Feeding monkeys, watching movies with cats, playing video games – sounds like a nice day off, right? If you’re a neuroscientist, though, it sounds a lot like the last half-century of research into connecting human brains to computers.
Elon Musk’s Neuralink, impressive as it is, has only been in the game since 2016, and it’s just part of an ever-building wave of neuroscientific accomplishments. Those monkeys? They’re feeding themselves with robot arms connected to their brains. The cats are watching videos, and we’re translating the resulting brain activity back into images. And the video games? You can play them with your mind.
What are brain-computer interfaces (BCIs)?
As complex as the actual implementation is, the concept behind a BCI is fairly basic. Neurons in your brain fire electrical signals to each other when you do things. Using electrodes either in or on the skull, we can detect these spikes, digitize them, figure out what the brain is doing, and translate that activity into some action or data.
This way, we can send interpretable electrical signals to things like computer systems, prostheses, and even robots, allowing us to potentially control the devices around us, and even our own bodies, with just a thought. This connection works both ways, too. We can encode digital information as brain-readable electrical signals and send them in, providing a type of sensory input.
BCIs are usually divided into three main categories: invasive, partially invasive, and non-invasive. As the names suggest, invasive BCIs are implanted directly into the brain, which gets them the best “reception” in terms of being able to pick up detailed information about which neurons are doing what.
Partially-invasive BCIs (that’s what Neuralink is) are placed in the skull, but not directly in the brain, which makes them safer, less surgically-intense, and still fairly efficient.
Non-invasive BCIs aren’t attached to the head at all. You’ve probably seen them before in the form of electrode-studded rubber hats. They’re the easiest to use, but because they’re further from the brain yet in the skull in the way, the waves generated by firing neurons don’t come through as well, meaning the data is noisier and doesn’t allow for control as precisely as in-brain implants.
What have we done with BCIs so far?
The primary goal behind this research isn’t to create an AI-enhanced human with telepathic Internet powers. That will probably happen, but the goal is more of a knock-on effect resulting from medical research into things like treating paralysis, blindness, deafness, muteness, seizures, and other sensory/central nervous system issues, which are driving a lot of the current brain-computer interface research.
Cochlear implants, for example, are already in use by hundreds of thousands of people. They’re not interfacing with a computer, per se, as their job is to take sounds from the outside world and translate them into electrical signals that can be processed by the brain, but they’re a good demonstration of how we can send brain-readable signals from machines.
Retinal implants, while nowhere near as common yet, are already restoring limited vision to those who have lost it. People have even successfully used brain signals to control robotic arms and legs, like Juliano Pinto, a paraplegic who made the kickoff at the 2014 Brazil World Cup using a mind-controlled exoskeleton.
When it comes to actually connecting brains to computers in a more traditional sort of read-write capacity, though, a lot of the biggest breakthroughs so far have been achieved with monkeys.
As early as 1969, scientists were able to get monkeys to move the needle on a dial using just brain signals. By 2008, they were using their brains to control robot arms that brought them food, and, in 2011, monkeys were moving robot arms while receiving feedback to the part of their brains responsible for arm control, as if they were picking up signals from one of their actual arms. In 2016, monkeys achieved twelve-words- per-minute typing speeds using neural implants, spelling out Shakespearean phrases like “A banana by any other name would smell as sweet.”
If these trends continue, we won’t even have to worry about AI taking over the world – telepathic monkeys will probably get there first.
We’ve also been able to pick up accurate image signals from the brains of both cats and humans. By monitoring brain activity in cats, each watching one of eight short films, researchers were able to construct enough images from the screen to identify which movie each cat was watching.
And that was in 1999 – by 2008, researchers had successfully done this in humans. Between those two trials, a teenage boy who had gotten brain implants for epilepsy was able to beat several levels of Space Invaders using nothing more than brain signals.
This list could easily go on: people are typing, moving mice on screens, driving cars, tweeting, sending silent messages to each other over the Internet, piloting drones, controlling smart TVs, and more. This technology is headed in some interesting directions, as evidenced by the wealth of startups rolling out big ideas.
This project grabs a lot of headlines, and for good reason: Neuralink is probably the smallest, most lightweight, easily-implantable partially-invasive BCI out there, and it doesn’t sacrifice much in terms of functionality. Here’s how it works:
- Neuralink uses a robotic tool to implant tiny threads (1/10 the width of a human hair) into the brain. Currently, this requires drilling tiny holes, but it may use a laser in the future.
- These threads are connected to chips, called “N1 Sensors” embedded in the skull, which can process the data they receive and also send electrical signals into the brain.
- Those chips are connected to a device called “The Link,” which is a wearable external device that can interact with the chip and wirelessly send and receive information from other devices.
Trials on rats and monkeys have already proved successful, according to Neuralink, and they’re aiming to start human trials soon. The first recipients will be people who have a medical need for the device, but Elon Musk has made no secret of wanting to eventually create human-AI cooperation through this sort of brain interface. The final product will be able to link to and control smartphones and other devices, and as the technology improves, we’ll likely see a lot of other applications.
A brain implant developed by Cyberkinetics, BrainGate is one of the earliest modern BCIs. Human test subjects were able to successfully connect computers with their brains as early as 2004, and by 2012 Cyberkinetics demonstrated a brain-controlled robotic arm that enabled paralyzed people to reach, grasp, and even drink from a bottle. They have a long list of publications and achievements, ranging from helping people control robot limbs to using brain signals to play the piano on a tablet.
Given that not many people are likely to line up to get holes drilled in their heads, it makes sense that a commercially-available non-invasive BCI would be a good first step. That’s exactly what Emotiv does. Their EEG (electroencephalograph) headsets can pick up signals from your brain, allowing you to analyze your own mental activity, control devices, or even get insights into how consumers are using products. It’s relatively inexpensive, produces research-grade data, and all you have to do is slip it on!
Like Emotiv, Neurable is taking the non-invasive route, marketing its headsets as hands-free, voice-free ways to control the digital world. Something that sets them apart is their focus on VR technology as a way to enable mind-controlled gaming, training, and digital control, which could be quite useful as a way to train your brain into quickly picking up new patterns, like adapting to virtual limbs.
These are just a few of the current projects going on in the brain-computer interface space. Like any cutting-edge tech field these days, there are lots of players vying to be first, and there have already been a lot of interesting applications put into the field, from BrainCo’s study-enhancing headband to Kernel’s AI-powered memory storage chips.
How far are we from BCIs being the norm?
You probably won’t be getting a Neuralink installed anytime soon. Despite its small size, putting any holes in your head at all seems like sort of a big jump. However, the market for non-invasive electroencephalogram headsets is starting to grow, and some commercially available headsets actually come at a somewhat reasonable price.
Eventually, these might be reliable enough for you to use as part of your normal technology stack, but it may take some time. BCIs are improving and miniaturizing faster than ever, though, as Neuralink demonstrates, so it’s entirely conceivable that in the future getting a brain-computer link installed will be something akin to, as Elon Musk puts it, getting LASIK. Not everyone will do it, but it will be a relatively trivial procedure if it’s something you need.
Image credits: Blausen 0244 Cochlear Implant, Emotiv Epoc+, EEG-based BCI, An integrated brain-machine interface platform with thousands of channels, Monkey using a robotic arm, Brain-computer interface (schematic)