The brain creates our movements, sensations, decisions, and much of what we recognize as personality. When a particular brain circuit fails, the consequences can be devastating. But understanding how those circuits work also creates opportunities to repair function, bypass an injury, or build an entirely new route for interacting with the world.
Brain-machine interfaces sit at that intersection of neuroscience, engineering, and medicine. Their immediate promise is not downloading knowledge or reading every thought. It is something more concrete: allowing a person with paralysis to operate a computer using the intention to move.
My discussion with neurosurgeon Dr. Matthew MacDougall explored both that practical goal and the much larger possibilities beyond it. The important distinction throughout is between what a device is designed to do, what has been demonstrated in research, and what remains a long-term aspiration. The Neuralink development plans described here reflect the stage of the program at the time of our conversation, rather than a current clinical-status update.
Key Takeaways
- Brain-machine interfaces can translate movement intentions into computer commands without restoring movement to paralyzed limbs.
- Robotic electrode placement addresses precision demands beyond the capabilities of a surgeon’s hands.
- Effective computer control depends on both software adaptation and the person learning to use the interface.
- Restoring bodily movement and expanding cognition remain distinct, longer-term goals beyond the initial computer-control application.
Table of Contents
- The Brain Is a Collection of Functions That Can Fail Separately
- How MRI-Guided Lasers Can Treat Difficult Brain Tumors
- Neuralink’s Initial Goal: Computer Control for People With Paralysis
- How the Brain and Software Learn to Work Together
- Implant Safety and the Difference Between Treatment and Enhancement
- Could an Interface Reconnect the Brain to the Body?
- Brain-to-Brain Communication Without the Mind-Reading Myth
- Why Pigs and Monkeys Were Used in the Research
- The Longer-Term Vision: Mental Health, Cognition, and Communication
The Brain Is a Collection of Functions That Can Fail Separately
Neurosurgery provides a particularly direct view of the relationship between brain tissue and behavior. The brain is not one undifferentiated organ performing a single task. It contains interconnected systems that support vision, sensation, hormone regulation, movement, impulse control, and many other functions.
MacDougall described a young patient who suffered severe injuries in a car accident. Surgery helped preserve his life, but damage to both frontal lobes left him with profoundly impaired impulse control. During recovery, he made an inappropriate sexual demand involving an elderly nurse.
The point of that story is not the unusual behavior itself. It is that the ability to regulate behavior depends on functioning brain circuitry. A person can survive a brain injury while losing capacities that are central to independent life and social interaction.
For background on the major brain regions and their functions, the National Institute of Neurological Disorders and Stroke provides an accessible guide to brain anatomy.
This is what makes neurosurgery both extraordinarily high stakes and deeply rewarding. A small injury can have enormous consequences. Conversely, treating the right problem can sometimes restore function in ways that once seemed impossible.
How MRI-Guided Lasers Can Treat Difficult Brain Tumors
One example of that progress is the use of laser treatment for selected brain tumors.
Traditionally, reaching a small tumor deep in the brain could require passing through healthy tissue. That risks damaging neurons and white matter, the connections that allow different brain regions to communicate. Sometimes the route to a tumor can be as concerning as the tumor itself.
MacDougall described a less invasive approach:
- A small opening, approximately 2 millimeters across, is made in the skull.
- A thin fiber-optic probe is directed toward the tumor.
- Laser energy heats the targeted tissue while the patient is inside an MRI scanner.
- Repeated MRI measurements allow the surgical team to monitor temperature and control the treatment.
The aim is to heat the tumor enough to destroy its cells while minimizing injury to surrounding brain tissue.
In selected cases, this can make treatment possible in locations where conventional surgery would carry substantial risk, including the brainstem or regions involved in movement and speech. It does not mean every tumor can be treated this way. It illustrates a broader principle: better tools can change what is surgically accessible without requiring a more destructive operation.
Neuralink’s Initial Goal: Computer Control for People With Paralysis
Neuralink tends to attract two extremes of expectation: extraordinary hopes about human enhancement and extraordinary fears about control of the mind. MacDougall’s description of the initial medical goal was considerably more specific.
Consider someone with a severe spinal cord injury who cannot move their arms or legs. The brain regions responsible for planning and initiating movement may still function, but the pathway carrying those instructions to the body has been interrupted.
The intended movement still produces activity in the motor cortex. A brain-machine interface attempts to record that activity and translate it into a useful output.
The first application MacDougall outlined was control of a computer mouse and keyboard. A person might attempt or imagine a hand or wrist movement, and software would translate the associated neural signals into cursor movement or a click.
That could allow someone to open a word processor, select letters on an onscreen keyboard, communicate, and use the internet without moving their hands.
This is not the same as repairing the spinal cord or restoring movement to the person’s limbs. It is an alternative output pathway from the brain to a digital device. Restoring access to communication and computing is itself a meaningful form of independence.
Why a Robot Places the Electrodes
Recording these signals requires electrodes small enough to interact with brain tissue on a very fine scale. The threads MacDougall described are thinner than a human hair, with tiny loops that a placement system must grasp and position.
At that scale, the limitations of a human hand become important. The brain’s surface also contains densely distributed blood vessels, and electrode placement requires precise control of location and depth.
A robotic insertion system is therefore not simply a convenience. It addresses a task that exceeds the steadiness and precision available to a surgeon working manually.
In the surgical workflow MacDougall described, the neurosurgeon still opens the skin and skull and prepares the brain surface. The robot performs the specialized electrode-placement portion. That is a narrower role than an autonomous machine carrying out an entire brain operation.
How the Brain and Software Learn to Work Together
An implanted electrode does not arrive with an instruction manual explaining what every recorded neuron means. Initially, the software receives patterns of electrical activity without knowing which patterns correspond to which intended movements.
Training provides that relationship.
For example, a person can be asked to move a cursor toward a target in the upper-right corner of a screen. Because the task specifies the goal, the software can look for neural activity associated with the intention to move in that direction. It repeats the process for other directions and actions.
The learning process has two sides:
- The decoder adapts: Software identifies patterns that help predict the intended movement.
- The person adapts: The brain learns how to use the feedback provided by the cursor’s behavior.
MacDougall likened the challenge to playing a piano whose keys gradually change which notes they produce. The mapping is not perfectly fixed, yet a sufficiently adaptable system can learn to operate it.
This matters because recorded signals can shift. Turning a device off overnight and returning the next day may not produce exactly the same set of signals. The goal is a working relationship between brain and decoder that can accommodate those changes.
At the time of our conversation, MacDougall described monkeys using neural signals to play computer games for smoothie rewards. He also reported high information-transfer performance in those experiments. Those results supported the functional approach in animals, but they should not be confused with a demonstration of every proposed human application.
Implant Safety and the Difference Between Treatment and Enhancement
Would a neurosurgeon who develops a brain implant be willing to receive one? MacDougall said he would, while making a useful distinction between confidence in a technology and having a good reason to use it.
If someone can already operate a mouse and keyboard with their hands, a device that reproduces those abilities may offer little additional benefit. For a person unable to move their limbs, the same capability can be transformative.
That is the difference between a medical application and elective enhancement. The potential benefit is not the same for every person.
MacDougall expressed strong personal confidence in the implant’s safety based on his surgical experience during development, comparing it favorably with some established neurosurgical procedures. That assessment is his professional judgment, not a guarantee of safety or a substitute for clinical evidence.
He suggested that enhancement would become a more compelling proposition if an interface eventually enabled something a person could not otherwise do, such as communicating faster than with existing typing or speech-based tools. That was a future use case, not a capability established in our discussion.
A Simpler Example: An RFID Implant in the Hand
MacDougall already had a small implanted technology in his hand: a writable radio-frequency identification, or RFID, tag.
At different times, he had used it to unlock his front door and gain access to doors at Neuralink. A nearby reader temporarily powers the passive tag and exchanges a small amount of data with it.
Unlike a battery-powered implant, it does not need periodic charging. He described the tag as enclosed in biocompatible glass, with an additional silicone layer he added because he was concerned about breaking the glass while rock climbing.
He expected it could remain in place indefinitely, although newer versions with more storage might eventually make an upgrade attractive.
An RFID tag is not comparable in complexity or surgical risk to a brain implant. Its relevance is more modest: it is a practical example of implanted technology providing an everyday function, and of MacDougall’s willingness to personally use a form of augmentation.
Could an Interface Reconnect the Brain to the Body?
Computer control bypasses the need for the limbs to move. A more ambitious goal is to restore communication between the brain and the body below a spinal cord injury.
The proposed architecture is straightforward to describe, even though implementing it is difficult:
- Record movement-related signals from the motor cortex.
- Transmit the relevant information around the damaged section of spinal cord.
- Use a second implant to stimulate spinal circuitry below the injury.
The idea is to make use of neural pathways that remain intact below the damaged area. Rather than repairing the injured segment itself, the system would create a bridge around it.
MacDougall described work on spinal cord stimulation as part of this longer-term direction. But he repeatedly distinguished it from the initial computer-control application. Moving a cursor and reconnecting someone to their own muscles are different technical and clinical problems.
That distinction is especially important when discussing paralysis. A promising route toward restoring movement is not the same as an established ability to make someone walk again.
Brain-to-Brain Communication Without the Mind-Reading Myth
The phrase “brain-to-brain communication” sounds like science fiction. One version, however, is conceptually much closer to ordinary texting.
If two people can each use a brain interface to operate a keyboard, they can exchange messages without speaking or moving their hands. That is communication through neural interfaces, but it does not require one person to gain unrestricted access to the other’s thoughts.
MacDougall also described a more ambitious possibility: recording signals from brain regions involved in speech, translating intended communication into language, and delivering it to another person through an audio device, perhaps using bone conduction.
These possibilities should not be collapsed into the idea that every private thought becomes available to someone else. Our discussion did not establish a method for reading arbitrary thoughts or explain a complete system for controlling access to them. The grounded starting point is deliberately communicating through a device, not involuntary mental transparency.
Why Bandwidth Matters for Future Brain Interfaces
The larger vision depends partly on scale. Recording from more neurons could provide a richer channel between the nervous system and a computer.
MacDougall’s interest was in technology that might eventually expand to very large numbers of neural connections. A high-bandwidth interface could, in principle, support more than cursor movement, including tools that assist cognition or provide access to artificial intelligence.
That remains a research direction. The possibility of scaling an interface is one reason to pursue it, but it is not proof that every imagined application will work.
Why Pigs and Monkeys Were Used in the Research
Animal research is one of the most emotionally and ethically charged aspects of developing brain implants. MacDougall’s position was that the team would prefer an effective alternative, but that animal testing was part of establishing safety and function before human use.
He also placed the issue in the broader context of society’s use of animals for food, arguing that the comparatively small number used in medical research should be considered against that background. That is an ethical argument, not a resolution of every concern about animal experimentation.
The scientific roles of the animals he described were different.
Pigs: Studying Implant Safety
Pigs provided a biological setting with skull dimensions and characteristics useful for testing the implant and surgical approach.
The main questions included whether the device could be implanted safely, remain intact, and avoid harmful tissue reactions. The goal was not primarily to decode complex cognition from pigs.
Some pigs were also trained to walk on a treadmill for work related to limb movement and spinal cord research.
Monkeys: Studying Functional Computer Control
Monkeys were used to test whether the recorded signals could support useful control of a computer task. Their role was to help validate the functional chain from neural recording to decoding to cursor movement.
MacDougall described behavioral experiments designed around voluntary participation. According to his account, animals had access to normal food and water, and the smoothie rewards were additional treats rather than a response to deprivation. If a monkey did not want to participate in a particular session, he said, it was not compelled to do so.
He explicitly excluded surgery from that description of voluntary participation. That distinction matters: an opt-in behavioral task does not make every part of an animal research program voluntary.
These welfare practices were his description of the program, rather than an independent assessment. They nevertheless identify a meaningful design question for research: how much agency can an animal retain while researchers obtain the information needed to evaluate a medical device?
The Longer-Term Vision: Mental Health, Cognition, and Communication
When asked about the future, MacDougall began with suffering rather than superhuman abilities.
He identified addiction, depression, suicide, and brain-related contributors to obesity as problems he would like to help address over the course of his career. His aspiration was to give people more control over damaging patterns of brain function.
He imagined progress over a roughly 10- to 20-year horizon, but that was a statement of hope and direction, not evidence that the interface discussed can treat those conditions or a promise of when effective treatments will arrive.
Further out, his vision included more direct access to information, AI tools integrated with cognitive activity, and communication less constrained by speech and typing.
There is a real bottleneck here: our thoughts must usually be translated into words, movements, or typed characters before another person or a computer can use them. Speech is powerful, but it is also limited and vulnerable to misunderstanding. A richer interface might eventually change that relationship.
MacDougall even entertained voluntary networks of connected minds, while acknowledging that such possibilities were far enough away that he could not specify how they would be implemented.
The useful way to hold these ideas is neither to dismiss them automatically nor to treat them as inevitable. They are possibilities that motivate research. The work in front of us remains much more concrete: safe electrodes, precise surgery, stable recordings, effective decoding, and meaningful benefits for people with serious medical needs.
The most important near-term promise of a brain-machine interface is agency. For someone who cannot move, converting an intention into a message, a selection, or an action on a screen can reopen a route to the world. That is a substantial goal on its own, and a foundation from which larger possibilities can be tested.
