How does the Neuralink brain chip work? Elon Musk’s implant explained

Would you allow a computer direct access to your thoughts? The question isn’t as far-fetched as it once was — and the news that Elon Musk’s Neuralink has embarked on its first human trial might just have brought mass-market mind-reading machines a step closer.

The idea is relatively simple. Neuralink, founded by Musk in 2016, has developed a coin-sized device that fits inside a hole in the skull. Just over a thousand flexible electrode “threads”, each much thinner than a human hair, extend from it into the grey matter of the brain, where they monitor the electrical activity of its cells. An AI will then decode that activity.

In a narrow sense, the system will try to read the patient’s thoughts and we already know that this kind of technology can deliver extraordinary results. In the early 2000s, American scientists showed that monkeys implanted with neural interfaces could control robotic limbs with their minds.

Since then, humans fitted with similar systems have played video games, manoeuvred robots, sent emails and made purchases on Amazon — all without lifting a finger, via a computer that was deciphering their brain activity.

• First human brain implant showing promising results, says Musk

Some of the most striking experiments have used neural implants to help people who have lost the use of their legs to stand and take steps again, albeit with the help of walkers. In 2021 scientists at the University of California San Francisco unveiled an neural implant that gave a paralysed man a voice for the first time in more than 15 years.

It seems, though, that Musk wants to go much further. He has spoken about restoring sight to the blind and allowing quadriplegics to regain “full-body functionality” — a feat that most researchers believe is decades away, if it’s possible at all.

More than that, Musk has said that his ultimate aim is “symbiosis with artificial intelligence”. He wants, eventually, to produce a mass-market “general population device” that would directly connect users’ minds with powerful computers.

Could he — or a successor — achieve those goals? So far, all the information we have about Neuralink’s first human trial appears to be a tweet saying that it’s happened. Scientists will want to pore over the details. How many electrodes have been implanted and how? How well is the system performing? Most importantly, how is the patient doing?

However, it’s worth knowing that Neuralink is already being taken very seriously by experts in the field, who say that it’s looking to build on ideas widely viewed as promising.

A robot arm built by Neuralink that inserts its electrode threads into the brain could prove to be a key advance. Making the threads flexible should limit the damage they do to the jelly-like brain. According to Neuralink, the robot can insert them while avoiding fine blood veins. Precision is crucial: how well these systems will work will depend, in part, on how detailed a picture of brain activity they are able to provide. Speed is also important if you want to build implants in the future that have many thousands of electrodes, each of which must listen in on a specific cluster of neurons.

• What’s it like to have a mind-reading microchip in your brain?

The most widely tested brain implant at the moment is known as the Utah array. It’s a hard silicon square that looks like a miniature bed of nails. It has 100 stiff protruding needles, each about a millimetre long. These electrodes can cause scarring, which degrades the signal they pick up. On paper, Neuralink’s alternative makes it look quaint — and other companies are building rival systems that might prove to be just as sophisticated.

Still, though, Musk’s vision of seamless, mass-market neural implants looks a very long way off. At the moment, we’re pretty good at inferring sensory stimuli from brain activity. An AI can predict, for instance, what your eyes are seeing by analysing MRI scans of your brain.

AIs can also take brain data and predict what body movements you want to make. That’s largely because the motor cortex — the bit of the brain that directs movement — is nice and accessible, a strip of matter that sweeps across the top of your head. We know its function, and that activity in different sub-regions precedes the movement of different limbs.

This makes AI-enabled neural implants potentially useful for people with medical conditions, such as the loss of movement in limbs. But if the aim was to share access to your inner monologue so that you could commune with a supercomputer — well, we wouldn’t really know where to put the implant. The networks of brain cells that give rise to abstract thought remain mysterious.

Another big question is whether all brains work the same. Voice recognition systems can be trained to recognise voices because all voices sound alike, more or less. Whether the same is true for brains we don’t know. For an AI truly to tap into your mind, perhaps it would have to be tailor-made for you. And perhaps it would take a human lifetime to train it to be useful.

Or perhaps not. Maybe an AI is being developed in a lab in California or Beijing that will crack what philosophers call “the hard problem of consciousness” and lay bare the mechanisms that render experience from matter. If it is, Neuralink might just turn out to be the partner it needs.