Neuralink brain implant: how it works and what it promises

A neurosurgeon who works at Neuralink joined a long conversation about the brain, surgery and the company's mission to build a brain implant. This summary covers the main ideas: what the technology does today, how it learns, why robots are needed and where the neurosurgeon thinks it could go.

The brain as a set of modules

The neurosurgeon describes the brain as a collection of functional modules duct taped together inside a bone box. When one fails, the results can be dramatic. A young patient with bilateral frontal lobe damage after a car accident woke up with essentially no impulse control. Everything the brain does, from hormone control to vision to moving a bicep, can go wrong, which makes neurosurgery high stakes but deeply satisfying when it works.

One new technique shows the progress. For a small, deep tumor, surgeons once had to cut through healthy brain and disrupt white matter. Now a 2 mm drill hole lets them pass a fiber optic cannula connected to a laser. The tumor is heated under direct MRI, with images taken every second or so, so the team can monitor temperature and kill tumor cells while sparing nearby brain. Even tumors in the brain stem, the primary motor cortex or Broca's area, once considered inoperable, sometimes come out without harm.

What Neuralink is building

The company is making a neural implant plus a robotic insertion device. Electrodes smaller than a human hair are placed across a small region of the brain. The first target is people with a bad spinal cord injury. Their motor cortex is perfect but no longer wired to the arms and legs.

The first goal is to let these patients control a computer. They imagine flexing a fist or moving a wrist, and on screen the cursor moves, clicks an icon and opens a word processor, while an on-screen keyboard lets them type. It is not about moving a prosthetic yet. The aim is digital freedom.

Why a robot places the electrodes

The electrode threads end in loops a couple of microns wide, and the blood vessels on the cortical surface are numerous and densely packed. A human hand is not steady enough to place them blindly, at the right depth, through all cortical layers. In the current submission to the FDA, the robot only does the electrode placement. A neurosurgeon still opens the skin and skull and presents a pristine brain surface.

How the device learns

The system couples simple adaptive software with a human brain. At first, the signals look random to the chip. The software correlates neural activity with targets on a screen, such as up and right or down and left, and builds a model of what the person intends. The person then gets better at using that model. Because signals shift overnight, the brain and the software keep adapting to each other. It is like playing a piano whose keys change sound over time, yet the brain solves the problem.

The team has tested this in monkeys, which play video games for smoothie rewards. Neuralink holds the world record for the information bit rate pulled from a monkey brain to control a cursor.

Reconnecting the body and sharing thoughts

Reconnecting the brain to a person's own limbs is further down the roadmap. One path is to put a second implant in the spinal cord, below the injury, so the two devices talk to each other and carry intact motor signals past the damaged area.

On the idea of hearing each other's thoughts, the neurosurgeon says it requires no magic. Texting by phone already works. Moving that function to an implant in a verbal brain region would let people speak thoughts that are rendered as speech, for instance through a bone conducting implant.

Safety, animals and personal commitment

The neurosurgeon says they would accept an implant, though the first version would add little for someone who can already use a mouse and keyboard. They vouch for its safety based on hundreds of surgeries, and they carry a passive RFID chip in their hand, coated in glass and silicone, which used to unlock their front door and the doors at work.

On animal research, the neurosurgeon notes that FDA approval first requires safety data from animals. Pigs serve as a safety platform because their skulls resemble ours, while monkeys validate that the signals work. The team says animals can opt out and are never deprived of food or water.

Conclusion

The near-term mission is to reduce suffering for people with no good options. Over 10 to 20 years, the neurosurgeon hopes people will gain control over addiction, depression and obesity, and eventually faster communication than speech. The ceiling is high because there is no obvious reason the interface cannot reach millions of neurons as the technology is refined.