The power of thought
Labs around the world are building machines that we can control with our minds. How long will we have to wait for an upgrade?
You get back from work, crash out on the sofa and pick a track from your favourite playlist. Without moving from that spot you start heating up the oven to cook dinner before beginning a conversation with your friend who lives on the other side of town. You do all this without ever saying a word or pressing a single button. How did anyone get anything done before brain interfaces?
The idea that we could run our lives from inside our heads is, obviously, a fantasy, but there are those who are attempting to make it a reality. In 2017, SpaceX and Tesla billionaire Elon Musk announced a new venture, Neuralink. Its aim: to build a high-bandwidth, implantable brain-computer interface that will put us permanently online and allow us to communicate wirelessly with anything that has a computer chip. The device could, theoretically, allow us to have thought conversations with our friends, share memories as if they were smartphone videos and ‘know’ anything we wanted by simply calling it down from the cloud.
Meanwhile, earlier this year, the US Defense Advanced Research Projects Agency (DARPA), announced plans to develop next-gen brain-computer interfaces, with the aim of enhancing the abilities of military personnel. A recently released document suggested a possible experiment for testing these devices: “a human subject controlling multiple drones in a virtual reality setup, while receiving sensory feedback to portray the status of each drone.” In other words, we might one day see soldiers controlling drones with their minds.
It sounds impressive, but is it possible? Primitive versions of brain-machine interfaces have already been used to help paralysed people move prosthetic limbs, but could we really see this technology making the leap to everyday use?
A brain-computer interface is a device that’s able to read the electrical impulses coming from the brain’s nerve cells (neurons) using electrodes and ideally also write to the brain, delivering information to the user by stimulating groups of neurons. Neuralink’s ultimate goal is to build an interface that interacts directly with each of the 86 billion neurons in our brains, and the company is apparently in the process of putting together a crack science team for its project. The finer details of exactly how Neuralink plans to do this remain under wraps, however.
“I’m still looking for more information on this,” says Dr Davide Valeriani, who studies brain-computer interfaces at the University of Essex. “Musk has announced these initiatives and then for a while hasn’t said anything else.”
Valeriani works with the kind of brain-computer interfaces that you might be more familiar with – electroencephalography (EEG) caps, those ugly skullcaps with all the sensors and wires attached to them. “You can imagine this as a system you can put in a backpack, with electrodes integrated into something we wear already, a hat or hairnet or whatever,” says Valeriani. All it takes to get this system working for a particular user is half an hour or so of training, not for the human but for the machine, which has to learn which patterns in the person’s brain are associated with certain thoughts.
Valeriani uses these EEG setups for group decision-making tasks. In one experiment from a 2017 study, his team asked groups of people wearing the caps to look at penguins and try to spot a polar bear in each image. Electrodes in the EEG caps monitored their brain signals and a computer delivered a collective answer. The computer learned to recognise signals associated with each person’s confidence in their decision and gave more weight to confident responses when coming up with the answer – whether there was or wasn’t a polar bear. Perhaps it’s not too much of a stretch to imagine similar technology being used by police officers to search for suspects on CCTV footage or by soldiers assessing warfare scenarios, the only downside being the EEG hairnets and backpacks full of electronics they’d need to wear.
The alternative is having electrodes implanted directly in your head, which is what Matthew Nagle did in 2004. Trials of implantable brain-computer interfaces have so far been mostly focused on paralysed people, because for them, the gain in function is worth the surgery and its risks. A quadriplegic, Nagle took the opportunity of a trial to get hooked up to a computer, allowing him, with practice, to control a cursor on a computer screen with his mind, operate a TV and send emails.
Last year, researchers used an updated version of this implanted ‘Braingate’ interface to give three paralysed people the ability to type up to eight words per minute with their brains. Unfortunately, the current state-of-art for this system requires roughly 100 electrodes and a thick set of cables to be plugged in directly through the top of your skull, risking infection and resembling something out of The Matrix.
“That’s one of the major issues,” says Prof Thomas Stieglitz, who’s developing brain-computer interfaces for medical applications at the University of Freiburg in Germany. “There are still these ugly connectors that are screwed into the skull and poke through the skin.” Scaling up to a whole-brain interface – à la Neuralink – would require millions or billions more electrodes, which currently can’t be detached from their connectors.
In Freiburg, Stieglitz’s team is trying to build an implant that can suppress the brain signals leading to an epileptic seizure – a step, perhaps, towards widespread use of brain-computer interfaces for the more able-bodied. “Our dream,” he says, “would be that the implant has a program that says ‘Okay, this seems to be a seizure event in six seconds and I know that I should stimulate this part of the brain to interrupt the seizure.’” In fact, he adds, there’s already one implantable device, a neurostimulator from the company NeuroPace, that’s approved as a medical product for this purpose.
Meanwhile, University of Freiburg spin-off company Neuroloop is developing a blood pressure implant that stimulates fibres in the vagus nerve that give the brain information about blood pressure. It sends a signal to the brain telling it that blood pressure is too high, triggering the body’s so-called ‘baroreflex’, which can rapidly lower blood pressure via changes in the heart muscles and blood vessels.
At the same time, however, Stieglitz is bogged down in some of the engineering problems that researchers face in creating the early incarnations of these implants – problems that will have to be solved whether we want to cure epilepsy or conduct thought conversations with our friends. “The challenge is to design the system such that it can interact with the human body for an envisioned lifetime,” says Stieglitz. That means finding a way to power it wirelessly inside the skull without having to remove it to charge the batteries, as well as making sure it doesn’t damage the nerves that it interacts with or corrode in the watery environment of the body. According to Stieglitz, the latter problem may be tackled by making ‘soft implants’ that mimic the floppiness of nerve tissue, but it would leave surgeons with a task akin to “implanting a jellyfish”.
As well as practical issues, there’s a minefield of ethical ones with implantable devices. DARPA considers the “burden of surgery” too great and the risks too high for testing in able-bodied soldiers, while Valeriani believes it’s better to invest in proven, non-surgical interfaces, which are cheaper. Valeriani admits, however, that placing electrodes on the outside of the skull can’t deliver anywhere near the level of detail that would be required for a whole-brain interface. External electrodes only allow neuroscientists to get a general idea of what regions of the brain are saying. Getting an accurate reading from a single neuron requires going inside the brain and that means some major surgery. Or does it?
Eat my (neural) dust
Five years ago, a team at the University of California, Berkeley, first described neural dust. Today, two of its inventors, Prof Jose Carmena and Prof Michel Maharbiz, are starting a neurotechnology company, Iota Biosciences, developing tiny neural implants that they imagine being installed in a simple outpatient procedure – “in the same way that you get a piercing or a tattoo,” explains Carmena.
Using implants the size of grains of sand, they’ve shown that they can record and stimulate nerves in rats. They picture a future where we would have a bunch of neural dust motes implanted to keep tabs on our health via fitness trackers, and treat everything from heart issues to asthma just by tweaking the right nerves. Iota’s motes would be wireless and batteryless, potentially doing away with cable connectors and solving the problem of providing power for a lifetime.
But how do you get the implants into the brain without opening the skull? One approach might be to wait until the technology scales down even further, so the motes could be injected, perhaps into spinal fluid. DARPA imagines something similar for its military devices. Its recent document covers nano-sized devices that would be delivered to the brain by “ingestion, injection or nasal administration”.
Maharbiz questions whether implants that small could do anything useful. In fact, the Iota pair believe it’s possible to achieve “mind-boggling” things without even tapping into the central nervous system. Instead, their dust motes could access the brain via its nerve branches in our limbs and organs, in a similar way to NeuroPace’s blood pressure device. “There are other places in the nervous system where we think you can actually put these ports,” says Maharbiz. “It won’t give you the same bandwidth as having a thousand channels in your cortex, but you’ll be surprised at how many things you can do – such as enhancement of your cognitive capabilities – by stimulating these peripheral nerves.”
It sounds as if neural dust could be the perfect solution for anyone afraid of a little craniotomy. But could it also be used to build Elon Musk’s dream machine? Can we imagine a whole-brain interface made up of millions of electronic dust motes?
Interestingly, one of the Iota team’s collaborators on the original neural dust paper was Dongjin Seo, who now works with Neuralink. While Musk remains uncharacteristically silent on his new project, Carmena and Maharbiz know a few of the team and say there’s “no hype at all” in the idea of them building a next-gen implantable brain interface. But regardless of Musk’s other ambitions, the first people to benefit will be those with certain medical conditions, says Carmena. “The reality is they’re going to build clinically viable devices and we need those by yesterday,” he says. “In terms of the use, it’s going to be medical for a long time. I can’t tell you how long, but it’s going to take some time.”
So for now, watch this space, but if anyone is going to up-end the world of brain-computer interfaces, you wouldn’t bet against it being Elon Musk.
This is an extract from issue 322 of BBC Focus magazine.
Subscribe and get the full article delivered to your door, or download the BBC Focus app to read it on your smartphone or tablet. Find out more