Early 2017 has been an exciting time for brain-computer interface (BCI) development. Major players in the tech industry have announced their involvement in the field- Elon Musk announced his new venture Neuralink, which is fueled by the fear of humans being left behind by A.I., and several weeks later Facebook announced their own efforts led by DARPA’s Regina Dugan, with the goal of enabling users to type 100 words per minute using their minds.
But what exactly are BCIs? What else could they be used for? How do they work and why all the hype?
What is a BCI?
Simply put, a brain-computer interface is a way to connect the brain to an external device in order to send and/or receive information directly from it. They’re nothing new as research started in the early 70’s, so why are they gaining so much attention lately? What makes them so exciting?
Meet Steven Hawking
He is one of the most brilliant theoretical physicist of our time, greatly contributing to the fields of cosmology, quantum gravity, and black holes. Due to a nasty neurodegenerative disease, he’s been trapped in his own body, unable to speak or move for several decades.
His mind is fully functional- he even writes books and gives lectures but unfortunately his way of communicating with the world has been reduced to typing through an mechanism embedded in his glasses using his cheek muscle movement, allowing him to type just one word per minute.
The Biological Bottleneck
Hawking’s mind is not the only one trapped in a body that has reduced it’s creative output. We’re all victims of the same phenomenon.
The process of getting thoughts out of our heads and into the real world is incredibly slowed down as they first have to be converted to mechanical motion, whether it’s our fingers to write or type, or our larynx to speak and pronounce.
As you can see, we listen 10x faster then we write. This holds true not just for language but any form of information- images, video, and audio.
This gets at the very problem BCI’s are aiming to solve: increasing our cognitive bandwidth, as our Input ability exponentially exceeds our Output ability.
So Where Are We Today?
Currently there are several commercial BCI’s that are routinely used by thousands of people, such as deep brain stimulators that alleviate Parkinson’s Disease symptoms, cochlear implants that restore hearing and and even retinal implants that restore vision.
Here are some of today’s BCI’s:
Scientific application aside, EEG BCI’s have become fairly affordable over the years. You can pick up a headset at Amazon for a hundred bucks:
There are even brain games…
Health aside, the exciting stuff is yet to come. Here are some possibilities:
Brain to Computer
Here are some potential advantages of why you would want to connect your brain to a computer:
- Being able to capture your ideas into words, images or even video in real time, without the need to “recreate” them with software
- Access information and data from the Internet on the fly
- Upload your memories
- Download knowledge and possibly skill sets (questionable as of now)
- Have a cognitive AI assistant to aid you in decision making and task management
- Adding extra sensory inputs
- countless more
Brain to Brain
Things get even more interesting when people are able to interact via BCIs, which would usher in a major paradigm shift in human communication.
For thousands of years, our language hasn’t evolved past writing. Sure, our methods have gotten more advanced, but it has essentially remained the same, as we still type out individual letters on our phones or speak.
BCI’s could turn telekinesis into reality- being able to send thoughts and even emotions to your friends and family wirelessly from anywhere in the world. And instead of describing your thoughts in words, you send them in a “raw” format; the abstract concept in your mind before you form it into a word. Multi-dimensional concepts that would otherwise take thousands of words to explain, could all the sudden become intuitive to others in seconds.
This all sounds great, but where do we even start?
It’s All in Your Head
“If the human brain was so simple that we could understand it, we would be so simple that we couldn’t”
In order to understand how to connect to the brain, we need to cover some fundamental neurophysiology principles…
The Three Brains
Let’s start with the big picture first-
This is a gross oversimplification of the configuration of the human brain, but you can think of it in 3 distinct layers that evolved over time:
The Reptilian Brain
The first and oldest layer that sits at the base. As the name suggests, it evolved before homo sapiens and is shared by all reptiles and mammals. It is responsible for basic involuntary functions like heart rate, breathing, blinking, etc as well as survival instincts such as fear and aggression.
The Mammalian Brain
Evolved on top of the reptilian brain houses the limbic system and is responsible for memories and emotions. It governs the fight or flight reflex, along with desire for food and sex.
Last but certainly not least, the neocortex is essentially what makes us human. It’s responsible for thinking, reasoning, planning and executive functions.
The Brain’s Building Blocks
The main cells of the brain are called neurons and they are cognitive highways on which electric signal travels to communicate with other neurons. There’s about 80 billion of them and they all connect to each other, forming a massive network of 100 trillion synapses (the connection between neurons.)
Neurons have to two sets of cables to to speak:
- dendrites: receive electric impulses from other neurons
- axons: transmit electric impulses to other neurons
How neurons “talk” to each other
When the axons of one neuron releases neurotransmitters to the dendrites of another, that chemical reaction changes the receiving neuron’s charge to positive, causing it to fire an electric signal which is called an “action potential”
These action potentials are binary- it’s either on or off with no middle state and could travel anywhere between 1 to 100 meters per second depending the axon layering of myelin sheath, which not only coats and protects axons, but speeds up the electric impulse.
Ok, now that we covered some basic neuroscience premises, and got the dull part out of the way, let’s move on…
As if the sheer number of neurons and synaptic connections is not challenging enough there’s another layer of complexity- neuroplasticity.
Neurons are always rewiring themselves to other neurons and even changing their structure and functionality based on our life experiences in order to optimize us to our environment.
By this point it should be apparent that we understand extremely little of the brain. Harvard professor Jeff Lichtman is quoted stating:
“If everything we need to know about the brain is a mile long, how far do you think we’ve walked that mile?
Depressing, I know, but here’s the good news…
Don’t Sweat the Small Stuff
The good news is that we don’t actually need to fully understand the brain in order to create BCI’s that work. We just need a really good grasp of it’s high level functionality. We don’t fully understand what happens inside of A.I. neural networks either, but that didn’t stop us from building an AI to beat one of the top Go players in the world.
Our next logical step towards achieving the potential wonders of BCI’s is to scale up our neuron capturing capabilities as we’ve already made significant progress in that area.
When it comes to capturing neural activity, there are two promising methods
- Record an electron’s electric activity when it fires (action potential)
- Record the oxygen spike (oxygenated hemoglobin) as blood flow increases to the neuron’s general area, which has a 2–4 second delay
Comparing BCI Capturing Approaches
When it comes to recording brain activity here are the 6 major devices. We’ll analyze them by the following attributes:
- Scale -How much information could be captured from the whole brain)
- Resolution -The quality of information captured
In order to have useful BCIs, we need an approach that ranks highly in both of those ares.
Let’s start with the non-invasive methods which don’t require surgery…
(functional magnetic resonance imagine)
-Neuroimaging procedure that captures neural activity by tracking changes in blood flow, which due to its nature is delayed by several seconds, not providing real-time information. It also requires strong magnetic activity.
Scale: high (shows info across the brain)
Resolution: medium-low spatial, very low temporal
One of the most ubiquitous method due to low cost and easy deployment that uses external electrodes to track electric activity.
Resolution: low spatial, medium-high temporal
Now, let’s move to the further end of the invasiveness spectrum that require some form of surgery…
Similar to EEG, except it’s deployed below the scalp, on surface of the brain.
Resolution: low-spatial, high temporal
(Local Field Potential)
Similar to ECoG except it’s pinned on the actual cortex of the brain using micro electrodes.
Resolution: medium-low spatial, high temporal
Similar to LFP, it uses a neural electrodes to capture single neuron activity.
Scale: Extremely small
Resolution: Extremely high
A suction cup that clamps a neuron and can record its activity with a super-high resolution with very little noise. What makes this technique unique is the fact that the clamp can actually stimulate the neuron as well with electric current.
Scale: Extremely small
Resolution: Extremely high
Sharp Electrode Recording
This is the only intracellular technique that actually penetrates the neuron to record individual electric activity. This however will never scale as it kills the neuron.
Resolution: As high as it gets.
Ian Stevenson, PhD has observed:
The number of neurons we can simultaneously records doubles every 7.4 years, if continued, end of this century will reach million, entire brain by 2225
It’s disheartening to hear that even with exponential progress we’re more than 200 years away from a full brain capture according to Stevenson, but a counter observation by Ray Kurzweil keeps our hopes alive to witness an entire brain capture in our lifetimes:
Exponential trends themselves have experienced exponential growth
“when the Internet went from 20,000 to 80,000 nodes over a two year period during the 1980s, this progress remained hidden from the general public. A decade later, when it went from 20 million to 80 million nodes in the same amount of time, the impact was rather conspicuous.”
State of the Art Instruments
We’ve covered a wide breadth of neuroimaging methods, but techniques by themselves are not enough- we need sufficient advancements in the instruments we use in order to increase overall bandwidth, less invasively.
To put things in perspective, today, a couple hundred electrodes measure the activity of ~500 neurons. In order to achieve meaningful advances, we need to be capturing 100k to 1 million, depending on who you ask.
Here are the most promising instruments that could take us to where we want to be…
A silk film embedded with probes that wraps around the crevices of the brain’s surface.
This electronic tattoo has the potential to be applied to organs like the heart and brain.
Soft mesh that’s so flexible that it could be injected through a needle and then unfold and wrap around the brain.
Transcranial Magnetic Stimulation (TMS)
A magnetic coil that stimulates the brain externally via magnetic waves.
As the name implies, tiny 100µm silicon sensors are sprinkled all over the brain to capture neural activity and transmit it wirelessly through your skull. It is highly promising as it could scale extremely high and cover the entirety of the brain, while providing very high spatial and temporal resolution. (I was actually present at the GF2045 conference when it was announced in 2013 *slight flex)
This could be the wildest and possibly most drastic paradigm-shifting mechanism. It involves injecting a virus into the brain which delivers light sensitive probing molecules which can be controlled through light. Optogenetics has perhaps the highest potential for resolution and scaling.
This concludes our brief introduction the wonders and possibilities of Brain-Computer Interfaces. Subscribe to the publication for future updates and insights, as we’ll cover major developments.