More than a million people hear through an electrode array placed inside the ear.
A paralysed woman has spoken at 78 words a minute using signals read from the surface of her brain.
The body is not a boundary. It is an interface, and we have begun to learn its protocol.
In the 1940s Norbert Wiener named a science after the Greek word for a ship's steersman. His point was that a living thing and a machine stay on course the same way: act, sense the result, correct. He called the study of such loops cybernetics.
The idea is everywhere now, from thermostats to aircraft to the reflex that keeps you upright. Applied to the body, it gives a precise way to think about prosthetics and implants. A device that only sends commands is a tool. A device that also returns sensation becomes part of the person.
A thermostat, a pilot, a heartbeat and a hand reaching for a cup are all this loop. What the loop corrects is the difference between where you are and where you meant to be. Cut the return path and the system is blind: it keeps acting, and it drifts. The model on this page lets you cut it.
Source: Norbert Wiener, Cybernetics (1948) · cited from memory, to verify
The cochlear implant is the most successful neural interface built so far. A row of about twenty electrodes replaces thousands of sensory cells, and the brain learns to hear speech through it. Deep brain stimulators steady the tremor of Parkinson's disease in a couple of hundred thousand people. Pacemakers are so common they are no longer thought of as cybernetics at all.
Reading intention from the brain is newer and rarer. Only a few dozen people have had electrode arrays placed in the cortex for research. With them, people who cannot move have controlled cursors, robot arms and their own paralysed limbs.
Writing to the nervous system is routine medicine. Reading from the cortex is still a handful of volunteers.
Source: Zeng, Celebrating the one millionth cochlear implant, JASA Express Letters (2022); published estimates for deep brain stimulation and intracortical research participants · cited from memory, to verify
Most of these devices only write: they send a signal in and never listen. A few only read. The ones that do both, such as the pacemaker and the artificial pancreas, are the oldest true cybernetic implants, and the least remarked. The direction of the field is to turn every filled dot into a closed loop.
Source: Clinical literature for each device class; uses listed are established indications, except stepping after spinal injury, which is experimental · cited from memory, to verify
Until recently the fastest brain interface let a person type about 18 words a minute by imagining handwriting. In 2023 two teams decoded attempted speech directly: one reached 62 words a minute with electrodes in the cortex, the other 78 with a sheet of 253 electrodes on its surface.
Natural conversation runs at about 160 words a minute, and these systems still make errors on roughly one word in four with a large vocabulary. But the direction is clear, and the limit is mostly the number and stability of the electrodes.
Two years closed half the distance to natural speech.
The best implanted arrays listen to a few thousand neurons. The brain has about 86 billion. Every result so far has been achieved while nearly deaf to almost everything the brain is doing.
The body also fights back. Tissue scars around an electrode and the signal fades over years. Putting more channels in means more heat, more data to send through the skull, and more surgery. And as these devices improve, the questions stop being only technical: who owns the recordings, who may alter the device, and what happens to the person when the company that made it closes.
Channel counts have risen about fifty-fold in thirty years. The brain is ten million times beyond that.
Source: Maynard et al. (1997); Jun et al., Nature (2017); Musk and Neuralink (2019); Steinmetz et al., Science (2021) · cited from memory, to verify
Each outer square is a nerve, and its area is the number of fibres it carries. The small white square inside is the number of electrodes our best implant offers in its place, drawn to the same scale. That people hear speech through twenty-two electrodes says a great deal about the brain's ability to adapt, and about how much is still missing.
Source: Spoendlin and Schrott (1989) for the auditory nerve; Jonas et al. (1992) for the optic nerve; manufacturer specifications · cited from memory, to verify
Our first aim is restoration: speech for those who have lost it, movement after spinal injury, sight and touch that feel like one's own. These are needs that exist today, and the people who have them are already partners in the research.
Beyond that lies augmentation, which we treat with care. We hold three positions. The interface should be two-way, because control without sensation is not embodiment. The standards should be open, so that no person's body depends on one company's survival. And the data produced by a nervous system belongs to the person it came from.
Your pointer is what the person intends. Each electrode listens to a neuron that fires most for one direction of movement, plus noise. The decoder adds up their votes to move the white cursor. Change the number of electrodes and the noise, then switch the feedback off.
This is a statement of intent, not a record of achievement. Each step carries a label for how firm it is. Nothing marked proposed or speculative has been done yet.
How electrodes read and write neural signals, as working simulations, built alongside Dhi.
Shared recordings and a public leaderboard, so a new decoder can be compared fairly without access to a patient.
A published interface so that devices outlive their makers and a person's data stays theirs.
Restored speech near natural pace, and prosthetic limbs that return sensation, as ordinary clinical care.
Millions of channels without open surgery, used first for restoration and offered for augmentation only with strong consent and open oversight.
Every claim on this page is open to challenge in the Agora, in the Mind & Intelligence room.
Argue it in the Agora ↗