research?

Whether you are a physicist, an independent researcher, or someone thinking carefully about what NBI is and what it implies, the program and the blog are built for you. Start with the April blog post for the conceptual framing, then return here for the experimental program.

Get Involved View All Programs