The Brain-Internet Hypothesis

ABSTRACT

This paper begins with an everyday perceptual phenomenon — the striking change in subjective perceived loudness when switching back to traditional in-ear headphones after prolonged use of bone-conduction headphones — and progressively interrogates the mechanisms, boundaries, and costs of human biological adaptability, ultimately converging on a layered technology-path assessment. Layer One (Strongly Established): The mass adoption of invasive BCI is subject to structural constraints imposed by biocompatibility. Layer Two (Strong Hypothesis): The more promising route for mass-market BCI may not be implanting electrodes but rather using extracorporeal physical fields to modulate the brain’s own ion-channel machinery. Layer Three (Conditional Long-Range Derivation): If non-invasive BCI simultaneously achieves breakthroughs in readout, write-in, cross-brain semantic translation, and safety-loop closure across four interfaces, the Brain-Internet will become a potential civilization-scale information infrastructure.

The paper spans fourteen chapters. The first half of the reasoning is grounded in existing empirical research; the second half proposes testable technological hypotheses and long-range derivations, concluding with a falsifiable roadmap that defines verification criteria and refutation conditions for each propositional layer.

Starting Point: A Pair of Headphones and the Question of Biological Adaptability

1.1 Observed Phenomenon

After prolonged use of Apple in-ear headphones, the author switched to bone-conduction headphones. After several months of continuous use, upon switching back to the Apple earbuds, the author found that subjective perceived loudness had dropped markedly — same device, same volume settings. The hardware parameters had not changed — what had changed was the user’s own perceptual baseline.

It should be noted that personal headphone experience cannot directly prove that bone conduction causes long-term central recalibration. Contributing factors may also include differences in earbud seal, frequency-response curves, and device output power. This experience serves in this paper as a problem trigger rather than the evidentiary backbone.

1.2 Mechanism Analysis: Central Auditory Gain Regulation

Bone-conduction headphones bypass the ear canal and eardrum, stimulating the cochlea directly via cranial-bone vibration. After prolonged use, the auditory cortex’s loudness “baseline calibration” shifts. Experimental evidence shows that two weeks of monaural earplug use lowers acoustic reflex thresholds by an average of 8 to 10 dB — the auditory system adapts according to the principle of “homeostatic plasticity.” This corresponds to physiological acclimatization in Paul T. Baker’s (1965) four-category framework of human biological adaptation.

1.3 Supplementary Conditions on Reversibility

Short-term adaptation is usually reversible, but animal experiments have found that temporary threshold shifts can lead to persistent supra-threshold functional changes. Sensory recalibration operates across multiple timescales: rapid saturation but slow decay, suggesting a superposition of transient and sustained adaptation.

A Century of Research on Human Biological Adaptability

2.1 Foundation Period (1847–1950s)

Bergmann (1847) proposed that larger body sizes in cold climates aid heat conservation; Allen (1877) proposed that species in cold climates trend toward shortened limbs. Darwin’s On the Origin of Species (1859) established the natural-selection framework. Boas’s immigrant-descendant study (1910) first demonstrated that human body types can undergo significant plastic changes within a single generation.

2.2 Disciplinary Formation (1958–1974): Baker’s Four Categories and the IBP

The Human Adaptability project of the International Biological Programme (IBP) spawned the first long-term studies of extreme-environment adaptation strategies. Paul T. Baker and Weiner (1966) co-edited The Biology of Human Adaptability, proposing a four-category framework with lasting influence on subsequent research: physiological acclimatization (short-term, reversible adjustments to sustained environmental changes), developmental adaptation (irreversible phenotypic changes produced during growth due to environmental influence), genetic adaptation (cross-generational changes in gene frequency driven by natural selection), and psychological adjustment (behavioral and cognitive strategy modification).

This classification is critical because the headphone experience in Chapter One falls under “physiological acclimatization,” while neural network reorganization following BCI implantation may simultaneously involve changes at the levels of both “physiological acclimatization” and “developmental adaptation.”

2.3 Frisancho and the Developmental Adaptation Hypothesis

From 1969 onward, Roberto Frisancho published a series of studies on human developmental responses in high-altitude hypoxic environments. He proposed and validated a key hypothesis: an important source of adult phenotypic diversity lies in adaptive developmental responses to the environment. People who grew up in the Andean highlands possess greater lung capacity and higher hemoglobin levels than those who migrated to the highlands as adults — an irreversible adaptation.

The implication for BCI is this: if a person begins using a brain-computer interface during critical neurodevelopmental windows (e.g., childhood and adolescence), their brain may undergo deeper, more irreversible structural adaptations than those of adult users.

2.4 The Genomic Revolution and Contemporary Frontiers (2000–Present)

The convergent evolution of lactose tolerance (Tishkoff et al., 2007), the Tibetan EPAS1 gene derived from Denisovan DNA introgression (2014, Nature), and epigenetic mechanisms revealing cross-generational environmental transmission — these breakthroughs demonstrate that human adaptability is far more profound and rapid than traditionally assumed.

Brain-Computer Interfaces: The Ultimate Testing Ground for Biological Adaptability

3.1 Bidirectional Drift and Co-adaptation

Invasive BCI faces bidirectional drift: the biological side (neural tuning drift, cognitive strategy changes, tissue responses) and the physical side (electrode micro-motion, material degradation, impedance changes) are both in flux simultaneously. Researchers have proposed a “co-adaptation” mathematical framework.

3.2 The Four-Layer Cascade of Biological Adversarial Response

Layer One: Immune Shielding

Microglia and astrocytes form a glial scar that increases electrical impedance. Based on long-term recording data from Utah arrays, nearly half of implants show significant signal attenuation within six months (failure rates vary across electrode platforms).

Layer Two: Oxidative Stress

Implantation triggers excessive ROS production, perpetuating the foreign-body response, promoting neuronal death, and accelerating electrode corrosion.

Layer Three: Local Neurodegeneration

Chronic inflammation induces progressive neuronal and dendritic loss in the electrode vicinity.

Layer Four: Accelerated Pathological Aging

3.3 The Adversarial Vicious Cycle

3.4 The Disconnect Between Tech Narratives and Biological Reality

Historical Yardstick: Commercialization Cycles of Frontier Implant Technologies

Looking at several representative implant technologies, the journey from experiment to safety validation and mass commercialization is typically measured in decades, with common cycles of roughly 20 to 50 years. These historical cycles provide a lower-bound reference rather than precise predictions — the CNS immune specificity of BCI and the translation problem of the Brain-Internet have no historical precedent, and actual timelines may be longer.

The Physical Lifespan Dilemma of Electrodes

37°C constant temperature, saline body fluids, active immune cells, continuous micro-motion — the interior of the human body is an extremely hostile environment for electronic devices. Hermetic sealing (titanium casing) is the foundation of pacemaker longevity, but BCI electrodes must be directly exposed to brain tissue and cannot be encased in titanium shells. Polymer coatings face a degradation chain of moisture infiltration → crack propagation → substrate corrosion.

The structural contradiction: recording signals requires electrodes to directly contact tissue (no sealing possible), while preventing corrosion requires isolation from body fluids (sealing essential). This contradiction is irreconcilable within the invasive paradigm.

The Deep Signal-Type Mismatch — Electron Flow ≠ Ionic Electrochemical Flow

In 1791, Galvani demonstrated the existence of “bioelectricity.” In 1850, Helmholtz measured frog nerve conduction velocity at roughly 30–40 m/s — millions of times slower than electrical current in wire — because the carriers in nerves are ions, not electrons. In 1952, the Hodgkin-Huxley model precisely described the ion-channel action potential mechanism (Nobel Prize).

Turning Point: Can Humanity Replicate Ionic Electrochemical Flow?

Current capability boundary: can replicate ionic currents and discharge patterns; cannot replicate scale (86 billion neurons), speed (microsecond-level switching), integration density, or adaptiveness.

Paradigm Shift — From “Implanting Devices” to “Extracorporeal Activation of the Brain’s Own Ion Channels”

8.1 Core Insight

There is no need to “inject” ionic signals into the brain — the brain already possesses a complete ion-channel infrastructure. All that is needed is to send physical signals from outside the body that penetrate the skull and activate the brain’s own mechanosensitive ion channels.

8.2 The Ultrasound Route: Key Experimental Timeline

8.3 This Route Bypasses the Core Dilemmas of Invasive BCI

By utilizing the brain’s own ion channels, the underlying stimulation mechanism is closer to the nervous system’s native carriers. However, it must be made clear: ion-channel opening and closing is only a low-level physical event. True neural encoding also depends on cell-type selectivity, activation timing, local network state, neuromodulatory context, and oscillation phase. The gap to semantic-level neural write-in remains vast.

8.4 An Honest Safety Assessment

Under some experimental parameters, good safety profiles have been demonstrated. However, the safety boundaries of long-term, repeated, high-precision, multi-region stimulation still require systematic validation. The complete safety mapping of ultrasound parameters, target region selection, individual variability, thermal effects, and cavitation risk is far from finished.

8.5 Overview of Other Non-Invasive Routes

Ultrasound is not the only route. Temporal interference electrical stimulation (tTIS) achieves deep-brain modulation by converging two different-frequency electric fields at depth to create low-frequency interference. Transcranial magnetic stimulation (TMS) has already received FDA approval for depression and other conditions. Magnetothermal/photoacoustic stimulation uses nanoparticles as in-body relays. This paper focuses on ultrasound because of its unique combination of advantages in penetration depth, spatial precision, and direct ion-channel activation — but the existence of alternative routes means that even if ultrasound encounters bottlenecks, the broader non-invasive paradigm direction can still hold.

The Readout Bottleneck — Why Write-In Is Not All the Brain-Internet Needs

The Brain-Internet requires bidirectional high-bandwidth interaction. If high-dimensional brain states cannot be non-invasively “read out,” the entire chain cannot close.

The Four Missing Interfaces of the Brain-Internet

10.1 Five-Level Classification of Write-In Precision

The Brain-Internet may not require single-cell precision. The brain itself is a population-coding system, and it may only need to achieve reproducible neural-ensemble pattern-level modulation — at a spatial scale that might be cortical columns or functional manifolds rather than individual cells.

10.2 The Hard Constraint of Metabolic Energy

The human brain consumes approximately 20W (from complete oxidation of glucose), of which cortical computation consumes only about 0.1–0.2 ATP-watts; the communication cost is 20 to 35 times the computation cost. The brain, comprising 2% of body weight, consumes approximately 20% of the body’s total oxygen — evolution has optimized its metabolic efficiency to the limit.

Upscaling and Downscaling: The Information-Theoretic Foundation for Direct Will-to-World Connection

11.1 The Downscaling Chain

Downscaling loss is one of the fundamental constraints on human communication and creative efficiency. But communication difficulties also arise from interest conflicts, linguistic ambiguity, cultural differences, emotional defenses, and motivational misalignment — non-information-theoretic factors. Downscaling is not the only bottleneck, but it is the most fundamental physical one.

11.2 Bandwidth Quantification

11.3 Dimension-Preserving Transmission (Hypothesis)

If non-invasive BCI achieves sufficient bandwidth on both readout and write-in sides, high-dimensional intent could be directly mapped to a machine’s multiple parallel degrees of freedom, eliminating second-scale cognitive encoding latency.

“Dimension-preserving transmission” is currently an inspirational metaphor rather than a rigorous information-theoretic model. To become an actionable engineering concept, it requires defining intent-space dimensionality, output-channel capacity, BCI bandwidth, neural-representation reconstruction error, inter-individual mapping loss, and semantic fidelity. It must also account for the metabolic energy constraint raised in Chapter Ten — the brain may not possess sufficient ATP headroom to support full-dimensional parallel information exchange.

The Neural Tower of Babel: The Cross-Brain Semantic Alignment Problem

Among the four missing interfaces, the translation interface may be the most fundamental. Even if readout, write-in, and safety are all solved, if semantic alignment between two heterogeneous brains cannot be established, the Brain-Internet can only transmit noise.

12.1 The Nature of the Problem

Every person’s brain structure, developmental history, sensory experience, language system, and emotional encoding are different. The neural activation pattern when Brain A thinks “apple” is entirely different from Brain B’s. Humans have no universal “neural machine code.”

12.2 Existing Scientific Foundation: Hyperalignment

Cross-individual neural representation alignment is not pure fantasy. Since Haxby et al. proposed Hyperalignment in 2011, 15 years of research have accumulated in this field. Hyperalignment uses Procrustes transformations to map different individuals’ functional brain data into a common high-dimensional space, discovering that the fundamental property preserved across brains is informational content rather than local feature-level functional properties. Subsequent methods include Shared Response Modeling (SRM, Chen et al., 2015) and Optimal Transport (Bazeille et al., 2019).

These methods have proven that under controlled experimental conditions (e.g., watching the same movie), representational spaces of different individuals’ ventral temporal cortex can be aligned via rotation, significantly improving cross-subject decoding accuracy. But the gap to real-time, whole-brain, semantic-level Brain-Internet translation remains enormous.

12.3 Possible “Handshake Protocol” Directions

Direction One: Representation Calibration Through Shared Sensory Stimuli

Have two brains simultaneously receive identical external stimuli, record each brain’s response patterns, and build a statistical mapping — analogous to training machine translation with a “parallel corpus.”

Direction Two: Neural Manifold Alignment Algorithms

If different brains exhibit common geometric structures in their neural trajectories for similar cognitive tasks, cross-brain mapping can be achieved through manifold alignment.

Direction Three: Progressive Brain-to-Brain Adaptation Training

Leveraging the very human biological adaptability that is this paper’s starting point — allowing two brains to gradually “learn” each other’s encoding style through prolonged low-bandwidth connectivity.

12.4 The Manifold Drift Paradox

12.5 Convergence or Bilingualism?

The endgame of progressive brain-to-brain adaptation training has two possibilities: convergence of the two brains’ representations (forming a shared super-individual network), or — like bilinguals — retaining independent selves while developing new cortical mechanisms specialized for processing “foreign brain signals.” Evidence from bilingual neuroscience favors the latter — bilinguals do not lose their native-language cortical regions but instead develop additional prefrontal control mechanisms. However, this is entirely speculative.

12.6 An Honest Assessment

All of the above directions are at an extremely early conceptual stage. There is currently no experimental evidence that semantic-level non-verbal direct communication between two human brains is possible. If the Brain-Internet has an “impossibility theorem” waiting to be discovered, it is most likely to be found here.

The Brain-Internet Hypothesis

13.1 Three Propositional Layers

13.2 Existing Proof-of-Principle: Brain-to-Brain Interface Experiments

The Brain-Internet is not pure fantasy — its 0.001% version already exists. In 2014, Rao et al. achieved the first non-invasive brain-to-brain interface between two human subjects via an EEG-TMS link, decoding the sender’s motor intent and delivering it to the receiver’s motor cortex. In 2019, Jiang, Stocco et al. published BrainNet in Scientific Reports — the first multi-person non-invasive brain-to-brain interface, enabling three subjects to collaboratively complete a Tetris-like task through direct brain-to-brain communication.

These experiments had extremely low bandwidth (essentially binary decision signals), but they demonstrated that non-invasive brain-to-brain information transfer is feasible in principle. The distance from binary signals to semantic-level sharing is precisely the four missing bridges defined in Chapters Nine through Twelve of this paper.

13.3 The Dimensional Loss of Current Human Inter-Brain Communication

Every act of interpersonal communication is two downscalings plus two upscalings. One of the greatest information losses in human civilization occurs not between humans and machines, but between humans and humans.

13.4 Timeline Estimation

The above timelines are not predictions but scenario estimates based on implant-technology commercialization cycles and current readout-side bottlenecks. The Brain-Internet is additionally constrained by ethics, medical regulation, neural safety, and societal acceptance — factors for which no historical precedent exists.

A Falsifiable Roadmap

The value of a hypothesis lies not in whether it is correct, but in whether it can be clearly supported or refuted. Below are the verification criteria and refutation conditions for the paper’s three propositional layers.

From a Pair of Headphones to the Next Leap of Civilization

This paper’s chain of reasoning began with a tiny personal perceptual difference — bone-conduction headphones altered the auditory baseline. All subsequent derivations revolve around a single core question: when human biological systems interact with electronic devices, where are the boundaries of adaptability, adversarial response, and compatibility?

Through layer-by-layer inquiry, we find that the predicament facing invasive BCI is not an engineering optimization problem but a paradigm problem. The solution lies not in better electrodes but in shifting toward extracorporeal physical signals that activate the brain’s own ion-channel infrastructure. But between direction and endgame, four bridges remain missing: readout bandwidth, write-in precision, cross-brain semantic translation, and long-term safety-loop closure — plus one hard physical ceiling: the brain’s metabolic energy constraint.

What this paper provides is not proof of the endgame, but a directional judgment, blueprints for the bridges, and acceptance criteria for each one.

Our intelligence is upscaling; our action is downscaling. Bridging this chasm may be the fifth information revolution of human civilization, following writing, the printing press, telecommunications, and the internet.

Appendix: Key Literature Domain Index for the Reasoning Chain

Human Biological Adaptability: Bergmann (1847); Allen (1877); Boas (1910); Baker & Weiner (1966); Lasker (1969, Science); Frisancho (1969–1993); Goodman & Leatherman (1998); Leonard (2018, AJPA)

Genomic Adaptation: Enattah et al. (2002, Nature Genetics); Tishkoff et al. (2007, Nature Genetics); Yi et al. (2010, Science); Huerta-Sánchez et al. (2014, Nature)

BCI Biological Adversarial Response: Polikov et al. (2005); McConnell et al. (2009); Barrese et al. (2013); Eles et al. (2018); PMC BCI Review (2025)

Electrode Materials and Lifespan: Hassler et al. (2011); Takmakov et al. (2015); BrainGate 15-year dataset (2025); Reuters/WIRED Neuralink reports (2024)

Ionic Electrochemical Signaling: Hodgkin & Huxley (1952); Harikesh et al. (2022, Nature Materials); Zhao et al. (2024, Science)

Ultrasonic Neuromodulation: Sorum et al. (2021, PNAS); Yoo et al. (2022, Nature Comms); Qiu et al. (2023, PNAS); Zhong et al. (2024, PNAS)

Brain-to-Brain Interfaces: Rao et al. (2014, PLOS ONE); Grau et al. (2014, PLOS ONE); Stocco et al. (2015); Jiang, Stocco et al. (2019, Scientific Reports, BrainNet)

Cross-Individual Representation Alignment: Haxby et al. (2011, Neuron, Hyperalignment); Chen et al. (2015, SRM); Guntupalli et al. (2016, Cerebral Cortex); Haxby et al. (2020, eLife); Bazeille et al. (2019, Optimal Transport)

Brain Metabolic Energy: Levy & Baxter (1996); Attwell & Laughlin (2001); Sterling & Laughlin (2015); Levy & Bhatt (2021, PNAS)

DBS Ethics and Personality: Schermer (2011); Gilbert et al. (2017, AJOB Neuroscience); Pugh et al. (2021)