Semantics of Brain-Machine Hybrids

Abstract

Brain-machine interfaces, also known as brain-computer interfaces, represent a rapidly advancing field at the intersection of neuroscience and technology, enabling direct communication pathways between the brain and external devices. This review charts the historical evolution of brain-machine interfaces, from fundamental discoveries such as electroencephalography and volitional single-neuron control to sophisticated decoding of neural population activity for real-time control of robotics and sensory reconstruction. Clinical breakthroughs lead to unprecedented success in restoring motor function after paralysis through brain-spine interfaces, enabling high-speed communication through thought-to-text/speech systems, providing sensory feedback for prosthetics, and implementing closed-loop neuromodulation for the treatment of neurological disorders such as epilepsy and depression. Beyond therapeutic applications, brain-machine interfaces drive innovation in neurotech art (neuroart) and entertainment (neurogames), allowing neural activity to directly generate music, visual art, and interactive experiences. In addition, the potential for human augmentation is expanding, with technologies that enhance physical strength, sensory perception, and cognitive abilities. These converging advances challenge fundamental concepts of human identity and suggest that brain-machine interfaces may enable humanity to transcend inherent biological limitations, potentially ushering in an era of technologically guided evolution.

1. INTRODUCTION

In “The Jameson Satellite,” a serialized story published in the science fiction magazine “Amazing Stories” in 1931, American author Neil Ronald Jones introduced the Zoromes, a fictional race of cyborgs. The Zoromes are described as mechanized beings with cubic metallic bodies, four articulated appendages for locomotion, and six metallic tentacles extending from their upper torso. Originally biological organisms, they achieved a form of immortality by transplanting their brains into mechanical chassis, allowing the biological brain to retain control of the mechanical construct. By transcending biological limitations and achieving perpetual existence, the Zoromes arguably represent the earliest documented conceptualization of a brain-machine interface (BMI) from a modern scientific perspective.

A BMI, also often called a brain-computer interface (BCI), represents a direct communication pathway established between the central nervous system—often the brain—and an external device, such as a computer or a prosthetic limb.¹⁾ The fundamental goal is to translate neural signals reflecting intent into actionable commands for controlling machines or computers.

The process begins with neural signal acquisition. This can be achieved through invasive methods, where surgically implanted electrodes, such as microelectrode arrays or electrocorticography grids, directly interface with cortical tissue. These provide high-fidelity signals, capturing individual neuron action potentials (spikes) or local field potentials (LFPs) from specific brain regions, often the motor or sensory cortex. Alternatively, non-invasive methods, such as scalp electroencephalography (EEG) or functional near-IR spectroscopy (fNIRS), record neural activity from outside the skull.

Sophisticated computational algorithms, frequently employing machine learning (or artificial intelligence, AI), filter the data, extract relevant neural features (e.g., firing rates, oscillatory patterns), and ultimately decode the user’s intention (e.g., desired movement direction, cognitive state). The decoded intention is then converted into output commands that operate an external device. Applications include controlling prosthetic limbs, navigating computer interfaces, operating wheelchairs, or using text generation software, offering pathways for interaction with the environment.

BMIs hold immense clinical promise for restoring motor, sensory, and communication functions in individuals affected by severe neurological conditions like spinal cord injury, stroke, or amyotrophic lateral sclerosis (ALS). They also serve as powerful research tools for investigating fundamental neuroscience questions about neural coding, motor control, sensory processing, and brain plasticity. Furthermore, closed-loop BMIs are being explored for neuromodulation therapies, potentially treating conditions like epilepsy or Parkinson’s disease by detecting aberrant neural states and delivering targeted interventions.

2. THE PAST OF BMI

2.1. Prehistory of BMI

Modern BMI research traces its origins to early 20th-century neuroscience. A critical foundation was the discovery of measurable electrical activity in the brain. Julius Bernstein, a student in the laboratory of the German physiologist Emil du Bois-Reymond, recorded the time course of an action potential around 1865; however, at that time, recording neural activity was significantly hindered by the limitations of experimental techniques. The development of EEG recording was crucial for enabling more accessible recording of neural activity. In 1875, Richard Caton published a paper titled “Electric Currents of the Brain,” describing the recording of electrical potentials from the surface of the heads of rabbits and monkeys.²⁾ Subsequently, in 1924, Hans Berger reported the first recordings of human EEGs and identified rhythmic oscillations such as alpha waves.³⁾ Berger acknowledged that his discovery of alpha waves in the human brain was inspired by Caton’s paper. This non-invasive method of recording brain activity would later become an important cornerstone for the development of BMI.

Another important advance in the field of BMI is the demonstration that neural activity can be volitionally controlled; note that a BMI can only function effectively if the user is able to control their neural activity voluntarily. In a landmark experiment in 1969, Eberhard Fetz demonstrated that unanesthetized monkeys (Macaca fascicularis) could learn to volitionally modulate the firing rate of individual neurons in the motor cortex.⁴⁾ The central method used in this research was operant conditioning. The monkeys received a reward (food pellets) for increasing the firing rate of a specific neuron, while auditory or visual feedback indicating the neuron’s firing rate was provided to aid the learning process. After training sessions, the monkeys were able to significantly increase the activity of target neurons (often by 50 to 500%) above baseline rates. Conditioning a decrease in firing rate was also successful. This control was largely independent of the neurons’ typical motor-related activity patterns. A later study demonstrated more robust neural control, with monkeys learning to control nearly all tested neurons, maintain target firing rates for several seconds, and exhibit improved performance over successive training sessions.⁵⁾ These works provided the first direct evidence that individual cortical neurons could be brought under volitional control. This landmark discovery suggests that if single neurons could be volitionally modulated, they could potentially serve as control signals for direct brain-machine communication.

2.2. Early Days of BMI

Several years before the terms BMI and BCI were formally introduced, composer Alvin Lucier created a piece titled “Music for Solo Performer” that utilized the performer’s neural oscillations (specifically alpha waves, 8–13 Hz) to generate sound.⁶⁾ EEG electrodes (typically three silver electrodes) positioned on the performer’s scalp detected alpha activity. These raw signals were substantially amplified using a differential amplifier and subsequently filtered to isolate the alpha frequency band. The amplified signals were then transmitted to speakers placed near or above various percussion instruments, including drums, cymbals, pianos, and other resonant objects such as trash cans. The low-frequency vibrations emanating from the speakers caused the instruments to resonate, thereby producing sound. The performer would sit still and attempt to enter a state conducive to generating alpha waves.

Lucier’s work is considered a landmark in experimental music and an early instance of artistic sonification. This work highlighted the potential of EEG as a signal source for controlling external systems within an artistic, open-loop context. It explored the direct translation of endogenous physiological signals into external artistic expression and resulted in performances described as “theatrically exciting.” Although not constituting a BCI in the contemporary sense of goal-oriented control, “Music for Solo Performer” represents one of the earliest documented examples of actuating an external device directly via extracted brain signals. This piece translated the abstract concept of measurable brain activity into the concrete realm of physical sound generation and predated the formal definition of BCI. It effectively demonstrated signal extraction and external actuation.

Jacques Vidal is widely credited with introducing the term BCI into the scientific literature through a 1973 publication.⁷⁾ He conceptualized computers as a potential “prosthetic extension of the brain” and initially employed the term extensively. His 1973 paper articulated the “BCI challenge”: controlling external objects using EEG signals, particularly slow cortical potentials like the contingent negative variation (CNV).

Vidal followed this in 1977 with the first published experimental demonstration of the BCI concept.⁸⁾ This system utilized non-invasive EEG recorded over the visual cortex, specifically visual evoked potentials (VEP). VEPs are neural responses elicited by visual stimuli. Vidal’s system detected VEPs to determine the user’s point of visual fixation. This gaze information was subsequently converted into commands to manipulate graphical objects, such as a cursor on a computer screen, enabling the user to navigate a maze.

Vidal’s contributions were highly significant. He established the term “BCI,” demonstrated the feasibility of non-invasive control using VEP, and explicitly linked BCI development to computer interaction and visual neuroscience, advancing beyond theoretical concepts and single-neuron studies. Although considered to possess limited accuracy at the time, it constituted an important proof of concept.

Bozinovski et al. reported the first control of a physical object (a mobile robot) using non-invasive EEG signals in 1988.⁹⁾ This robotic control system employed changes in the EEG alpha rhythm (8–13 Hz)—specifically, the increase associated with relaxation or closed eyes (termed contingent alpha variation, CαV, by the authors)—to generate start/stop commands. The EEG was recorded from the Pz scalp location and processed using software designed to recognize the presence or absence of predominant alpha activity. Increased alpha power (associated with closed eyes) triggered a “start” command, while decreased alpha power (associated with open eyes) triggered a “stop” command. The robot, called Elephobot Line Tracer, possessed autonomous line-following capabilities, with the BCI providing high-level start/stop control. The system incorporated machine learning elements for calibration (learning probability density functions for closed/open eye states) and real-time pattern recognition.

In 1990 (though some sources indicate research dating back to 1988), the same group reported a closed-loop BCI utilizing CNV potentials.¹⁰⁾ The CNV is a slow negative potential that develops in anticipation of a stimulus or response. Their system employed the CNV, reflecting the brain’s state of expectation, to control a buzzer within a feedback loop. They termed the cognitive component representing anticipatory learning the electroexpectogram. This research also incorporated adaptive filtering techniques.

The 1988 robot control experiment was significant as the first demonstration of controlling the movement of a physical mobile object using EEG, thereby directly linking EEG research with robotics. It addressed the psychokinesis challenge from an engineering perspective and introduced machine learning methodologies into the BCI field. Their subsequent CNV research explored the use of anticipatory brain potentials and closed-loop adaptive control. These contributions expanded the repertoire of EEG signals and control paradigms investigated in early BCI research.

2.3. Toward Practical Application of BMI

A key question in motor neuroscience involves understanding how the brain achieves precise motor control, such as arm reaching, given that individual motor cortex neurons exhibit broad tuning to a particular preferred direction. Investigating this issue, Georgopoulos and colleagues studied primate motor cortex during reaching tasks in three dimentional (3D) space. In 1986, they discovered that the direction of movement is accurately encoded not by single neurons, but by the collective activity of a neuronal population.¹¹⁾

They proposed the population vector algorithm. In this model, each neuron’s contribution is represented as a vector pointing in its preferred direction, which was determined experimentally, e.g., via a center-out task. The magnitude of this vector is weighted by the neuron’s firing rate (or change in firing rate) during a specific movement. The vector sum of all these individual contributions yields the “population vector,” the direction of which accurately predicts the actual direction of arm movement. The magnitude of the population vector can correlate with movement speed.

This finding represented a pivotal advance in understanding motor control and neural coding. It demonstrated how precise information could emerge from the combined activity of broadly tuned elements. This work shifted the focus from the capabilities of single neurons, as explored by researchers like Fetz, to the computational power inherent in population codes, and provided a crucial conceptual and mathematical bridge between the inherent variability of individual neuronal activity and the brain’s capacity to generate precise, directed motor output, showing how a population could reliably encode information. The population vector algorithm established the theoretical groundwork for sophisticated invasive BMIs aimed at controlling prosthetic limbs or cursors based on decoded population activity, providing a powerful mathematical framework for decoding motor intent from neural ensembles.

With the goal of “reading the neural code,” Dan and colleagues sought to decode visual information from neural firing patterns within the early visual system.¹²⁾ Specifically, they reconstructed visual stimuli from ensemble activity in the cat lateral geniculate nucleus, a critical thalamic relay nucleus for vision. Methodologically, they recorded spike trains from ensembles of lateral geniculate nucleus neurons (up to 177 cells) while anesthetized cats viewed movies depicting natural scenes. They employed linear decoding techniques (multi-input, multi-output reverse correlation/reconstruction filters) to estimate the spatiotemporal visual input likely responsible for the observed spike patterns. These filters accounted for individual cell properties and inter-neuronal correlations. Reconstruction quality was assessed by comparing the reconstructed visual information to the actual stimuli using metrics such as correlation coefficients and spectral analysis. The results demonstrated the successful reconstruction of recognizable, spatiotemporal natural scenes (including moving objects and faces) from neuronal ensemble activity. Reconstruction quality improved with the number of neurons included in the analysis, beginning to saturate around 12–16 cells for natural scenes, suggesting neuronal adaptation to natural stimulus statistics. Linear decoding methods proved capable of capturing significant visual information, particularly at higher temporal frequencies, with neuronal noise identified as a major limiting factor for fidelity.

This work constituted a pioneering demonstration of reconstructing dynamic, natural sensory experiences from neural activity. It showed that linear methods could extract substantial visual information from population codes in the early visual pathway. These findings paved the way for developing sensory neuroprosthetics (e.g., visual prostheses) and provided valuable insights into how sensory information is encoded and represented by neural ensembles. This research complemented motor decoding efforts by addressing the sensory input aspect of brain interfaces.

In response to these initial attempts, Nicolelis and colleagues investigated whether population activity recorded from the motor cortex could be utilized for real-time control of external robotic devices. In a 1999 study using rats, they trained the animals to press a lever, which controlled a robot arm delivering a water reward. Ensembles of motor cortex neurons were simultaneously recorded¹³⁾; note that more than 25 neurons are required for successful control. Mathematical transformations, including neural networks, were developed to generate ‘neuronal population functions’ capable of predicting lever trajectory from neuronal firing rates. These functions were then converted into command signals to directly control the lever in real-time, establishing a ‘neurorobotic’ mode of operation. It is worth noting that machine learning called recurrent neural networks was used for the first time in BMI.

A subsequent study published in 2000 extended this paradigm to owl monkeys.¹⁴⁾ Larger electrode arrays (up to 96 electrodes) were implanted in multiple cortical areas, including motor, premotor, and somatosensory cortex. Population activity was recorded while the monkeys performed reaching tasks. Computational algorithms, initially employing linear models and other mathematical methods, were used to predict hand trajectory from neural signals in real-time. These decoded signals subsequently enabled the monkeys to control a multi-jointed robot arm in three-dimensional space; the monkeys successfully controlled the robot arm using decoded brain signals derived from distributed cortical populations. Neural signals related to trajectory were observed broadly across the recorded cortical areas. Crucially, population-based decoding consistently outperformed predictions based on single neurons. The feasibility of remote control over the internet was also demonstrated.

These experiments were significant achievements, demonstrating—first in rats and subsequently in primates—that brain activity could be decoded in real-time to provide continuous control of complex robotic devices. This work validated the potential of invasive BMIs for restoring movement in individuals with paralysis by enabling control over sophisticated prosthetics. The findings also reinforced the principle of distributed population coding, consistent with Georgopoulos’s work, and highlighted the brain’s plasticity in adapting to neuroprosthetic control. These advancements around 1999 marked a pivotal moment in the field. Research progressed beyond the theoretical possibilities outlined by Georgopoulos or the basic control demonstrated by Vidal and Bozinovski, showcasing the ability to perform complex, high-fidelity decoding of both sensory information and continuous motor commands in real-time. This considerably advanced the benchmark for BMI capabilities.

2.4. Recent Application of BMI

Restoring movement after paralysis has been a major focus of BMI studies. In 2023, researchers in Switzerland (EPFL/CHUV) and France (CEA-Clinatec) reported a breakthrough “brain-spine interface” for a man with chronic tetraplegia.¹⁵⁾ Two 64-electrode implants were placed on the surface of his motor cortex and connected wirelessly to an implanted pulse generator on his spinal cord. When the participant thought about walking, the system decoded the brain signals in real time and stimulated his leg muscles accordingly. As a result, he regained the ability to stand, walk, and even climb stairs after years of paralysis. The patient described the implant as “life-changing,” allowing him to perform daily activities such as getting up to paint and enjoying a drink with friends while standing. This digital bridge between the brain and spine demonstrates how BMIs can bypass damaged nerves and restore voluntary movement. It builds on earlier trials of epidural electrical stimulation for spinal cord injury, but now with direct brain-controlled movement rather than pre-programmed steps.

Other BMI trials have helped paralyzed patients control external mobility aids. For example, researchers have enabled people with tetraplegia to drive brain-controlled wheelchairs through thought alone in real-world environments using noninvasive EEG-based BCIs with training.¹⁶⁾ Such studies report that users can safely navigate obstacle courses after iterative learning, highlighting BMI potential for independent mobility in the home or hospital settings. In another dramatic demonstration funded by Defense Advanced Research Projects Agency (DARPA) in the United States, a 55-year-old paralyzed woman was able to pilot an F-35 jet simulator using only her mind—she had implanted electrodes originally for a robotic arm study, which were repurposed to control a flight simulator.¹⁷⁾ This feat, while experimental, showed that high-dimensional motor tasks (like aircraft controls) can be mapped from neural signals, an application of interest to the military for remote vehicle operation and to aviation medicine.

Brain-controlled prosthetic limbs have seen major improvements in the last few years, especially with the addition of sensory feedback. Earlier BMI prosthetic trials allowed paralyzed individuals to control robotic arms to perform tasks like grasping objects, but without sensation, it was challenging to manipulate fragile or unseen items. Recent trials have integrated bidirectional BCI systems that not only read motor commands from the brain but also write sensory information back to the brain (or peripheral nerves). For example, in a 2021 University of Pittsburgh study, a participant with tetraplegia (implanted with motor cortex electrodes) controlled a robotic arm that could “feel” via electrical stimulation of his sensory cortex.¹⁸⁾ Whenever the robotic hand touched an object, signals were fed back into the brain, creating an artificial sense of touch. This tactile feedback enabled the man, Nathan Copeland, to perform tasks like pouring water between cups much more fluidly. With vision alone, it took him about 20 s to grasp and move an object, but with added touch feedback he did it in 10 s. He also found that objects did not slip from the robot hand as often, since he could feel how firmly he was grasping them. This closed-loop BCI demonstrates how neuroprosthetics with sensory feedback vastly improve performance and user confidence.

Restoring communication is another critical area of BMI clinical research, especially for patients with ALS, brainstem stroke, or other conditions that leave them “locked in.” Recent trials have achieved unprecedented communication speeds by decoding neural activity into text or speech. In 2021, a Stanford University-led team, as part of the BrainGate consortium, implanted microelectrode arrays in the motor cortex of a man with high spinal cord injury and demonstrated a brain-to-text BCI that decoded his attempted handwriting motions.¹⁹⁾ By imagining himself writing letters, the participant was able to output text on a screen at about 90 characters per minute (roughly 18 words per minute)—a record at that time for BCI communication, comparable to average smartphone typing speeds in able-bodied peers. This “mental handwriting” approach was more than twice as fast as previous BCI typing methods and underscored how machine learning can extract rapid, fine-grained patterns from neural signals.

Building on such advances, researchers have begun to tackle direct speech prosthesis via BMI. In 2023, a team at University of California, San Francisco implanted a high-density electrocorticography array in a woman who had lost the ability to speak due to a brainstem stroke.²⁰⁾ Using AI-based decoding, they achieved the first instance of synthesizing both whole words and facial expressions from brain activity. In this trial, the participant’s attempted speech signals were converted in real time into a digital avatar speaking with her intonation and even smiling or scowling appropriately. The system could decode brain signals into text at nearly 80 words per minute, vastly outpacing her prior assistive device. While still experimental, the technology points toward a future officially approved BCI. Indeed, an earlier case in 2021 had already shown that a mute, paralyzed patient could produce dozens of words (from a limited vocabulary) via a speech BCI at approx. 75% accuracy, using an implant in speech motor cortex. By 2023, accuracy and vocabulary had improved; one BrainGate study reported decoding internal speech with up to 97% accuracy (with autocorrect) for an ALS patient, allowing him to communicate words he was “thinking” of saying.²¹⁾

Notably, even patients in completely locked-in states (no eye or muscle movement) have begun to benefit from BCIs. In one 2022 case, a late-stage ALS patient received an implanted electrode system and learned to respond to auditory cues via modulating his brain signals.²²⁾ Over months, he was able to slowly spell out messages (around one character per minute) by thinking “yes” or “no” in response to an auditory speller, thus regaining a basic communication channel. While extremely slow, this study proved that some communication is possible even after all muscle control is lost, using only the brain’s electrical activity. Such cases remain rare, but they highlight the importance of BMI for preserving the ability to communicate in the most severe paralysis conditions.

While many BMIs focus on restoring lost motor or communication function, an important subset of clinical trials involves implanted brain interfaces that treat neurological disorders like epilepsy, tremor, or Parkinson’s disease. These are often closed-loop systems–they sense abnormal brain activity and deliver targeted stimulation or other interventions. A prime example is the NeuroPace Responsive Neurostimulation system for epilepsy, which, although approved in 2013, saw wider adoption and follow-up studies in the past five years. The NeuroPace Responsive Neurostimulation is a BMI that monitors EEG activity from electrodes on the brain and automatically applies electrical pulses when it detects epileptic seizure patterns, often aborting seizures before symptoms escalate. Long-term trial data showed significant median seizure frequency reductions (around 75% in many patients at 6+ years of use) and a subset of patients becoming nearly seizure-free.²³⁾ These outcomes underscore that closed-loop BMIs can provide lasting therapeutic benefits in a real-world setting. Moreover, cognitive and mood improvements were noted in some patients, likely because better seizure control improves overall brain network stability.

Researchers are now extending the closed-loop neuromodulation concept to other conditions. Adaptive deep brain stimulation (DBS) for Parkinson’s disease is a key trend. Next-generation DBS devices (e.g. Medtronic Percept PC, AlphaDBS) can record brain signals such as local field potentials and adjust stimulation in real time. Clinical studies have tested “adaptive DBS” algorithms that raise or lower the stimulation intensity based on neural feedback, aiming to deliver therapy only when needed. Early trials in Parkinson’s patients have shown that adaptive DBS can maintain symptom relief (tremor, rigidity control) with less stimulation on time, thereby potentially reducing side effects. In 2021, a breakthrough case at University of California, San Francisco applied a personalized closed-loop DBS system for severe depression.²⁴⁾ Investigators identified a specific brain activity pattern that corresponded to the patient’s depressive symptoms; they then programmed a neurostimulator to detect that biomarker and stimulate the ventral striatum whenever the pattern appeared. The result was a dramatic and sustained alleviation of depression in this one patient. This was the first in-human demonstration of a closed-loop BMI for a psychiatric disorder, and it paves the way for larger trials in depression and obsessive–compulsive disorder using responsive brain implants.

As BMI technology matures, a growing ecosystem of companies and investors is pushing the field toward commercial products. The past decade has seen an influx of startup activity, venture capital funding, and even big corporate partnerships in neurotechnology. At the same time, regulators like the Food and Drug Administration (FDA) in the United States have begun creating pathways to evaluate and approve these novel devices. To summarize the competitive landscape, Table 1 compares a few leading BMI efforts and their recent progress.

Table 1. Leading BMI Developers and Their Progress in 2020–2025

Organization	Approach & Device	Recent Milestones (2020–25)	Funding/Partnerships
Neuralink (U.S.A.)	Fully implanted 1024-channel microelectrode array (“N1”) with robotic surgical insertion.	Demonstrated a monkey controlling a video game via implant (2021); won FDA approval for first human trial (May 2023); reportedly performed first human implants by late 2023.	Approx. $363M venture funding (2016–2022); backed by Elon Musk. Collaborations kept mostly in-house; high public profile.
Synchron (U.S.A./AUS)	Stentrode endovascular neural interface (16-electrode mesh delivered via blood vessels).	First BCI implanted without open brain surgery; completed first human study in AUS enabling texting and e-shopping by thought. Gained FDA IDE and implanted 6 U.S. patients (2021–23); preparing larger trial toward approval.	Approx. $75M Series C funding in 2022 (total approx. $145M); investors include Bezos & Gates. Partnerships with Mount Sinai Hospital, DARPA (funding).
Blackrock Neurotech (U.S.A.)	Utah array-based implant (NeuroPort/MoveAgain); developing high-density “Neuralace” (>10k electrodes).	BCI tech used in 30+ research patients via BrainGate; MoveAgain system received FDA Breakthrough designation (2021). Aiming to launch first commercial BCI for home use in paralysis in mid-2020s. Secured major investment to scale up (2024).	$200M investment from Tether in 2024 for majority stake (valuation approx. $350M). Long-term supplier to academic labs; partnered with AE Studio on software.
BrainGate Consortium (U.S.A.)	Academic multi-center trial using surgically implanted Utah arrays + custom decoders.	Ongoing trial with continuous improvements: published wireless transmitter results (2021); showcased 90%+ accuracy speech decoding for an ALS patient (NEJM 2023). Demonstrated safety over 15+ years of implants.	Funded by NIH, Dept. of Veterans Affairs, and others. Not a company; collaborations with Brown, Stanford, MGH, Case Western, etc. Provides foundational research for industry.
Paradromics (U.S.A.)	High-channel-count ECoG implant (thousands of electrodes) and data hub.	Achieved high-bandwidth recording in animals; awarded FDA Breakthrough Device status (c.2022). Building “Neuroport” for speech BCI and planning first human studies (2024).	Approx. $38M venture seed funding + DARPA contracts; part of DARPA’s NESD program. Recently raised total approx. $88M.

Approaches vary between invasive and minimally invasive, and each has reached important milestones toward clinical translation. Funding has accelerated, with private capital complementing earlier government-funded research.

2.5. Neurotech Art and Entertainment

In recent years, artists and scientists have collaborated to transform neural activity into visual art and interactive experiences (Table 2). One striking example is a pair of brain-powered dresses presented by engineer Leonhard Schreiner and designer Anouk Wipprecht. In the Screen Dress, the wearer’s brain engagement level (measured via a 4-channel EEG headband) is visualized by animated eye graphics on the dress that open and move with the wearer’s mental state.²⁵⁾ In the companion Pangolin Scales Dress, a full 1024-channel “ultra-high-density” EEG system drives motors and light-emitting diodes on a 3D-printed exoskeletal dress; the dress’s scale-like flaps and lights respond in real time to the wearer’s brainwaves. For example, calm theta waves create soft purple movements, while alert beta waves trigger rapid white flickers). These projects transform brain signals into dynamic fashion statements, demonstrating how neural data can directly animate physical art on the human body.

Table 2. Representative Neurotechnology-Driven Art and Entertainment

Project/Work	Creative Domain	Brain Source	Interface/Method
Music for Solo Performer (1965)	Live music performance	Human (EEG alpha waves)	Analog EEG amplification to trigger percussion instruments in real time.
Brain Painting BCI (2010s)	Visual art (digital)	Human (EEG P300 focus)	Non-invasive P300 speller interface to select colors/shapes on a virtual canvas.
Paramusical Ensemble (2015)–“Activating Memory”	Music composition/performance	Human (EEG–locked-in patients)	EEG cap with visual P300 selection; patients choose musical phrases played by live string quartet.
Encephalophone (2017)	Musical instrument	Human (EEG–alpha/mu)	EEG-based instrument translating brain rhythms into synthesizer notes in real time.
Screen Dress (2025)	Interactive fashion art	Human (EEG–4 channels)	EEG engagement level drives animated eye graphics on dress’s embedded screens.
Pangolin Scales Dress (2025)	Interactive fashion art	Human (EEG–1024 channels)	Ultra-HD EEG controls servomotor “scales” and LED lights reflecting different brainwave frequencies.
Brainlight (2015)	Light sculpture	Human (EEG)	Wireless EEG headset controls illumination in a brain-shaped perspex sculpture.
DREAM 2.2 (2018)	Immersive installation	Human (EEG–audience)	Custom EEG-driven audio-visual system; brainwaves modulate projected visuals and sounds in a maze.
MEART–Semi-Living Artist (2001–04)	Robotic art installation	Rat neurons in vitro (MEA)	Cultured cortical cells on electrode array send signals to robotic drawing arm, creating art.
Neuralink “MindPong” (2021)	Gaming (demo)	Monkey (implant BMI)	1024-electrode wireless implant in motor cortex; neural activity moves cursor to play Pong.
DishBrain Pong (2022)	Gaming (experiment)	Human & mouse neurons (in vitro)	Microelectrode array with 800k neurons; electrophysiological feedback enables Pong paddle control.
Mindflex Toy (2009)	Game/Toy	Human (EEG–Focus)	EEG headset measures concentration to control airflow that levitates a foam ball through obstacles.

From brainwave-driven dresses and paintings to thought-composed music and neuron-enabled gameplay, these projects demonstrate the expanding boundaries of art and entertainment powered by brain-machine interfaces. They illustrate how EEG signals, brain implants, and even cultured neurons can be used to generate music, control visuals, and animate objects in real time.

An interactive light sculpture by artist Laura Jade makes the activity inside your head visible. The piece is a brain-shaped Plexiglas sculpture engraved with neural patterns that glows in real time in response to the participant’s EEG signals.²⁶⁾ Wearing a wireless Emotiv EEG headset, viewers see the sculpture glow in different colors and intensities as their brain activity changes, essentially using their minds to “paint” the sculpture with light. This project, created during Jade’s art-science residency, merges lighting design with neuroscience to provide a personalized visual display of one’s own neural oscillations.

Brain Painting is a P300-based BCI art tool that allows people to create digital paintings using only their brain signals. It presents a palette and shapes on a screen and uses EEG to detect where the user is focusing attention to select colors or brush strokes.²⁷⁾ By cycling through options with visual flashes and using the brain’s response, even people with locked-in syndrome have been able to paint pictures without any muscle movement. This innovative art-making method turns brain activity into brush strokes, demonstrating a non-medical use of BCI to facilitate creative expression for those who otherwise cannot physically paint.

In collaboration with Steve Potter’s lab, SymbioticA, an Australian bioart group, is bringing neural-driven art into the animal lab. It was essentially a primitive cybernetic organism—an amalgam of wetware and hardware—that produced evolving drawings. In this art-science installation, living brain matter literally became the artist.²⁸⁾ A culture of rat cortical neurons grown on a multi-electrode array served as the “brain” to control a drawing machine. The neuron signals were sent over the Internet to a robotic arm that drew on paper, creating abstract art based on the neural activity. This provocative piece demonstrated an unexpected form of artistic creation: a hybrid animal-machine system in which the spontaneous activity of a non-human brain resulted in tangible art.

In PluginHUMAN, an immersive art installation by Sargeant and Dwyer, the audience’s brainwaves drive a maze of audiovisuals.²⁹⁾ Visitors wear EEG headsets and walk through a maze of hanging panels with projection-mapped visuals and surround sound. A custom brain-computer interface translates each participant’s brainwave patterns into changes in the projected abstract imagery and ambient soundscape. The result is a mesmerizing “brain forest” environment that changes its colors, lights, and audio in response to the minds of its explorers, making each visitor’s experience unique to their own neural data.

We also presented a novel BCI technique that uses latent diffusion models, a type of deep neural network, to generate images directly from continuous brain activity.³⁰⁾ Unlike previous BCI art methods that rely on decoding specific user intentions (e.g., via P300 signals) to control drawing tools, this end-to-end approach bypasses conscious control. By recording local field potentials from the neocortex of freely moving rats, this system continuously transformed the neural signals into sequences of morphing images using a pre-trained Stable Diffusion v1.5 model without text input. The results showed that the generated images successfully reflected the statistically dynamic nature of the underlying neural activity, showing continuity between successive frames (Fig. 1). This method provides a unique way to visualize brain function and opens new possibilities for creative expression independent of conscious intent, with potential future applications in art.

Fig. 1. Diffusion Model-Based Image Generation from Rat Brain Activity

A novel BCI using a latent diffusion model was developed to generate images directly from continuous rat brain activity. This system bypasses conscious control, creating visuals that reflect raw neural dynamics. This offers new paths for unconscious artistic expression and brain function visualization. (A) A photograph of an experimental setup. (B) Detailed illustration of the steps taken to generate an image from rat local field potentials. The recorded local field potentials were mapped onto a noisy latent vector z_T, which was then denoised to produce a denoised latent vector z. This latent representation of the denoised image was then processed by a latent-to-image decoder to produce image X. (C) Noisy latent vector z_T and the corresponding generated image obtained by the diffusion process and the latent-to-image decoder. This figure is modified from a paper by Yamashiro et al.³⁰⁾

The paramusical ensemble “Activating Memory” is a unique musical performance in which four paralyzed patients became composers by using their EEG signals. Developed by Eduardo Miranda’s team at the University of Plymouth, the system allowed each participant (wearing an EEG cap) to select short musical phrases on a computer by directing their attention to specific flashing icons using a P300 brain-computer interface.³¹⁾ The selected passages were then played in real time by a live string quartet, effectively turning the patients’ thoughts into a collaborative composition. The resulting piece, “Activating Memory,” was premiered at a contemporary music festival in the United Kingdom and demonstrated a powerful fusion of neuroscience and art: bedridden individuals creating and performing complex music using only brain signals.

The Encephalophone is also a mind-controlled musical instrument invented by neuroscientist/musician Thomas Deuel and colleagues. The Encephalophone is played without any physical movement: the performer wears an EEG cap and generates musical notes by modulating specific brain rhythms, such as alpha waves.³²⁾ These brain signals are translated into pitches on a synthesizer in real time, allowing the user to play melodies using only his or her thoughts. Originally developed to enable disabled musicians to continue making music, the Encephalophone also opened up a new musical realm in which brain waves are the instrument. This device illustrates how brain-machine interfaces can serve creative expression—here, the brain becomes a musical controller, producing real-time compositions from neural activity.

In a demonstration by Elon Musk’s Neuralink, a macaque monkey named Pager played a game of Pong using only his brain implants.³³⁾ Researchers implanted tiny electrode arrays in the monkey’s motor cortex and trained Pager to move a cursor with a joystick. Once calibrated, the joystick was disconnected—yet the monkey could continue to move the on-screen paddle telepathically via neural signals. A video released in 2021 showed Pager successfully collecting the Pong ball using only his mind, without a physical controller. This demonstration highlighted an entertaining (and viral) example of a brain-machine gaming interface with an animal, underscoring that even a primate’s neural activity can directly control a video game in real time.

Going a step further, scientists have taught a collection of living neurons in a dish to play a video game. In an experiment led by Cortical Labs, about 800000 human and mouse neurons grown on a microelectrode array were connected to a computer running Pong.³⁴⁾ The system, dubbed “DishBrain,” fed electrical signals to the neurons indicating the position of the Pong ball; remarkably, the neural cluster learned to adjust firing patterns to move a virtual paddle and hit the ball back. Within minutes of feedback training, the dish of brain cells demonstrated goal-directed learning and successfully played a rudimentary game. This sci-fi-like project was a research effort to study learning and intelligence, but it doubled as an unexpected entertainment application—effectively an arcade game played by a petri dish of neurons.

Brainwave games and toys—Brain-computer interfaces have also entered the playful realm of consumer products. For example, Mindflex is a toy in which players guide a foam ball through an obstacle course using only their concentration. An EEG headband measures the player’s level of concentration and adjusts a small fan that floats the ball—the more you concentrate, the higher the ball floats.³⁵⁾ Similarly, interactive exhibits such as Mindball have pitted the relaxation levels of two players against each other: EEG sensors detect who is calmer and move a ball toward the more stressed opponent. Even wearable fashion gadgets like the Necomimi (cat ears) have become popular, with motorized cat ears that perk up or droop down depending on the wearer’s brain state (alert or relaxed). These whimsical games and devices show how neural signals can be harnessed for fun, allowing people to control physical objects or avatars using their mind as a game controller.

2.6. BMI-Augmented Humans

Bridging humans with machines or computers can be used as human enhancement technologies, which aim to augment natural human abilities through integration with advanced devices and systems. Unlike medical prosthetics or therapies that restore lost function, non-medical enhancements are designed to push capabilities beyond typical human limits—enabling stronger bodies, sharper senses and faster minds. They cover a wide range of technologies, from exoskeleton suits and BMI to AI-powered wearables and sensory augmentation devices (Table 3).

Table 3. Representative Technologies for Human Augmentation

Technology	Status	Purpose/capability	Origin (Country/region)
Shoulder/Back Exoskeletons (e.g., SuitX, German Bionic)	Implemented in industry	Augment lifting strength; reduce worker fatigue/injury	U.S.A., Germany, Japan (global adoption)
Military Exosuit (Lockheed Martin Onyx)	Field-tested	Assist soldiers in carrying heavy gear with AI-aided movement	U.S.A.
Jetpack Rescue Suit (Gravity Industries)	Demonstrated (2020)	Enable rapid personal flight for emergency responders	U.K.
AR Helmet for Pilots (F-35 HMDS)	In service	360° vision and data overlay for enhanced situational awareness	U.S.A. (used by various air forces)
AI Facial Recognition Glasses	Deployed	Real-time identification of individuals in crowds (augmented vision)	China
EEG Mind-Control Interface (NextMind)	Demo/Dev Kit (2020)	Hands-free control in VR via brain signals (attention-based BCI)	France
AlterEgo Silent Speech Wearable	Prototype tested	Enable silent communication with AI by reading internal vocal signals	U.S.A.
Brain-to-Brain Interface (BrainNet experiment)	Research trial	Direct brain signal sharing for group problem-solving	U.S.A./Europe collaboration
Robotic Third Thumb (UCL)	Research trial	Extra thumb controlled by foot–enhances one-hand dexterity	U.K.
Neuralink/Synchron BCI Implants	In trials (since 2022)	Brain implants enabling thought-based device control (paralysis aid, future augmentation)	U.S.A./Australia (global research)
DARPA TNT Neurostimulation	Research phase	Accelerate learning in healthy individuals via nerve stimulation	U.S.A.
Future Memory Chip Implant	Theoretical/Early dev.	Boost memory retention and recall beyond normal human limits	U.S.A. (academia, DARPA)
Bionic AR Contact Lens	Prototype (early tests)	Augmented reality display in the eye; potential IR/UV vision	U.S.A. (Silicon Valley startups)
Powered Full-Body Exosuit	Early development	Drastically amplify strength/speed for soldiers and workers	U.S.A., China, EU (defense R&D)
Brain-Cloud Interface	Conceptual (R&D ongoing)	Direct wireless link between brain and cloud AI for instant information access	Global (tech visionaries)

Human enhancement technologies are rapidly moving from concept to reality around the world. Table lists a number of these technologies, their status, purpose, and origin.

Factories and warehouses have started to equip workers with wearable exoskeleton suits to reduce fatigue and prevent injuries during physically demanding tasks, although these are not based on BMI. For example, car manufacturers such as Mazda Toyota Manufacturing in the U.S. have integrated shoulder-supporting exoskeleton vests into their assembly lines. After trialling the devices in 2022, the company deployed dozens of exoskeleton units to assist workers with overhead tasks (such as installing components under car frames), resulting in a reported 55% reduction in shoulder fatigue among workers. European companies have also adopted devices such as the German Bionic Cray X suit, a powered exoskeleton that can help lift up to 30 kg repeatedly by supporting the user’s back and legs. The Cray X even uses AI algorithms and an early warning system to monitor posture and provide feedback to the wearer—essentially anticipating strain and reducing the risk of injury during heavy lifting. These real-world applications illustrate how augmentative wearables are enhancing human strength and endurance in industry.

Real-world applications of augmented reality and sensor fusion are giving people superhuman sensory awareness. One high-profile example is the helmet-mounted display on the F-35 fighter jet, which gives pilots 360-degree vision and fused sensor data. The advanced helmet uses cameras around the aircraft to allow the pilot to literally ‘see through’ the body of the aircraft, displaying a panoramic view on the visor. It overlays critical information and target markers onto the pilot’s vision as an augmented reality feed. The F-35’s helmet even integrates night vision and IR inputs, allowing the pilot to see in the dark without separate goggles. This operational system effectively enhances human vision and situational awareness beyond natural limits, and is in service with U.S. and allied militaries.

While invasive brain implants for augmentation are still experimental, non-invasive BCI devices have been publicly demonstrated. EEG headsets are being used in new ways for entertainment and communication. At the 2018 Consumer Electronics Show, a startup showed off a consumer BCI called NextMind that allows users to control VR games using only their thoughts-by focusing their attention on targets, players could perform actions like exploding virtual objects without using their hands. Similarly, academic groups have held “drone races” in which participants control drones through EEG headsets, effectively moving an object through thought alone. These demonstrations are early steps toward everyday mind-machine integration. Another example is the MIT Media Lab’s AlterEgo prototype, a wearable device that enables silent, internal communication with a computer. The system detects neuromuscular signals from the wearer’s jaw and throat as they verbalize words in their head, allowing them to ask questions of an AI assistant without speaking and hear answers through a bone-conduction earpiece. In one demonstration, users were able to silently report their opponent’s moves in a chess game and receive recommended responses from a computer, all without making a sound. This type of intelligence augmentation device effectively acts as a private, thought-driven “smart assistant,” pointing to future everyday applications of BCI technology.

Beyond the early implementations above, many human enhancement technologies are currently in the experimental or trial phase. Scientists, companies, and military research agencies across the globe are actively testing new ways to push human abilities further. These projects provide a glimpse of what may soon become feasible.

One fascinating line of research is adding extra functional limbs to the human body. In a recent trial at University College London, researchers equipped volunteers with a robotic “Third Thumb”—a 3D-printed thumb extension worn on the hand and controlled with pressure sensors under the big toes.³⁶⁾ Over five days of training, participants learned to use the extra thumb to carry out dexterous tasks one-handed, such as building a tower of blocks, and even multitasking (manipulating objects with the extra thumb while doing mental math). Impressively, users quickly began to feel the robotic thumb as if it were part of their body, demonstrating the brain’s adaptability to augmentation. This experiment—confirmed by neural imaging—shows that the human brain can incorporate a prosthetic enhancement and suggests future interfaces for adding entirely new limbs or tools to able-bodied users. These supernumerary limb experiments are expanding the limits of human multi-tasking and coordination.

Pushing the envelope of human communication, scientists have conducted experiments directly linking brains of multiple people. In a notable 2019 study, an international team created a Brain-to-Brain Interface that let three people collaborate using brain signals alone.³⁷⁾ Two “sender” participants wearing EEG caps conveyed a yes/no instruction (whether to rotate a block in a Tetris-like game) by just thinking it, which was then transmitted via magnetic stimulation to a third person’s brain. The third participant, not seeing the game, successfully received the instruction in their mind and completed the task, demonstrating direct brain-to-brain communication of information. This and similar trials (sometimes dubbed “BrainNet”) hint at a future where teams could exchange information telepathically or solve problems together by sharing cognitive load between brains. Such capabilities, while rudimentary now, represent a form of collective human enhancement—essentially creating a networked “hive mind” for collaborative thinking.

Researchers are also experimenting with technologies to improve cognitive functions such as learning, memory, and concentration in healthy people. DARPA has a program called Targeted Neuroplasticity Training to accelerate learning by stimulating the nervous system. This project is investigating methods such as non-invasive electrical nerve stimulation to increase brain plasticity during training, effectively accelerating skill acquisition. By triggering the release of neurotransmitters that strengthen neural connections, this approach aims to improve how quickly someone can learn a foreign language or interpret complex images, for example. Several human trials are underway to test whether techniques such as stimulating the vagus nerve during training can improve cognitive performance without significant side effects. Similarly, companies have marketed transcranial brain stimulation headsets (using mild electrical or magnetic pulses) that claim to improve everything from athletic performance to video game concentration. Preliminary military research has suggested that transcranial direct current stimulation may modestly accelerate learning in tasks such as target recognition or language proficiency.³⁸⁾ These studies treat the human body as a system that can be tuned for better performance, blurring the line between biology and machine optimization.

Another experimental frontier is the synergy of human and artificial intelligence working as a combined unit. Military research programs in the U.S., China, and Europe are testing “cognitive co-pilot” AI systems that support humans in complex tasks. For instance, the U.S. Air Force has tried an AI co-pilot in a fighter jet simulator that can autonomously handle tactical maneuvers, freeing the human pilot to focus on strategy. In one 2023 test, a live flight of a training jet was controlled for a time by an AI agent, demonstrating a form of shared control between human and machine. Meanwhile, experimental command interfaces allow a single soldier to coordinate multiple robotic drones with high-level brain signals or simple gestures, with AI managing the low-level details. The goal of these trials is to augment a person’s effective reach and decision-making—essentially multiplying what one human can do through intelligent automation. Early results show that humans can adapt to trusting an AI partner for critical split-second tasks (e.g., air combat maneuvers) and that brain-control of swarms is feasible on a basic level. As this field progresses, we could see operational scenarios where a human operator, equipped with a BCI and AI assistance, performs the work of an entire team—a dramatic leap in human capability empowered by machines.

3. THE FUTURE OF BMI

3.1. BMI-Induced Metamorphosis of Humans

Numerous innovative enhancement technologies are in the theoretical or early development stage. While not yet realized, they are grounded in current research trends and could become feasible in the near future. Here, I would like to predict the possible future forms of humans by listing several possibilities.

Future human augmentation will not be limited to existing senses; it will likely involve granting humans entirely new ways to perceive the world. Engineers are developing bionic contact lenses that could give users continuous augmented reality vision or even the ability to see parts of the electromagnetic spectrum we normally cannot (IR, UV). Prototypes of AR contact lenses (such as those by Mojo Vision in the U.S.) have already been tested, foreshadowing discreet vision enhancements. Researchers also imagine wearable sensor suites feeding into the nervous system to provide novel senses—for example, a vibrotactile vest that gives a 360° proximity sense (alerting the wearer to movements or obstacles all around them), or an implant that vibrates to indicate cardinal direction (essentially a built-in compass). Some biohackers have experimented with magnetic implants to feel electromagnetic fields, hinting at how an added sense could work. Indeed, in my laboratory, we have successfully implanted a geomagnetic sensor into the brain of an animal and had it solve a maze relying on its absolute sense of direction³⁹⁾ (Fig. 2).

Fig. 2. Visual Cortical Prosthesis with a Geomagnetic Compass Restores Spatial Navigation in Blind Rats

Spatial navigation was restored in blind rats via artificial head-direction cues. Microstimulators connected to a digital compass were implanted in the rat visual cortex. The blind rats learned to use this real-time directional feedback to successfully navigate mazes, performing similarly to sighted rats, demonstrating sensory substitution’s potential. (A) Wiring diagram of the magnetic sensor brain chip. Two brain stimulation electrodes are connected to the digital magnetic sensor. The stimulation intensity can be changed in the range of 0–4V (volts), and it is equipped with a power switch and a rechargeable lithium battery. (B) External image of the magnetic sensor brain chip. (C) Image of the chip attached to the head of a rat. It is 25 mm long, 10 mm wide, 9 mm thick, and weighs 2.5 g. This figure is modified from a paper by Norimoto and Ikegaya.³⁹⁾

Other researchers have used sensory substitution interfaces to give blind individuals rudimentary “vision” through touch or sound; the same concepts could apply to sighted users to extend their perception. A plausible future device might let a user “feel” Wi-Fi signals or chemical traces in the air, analogous to how some animals sense magnetic fields or pheromones. These sensory extensions would fundamentally expand human experience and capabilities—a person could navigate in pitch darkness via IR sense or detect health-related signals from their body in real time. As sensor technology and brain interfacing improve, the boundary of human senses will broaden dramatically.

The coming years will likely see today’s exoskeletons evolve into more powerful, agile, and widely used gear. Researchers aim to create full-body powered suits that substantially increase strength, speed, and endurance while remaining practical for regular use. Advances in lightweight materials, energy-dense batteries, and soft robotics suggest that by 2030, a wearable exosuit could allow a person to run at Olympic sprinter speeds or carry a few hundred pounds with ease. Military organizations are funding next-generation combat suits that integrate ballistic protection with strength augmentation—essentially an “Iron Man” style armor that gives soldiers both defense and amplified power. For instance, China has publicly shown concept videos of powered armor for soldiers, and European projects (like in Germany and France) are exploring exoskeletons to help infantry move faster with heavy kit. In civil applications, future exoskeletons might enable construction workers to handle heavy tools and materials single-handedly, or allow a firefighter to carry multiple people out of danger. As these suits become more autonomous, AI will likely be embedded to coordinate the suit’s movement with the wearer’s intent, providing seamless assistance. The trajectory of development points toward a time when wearing a robotic suit could be as common for certain jobs as using a power drill—effectively making superhuman strength and endurance part of the standard human toolbox.

Farther on the horizon, futurists and some researchers envision a world where human brains interface directly with cloud computing and AI systems. In theory, a high-bandwidth wireless BCI could allow real-time communication between the brain and external databases or AI “co-processors.” This means a person could potentially access information instantly (without typing or even formulating a query in words)—for example, recalling specific data or images from the cloud as if it were their own memory. Such a brain-cloud link could also enable “software upgrades” to the human mind, like downloading a new language or skill set directly into neural patterns (a concept popularized in sci-fi, but being considered in brain simulation research). At a simpler level, a direct brain-AI connection might function as an ever-present cognitive assistant, monitoring one’s neural activity for signs of confusion or fatigue and injecting just-in-time knowledge or alertness boosts. Some experimental interfaces have already achieved one-way information transmission (e.g., sending visual data to the brain via neural implants for artificial vision), so extending this to bidirectional flow is a matter of overcoming technical challenges of bandwidth and decoding thoughts reliably. If achieved, the result would be a profound leap in human capabilities—effectively melding human intelligence with machine intelligence. A person could think of a complex problem and have an AI parallel-process it, returning suggestions directly into the person’s conscious train of thought. While still speculative, research by leading institutions and companies is steadily moving in this direction, indicating that brain-cloud interfaces are considered a plausible evolution of both AI and human augmentation technologies.

As humanity pushes further into space and extreme Earth environments, augmentation tech is expected to play a key role. Concepts are being explored for astronaut augmentation suits that might protect against radiation or muscle/bone loss by electrically stimulating the body during long missions. Exoskeletons and augmented reality will assist astronauts in performing complex repairs or construction in space by providing extra strength and real-time instruction overlays. Even neural control of robotic avatars is considered—for instance, an astronaut on Mars controlling multiple robotic rovers or drones via brain interface to multiply their reach across a wide area. On Earth, deep-sea divers or miners might wear exosuits that recycle breathing air and withstand pressure, effectively enlarging the habitable zone for humans. Enhancements like expanded spectral vision could allow disaster responders to see through smoke or detect vital signs behind walls. These future applications, while specialized, demonstrate the broad potential of human augmentation: to adapt our bodies and minds to environments and tasks previously beyond our capacity.

3.2. Dreamlike Futures Driven by BMI

I would like to predict the future neurotechnology I would like to see in 100 years. These technologies are unlikely to be realized in the near future. But the progress of science and technology has always exceeded our expectations. Therefore, with the hope that future progress will exceed expectations in a non-linear way, I predict the future world.

3.2.1. Dream Theater

Projected advances in non-invasive neural interfacing and artificial intelligence may enable the development of systems capable of recording, analyzing, and replaying human dream experiences. Such “dream theater” technology would likely use sophisticated sensors, perhaps integrated into sleep environments, to capture the complex patterns of neural activity, particularly during REM sleep, along with associated physiological markers. AI algorithms trained on large neuro-phenomenological datasets would then translate this bio-signal data into reconstructible multi-sensory formats, allowing users to retroactively experience their dreams in immersive virtual environments. Potential applications are many, ranging from therapeutic uses where individuals could explore unconscious content under clinical guidance, to significant advances in neuroscience research on consciousness and memory. In addition, the ability to archive and even share dream content (with explicit consent) could create new avenues for artistic expression, interpersonal connection, and potentially educational simulations based on symbolic dream structures.

3.2.2. 100% Synchronized Mobility

Personal transportation could evolve into a highly integrated, autonomous system of multimodal mobility pods capable of both ground and air travel. This “synchronized mobility” paradigm envisions personal devices seamlessly connected within a larger network managed by sophisticated AI. These pods would use advanced propulsion systems to enable transitions between, for example, ground-level magnetic conduits and low-altitude flight paths. Crucially, the system would operate with near-total automation, optimizing routes, speed, and mode of travel based on real-time network traffic data, environmental conditions, user destination inputs, and potentially even the user’s physiological state as monitored by integrated biometric sensors. The goal is to provide efficient, safe, and comfortable transportation tailored to individual needs, eliminating traditional traffic congestion and minimizing travel time. Potential societal benefits include dramatically improved transportation efficiency, increased accessibility for diverse populations, reduced accident rates, and the possibility of significant urban redevelopment by reclaiming space currently dedicated to traditional road infrastructure.

3.2.3. Delivering the Five Senses

Future communication technologies may transcend current audio-visual limitations to incorporate the transmission and reproduction of the full range of human senses—sight, sound, touch, smell, and taste—over distances. This concept requires the development of advanced sensor suites capable of capturing detailed haptic information (pressure, texture, temperature), complex olfactory and gustatory data (via sophisticated chemosensors), and high-fidelity audio-visual streams. This multi-sensory data would require high-bandwidth transmission and sophisticated receiver interfaces, including full-body haptic suits, spatial audio systems, high-resolution visual displays, and controlled olfactory and gustatory simulators capable of replicating specific smells and tastes based on the transmitted chemical analyses. Such technology would enable true telepresence, allowing individuals to experience remote environments or share sensory experiences with others with unprecedented realism. Applications include remote collaboration requiring tactile feedback (e.g., telesurgery, complex repairs), deeply immersive education and training, highly realistic virtual tourism, enhanced social interaction for geographically separated individuals, and new opportunities for accessibility.

3.2.4. Superhuman Sports Competition

The next century could see the emergence of novel sports competitions that are fundamentally different from current sports, enabled by virtual reality (VR), augmented reality (AR), accessible zero-gravity environments, and human-machine integration. These “superhuman sports” could use fully immersive virtual arenas where physics can be altered, or use AR overlays on physical fields to add dynamic information or interactive elements. Dedicated facilities that provide microgravity or zero gravity conditions would enable sports that emphasize three-dimensional movement and unique biomechanics. In addition, athletes could use performance-enhancing technologies, such as sophisticated exoskeletons for increased strength or agility, or neural interfaces for faster reaction times, and compete in specific regulated classes. A significant development could be the integration of advanced robotics and AI, leading to formats where humans work with robotic teammates, requiring complex inter-agent strategies, or compete against highly capable AI opponents. These competitions would push the limits of human physical and cognitive performance, offer new forms of spectacular entertainment, potentially experienced through immersive spectator interfaces, and serve as platforms for research in areas such as robotics, AI strategy, and human augmentation.

3.2.5. Instantaneous Skill Acquisition

Developments in BCI technology and cognitive neuroscience could potentially lead to methods for transferring complex skills and knowledge directly into the human brain, bypassing traditional learning processes. This “instantaneous skill acquisition” would hypothetically involve BCIs capable of precise neural stimulation, guided by AI systems that translate curated skill datasets (e.g., machine operating procedures, linguistic knowledge) into specific patterns of neural activity. The goal would be to induce or accelerate the formation of the neural correlates associated with the desired skill or knowledge base. If realized, this could drastically reduce the time required to achieve proficiency in complex tasks, facilitate rapid workforce adaptation, enhance emergency response capabilities by enabling personnel to quickly learn to use unfamiliar equipment, and potentially democratize access to specialized expertise.

3.2.6. Telepathy

BCIs combined with sophisticated AI could enable a form of technologically mediated telepathy, facilitating direct brain-to-brain communication. This system would aim to decode the neural correlates of thoughts, intentions, and possibly emotional states in an individual and transmit this information, possibly via an intermediary AI translator that converts it into a standardized neural code or directly stimulates the recipient’s BCI to evoke the intended meaning. Such a technology could transcend language barriers and allow for richer, more nuanced, and faster communication. In theory, research into mapping the neural or analog signaling systems of non-human organisms could enable rudimentary forms of interspecies communication, allowing humans to interpret basic states or intentions in animals, or even to analyze collective signals from plants or microbial communities. Potential applications include universal translation, enhanced collaboration, novel therapeutic tools for communication disorders, and the promotion of deeper ecological awareness.

3.2.7. Virtual Space-Time Warp

Leveraging mature VR and comprehensive sensory simulation technologies, it may become possible to create highly immersive and interactive reconstructions of past events, environments, and individuals. This “Virtual Time Machine” concept relies on powerful AI engines to process massive data sets—historical records, archaeological findings, paleontological data, climate models, potentially even digitized personal archives—to generate high-fidelity simulations. Users equipped with advanced VR displays and full sensory feedback systems (haptic, olfactory, gustatory) could experience these simulations with a strong sense of presence. Functionality could include exploring historical settings such as ancient cities, witnessing significant events of the past, encountering realistic simulations of extinct creatures based on scientific data, or even interacting with AI-driven reconstructions of deceased individuals modeled from their available data. Primary applications are in education, providing embodied learning experiences far beyond traditional methods, and potentially in therapeutic contexts such as grief counseling (though ethically sensitive). It could also be used for entertainment, cultural preservation, and professional training.

3.3. Obsolescence of Man

The relentless pace of scientific advancement, particularly in the fields of neurotechnology, BMI/BCI, and human augmentation, compels a perspective beyond immediate clinical or practical applications. While detailed empirical findings are the foundation of scientific progress, synthesizing the trajectory indicated by these converging fields reveals profound implications that challenge our most fundamental understanding of human identity and potential. An analysis of current capabilities-restoring complex motor function via neural bypasses, decoding neural correlates of language, augmenting sensory input and physical capacity, achieving rudimentary brain-to-brain information transfer-points to a conclusion of startling significance: the biological blueprint of Homo sapiens no longer dictates the necessary or ultimate limits of human existence. The assumption that we must remain confined to our current form is being actively dismantled by our own technological ingenuity.

For centuries, the study of the human condition has sought to identify an immutable “human nature”—a core set of characteristics defined by our biology. For example, reason, complex language, the use of tools, certain modes of consciousness, and the inescapable realities of aging and mortality might have been considered definitional constraints. Based on these biological limitations and possibilities, complex social, ethical, and political frameworks have been constructed. However, the data presented from BMI and augmentation research does not merely refine these frameworks; it fundamentally destabilizes them. It empirically demonstrates that “human” functionality is not a static property, but a remarkably adaptive system capable of radical expansion and modification when integrated with technology. Our current biological state increasingly appears to be a contingent starting point, not a fixed destiny.

Consider the strong evidence for the decoupling of intention and cognition from their native biological substrate. When cortical signals representing the intention to walk are successfully decoded and used to drive spinal cord stimulators or prosthetic limbs, bypassing damaged neural pathways, we are witnessing more than therapeutic success. This demonstrates the principle of functional substrate independence in action. The cognitive process achieves its functional goal via an artificial medium, proving that the execution of intention is not inextricably bound to the original organic hardware. The central nervous system turns out to be an astonishingly adaptable control system, capable of interfacing with and directing non-biological components. This empirical reality forces us to confront the core idea: there is no intrinsic biological mandate that requires humans to remain solely within the operational boundaries of their birth-assigned physiology.

The trajectory of human augmentation reinforces this conclusion. Exoskeletons that augment strength, augmented reality systems that fuse digital information directly into perception, BCIs that translate thought into control signals, and even experimental integrations such as supernumerary robotic limbs demonstrate the profound morphological and cognitive plasticity achievable through technological mediation. These are not merely external devices; they represent nascent forms of human-machine integration that extend embodiment, sensory range, and cognitive capabilities in ways that redefine the user’s interaction with the world. The boundary between the biological “self” and the artificial “tool” becomes functionally blurred, potentially irrelevant. This leads to a necessary reconceptualization of human potential. It is no longer just about optimizing innate biological talents or pushing the limits of unaided physiology. Instead, human potential becomes an open-ended design and engineering challenge—a question of what capabilities can be achieved through the synergistic integration of biological systems with engineering technologies, limited more by ingenuity and physical laws than by basic biology. The potential is no longer just within us; it is constructible.

This demonstrated malleability forces a radical reassessment of what constitutes the meaning of human existence from a functional and operational perspective. If core biological limitations—physical frailty, sensory limitations, cognitive processing speeds, even the barriers of communication—can be systematically overcome or augmented by technology, then the very challenges that have historically defined the human struggle lose their centrality. What defines our existence when the timeline of mortality becomes elastic, when knowledge acquisition bypasses laborious learning (“Instant Skill Acquisition”), when communication transcends linguistic ambiguity (“Telepathy”), or when subjective experience can be curated and shared with unprecedented fidelity (“Delivering the Five Senses,” “Dream Theater”)? The operational definition of being human is shifting. The “meaning” derived from navigating a specific set of biological constraints is replaced by the meaning found in exploring a vastly expanded state space of possible functions, experiences, and interactions enabled by technology.

As a result, the meaning and purpose of human life has changed—or, more accurately, the parameters that define possible meanings and purposes are undergoing a revolutionary expansion. Historically, purpose was often framed by biological imperatives (survival, reproduction, adaptation to the environment) and the pursuit of knowledge or fulfillment within those constraints. Technology now introduces the possibility of designed purpose. Meaning can increasingly be derived from participating in technologically mediated realities, contributing to hybrid cognitive networks (‘BrainNet’, AI co-pilots), exploring novel sensory inputs or engineered environments (‘Virtual Space-Time Warp’), or even deliberately shaping our own ongoing evolution. The narrative arc of human life, traditionally bounded by biological milestones, is subject to technological revision. Purpose is transformed from a found condition to a designed goal, potentially selected from a menu of technologically enabled possibilities that far exceeds biological precedent.

What, then, will be the future operational role of entities derived from the Homo sapiens lineage? Extrapolation of current trends—toward deeper brain-computer interfacing, enhanced physical and cognitive augmentation, possibly even brain-cloud integration—suggests that biological humans may be a transitional form. We may be precursors of successor entities that seamlessly integrate biology and technology and possess capabilities (augmented senses, radically enhanced cognition, networked consciousness, indefinite operational lifespans) that make them functionally distinct from us. This is not a dystopian prediction, but a plausible trajectory based on observed scientific progress and capability expansion.

Acknowledging this trajectory requires intellectual courage that goes beyond reflexive fears of “losing humanity. What specific aspects of our current biologically constrained existence do we implicitly defend as sacrosanct? Vulnerability to disease and decay? Cognitive biases hardwired by evolutionary pressures? The inherent limitations of biological computation and communication? Emerging technologies offer potential pathways beyond these limitations. A rigorous scientific perspective compels us to analyze this potential without premature ethical closure, focusing instead on the functional and evolutionary implications.

Ongoing research into brain-to-brain communication, cognitive augmentation through AI partnership, and targeted neuroplasticity represent more than technological development; they are explorations into the fundamental nature of cognition, individuality, and cooperation. They explore the possibility that intelligence will become a networked, hybrid phenomenon, shifting the primary unit of agency from the individual biological organism to integrated human-machine systems. The future role of humans may be less about individual capabilities and more about contributing specialized biological processing or subjective experience within a larger cognitive architecture, tackling challenges on a scale previously unimaginable.

While highly futuristic concepts such as widespread telepathy or instantaneous skill downloads remain speculative, they represent logical extensions of demonstrated principles: increasingly precise neural decoding and stimulation, sophisticated AI pattern analysis, and growing understanding of the neural correlates of cognition and skill. To dismiss such possibilities is to ignore the consistent historical pattern of technological acceleration, where today’s theoretical frontier becomes tomorrow’s technological reality.

Thus, the central insight-that there is no immutable biological imperative for humans to remain as they are—is not an abstract philosophical notion, but an emergent consequence of scientific and technological progress. It represents a potential turning point in the history of life, in which a species develops the capacity to actively and consciously redesign its own functional golem, transcending the limitations imposed by its evolutionary heritage. We are potentially witnessing the dawn of conscious, technologically mediated evolution. The critical task now is not merely to refine these technologies, but to grapple with the profound transformation they imply for human potential, the operational meaning of human existence, and the radically altered landscape of the meaning and purpose of human life. Rejecting the notion of a fixed human essence allows us to confront the possibility that we are not the endpoint of evolution, but perhaps the initiators of a new, uncharted phase-one defined not by the constraints we inherit, but by the capabilities we choose to build. This perspective, grounded in scientific reality while acknowledging its paradigm-shifting implications, offers a truly breakthrough view of our potential future.

4. CONCLUSION

In this review, I have described the history of brain-machine fusion from its earliest days and made predictions about the future. I have been fully aware that the realization of these boldly drawn future scenarios—ranging from deeply personal dream theaters and hyper-efficient synchronized mobility to the transmission of full sensory experiences, the emergence of superhuman sports, the possibility of instantaneous skill acquisition, technologically mediated telepathy, and immersive virtual journeys through time—would undoubtedly trigger a cascade of events. These technologies would undoubtedly trigger a cascade of profound and complex ethical questions requiring rigorous societal debate, careful governance, and the establishment of entirely new moral frameworks. This review article deliberately focuses primarily on articulating the sheer technological potential and visionary scope inherent in these concepts. In other words, the intention here is first to map out the frontiers of what might be possible from an innovation standpoint, stimulating thought about future capabilities and their transformative applications in fields as diverse as therapy, transportation, communication, entertainment, education, and even basic human connectivity (I personally look forward to the future that the Zoromes in “The Jameson Satellite” will realize safely and beneficially). Therefore, deliberately reserving the equally crucial, and necessarily nuanced analysis of their multifaceted ethical implications—ranging from privacy and consent to equity, autonomy, and the very definition of human experience—for a separate, dedicated, and more appropriately detailed examination, rather than diluting the current exploration of forward-looking technological trajectories.

Acknowledgments

This article was prepared in commemoration of receiving the Award of the Pharmaceutical Society of Japan. The author wishes to express sincere gratitude to the members of the award selection committee, all relevant parties, and everyone who contributed to this research. This work was supported by JST ERATO (JPMJER1801), the Institute for AI and Beyond of the University of Tokyo, and AMED Brain/MINDS 2.0 (24wm0625207s0101; 24wm0625401h0001; 24wm0625502s0501). In preparing this manuscript, the author partly utilized several models of generative artificial intelligence, including their “deep research” functions, to search for literature, write the draft, and proofread.

Conflict of Interest

The author declares no conflict of interest.

Notes

This review of the author’s work was written by the author upon receiving the 2025 Pharmaceutical Society of Japan Award.

REFERENCES

• 1) Lebedev MA, Nicolelis MAL. Brain-machine interfaces: From basic science to neuroprostheses and neurorehabilitation. Physiol. Rev., 97, 767–837 (2017).
• 2) Caton R. The electric currents of the brain. Chicago J. Nerv. Mental Disease, 2, 610 (1875).
• 3) Tudor M, Tudor L, Tudor KI. Hans Berger (1873–1941)—the history of electroencephalography. Acta Med. Croatica, 59, 307–313 (2005).
• 4) Fetz EE. Operant conditioning of cortical unit activity. Science, 163, 955–958 (1969).
• 5) Moritz CT, Fetz EE. Volitional control of single cortical neurons in a brain-machine interface. J. Neural Eng., 8, 025017 (2011).
• 6) Lucier A. Music for solo performer. Lovely Music, Lovely Communications, Ltd. (1982). https://www.discogs.com/release/758828-Alvin-Lucier-Music-For-Solo-Performer-For-Enormously-Amplified-Brain-Waves-And-Percussion?srsltid=AfmBOorOmR6-UsheIJkd8VCee2uIfZJQIu0kb5yW3piHtIqdwzkGgRrX
• 7) Vidal JJ. Toward direct brain-computer communication. Annu. Rev. Biophys. Bioeng., 2, 157–180 (1973).
• 8) Vidal JJ. Real-time detection of brain events in EEG. Proc. IEEE Inst. Electr. Electron. Eng., 65, 633–641 (1977).
• 9) Bozinovski S, Sestakov M, Bozinovska L. Using EEG alpha rhythm to control a mobile robot, Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, New Orleans, LA, USA, on 1988/11/4-1988/11/7, 1988, IEEE.
• 10) Bozinovska L, Stojanov G, Sestakov M, Bozinovski S. CNV pattern recognition: step toward a cognitive wave observation. Proc. Eur. Signal Process. Conf., 1659–1662 (1990).
• 11) Georgopoulos AP, Schwartz AB, Kettner RE. Neuronal population coding of movement direction. Science, 233, 1416–1419 (1986).
• 12) Stanley GB, Li FF, Dan Y. Reconstruction of natural scenes from ensemble responses in the lateral geniculate nucleus. J. Neurosci., 19, 8036–8042 (1999).
• 13) Chapin JK, Moxon KA, Markowitz RS, Nicolelis MA. Real-time control of a robot arm using simultaneously recorded neurons in the motor cortex. Nat. Neurosci., 2, 664–670 (1999).
• 14) Wessberg J, Stambaugh CR, Kralik JD, Beck PD, Laubach M, Chapin JK, Kim J, Biggs SJ, Srinivasan MA, Nicolelis MA. Real-time prediction of hand trajectory by ensembles of cortical neurons in primates. Nature, 408, 361–365 (2000).
• 15) Lorach H, Galvez A, Spagnolo V, et al. Walking naturally after spinal cord injury using a brain-spine interface. Nature, 618, 126–133 (2023).
• 16) Rebsamen B, Guan C, Zhang H, Wang C, Teo C, Ang MH Jr, Burdet E. A brain controlled wheelchair to navigate in familiar environments. IEEE Trans. Neural Syst. Rehabil. Eng., 18, 590–598 (2010).
• 17) Phillip A. A paralyzed woman flew an F-35 fighter jet in a simulator—using only her mind, 2015.
• 18) Flesher SN, Downey JE, Weiss JM, Hughes CL, Herrera AJ, Tyler-Kabara EC, Boninger ML, Collinger JL, Gaunt RA. A brain-computer interface that evokes tactile sensations improves robotic arm control. Science, 372, 831–836 (2021).
• 19) Willett FR, Avansino DT, Hochberg LR, Henderson JM, Shenoy KV. High-performance brain-to-text communication via handwriting. Nature, 593, 249–254 (2021).
• 20) Metzger SL, Littlejohn KT, Silva AB, Moses DA, Seaton MP, Wang R, Dougherty ME, Liu JR, Wu P, Berger MA, Zhuravleva I, Tu-Chan A, Ganguly K, Anumanchipalli GK, Chang EF. A high-performance neuroprosthesis for speech decoding and avatar control. Nature, 620, 1037–1046 (2023).
• 21) Barrie R. Three companies to rival Neuralink in the BCI clinical trial landscape, 2024.
• 22) Chaudhary U, Vlachos I, Zimmermann JB, Espinosa A, Tonin A, Jaramillo-Gonzalez A, Khalili-Ardali M, Topka H, Lehmberg J, Friehs GM, Woodtli A, Donoghue JP, Birbaumer N. Spelling interface using intracortical signals in a completely locked-in patient enabled via auditory neurofeedback training. Nat. Commun., 13, 1236 (2022).
• 23) Rao VR, Rolston JD. Unearthing the mechanisms of responsive neurostimulation for epilepsy. Commun. Med. (Lond.), 3, 166 (2023).
• 24) Scangos KW, Khambhati AN, Daly PM, Makhoul GS, Sugrue LP, Zamanian H, Liu TX, Rao VR, Sellers KK, Dawes HE, Starr PA, Krystal AD, Chang EF. Closed-loop neuromodulation in an individual with treatment-resistant depression. Nat. Med., 27, 1696–1700 (2021).
• 25) Schreiner L, Wipprecht A, Olyanasab A, Sieghartsleitner S, Pretl H, Guger C. Brain-computer-interface-driven artistic expression: real-time cognitive visualization in the pangolin scales animatronic dress and screen dress. Front. Hum. Neurosci., 19, 1516776 (2025).
• 26) Jade L, Gentle S. New ways of knowing ourselves. BCI facilitating artistic exploration of our biology. Brain Art. Springer International Publishing, Cham, pp. 229–262 (2019).
• 27) Fazel-Rezai R, Allison BZ, Guger C, Sellers EW, Kleih SC, Kübler A. P300 brain computer interface: current challenges and emerging trends. Front. Neuroeng., 5, 14 (2012).
• 28) Bakkum DJ, Gamblen PM, Ben-Ary G, Chao ZC, Potter SM. MEART: The semi-living artist. Front. Neurorobot., 1, 5 (2007).
• 29) Sargeant B, Dwyer J, Mueller F. ‘floyd’. The storytelling machine: A playful participatory automated system featuring crowd-sourced story content, Proceedings of the 2018 Annual Symposium on Computer-Human Interaction in Play Companion Extended Abstracts, Melbourne VIC Australia, on 2018/10/28–2018/10/31, 23 October, 2018, ACM, New York, NY, U.S.A.
• 30) Yamashiro K, Matsumoto N, Ikegaya Y. Diffusion model-based image generation from rat brain activity. PLOS ONE, 19, e0309709 (2024).
• 31) Scharping N. Disabled musicians make music with their minds, 2016.
• 32) Deuel TA, Pampin J, Sundstrom J, Darvas F. The Encephalophone: a novel musical biofeedback device using conscious control of electroencephalogram (EEG). Front. Hum. Neurosci., 11, 213 (2017).
• 33) Neuralink monkey plays MindPong telepathically, 2021.
• 34) Kagan BJ, Kitchen AC, Tran NT, Habibollahi F, Khajehnejad M, Parker BJ, Bhat A, Rollo B, Razi A, Friston KJ. In vitro neurons learn and exhibit sentience when embodied in a simulated game-world. Neuron, 110, 3952–3969.e8 (2022).
• 35) Stein S. Moving objects with Mattel’s brainwave-reading Mindflex, 2009.
• 36) Kieliba P, Clode D, Maimon-Mor RO, Makin TR. Robotic hand augmentation drives changes in neural body representation. Sci. Robot., 6, eabd7935 (2021).
• 37) Jiang L, Stocco A, Losey DM, Abernethy JA, Prat CS, Rao RPN. BrainNet: A multi-person brain-to-brain interface for direct collaboration between brains. Sci. Rep., 9, 6115 (2019).
• 38) Krause MR, Zanos TP, Csorba BA, Pilly PK, Choe J, Phillips ME, Datta A, Pack CC. Transcranial direct current stimulation facilitates associative learning and alters functional connectivity in the primate brain. Curr. Biol., 27, 3086–3096.e3 (2017).
• 39) Norimoto H, Ikegaya Y. Visual cortical prosthesis with a geomagnetic compass restores spatial navigation in blind rats. Curr. Biol., 25, 1091–1095 (2015).