The new age of the cyborg?

In May, Noland Arbaugh became the first recipient of a Neuralink brain-computer interface. The 29-year-old had suffered a spinal injury while swimming – an accident that left him paralysed from the neck down. For the last eight years he had been using a wheelchair and a stick held in his mouth to handle electronic devices. But now there’s another way. Brain-computer interfaces, or BCIs, can translate intended movement signals in the brain into computer commands. With the interface, named Telepathy, Noland is now able to browse the web, play computer games and generally have far greater control over his computer and other devices. It may also help him to regain some physical movement.

Neuralink, owned by the tech billionaire Elon Musk, is one of dozens of companies and research groups currently developing BCIs. These devices are progressing rapidly and their clinical applications are increasing. They have the potential to give new life and independence to the many people suffering from spinal cord injury, estimated to be more than 100,000 in the UK alone. Today, most of the recipients of implantable BCIs are people who have been paralysed by injury or neurological disease. The devices serve the immediate function of facilitating communication or movement. In the long term, BCIs can aid rehabilitation in some recipients, by encouraging the growth of new nerve pathways.

But trials are also underway to assess their application for a range of physical and mental health conditions, including depression and Alzheimer’s – and Musk’s ultimate ambition is to use them for augmentation, to increase the power of the human brain. Meanwhile, wearable BCIs, which don’t require a surgical procedure, are increasingly available on the market, often with ambitious claims and little regulation. So while the widening use of BCIs has monumental potential, it also poses major technological and scientific challenges and raises a host of ethical issues.

The history of BCIs

BCIs are devices that provide a direct link between the brain and a computer, using the brain’s electrical activity. Signals can be recorded by either sheet-like implants that sit on the surface of the brain and record the collective activity of large groups of cells, or arrays of microelectrodes, each of which penetrates and records from an individual cell. Most such implants target a region called the motor cortex, which is involved in planning and executing voluntary movements. Cells in the motor cortex send signals down through the spinal cord to “secondary” motor neurons, which relay the information to the muscles.

In a spinal cord injury like Noland’s, the nerve fibres connecting the motor cortex to the spinal cord are damaged, or severed completely, so that these signals cannot reach the muscles. BCIs bypass the injury by translating the signals into instructions for controlling the output device. Thus, with a little training, a paralysed patient implanted with a BCI can control a wheelchair, a prosthetic limb or a cursor on a computer screen, simply by thinking about performing the necessary movements.

The history of brain-computer interfaces can be traced back to the 1920s, when the German psychiatrist Hans Berger developed electroencephalography (EEG), a technique that recorded the brain’s electrical activity using wire electrodes placed onto the scalp. He was the first to record brain waves – the rhythmic patterns produced by the electrical discharges of hundreds of thousands of nerve cells. Today, EEG is an indispensable clinical tool, used widely to diagnose and monitor epileptic seizures, brain tumours and sleep disorders, among other things.

In 1973, at the dawn of the age of personal computing, the term “brain-computer interface” first appeared in a paper published by Jacques Vidal of the University of California, Los Angeles. The paper described a project “to evaluate the feasibility and practicality of utilising the brain signals in a man-computer dialogue”. What then was a highly speculative idea was finally developed in 2004 by John Donoghue, an American neuroscientist working at Brown University as part of a trial dubbed BrainGate. After successfully testing the technology on monkeys, a device for use on humans was developed and implanted into tetraplegic Matthew Nagle, who was paralysed from the neck down after being stabbed in a fight. The 96-electrode device enabled Matthew to move a cursor on a computer screen. He continued using the device until his death in 2007.

BrainGate is now a consortium of neurologists, neuroscientists, engineers, computer scientists, neurosurgeons, mathematicians and other researchers. Their trials continue to this day – 14 more participants have been implanted with advanced versions of the same device. Musk’s Neuralink is now one of its major competitors. Launched in 2016, it quickly became one of the leaders in the field, benefitting in part from being owned by one of the world’s richest men. Among its projects is an experimental implant that would restore vision – even in “those who have lost both eyes and their optic nerve”, Musk claimed recently.

Worldwide, 21 research groups have conducted 28 trials involving a total of 67 mostly male participants, testing BCIs not just for restoring movement and speech, but also for a wider range of neurological conditions, including depression, epilepsy and obsessive-compulsive disorder. Eventually, they could be used to alleviate memory loss in patients with Alzheimer’s disease, and to treat a whole host of other neuropsychiatric conditions. But these applications are in the early stages and are producing mixed results. Treating psychiatric conditions will be hugely challenging, as most such conditions have multiple, complex causes, and effective treatment will first require a deeper understanding of their neurobiological bases.

An unregulated market

BrainGate, Telepathy and other similar devices are referred to as invasive BCIs, because implantation involves a neurosurgical procedure. This typically involves removing part of the skull and having a surgeon implant the chip. The procedures carry major risks. But non-invasive, wearable interfaces also exist.

While both types of BCIs are used for research purposes, the wearable devices have been commercially available for non-medical use for some time. Various private companies now offer relatively cheap wearable devices for video gaming and self-health monitoring. For under $1,000, you can buy a device to replace a joystick or keyboard in controlling a video game, or to provide real-time brain data to alert you to, for example, an imminent epileptic seizure.

Most commercially available BCIs use EEG to read brain waves. However, more controversially, some employ techniques such as transcranial magnetic stimulation and transcranial direct current stimulation, which modulate brain activity by applying small magnetic fields or electric currents to the surface of the cortex. Some manufacturers claim that their devices can improve cognitive performance. Multiple lab experiments have indeed found that these devices can strengthen performance on a variety of cognitive tasks, and enhance attention and memory, under controlled laboratory conditions.

These conditions, however, are much harder to replicate in the real world, so most of the claims made by commercial companies remain unfounded. Wearable BCIs may also be unsafe to use unsupervised outside a lab setting. An untrained user may burn their scalp, for example, or alter their brain activity in unintended ways. Worryingly, the market for wearable BCIs is still unregulated.

If BCIs are going to be taken up at scale, the safety concerns need to be addressed. At present, many of the implanted devices are built with stiff metallic electrodes, which typically cause bleeding and then scarring of the brain tissue, and also carry the risk of infection in and around the implantation site. Minimising this reaction is one of the challenges of manufacturing implantable BCIs. Additional surgeries are usually also required to recalibrate the devices or replace batteries.

All surgery carries risks, and thus far we have very little data on the possible adverse effects of implanted BCIs. According to a 2021 interim safety report giving a snapshot of the progress of the ongoing BrainGate trial, the 14 participants had a total of 68 “device-related adverse events”. The most common was skin irritation. However, there were also six serious events, including a rare form of epilepsy that appears to have been triggered by the device.

The participant fully recovered within four days, after which he continued participating in the trial for another eight months – but the unforeseen side effect highlights the possibility of life-changing consequences from what is still an emerging technology. Long-term implantation of a BCI could alter brain function in unpredictable – and undesirable – ways. It could alter a person’s sense of control over their own behaviour, or even their sense of identity.

Since the report was published, the average longevity of implantable BCIs has increased from just under 2.5 years to about 3.5 years. We will have to wait for data on longer-term use, but the relatively low rate of serious adverse events led the report’s authors to conclude in favour of more research and development, stating that the BCI being tested “has a safety record comparable with other chronically implanted medical devices”, such as artificial joints and cardiac pacemakers.

Meanwhile, there is progress on the use of electrodes made of soft materials. These cause less tissue damage and are less likely to cause infection than metallic ones. To this end, an Anglo-Italian team led by researchers at the University of Cambridge have developed an origami-inspired electrode array that can fold up into a compressed state and expand on the surface of the brain to cover a large area of the cortex. Currently, implantation of a BCI requires removing a section of the skull that is larger than the device itself. This new device can be implanted via burr hole surgery, a far less invasive procedure. So far, it has only been tested in pigs, however. “The brain has a consistency like jelly, so if you can make the electrodes flexible and soft and try to match its properties, it’ll be happier,” says Timothy Constandinou, a professor of bioelectronics at Imperial College London. “The challenge here is how to get the electrodes into the brain, and how to make them last for decades.”

Perhaps the biggest innovation in the Neuralink approach is the use of a surgical robot for BCI implantation. After the surgeon has cut open the scalp, removed a section of the overlying skull bone, and cut through the three meningeal membranes enveloping the brain, the robot is used to insert the electrodes into a precise location in the brain tissue. The surgeon then attaches the rest of the implant to the electrodes and closes the opening in the skull. “Millions of people can benefit from this, and there just aren’t that many neurosurgeons out there,” Neuralink president and co-founder Dongjin (“DJ”) Seo said, “so we hope that robots can do large parts of the surgery, and this is another category of product that we’re working on.”

The latest research

Meanwhile the application of BCIs is expanding. The most sophisticated implantable BCIs can not only read brain activity, but also “write” it. For example, as well as enabling a paralysed user to control a prosthetic limb, they can provide feedback about pressure and muscle position from sensors in the limb. This not only allows for better control of the prosthesis, but enhances its integration, making it feel more like a natural body part.

Various research groups are also developing devices that restore communication to paralysed patients – not through enabling typing this time, but through a direct link to the regions of the brain responsible for formulating speech. Ann, a 48-year-old who was severely paralysed by a brainstem stroke, was implanted with a speech prosthesis last year by surgeons at the University of California, San Francisco. The device consists of a paper-thin, 253-electrode array implant that reads motor cortical activity controlling her tongue, jaw, larynx and face, and translates it into commands for an avatar on a computer screen, at a rate of 78 words per minute, with 72 per cent accuracy. Previously she had used movement-tracking technology to select individual letters to communicate at about 14 words per minute.

One of the latest such devices, described in 2023 in the journal Nature by researchers at Stanford University, is a high-performance speech prosthesis that can decode speech with an accuracy of 91 per cent and display it on a computer screen in real time. More recently, a team at Johns Hopkins University Medical School reported long-term implantation of a similar device in a patient with motor neurone disease, who could use it to control a communication board for five-and-a-half months without retraining or recalibration.

Progress in this field depends on advances in both technology and our knowledge of how the brain works. For example, one recent study showed how individual neurons in the frontal cortex encode the meaning of words during speech comprehension. Instead of decoding the activity associated with producing individual words, future speech prostheses could possibly decode longer trains of thought, making them easier and quicker to use.

Most research is currently aimed at treating physical and mental health conditions, and restoring lost bodily function. However, BCIs are also increasingly being used to improve and augment the brain and body.

Three months after the successful implant into Noland Arbaugh, Neuralink announced success with another interface, implanted into a second patient. “Alex”, who is also paralysed due to a spinal cord injury, enjoyed similarly positive results. He’s not only able to control a computer; he has also been exploring more creative applications, like learning to use the chip to design 3D objects.

Afterwards, Neuralink CEO Musk appeared on a podcast hosted by computer scientist Lex Fridman to discuss the procedure and the long-term aims of the company. He said they are hoping to implant the devices in eight more patients by the end of the year. “The logical thing to do is start off solving basic neuron damage issues, but once the risk is minimal, we’ll aim for augmentation,” he said, adding that BCIs could “probably solve schizophrenia” and “help with memory”.

Musk eventually aims to commercialise his devices so that they can be used by millions of healthy people to enhance their mental and physical abilities – or, according to Neuralink’s mission statement, to “restore autonomy to those with unmet medical needs today and unlock human potential tomorrow.” He has also stated that he wants to implant millions of the devices within the next 10 years.

Some commercially available wearable BCIs are already marketed for augmentation purposes, but as we have discussed they have limited use and efficacy outside of the laboratory. However, other methods are being explored to connect more directly to the brain, without the need for drilling into the skull or other invasive surgical procedures. Emerging technologies include electrodes that can be delivered to the brain through the bloodstream, and free-floating wireless electrodes in the form of “neurograins” or “neural dust”. If these technologies are successful, they would allow for minimally invasive implantation, making it more likely that the devices will be used widely and for non-medical purposes. “Shrinking” components will also help to make implantation easier and less risky. Researchers at the Swiss Federal Institute of Technology Lausanne have fabricated an attempted speech-to-text BCI chip measuring just 2.46mm2, significantly smaller than the Neuralink device, which measures about 23 x 8mm, or 184mm2.

But it’s not as simple as making implantation easier and less risky. There’s usually a trade-off between safety and stability. Soft electrodes are less likely to damage the delicate brain tissue, for example, but they are also less likely to remain in place. The current iteration of the Neuralink device contains 1,024 wire electrodes, which are soft and flexible. But more than half retracted soon after implantation, leaving only about 400 to read signals from the patient’s brain. The movement of electrodes also causes the signal to change or degrade over time so that it needs to be recalibrated, or even replaced.

And while the devices need to remain small, the number of electrodes needs to increase if the product is to improve. “Increasing the complexity of the devices, in terms of how many electrodes they have, would provide access to more of the neural signal,” says Constandinou. “A second thing is making the implant last longer. Nobody wants to go through invasive surgery to be told that the device performance will degrade a year later.”

The many ethical concerns

We know that brain-computer technology will continue advancing rapidly, given the expected progress in engineering and materials science. The main obstacle to widening the application of these devices is likely to be our understanding of the brain. Meanwhile, whether used to control a wheelchair, produce speech or treat Alzheimer’s disease, BCIs present multiple ethical concerns. While the field progresses in leaps and bounds, there are questions about how the research is being conducted. Neuralink in particular has been criticised for the lack of transparency in its research process, while Musk and his team tend to share progress updates via social media, rather than the traditional route of scientific papers.

Another ethical issue concerns the participants. It’s vital that appropriate participants are selected, and that fully informed consent is given – but this can be a difficult thing to navigate. BCIs are experimental treatments, many of which are designed specifically for people with severe disabilities. Such people might be suffering severely in their day-to-day lives and volunteer for clinical trials out of desperation. They may have high expectations – restoring their previous quality of life or their ability to work, for example – which may, in the end, not be met.

Then there is the question of privacy. BCIs could record highly sensitive personal data, such as information about mental states, one’s propensity for certain patterns of behaviour, along with predisposition to, or presence of, neurological conditions. This raises questions about who can gain access to this data, and what they might do with it. Academic and medical researchers are obliged to anonymise the data they collect from BCIs but can, with consent, identify study participants. Private companies entering the space could potentially profit by giving sensitive brain data to third parties – influencing health insurance claims, for example.

The use of BCIs has grown rapidly in recent years, as private companies and academic institutions discover more about their potential uses, for everything from improving attention span to controlling a prosthetic limb. As wearable devices become cheaper, their use for non-medical purposes is likely to continue growing, making the need for regulatory processes more urgent. The cost of implantable devices is also likely to decrease dramatically in the near future.

But both wearable and implanted BCIs raise serious ethical concerns, around the safety of recipients, the transparency of the research process, and the security of the data obtained through direct access to a person’s brain activity. And if Neuralink achieves its aim “to bring this technology from the lab into people’s homes,” it would gain access to sensitive brain data from millions of users.

That aim also comes with longer-term concerns. Widespread use of BCIs could cause serious social disruption. Imagine a world where some people can afford to use these devices to enhance their mental and physical capabilities, while others are left behind. It could increase existing social inequalities, and perhaps create new ones. So while BCIs have revolutionary potential, strict regulation and proper data governance are urgently needed to ensure that the technology is used safely and equitably.

This article is from New Humanist’s winter 2024 issue. Subscribe now.