James Vornov, MD PhD
Neurologist, drug developer and philosopher exploring the neuroscience of decision-making and personal identity.
We know that BCIs work and hold great promise, but lets see what history tells us about the journey
In my last post, I described the current state of brain computer interfaces (BCIs). I was surprised to realize that we’ve had working devices for twenty years now. So it’s very clear that electrodes can record motor intent from the cerebral cortex, and this can be used to control devices like computer interfaces, keyboards, or robotic mechanisms. And remember that we don’t need to read the real motor intent; we can just record patterns, and the brain is adaptable enough that the intent can be remapped onto a completely different use. We don’t need to find the index finger control region; a spot on the cortex connected to controlling the tongue is easily repurposed to the index finger or even controlling a cursor.
The technology we have is relatively simple. We have an electrode either on the surface of the cortex picking up local activity or, more invasively, in the depths of the cortex near the neurons themselves, recording ensembles of spike trains. They seem to work more or less the same when we want to detect a motor intent under conscious control. The signal comes out via wires attached to an amplifier specialized for very low amplitude signals.
The practical challenge
There are lots of obstacles to implementation. The signals from the electrodes are tiny. Just 50 to 100 microvolts. And we’re seeing arrays of 1024 electrodes. Implanted multiply. Thousands of channels of tiny signals that need to be amplified. And protected from noise and electrical interference. After all, we don’t want the blender or vacuum cleaner to control the robotic arm. Clearly, shielding and high-performance, multichannel amplification is key. Which is why we see the patients in the current trials with backpacks and big power supplies. That’s a lot of electronics and amplification. And yes, that’s just to get the signal out, it still needs to be analyzed and transformed by deep neural network to control the physical robotic interface.
Are we anywhere close to the marketing picture of the wire going to a little puck under the scalp? My assumption is that the puck is the amplifier unit, and it would transmit to the control unit wirelessly.
Pacemakers from shoe polish tins to Bluetooth
Let’s look at where we are today with devices that help patients hear and see.
The story starts in 1958 with the first pacemaker. It’s an incredible story. At the Karolinska Hospital in Stockholm, the Swedish doctors had a 43-year-old patient whose heart was stopping up to 30 times a day because of Stokes-Adams, a complete block of the heart’s conduction system from the atrium to ventricle. From time to time, the heart just stops beating. His wife sat at his bedside, thumping his chest to revive him every time he passed out. Not a tenable situation to say the least.
Amazingly, she read an article in the newspaper about doctors at the hospital working on experimental heart stimulation in animals. She tracked them down and apparently wouldn’t leave them alone until they agreed to try their device on her husband. Presumably, she got someone else to watch her husband while she convinced them to try and save his life. They scheduled the surgery for that night when the OR staff had gone home. No witnesses.
They cast the circuit in epoxy in a Kiwi shoe polish tin. An actual tin, cleaned out and about the size of a hockey puck. It was based on a single transistor (new technology at the time; some of the first in Sweden were used) to trigger a 2V pulse every 1.5 milliseconds, driving the heart at 70 BPM. Let me just tell you from my perspective: used to on-demand pacing, this is terrifying. Bad timing with a spontaneous beat would send the heart into fatal ventricular fibrillation. It ran on rechargeable NiCad batteries, which would need to be recharged every week with an induction coil.
This was incredible engineering for a first human attempt. It was installed with open-chest surgery, putting the electrodes on the surface of the heart. And the first failed in 3 hours. The next morning, it had to be replaced and lasted six weeks. The patient had 26 pacemakers over the rest of his life. He died at age 86. By the time he died, pacemakers had batteries lasting ten years. Now we have pacemakers with Bluetooth communication that can be monitored and updated from a phone app. The care team can see battery status, alerts, and symptom journals.
The BCI puck?
When you see the BCI puck in those marketing images, they’re using an analogy to a pacemaker and this kind of technology. We have a similar story with cochlear implants which have restored hearing to over a million hearing-impaired people. The timeline and technology are pretty similar. The first attempts in patients were in the late 1950s with simple electrodes stimulating the cochlea with a single signal— loudness but no pitch. It wasn’t until 1978 that the first multichannel implant was trialed and became commercially available in 1984. Now we have fully implantable devices with no external parts, upgradable software, Bluetooth connectivity, real-time environment detection, and noise cancellation. Pretty much just like your Bluetooth noise-cancelling headphones but completely under the skin with a battery that lasts four or five days. We’re driving up to 20 electrodes, processing in real time continuously. With a chest battery that’s pretty small. Not far off from that Swedish Kiwi shoe polish tin.
And yes, if you were wondering, these cochlear implants can stream music from your phone, but apparently, it doesn’t sound great because the whole system is optimized for speech comprehension.
The retinal implant to restore sight
Let me finish this section on current technology with a very relevant cautionary tale, the Argus/Second Sight device to restore vision to the blind. This was a 60-electrode stimulating unit placed on the surface of the retina to provide very basic visual guidance. You can imagine with a 6 x 10 grid you wouldn’t be forming images, just providing information on light/dark status of the environment and some ability to find doorways and windows, outlines of people standing nearby. But don’t underestimate the value of just this level of visual awareness to a blind person. Even though this is basically a single numeral on a 1980s calculator, it provides the basic human affordance for moving around the environment and social engagement.
The device consisted of a pair of glasses with cameras, an implant, a belt-worn video processing unit, battery-operated, which would be kind of a medical-grade Raspberry Pi with custom chips and software for real-time visual processing with edge detection and contrast enhancement to get the most out of that 10 x 6 grid for the patient. It also had to have a simple zoom and some contrast adjustment to be usable. This is not just a camera and simple display; it was real-time adaptive visual processing, the kind of effort we now leave to deep neural networks and machine learning. The system would need developmental and maintenance software updates. And of course, the stimulus unit to tickle the retina with precise pulses.
Choosing the right patients limits utility
The patients were carefully selected based on what was known from restoring sight in the blind with surgical approaches, like removing cataracts. There is a critical period for postnatal formation of the neural networks for vision that requires environmental input to develop. Restoring sight in someone who has a fully developed visual system but lost vision due to accident or disease quickly provides full functional sight. But when children in India with congenital cataracts were removed, they struggled to gain fully usable vision, having trouble decoding the visual world into meaningful shapes in three dimensions. Similarly, with cochlear implants, initially the best results were with restoring hearing to those who had developed the ability to hear and decode speech. Of course, as the technology developed, implants were placed in infants, enabling the development of hearing, albeit a new version based on the cochlear stimulator, not acoustic wave transformation.
A major limitation was that these patients had to have had sight previously with bare or no light perception. This is actually a tiny fraction of those who we consider blind medically since many patients have some light perception or slivers of vision. Many blind patients read with magnification devices or navigate with a cane using their remaining visual cues. This was for the rare patient with an intact retina and no light perception. Plus they had to know what they were getting into. A retinal implant to provide just that 6 x 10 grid. Yet there were patients who were bitterly disappointed, expecting much more functional vision than they actually got from the device.
Second Sight fails and abandons its device
After a single clinical trial that took 7 years to implant just 30 patients, the FDA approved the device for marketing. A total of 350 patients had the device implanted after approval at a price of $150,000 all in for the device, surgery, and rehabilitation to learn to use the device. Even if I didn’t know the outcome of the story, with my years of experience in drug and device development, I could have predicted what would happen next.
I often tell companies that their first clinical trial is their first marketing test. If patients and investigators are lining up to help with the trial, if you have a waiting list or a lottery at sites for enrollment, you have identified an unmet medical need for a real population of patients with an attractive treatment. If you struggle for 7 years to complete a trial in 30 patients, the world is telling you that the patients aren’t there and/or your approach is not what patients and doctors want. The limited number of patients for the trial was an indicatin that this was going to be a tiny market providing an expensive treatment that probably never would become profitable considering the infrastructure needed to keep the devices up to date in the literally installed user base.
Unsurprisingly, after 350 patients were implanted over 7 years, the company discontinued the product, pivoting to — you guessed it — brain computer interfaces. While they promised to support the retinal implants for years to come, you might imagine that with that legacy overhead, the company wasn’t viewed as the best BCI investment by VCs, and they laid off most of their employees in 2020. These patients were left with unsupported hardware implanted in their eyes. It happened that when one of the belt units was damaged, the patients worked together to help find a solution. With hacker solutions and a community, many of these patients still have functioning implants. In fact, 24 of the original 30 patients in the trial still have working systems.
This is just the reality of device implants
We should recognize that manufacturers abandoning implants is not unique here. Pacemaker manufacturers declare end of life for products and stop supporting them. But like our original Swedish patient, pacemakers can be swapped out when they fail. And there have been recalls of devices due to problems like fragile wires prone to breaking. Which in the case of a pacemaker could be fatal. And in general, we leave non-functional hardware in place because the risks of removal are almost always greater than something going wrong as long as the presence of the device has been tolerated so far.
Prefer to follow via Substack? You can read this and future posts (and leave comments) by subscribing to On Deciding… Better on Substack: Brain, Self, and Mind
© 2025 James Vornov MD, PhD. This content is freely shareable with attribution. Please link to this page if quoting.
