Can brains and computers co-exist?
“Would you let Elon Musk tinker with your brain?” That’s the question Matt Novak asked in his article in which he covered the departure of Neuralink’s co-founder from the company. Exactly three days later all major news outlets reported the malfunction of Neuralink’s first implant. Perhaps most of us are asking the same question now. But what is this implant anyway and why are we inexorably moving towards a future of brain computer interfaces?
CHAPTER 1
After Stephen Hawking was paralysed due to amylotropic lateral sclerosis (ALS), the world was fascinated by how he used a speech synthesiser to communicate. In the beginning he used his fingers to generate sentences on a computer, then as the ALS progressed and he lost the use of his fingers a cheek switch was devised that could detect cheek movements using an infrared beam. Perhaps if he were alive now, he would have been excited about the prospects of controlling a computer by the use of his brain alone.
In March we saw a video of Noland Arbaugh who was paralysed from the shoulders down after a diving accident, moving a mouse on a computer screen by just using his thoughts. He was playing chess, and later he also played Civilization VI, a strategy game for a whole night. He was able to do this with the help of a chip that was put in his brain by Neuralink, a brain computer interface (BCI) company owned by Elon Musk. We got a first peek into the tech in 2019 when Neuralink published a small pre-print paper to demonstrate proof of concept. This type of brain computer interface has been around for a couple of decades though and requires special surgery for it to be implanted. What Elon Musk and co achieved was scaling up and miniaturising the electrodes so that more electrodes could be implanted to stimulate potentially single neurons. They also made the technology more physically flexible which aided compatibility with brain tissue.
Two weeks after the Neuralink news, there was news of another brain chip company called Synchron which is now being backed by Jeff Bezos and Bill Gates. Their device known as Synchron BCI has had a reasonable presence in the academic circles with multiple peer-reviewed publications of the technology between 2016 and 2021 as well as a successful published clinical trial in 2022. But it did not share the same media exposure as Musk’s Neuralink. The Synchron BCI method was different to Neuralink in that patients did not need to undergo special surgery for the electrodes to be fitted in the brain. Instead, the electrodes were being delivered into the brain using a form of a ‘stent’. Stents are tiny tubes that have been used for decades to keep open blood vessels blocked by plaque formation. By attaching electrodes to a stent and inserting into a blood vessel from a specific region of the brain, the electrodes are able to pick up electrical signals from that specific brain region. This method has proved to be a less invasive way to reach the motor cortex to enable control of a computer. Due to its less invasive nature and simpler procedure experts feel this technology has the potential of reaching more patients worldwide compared to Neuralink.
CHAPTER 2
The interface between brain and computer is not limited to brain chips that restore brain function. Knowledge from how the brain works is being used to train the next generation of computers. Artificial neural networks (ANN) is one such computational model inspired by the brain. This model consists of ‘nodes’ which resemble the clusters of neurons in our brain that process information and send them to connected nodes. By connecting nodes in various patterns ANNs can learn and model complex relationships. ANNs are widely used models in machine learning and artificial intelligence. However, they are not very efficient. ANNs need very large training samples to undertake real world tasks, and effective neural networks require enormous amounts of computing resources.
Brain on chip
In order to make ANNs more efficient, a more radical approach was conceived. This approach is known as ‘reservoir computing’. In this approach, electrical signals are passed through a fixed non-linear system called the ‘reservoir’. After electrical stimulation, the state of the reservoir is then interpreted as a simple readout. The advantage of this method lies in the fact that naturally available systems can be used as reservoirs — such as biological systems, and this significantly reduces computational costs. One such biological system being tried out is brain organoids.
Brain organoids are three-dimensional aggregates of neurons that develop and connect into networks in a natural biological process similar to how our brain development. When the neurons in the brain organoids are stimulated by electrodes not unlike those used in the brain chips, the neurons ‘learn’ the patterns of electrical stimulation and create a ‘memory’ which can then be used to generate a response that can be recorded by an attached computer. Using this method researchers found that brain organoids could be programmed to recognise audio clips and generate appropriate responses to different audio clips, thus enabling the computer attached to it to learn patterns more efficiently. It is believed that this method may be able to overcome the main caveats of training artificial intelligence models, i.e., long training times and limited data availability for training. By harnessing a biological system this method also uses low energy.
To conclude
We are witnessing the slow but inevitable march towards creating better computer chips to aid human brain function and creating better biological models of the human brain to aid efficient computing. We also now fully understand that our brain and our computers can co-exist in the same physical space. Perhaps in an ideal world of the future, physical disability will no longer be a lifelong struggle, and life changing injuries will be a thing of the past. But we are off to a slightly dodgy start as we watch ethically unreliable silicon valley billionaires freely tinkering with the very fabric of our existence, our brains.
Selected references
Opie, Nicholas L., et al. “Chronic impedance spectroscopy of an endovascular stent-electrode array.” Journal of neural engineering 13.4 (2016): 046020.
Mitchell, Peter, et al. “Assessment of safety of a fully implanted endovascular brain-computer interface for severe paralysis in 4 patients: the stentrode with thought-controlled digital switch (SWITCH) study.” JAMA neurology 80.3 (2023): 270–278.
Musk, Elon. “An integrated brain-machine interface platform with thousands of channels.” Journal of medical Internet research 21.10 (2019): e16194.
Cai, Hongwei, et al. “Brain organoid reservoir computing for artificial intelligence.” Nature Electronics 6.12 (2023): 1032–1039.
