What are Brain-Computer Interfaces and How do They Work?
This week, Elon Musk's Neuralink has captured global attention by successfully implanting their wireless brain chip, named 'Telepathy', into a human for the first time. This milestone marks a significant leap forward in the realm of neurotechnology, specifically the field of brain-computer interfaces (BCIs).
BCIs are a technology that enables direct interaction with computers and devices solely through thought. They use advanced neuroimaging technologies and machine learning algorithms to convert brain activity into machine commands. Neuralink’s ambitious project aims to help people with motor impairments by allowing them to control computers with their minds, restoring previously lost functions and helping them live a better life. (For more information on current ethical considerations, see Lea Schmid's article here.)
Musk recently shared Neuralink’s vision, stating, “Imagine if Stephen Hawking could have communicated faster than a speed typist or auctioneer. That is the goal.”
This article will share the simplified workings of BCIs, their potential applications, and the pressing safety concerns as neurotechnology companies like Neuralink move swiftly towards human trials.
How do BCIs work?
Although the communication from thought to computers may sound like something out of a Sci-fi movie, the underlying logic behind it can be understood through a simple acronym, MIND: Measure, Interpret, Encode, Deploy. By the end of this article, you will hopefully understand the basics of this technology, using this acronym to remember the foundational steps in turning thought into machine action.
Measure
Understanding the science behind the brain’s 'language' system and learning to measure it has been pivotal in developing BCIs. Our brain's mode of communication involves transmitting electrical impulses along neurons, creating a measurable electrical charge. This electrical activity is the signal that BCIs capture.
The foundation for measuring this neural 'chatter' was laid in the 1920s with the invention of the electroencephalograph (EEG). An EEG functions by utilizing a series of electrodes placed on the scalp or, in some cases, inside the skull. These electrodes are sensitive to the millisecond electrical charges resulting from the brain's neuronal communication. The unique electrical activity at different electrode points, captured by the EEG, is thus, indicative of specific cognitive states, processes or intentions.
Today, EEG devices have evolved into complex tools that allow for a more accurate measurement of our brain's data. Advances in EEG hardware and neurosurgery are enabling the surgical implantation of extremely thin electrodes inside the brain, offering more precise brain information (picture below). Different types of brain-measuring tools enable us to customise BCI designs according to the specific thoughts or actions that need to be captured.
Interpret and Encode
After measuring the brain “data” using EEG - capturing the electrical activity for a specific brain state, we bring in the Machine Learning (ML) models for interpretation and encoding. These models act like computational translators, turning the complex language of brain waves into understandable actions or thoughts that can be sent out to a device.
Here's an example of the steps of a basic BCI ML model.
Feature Extraction: The ML model first identifies key features in the EEG data related to the action/thought we are interested in. These features might include specific electric frequency bands (like alpha or beta waves), event-related signals, or even more complex features such as connectivity patterns between different electrodes (brain regions) placed on the head. Sophisticated algorithms parse through the EEG data, extracting these features while filtering out irrelevant information. By zeroing in on the most informative aspects of the EEG signals, we set the stage for more precise and meaningful pattern recognition.
Training the Model: We teach the model with a dataset where these patterns are linked to known actions or thoughts. Over time, the model learns to correlate certain brain wave patterns with specific actions or thoughts, so next time it sees that specific brain pattern, it will know the person is thinking about moving his left hand up, for example. Once trained, the model can interpret new EEG data in real-time.
Deploy
This is where the magic happens. The processed and encoded thoughts by the machine learning models are “deployed” to a computer or device to manifest the thought into action, whether this means moving a robotic arm or having a word appear on your screen after having thought about it. The method of transmission to the device varies, with some being done with wired connections, Bluetooth, or even Wifi.
What can BCIs be used for?
The potential applications of this technology are limited only by human imagination. For example, BCIs can interpret the brain signals of a person with paralysis, translate these into commands that a computer can understand, and then prompt a machine to move accordingly. Leading companies in this field, such as Neuralink, Blackrock Neurotech, and Precision Neuroscience, are demonstrating promising clinical applications for individuals with conditions like quadriplegia, paraplegia, upper spinal cord injuries, stroke, ALS, and muscular dystrophy. Some other creative commercial applications for BCIs include translating brain signals into text for computers for everyday use, identifying if you are in the right mental state to drive, or even the implementation of “passthoughts”, where one signs in to personal devices with their unique brain activity.
Future Directions
BCIs are exciting technologies that have the potential to significantly improve and change lives. People who haven't been able to speak, walk, move, or interact with technology can live more fulfilling lives by bridging our advancing devices with our thoughts. Scientific research in the BCI field is advancing rapidly, as evidenced by a more than 600% increase in publications featuring the keyword 'BCI' from 2005 to 2020 and a predicted 15% annual market growth from 2022 onwards. It's clear that BCIs are here to stay, and considering safety and data privacy issues, so contemplating where and how this technology could be applied is important.
I recommend a new book by Professor Nina Farahany called Battle for Your Brain - Defending The Right to Think Freely where she emphasises the need for meticulous ethical considerations whilst approaching the age of neurotechnology. She states “Neurotechnology will soon become the ‘universal controller’ for our interactions with technology. This can benefit humanity immensely, but without safeguards, it can severely threaten our fundamental human rights to privacy, freedom of thought, and self-determination.”
I agree with this statement and believe that, for the time being, whilst we research the psychological effects of this tech, BCI applications should remain limited to those who need it for clinical purposes. We want this technology to improve our well-being and our human experience at an individual level. The careless adoption of BCIs in the general population could further submerge us in a codependence on technology that we might not be ready for.
Nevertheless, be on the lookout for advances and novel applications of these fascinating technologies. Below are some fascinating applications for BCIs from top neurotech companies, showcasing their immense potential to significantly improve people's lives.
If curious, here are links to more information: