Almost two years ago, Dennis Degray sent an unusual text message to his friend. “You are holding in your hand the very first text message ever sent from the neurons of one mind to the mobile device of another,” he recalls it read. “U just made history.”
Degray, 66, has been paralysed from the collarbones down since an unlucky fall over a decade ago. He was able to send the message because in 2016 he had two tiny squares of silicon with protruding metal electrodes surgically implanted in his motor cortex, the part of the brain that controls movement. These record the activity in his neurons for translation into external action. By imagining moving a joystick with his hand, he is able to move a cursor to select letters on a screen. With the power of his mind, he has also bought products on Amazon and moved a robotic arm to stack blocks.
Degray has been implanted with these devices, known as Utah arrays, because he is a participant in the BrainGate programme, a long-running multi-institution research effort in the US to develop and test novel neurotechnology aimed at restoring communication, mobility and independence in people whose minds are fine but who have lost bodily connection due to paralysis, limb loss or neurodegenerative disease.
But while the Utah array has proved that brain implants are feasible, the technology has a long way to go. Degray had open brain surgery to place his. The system is not wireless – a socket protrudes from his skull through which wires take the signal to computers for decoding by machine-learning algorithms. The tasks that can be done and how well they can be executed are limited because the system only records from a few dozen to a couple of hundred neurons out of an estimated 88bn in the brain (each electrode typically records from between one and four neurons).
And it is unlikely to last for ever. Scar tissue, the brain’s response to the damage caused by inserting the device, gradually builds up on the electrodes, leading to a progressive decline in signal quality. And when the research sessions – which take place twice a week for Degray in his living facility in Palo Alto, California – come to an end, it will be disconnected and Degray’s telepathic powers cease to be.
Barely a couple of dozen people have been implanted with Utah arrays worldwide. Great progress has been made, says Leigh Hochberg, a neurologist at Massachusetts general hospital and an engineering professor at Brown University who co-directs the BrainGate programme, but “a system that patients can use around the clock that reliably provides complete, rapid, intuitive brain control over a computer does not yet exist”.
Help may be at hand. An injection of Silicon Valley chutzpah has energised the field of brain-computer or brain-machine interfaces in recent years. Buoyed by BrainGate and other demonstrations, big-name entrepreneurs and companies and scrappy startups are on a quest to develop a new generation of commercial hardware that could ultimately help not only Degray and others with disabilities, but be used by all of us. While some, including Facebook, are pursuing non-invasive versions, wireless neural implant systems are also being worked on.
In July Elon Musk, best known as the CEO of the electric car company Tesla, presented details of an implantable wireless system that his company Neuralink is building. It is already being studied in monkeys, Musk revealed, and it is hoped that human trials will start before the end of 2020. To date, Neuralink has received $158m in funding, $100m of it from Musk.
Elon Musk worries about the threat posed by artificial intelligence and claims BCIs may provide a way of keeping up with it
While the implant being developed is still the same size as one of the Utah arrays in Degray’s brain, it has far more electrodes, meaning it can record from far more neurons. While a Utah array – of which up to four or five can be inserted – typically has 100 electrodes, Neuralink says its version will have more like 1,000. And the company thinks it is feasible to place up to 10. Very thin threads of flexible biocompatible polymer material studded with electrodes would be “sewn in” by a robot to avoid piercing microvessels, which Neuralink hopes would ameliorate scarring, thereby increasing how long the device lasted. “Our goal is to record from and stimulate spikes in neurons in a way that is orders of magnitude more than anything that has been done to date and safe and good enough that it is not like a major operation,” said Musk in his presentation, adding that the procedure would be more like laser eye surgery than brain surgery. Medical concerns drive the device’s development, according to Musk, but he also worries about the threat posed by artificial intelligence and believes this could provide a way of keeping up with it.
There are smaller rival startups too. Paradromics, like Neuralink, is focused on many more and smaller electrodes but is aiming for an even higher density of probes over the face of its neural implant. In form, their device would look closer to the Utah array – a bed of needles with metal electrodes – and there would be no robotic surgery. “We want to hit the market as soon as possible,” says founder and CEO Matt Angle adding the hope is to begin a clinical trial in the early 2020s. The company has raised about $25m in funding to date including significant amounts from the Pentagon’s research agency, Darpa, which grew interested in BCIs after it realised the sophisticated robotic limbs it was building for injured soldiers returning from overseas needed brain control.
Synchron, based in Australia and Silicon Valley, has a different approach. The company, which has received $21m in funding to date, including some from Darpa, last week revealed that the first clinical trial of its Stentrode device had begun in Australia – ahead of both Neuralink and Paradromics.
The device avoids open brain surgery and scarring because it is inserted using a stent through a vein in the back of the neck. Once in position next to the motor cortex, the stent splays out to embed 16 metal electrodes into the blood vessel’s walls from which neuronal activity can be recorded. So far in the trial one patient – paralysed with motor neurone disease – has been implanted, with four others set to follow. The device’s safety will be studied along with how well the system allows brain control of a computer for typing and texting. While it can only read the aggregate activity of a population of neurons, of which it will take in about 1,000, there is enough data to make a system useful for patients – and less nuance in the signal actually makes it more stable and robust, says founder and CEO Tom Oxley.
Meanwhile, challenges remain for Neuralink and Paradromics. Whether scarring can be mitigated by very small electrodes is yet to be seen. There is also the issue of the electrodes being dissolved and corroded by the body – a problem that gets worse the smaller they are. How long Neuralink’s new polymer probes will last is unknown.
Beyond typing, no one is too specific. Brain commands to smart speakers? Brain-to-brain communication? Enhanced memory?
“No one is going to be super impressed with the startup companies until they start recording their lifetimes in years. The Utah array has a lot of issues – but you do measure its lifetime in years,” says Cynthia Chestek, a neural interface researcher at the University of Michigan. Then, even if we are able to record all these extra neuron signals, could we decode them? “We have no idea how the brain works,” says Takashi Kozai, a biomedical engineer at the University of Pittsburgh who studies implantable technologies. “Trying to decode that information and actually produce something useful is a huge problem.” Chestek agrees that more understanding of how neurons compute things would be helpful, but “every algorithm out there” would suddenly just start doing better with a few hundred extra neurons.
None of the three companies sees nonmedical applications in the short term, but argue that the implant technology could gradually branch out into the general population as people start seeing how transformational it can be.
The most obvious application may be brain-controlled typing. Oxley imagines a scenario where people who have grown up texting and typing – and are wholly dependent on their fingers for that – lose functionality as they age. Frustrated that they can’t maintain their speed, they may seek other ways to preserve their technological capability. Eventually a tipping point will occur as people see BCIs working better than the human body. “If the technology becomes safe, it’s easy to use and it provides you with superior technology control, there will be people who will want to pay for that,” says Oxley.
Of uses beyond that, no one is being specific. Brain commands to smart speakers? Brain-controlled car driving? Brain-to-brain communication? Enhanced memory and cognition?
If the technology were to make it outside the medical domain, the military is where we might see it first, says Dr Hannah Maslen, deputy director of the University of Oxford’s Uehiro Centre for Practical Ethics. For example, it might provide silent communication between soldiers or allow activation of equipment by the thinking of certain commands. It is hard to see most people opting to undergo a surgical intervention for recreational or convenience uses, she adds. But at a recent neurotechnology meetup in San Francisco of about two dozen tinkerers, Jonathan Toomim argued it was a logical next step. “We already use devices – our smart phones – that offload a lot of our cognition and augment our memory. This is just bringing the bandwidth between the human brain and those to a higher level,” said the self-described neuroscientist, engineer, entrepreneur and environmentalist, who makes his own neurofeedback gear.
The public should have a clear voice in shaping how neural interface technology is used and regulated over the coming years, concluded a report this month on the topic from the UK Royal Society. One concern is data privacy, though Maslen says this should be tempered by the fact that while BCIs may be portrayed as being able to “mind read” and “decode thoughts” – stoking fears that they will unearth innermost secrets – they are recording from very small areas of the brain mostly related to movement, and require the user’s mental effort to make them work. “Ethical concerns around privacy … don’t apply in such a full way,” she says.
Nonetheless, questions remain. Who owns the brain data and what is it being used for? And “brainjacking”, where a third party could gain control of the system and modify it in ways the brain’s owner has not consented to, is rooted in reality rather than science fiction says Maslen – pacemakers have been hacked before. Paradromics’ Matt Angle wonders to what extent data from BCIs could be used as evidence in court – for example to incriminate someone in the same way a diary or a computer might.
Further ethical issues arise around control and agency. If a brain implant doesn’t get your intention right, to what extent are you as the user of the device responsible for what is “said” or done? And how do we ensure that if a technology confers significant benefits, it is not just the rich who get it?
Society still has a few years to ponder these questions. Neuralink’s aim of getting a human clinical trial up and running by the end of next year is widely considered too ambitious, given what remains unproved. But many experts anticipate that the technology will be available for people with impairments or disabilities within five or 10 years. For nonmedical use, the timeframe is greater – perhaps 20 years. For Leigh Hochberg, the focus has to be on helping those who need it most. Says Degray of Neuralink’s device: “I would have one implanted this afternoon if I could.”
Is there an alternative to implants?
A worn, non-invasive brain computer interface which doesn’t involve brain surgery and can always be taken off may seem attractive. But the skull muffles the reading of neuronal signals. “The physics [of a non-invasive device] are just extremely challenging” says Cynthia Chestek of the University of Michigan.
Some companies are trying anyway. Facebook announced in 2017 it wanted to create a wearable device that would allow typing from the brain at 100 words per minute (as a comparison, Neuralink is striving for 40 words per minute – which is around our average typing speed – and Dennis Degray with his Utah array implant clocks about eight words per minute). This July, researchers at the University of California funded by the social network showed decoding of a small set of full, spoken words and phrases from brain activity in real time for the first time – though it was done with so-called electrocorticography electrodes laid on the surface of the brain via surgery. Meanwhile the company continues to work on how it might achieve the same thing non-invasively and is exploring measuring changing patterns in blood oxygenation – neurones use oxygen when they are active – with near-infrared light.
Also on the case is Los Angeles-based startup Kernel, founded by entrepreneur Bryan Johnson who made millions selling mobile payments company Braintree to PayPal. Kernel, into which Johnson has put $100m, started as a neural implants company but then pivoted to wearables because, Johnson says, the invasive road looked so long. Plenty of non-invasive methods exist for sensing and stimulating brain activity (indeed they form the basis of a large consumer neurotechnology industry). But none, says Johnson, is equal to being bridged into a next-generation interface. New ways are needed, and he believes Kernel has found one others have missed. “We will be ready to share more in 2020,” he says.
But assuming the technical challenges can be surmounted, social factors could still be a barrier, says Anna Wexler, who studies the ethical, legal and social implications of emerging neurotechnology at the University of Pennsylvania. Google Glass failed not because it didn’t work but because people didn’t want to wear a face computer. Will anyone trust Facebook enough to use their device if it does develop one?
This content was originally published here.