A Speculation on the Future of Brain Computer Interfaces

Warning: this article contains surgical and medical imagery that some readers might find upsetting or distressing. Please proceed with caution.

Published 19:19 11/7/2020, updated 17:45 12/05/2020

While technology is in the process of redefining every last aspect of our lives, there is still one last domain that big tech cannot directly touch — our brains. Phones and computers may manipulate our thought processes and influence our innate reward systems, but every interaction that we have with our devices is buffered by our fingers, our voices, our eyes. Our brains output a set of signals that command our bodies to execute a certain behavior, which is then received by the device of your choice; the device sends a set of signals back into our brain through our perceptual systems, and the cycle repeats.

How we interact with computers now. Note: I/O (input/output) devices are the things that allow us to communicate with our computer (think screens, keyboards, mouses, microphones, etc.).

But, the explicit boundary that blocks computers from directly interfacing with our brains is becoming more and more tenuous as computing power increases and our understanding of the brain deepens. Aptly named brain-computer interfaces (BCIs), this new technology makes it possible for scientists and engineers to cut out the middle man. BCIs would enable our brains to output signal directly to devices, and for those devices to then send signals directly back to our brains.

How we would interact with computers using BCIs.

For decades, science fiction writers and directors have pondered how such a technology might impact our lives and change our societies. Their works predict a wide array of outcomes, ranging from the dystopian society of Sleep Dealer, where BCIs allow for the mass exploitation of migrant workers and for the streamlining of military operations, to the utopian world of The Last Question, where humanity’s merging with a omniscient machine allows us to survive the heat death of the universe. But, this future society revolutionized by BCIs is still relegated to the realms of fiction and dreams. BCIs are currently just a budding technology in its infancy, a few small steps into the long journey towards changing the world.

The Science Behind BCIs

As the name "brain-computer interface" suggests, there are three main components in the BCI, each equally important to the technology. In this section, we will take a rigorous look at the brain, the computer, and the interface.

The Brain

The brain is often cited as the most complex object in the known universe. In your head sits a small but incredibly powerful biological computer capable of computation and inference that modern AI can only dream of. With hundreds of billions of individual cells and even more dynamic connections, our brains are unimaginably intricate and thus nearly impossible to fully understand with our current technologies. Below is a visualization by the University of Southern California's Laboratory of Neuro Imaging (LONI) of the neuron tracts of a person’s brain, viewed from the side — it’s important to keep in mind that each “strand” here represents not a single neuron, but a bundle of neurons.

To fully cover every facet of neuroscience would take thousands of pages, so rather than doing that, let's take a look at the brain at two different levels. Just a warning though, to all the neuroscientists out there: this is going to be way oversimplified, so plug your ears, cover your eyes, and skip to the computer section.

The Cells of the Brain

There are hundreds of billions of cells in the brain, separated into two main categories — neurons and glial cells. Glial cells, which I won't be covering in much depth for the sake of space, take on an auxiliary role, supporting neurons through clearing debris, protecting neurons from death and infection, and maintaining efficiency in neuronal communication. The star of the show in the brain are the neurons, specialized cells responsible for all of the computation and communication that goes on inside of your brain. Here is a drawing of a neuron made by an incredibly influential neuroscientist and artist named Santiago Ramón y Cajal (he also essentially robbed graves with his father in order to study anatomy, but that's a story for another time).

Since the above image is quite overwhelming to look at, let's instead take a closer look at a more simplified version of a neuron.

The soma, or cell body, is fairly similar to all other cells in the body — it has a nucleus that contains all the genetic information, and organelles to help the neuron produce energy and regulate cell function. The dendrites, axon, and axon terminals are more specific to neurons, and play a highly specialized role in a process called an action potential. For reference, dendrites are the part of the cell that receives input from another neuron, and axons are where the neuron sends output from to other neurons. More information can be found here.

Like all cells, neurons are just complicated sacks of charged particles, or ions. The neuron carefully guards the balance of ions that exist inside the cell and outside the cell, which allows it to maintain a resting membrane potential of -70 mV. You can think of the membrane potential as how negative the neuron is in comparison to the environment. During an action potential, the neuron's carefully calibrated balance of ions shifts dramatically, causing the membrane potential in a small portion of the axon to shoot up dramatically before falling again. This shift propagates down the axon, away from the soma and towards the terminals.

Action potentials are an all or nothing event, so the neuron either fires or doesn't. If the neuron fires, the axon terminals release neurotransmitters that then stimulate the dendrites of another neuron, repeating the cycle all over again. When one neuron fires, it will stimulate the neurons that it is connected to; this can create beautiful patterns of activity over swaths of hundreds or thousands of neurons.

In a brain with billions of neurons, deciphering action potentials can be difficult. Luckily smart neuroscientists have figured out that it's not actually impossible. In 1986, a paper was published that showed that arm movement could be mathematically modelled directly from neuronal output. While the science has since shown that the specific conclusions made in the paper were not correct, ultimately this paper was foundational in showing that neuronal output can be used to predict physical behavior. Shown below is a rendering of the model that the paper proposed.

The Systems of the Brain

Now that we have a basic understanding of how individual neurons function, let's take a look at higher level structures in the brain. Keep in mind, though, that everything in the brain is composed of neurons, and all functions of the brain depend on action potentials. Here is a labelled diagram of the brain.

A useful way of thinking about the brain is that the more surface level the structure is, the newer it is in an evolutionary sense. Deep structures such as the brain stem or the basal ganglia have been around for millions of years, and typically govern unconscious processes like breathing, sleep, muscle tension, and habits. Surface level structures like the cortex are quite new, and are responsible for processes like sensory processing, cognition, and attention.

There is a concept in neuroscience called "the localization of function" that underlies pretty much all of neuroscience research; it is the idea that a particular area of the brain is responsible for a particular function. Perhaps the most famous examples of this are Broca's and Wernicke's area. Both of these areas are important in speaking, but Broca's is mostly for the production of language, while Wernicke's is mostly for comprehension of writing or speech. If one of these areas is damaged, a very particular type of speech disorder, or aphasia, will result — damage to Broca's area will cause non-fluent aphasia, and damage to Wernicke's area will cause fluent aphasia. Here are videos that serve as examples of these kinds of aphasia, but be warned that this might be upsetting to some readers.

Click to reveal Broca's aphasia video

Click to reveal Wernicke's aphasia video

The way that scientists discovered this is quite gruesome, but the method still persists to this day. They take electrodes and stimulate different areas of the brain, carefully watching and logging what areas corresponded to what responses. In many brain surgeries, the doctors will keep the patient awake and have them do something that they want to continue to be able do after the surgery, and they will test different areas of the brain to map out what is safe to cut. Here is a video of an opera singer who underwent brain cancer surgery, and wanted to ensure that his singing would be intact post-surgery. The surgeon carefully stimulates different parts of the brain, while the patient sings; if the patient experiences any difficulties in singing or processing musical information at a stimulation site, the surgeon knows not to cut there.

The video below contains surgical and medical imagery that some viewers might find upsetting or distressing. Press the button below to reveal the content.

Click to reveal video

After decades of experimentation, scientists today have a fairly good idea of where certain functions are localized to in the brain. Being able to look at a portion of the brain and know what it does is extraordinarily important in BCIs. This is a diagram that shows where some (but obviously not all) functions are located.

Hopefully that was a broad enough overview of neuroscientific research, but if not, there are lot's of videos and write-ups of concepts that you're curious about. For now, let's move onto the "interface" part of BCIs.

The Interface

The primary job of the interface is to take information coming out of the brain in order to transmit it to the computer. There are many of complicated methods of neuro-imaging that scientists employ to get information out of the brain, each with its own benefits and deficits. To discuss the relative merits of these brain imaging techniques, we can rate each on two different scales: temporal resolution and space resolution. Below is a graph rating the major imaging methods ranked on these scales.

On this graph, the y-axis represents spatial resolution, where the higher you go, the worse the resolution gets. Think 240p vs. 1080p. The x-axis represents temporal resolution, where the more right you go, the worse the resolution gets. Think stopwatch vs. sundial.

The method with the best time resolution and space resolution here is labelled "neuron activity." This type of imaging is basically just recording the membrane potential and membrane current of single neurons or small groups of neurons. It's very invasive, as an electrode needs to be inserted into the brain.

Moving up in the spatial resolution dimension, we arrive at magnetoencephalography (MEG). Do you remember when we covered action potentials above? When ions move, they generate an electric current, which means that we also get to deal with electricity's evil twin sister, magnetism. MEGs measure the magnetic fields of the currents generated by action potentials.

Right above MEG is electroencephalography (EEG). EEGs directly measure electrical activity of the brain. This method has less spatial resolution than MEG, but it is able to measure activity from more areas of the brain. EEGs are an incredibly important way of measuring brain activity due to its incredibly good temporal resolution.

Moving over to the other category, let's look at single photon emission computed tomography (SPECT) and positron emission tomography (PET). This type of imaging involves injecting patients with a radioactive isotope, which then allows scientists and doctors to trace blood-flow in the brain.

The last neuroimaging technique to cover is functional magnetic resonance imaging (fMRI), which is probably the flashiest and most famous of all the methods. fMRI uses strong magnets to align certain atomic nuclei to measure blood flow. Because it is so sensitive to noise, though, scientists use statistical methods to tell whether there was actually activity in a certain region. Unfortunately, fMRI is quite controversial because of this, as you can almost always manipulate statistics to get what you want. Scientists once sent a dead salmon through a fMRI scan and found statistically significant activity. Despite this, fMRI remains extraordinarily popular in research and in medicine.

If you were to design a system that could serve as an interface between the brain and a computer, what would you prioritize? You'd probably want a way to capture as much of the brain as possible, with as much spatial and time resolution as possible. In a perfect world, we would have a way of collecting all of the information that the brain produces with the very best spatial and time resolution. Sadly, we do not occupy such a world, and so we have to weigh our options and find an option that works. Looking at spatial and temporal resolution, which one is more important? Since so much of our interaction with computers is dependent on timely input and feedback, temporal resolution is probably more important. With those constraints in mind, our best option is probably EEGs. EEGs provide excellent temporal resolution at the expense of spatial resolution, but there are ways of remedying that low spatial resolution.

EEGs use electrodes to read electrical activity coming out of the brain. An obvious way of increasing the amount of information that EEGs can capture is through increasing the number of electrodes. The cheapest EEGs have 1 electrode, and are fairly useless beyond capturing the most basic of brain information.

You could probably fit around 512 electrodes on an EEG cap, which would provide a lot more information than 1 electrode. Here is a cap with 256 electrodes.

BCIs made with EEGs are called "non-invasive BCIs." Unfortunately, for most use-cases, even these high electrode density EEG would not be good enough for BCIs. So, beyond just cramming more electrodes on top of someone's head, scientists can also start putting electrodes under the scalp to increase the amount of useable data. This type of neuro-imaging would no longer be called EEGs, though, as EEGs are by definition from outside the scalp. As you can imagine, placing electrodes under the scalp is quite invasive.

Putting electrodes directly into someone's brain yields the best information resolution, and is known as a local field potential (LFP) recording device. Because your brain generally does not enjoy having needles stuck in it, though, these electrodes can only ever be implanted in the short term. Also, while spatial resolution is good, it cannot capture information from a large portion of the brain. BCIs that use LFPs are called "invasive BCIs."

Rather than sticking electrodes into someone's brain, scientists can place electrodes onto the surface of someone's brain; this technique is called electrocorticography (ECoG). ECoGs can measure many LFPs all at once, so it can measure activity from a larger area of the brain. While less dangerous and invasive than LFP recording devices, ECoGs still require brain surgery. BCIs that use ECoGs are called "partially invasive BCIs."

The image below contains surgical and medical imagery that some viewers might find upsetting or distressing. Press the button below to reveal the content.

Click to reveal image

Image of ECoG. Image from: Nanda et al.

Most BCIs are made using LFPs, ECoGs, or EEGs, but there are certainly other methods that scientists are working on. In many cases where brain activity measurements are not really needed, neuro-imaging techniques like the ones shown in this section are unnecessary; but, these are not known as BCIs.

The Computer

The last component of the BCI is the computer. Really, though, there are two meanings to "computer" here. One is actually an extension of the interface, as it is the algorithm used to decode brain information. The other is the actual computer that is being controlled.

Getting Useful Information from Electrical Activity

As you saw in the explanation for what an EEG is, the information coming out the interface is a time-series set of voltage potential values. To get relevant information out of the electrical activity, we need to process the signal into something we can easily use and then translate the processed signal into meaningful data.

Processing the signals

Warning to all the electrical engineers out there: this is way oversimplified, so cover your eyes, plug your ears, and head to the classification section. There are a great many methods to process signals, but all of these techniques involve some sort of converting time data into frequency data. Using Fourier transforms, eigenvectors, or autoregressive methods, it is possible to turn messy, difficult-to-work-with time data into frequency data. We can then extract useful features from this frequency data.

Translating Features into Meaningful Information

Warning to all the computer scientists out there: this is way oversimplified, so cover your eyes, plug your ears, and head to the next section. Now that we have a manageable signal to work with, we need to translate brain activity into actual information. For example, based on this EEG data, is this person thinking about moving their left or right hand? The state-of-the-art way to do this would be a machine learning concept called classification. What classification algorithms do is take sets of features and categorize them into classes. Usually this take quite a lot of data, the algorithm needs training before it can do anything. We can look at this graph for a simple example of a classification algorithm; the red and green dots are what the algorithm were trained on, and the algorithm was able to successfully determine how to classify the data. Given this new data point, the algorithm would look at the features of this new data and make a determination for what class it belongs to.

Here is an overview of this section just for review.

What We Can Do with This Useful Information

The short answer here is that the world is your oyster. You can use the information to control a robot, move a computer cursor, or type on a keyboard. If you can get useable signal and convert it into meaningful information, then you can do anything you can think of. If you can make it work, go for it. In the next section, we'll look at some existing applications of BCIs to show the wide range of possible "computers" you can control with brain data.

What's Missing?

You maybe have noticed that this part of the essay focuses almost entirely on how to get information out of the brain and not how to put information in the brain. I kept this separate from the Brain, Interface, and Computer sections because really it's an entirely different problem. I wish with all my heart I could do a whole other long-winded section about this, but truthfully, this aspect of BCIs and neuroscience is still in its infancy. I will try my best to give information about it, but this input process is a complicated issue many magnitudes harder than the ouput process shown above. Maybe we can get away with a just-good-enough knowledge of the brain to extract information, but without a real, robust model for how exactly the brain works, inputting information into the brain is dangerous and difficult. With that said, there are exciting developments in the field that may or may not one day prove relevant to BCIs.

Optogenetics

So what if I told you there was a way we could shine a light at your neurons and they would activate? And with specific neuron activation we can semi-replicate actual brain function? You're in luck, this actually exists. Optogenetics is a method of using light-sensitive ion channels to cause neural activation. Scientists successfully "implanted" memories into rats using this method, and optogenetics are being used to study Parkinson's, autism, and a whole host of other neurological and psychiatric disorders. There are significant barriers (dangerous, invasive, etc.) associated with using optogenetic methods in BCIs, though, so don't hold your breath for it any time soon.

Good Ol' Electrical Stimulation

You're probably already familiar with electrical brain stimulation, and may have even seen it in everyday life. Cochlear implants electrically stimulate the cochlear nerve, bypassing the ear entirely. The brain surgery video above is also an example of electrical brain stimulation. This method is probably more relevant to BCIs, at least for the near future — recently, a group of scientists were able to use intracortical microstimulation to give monkeys artificial tactile feedback. Inputting information into the brain in this way is a tried-and-true method that has been done for quite a while already, but to get to the specificity and safety necessary to stimulate brain areas with high spatial resolution would take quite a bit more research.

Disclaimer

Science has a great way of progressing, even in the face of seemingly impossible obstacles and unsolvable problems. There is so much to explore and discover in the BCI field, and it is not impossible that some novel way to get signals out of the brain or to give input to the brain could be discovered, rendering much of the information presented here incorrect or obsolete. Perhaps the BCIs of the future will use chemical signals instead of electrical ones, or maybe some insane machine learning technique will arise that will be able to decipher brain activity without the need to electrode implantation. Anything is possible, so take everything written here with a grain of salt. Also, I have vastly oversimplified pretty much everything here so assume that all the information here is just 1% of 1% of the current knowledge about the topic.

The Frontiers

The previous section covered what we already know about the brain and the ways that we can get information out of the brain, but as you probably gathered, there is much to discover and learn. In the ever-evolving world of science, the frontiers are always expanding.

Current BCIs

Here are some examples of current BCI implementations, but of course there are many more implementations that are not included here.

Confirmation Bias Machine

The Confirmation Bias Machine was a project that I got the opportunity to work on with a student organization here at USC. I want to talk about this not only to polish my own ego a bit, but also to show just how accessible BCIs can be. You may not be able to do brain surgery, but there are ways of exploring BCIs in fun and exciting ways. With this project, we wanted to poke fun at current (2017) social media algorithms and how they show you only what you want to see. We placed a 1 electrode EEG on a participant and instructed them to watch a series of videos classified into different genres. Using a very simple machine learning algorithm, we captured what genre different participants liked and then gradually showed them that genre more and more until that was the only genre they saw.

A participant in the middle of a demonstration of our Confirmation Bias Machine.

MindBall

This BCI is an interactive game where two participants try to out-relax each other. When one participant is more relaxed, the ball on the table rolls towards their opponent, and the goal is to get it all the way in the other opponent's area. It's pretty amazing to watch two people sitting completely still apparently doing absolutely nothing but still completely engaged in a game. Here is a video from two lucky participants demonstrating MindBall.

Neuroprosthetics

Moving beyond just toy implementations of BCIs, here are two examples of a particular category of BCIs called neuroprosthetics.

Caltech Andersen Lab

In a joint project with USC and Los Amigos Rehabilitation Center, Caltech's Andersen Lab implanted a tetraplegic volunteer with an array of micro electrodes (an example of an invasive BCI) and helped him train for months to control a robotic arm. The science behind this was revolutionary, and involved recording from a part of the brain that encoded action planning rather than a part of the brain that encoded specific movements. The results speak for themselves, and a video demonstration of the BCI is shown below. The video below includes medical imagery that may upset or distress some viewers, so click the button to reveal the video and watch it.

Click to reveal video

JHU Applied Physics Lab

Researchers at JHU's Applied Physics Lab implanted electrodes into a patient using advanced techniques that mapped brain activity in real time during surgery. After extensive training, the patient was able to control two robotic arms and manipulate objects with them. The scientists also implanted electrodes in the somatosensation ("touch") part of the brain in hopes of one day allowing the patient to also feel from the robotic arms. A demonstration of the BCI is shown in the video below. The video below includes medical imagery that may upset or distress some viewers, so click the button to reveal the video and watch it.

Click to reveal video

Neuralink

Because of the splash surrounding Elon Musk's new company Neuralink, I saw fit to include a whole section about it here. This was really the whole motivation behind the essay, so I have inserted a whole supercut of their most recent demo here. The video below includes animal and medical imagery that may upset or distress some viewers, so click the button to reveal the video and watch it.

Click to reveal video

Personally, I think the true innovation of the Neuralink team and what Elon Musk brings to the table is the implantation process. Elon Musk is known for his engineering prowess and not necessarily for his scientific achievements, so this makes quite a lot of sense. I am excited for what they will be able to do in the future, but I also think that they are downplaying some of the challenges associated with BCIs and exaggerating some of the possible future applications.

Challenges

The scope of the problem here is large, and so before BCIs get anywhere, there are certain challenges that will need to be overcome.

Basic Science

While to a certain extent BCIs can view the brain as a black-box and just examine inputs and outputs, ultimately a good BCI implementation needs to build upon knowledge of the brain. Without adequate basic science, there can be no advanced BCIs. And to be quite honest, the little that we do know about the brain is absolutely dwarfed by what we do not know.

Difficult Signals

Even if we did know all there is to know about the brain, there is still the problem of the kinds of signals we are getting from the brain. There is a concept called the inverse problem, which is where you have an output and you need to deduce the cause. Here is a video that demonstrates what the inverse problem could look like.

This is notoriously difficult, especially in the BCI field. Sure, you could implant electrodes into every corner of the brain, but that is really invasive and dangerous. So, we essentially need to make predictions about what we are seeing based off of limited data. ECoGs decrease the magnitude of this problem somewhat, but it is still enormously difficult.

Another issue is that any biological data is very noisy, and brain data can be worse than other biological data. Not to mention that the noisy data problem would be exponentially worse in a realistic environment. Overcoming this would take massive leaps in sensor technology and much better signal processing techniques.

Invasiveness

Currently, you need brain surgery for a good BCI. Do you want brain surgery? Personally, I'd pass. I think with better sensors and more advanced computational techniques, this could be somewhat mitigated, but it would take some truly next level technology.

The Future

Debunking Myths and Exaggerations

As said in the previous section, our BCIs will forever be limited by our understanding of the brain. Some of the more far out speculation that is all over social media like Twitter and Reddit that I have seen include a possibility of curing mental illnesses, eradicating neurodegenerative diseases, uploading skills to your brain, and even complete merging with computers — one post I saw wanted to use BCIs to get high. If we do not currently understand how mental illnesses, neurodegenerative diseases, or skill-learning work, how do we even know what to target in the brain with BCIs? There is always room for optimism and hope in a seemingly more and more unstable world, but I think it's important to remember where we really are in terms of basic science and technology, and to form your expectations from there.

Realistic Expectations

Looking into the future, we can only guess how BCIs may evolve, and what they may be used for. With the added uncertainty of how advanced artificial intelligence will become and how it will interact with BCIs, it is no wonder that social media has exploded into wild, and sometimes outrageous, speculation.

Amid all of this hype, though, there are pockets of skepticism filled with people doubting that BCIs will gain traction at all. One such skeptic, named Joseph Han, is a successful entrepreneur, interdisciplinary researcher, and published engineer who has tackled some of the world’s most difficult problems. His latest company, Deep Valley Labs, focuses on revolutionizing artificial intelligence and machine learning through applying neuroscience concepts to computer architectures and algorithms — his goal is to one day create computers that surpass human intelligence.

He fears that BCIs might go the way of the jetpack, a technology that once pervaded popular culture as a symbol of modern innovation but has persistently failed to materialize as a consumer good. During our conversation, he was careful to delineate between research technologies and consumer products, stating that while BCIs will undoubtably continue to make a splash in the research world, ultimately the BCIs will likely not be adopted in everyday life due to high cost and low incentives. After all, he muses, “to actually actively put something in your body requires appropriate pay-off,” and for him, there is no clear answer for what that pay-off could be. Rather than BCIs, Dr. Han is most excited about other aspects of human-computer interaction (HCI), and looks forward to the day that natural language processing will allow us to talk to computers like we talk to each other. As for “the actual interface between the human and the computer, [he's] not sure if it requires a brain-computer interface.”

This is just one perspective in a sea of viewpoints about the future potential of BCIs, but this is probably a more likely scenario than a full merging with computers through BCIs. I'm sure Dr. Han and I would both like to be proven wrong, but if BCIs continue to be a invasive, dangerous procedure with no revolutionary benefits, then I can't see many people willingly choosing it.

What to Look Forward To

Just because BCIs might not be adopted as a everyday consumer good any time soon does not mean that there are no exciting applications for it. Sook-Lei Liew is a professor here at USC, and her research interests are in helping people recover function after stroke. Dr. Liew has successfully used BCIs to help patients recover motor function after stroke, and she sees rehabilitation as a major application for BCIs in the future. She explained the different types of rehabilitative BCIs to me: compensatory BCIs, which use BCIs to help compensate for loss of function, and restorative BCIs, which use BCIs to regain that function back. Because of the major challenges associated with understanding the brain, she stated that "in 5-10 years, I think [BCIs will be used in the] same functions as we have now — as spellers or maybe still in the research stage... I hope in 50 years, with technology improving, they are used in the mainstream for rehabilitation, pain relief, and similar functions."

She emphasized the role that scientists and the media will have to play in managing expectations about BCIs. When I asked about media coverage regarding BCIs, she said that "as scientists we have worked hard to try to present limits on what can be done with BCIs, but companies like Elon Musk's Neuralink attempting flashy demos doesn't really help us."

Sure, there are future scenarios where a technology will pop onto the scene and suddenly make the really insane applications for BCIs possible, but I think that the future that Dr. Liew presents here is much more realistic, and in some ways, reassuring. Personally, after having worked with Dr. Liew's lab as a student and after talking with her about BCIs, I think having people like her spearhead the effort to change the world with BCIs is much more comforting than having someone like Musk. Her vision will allow a technology like BCIs to be used to help people after brain injuries, which is as worthy of a cause as anything else.

Some Final Words

As someone who has always been interested in the intersection of computer science and neuroscience, I always feel torn in two about BCIs. One part of me is salivating for the future that Musk is pushing for, with the storing of memories and the uploading of skills, but the other part of me recognizes that such a thing would be nearly impossible with our limited understanding of the brain and with our current technology. The world imagined by science-fiction writers is far, far away, but I don't think that we really need to achieve that for BCIs to have made an impact in our society. At this point in time, it is important to be reasonable in examining the current progress in BCIs and to be realistic in making predictions about the future. But, even without the hyperbole or the wild speculation, BCIs are an amazing technology that represents humanity's incredible scientific and engineering achievements — with time and effort, there is little doubt that BCIs can one day make a meaningful impact in the world.

If you have any questions, comments, or suggestions, feel free to contact me at lauracao@usc.edu.

Click to see references