May 11, 2020
H2M Interfaces Bridging Humans With Machines
Recording dreams. Communicating through mental telepathy. Sending messages directly from our thoughts. These all sound like science-fiction, but is it going to become a reality soon when the “singularity” arrives? Perhaps to understand it, let’s take a look at H2M (Human-to-Machine) interfaces specifically with the BMI (Brain-Machine Interface).
The integration of humans with machines has been happening, but many may not be aware of it. This is because when the news is not sensational enough, it doesn’t get as much interest. Hearing aids, pacemakers and electronic implants enhanced human sensory perception and provide treatments for certain medical conditions. Hearing aids provide amplification of sound to improve hearing loss. Pacemakers deliver precise electrical signals for patient’s with heart conditions. These are tiny machines that provide treatment for patients. The development of devices that can interact with human cognitive processes via the brain would be a new class.
The thing they have in common is an interface to the human sensory system via the body’s very own biological electro-chemical system. Hearing aids amplify signals to the auditory canal of the ear. These are then converted into signals that reach the brain for further processing. Pacemakers use wires called leads that detect and correct abnormalities in a patient’s heart rate by regulating contractions. The basic function of each device requires an interface that can send or receive electrical signals to/from the body. These are electro-chemical in nature, and processed by the nervous system. There are also devices that can connect to the brain to perform tests, like an EEG (Electroencephalogram), in order to gather data on the brain’s electrical activity.
Sensations are perceived as the firing of electrical signals in the body by neurons. These occur at the synapses of nerve cells that are found throughout our body. The nerve cells fire pulses to be interpreted by the brain using neurotransmitters. When the brain receives these signals, it interprets them as reactions that are a mechanism for normal human function. This is meant for survival and safety, like a sensor sending a warning message back to the brain. Humans then process the feedback. For example when we come into contact with a hot surface, the body’s pain receptors are triggered. They fire off neurons and the brain causes a reaction to avoid the hot surface on a stove. The hot surface causes harm to the body and this sort of feedback is what keeps us safe. If there were no pain receptors, then the body would not feel anything, but damage would be done to the skin and tissue in abnormal circumstances.
Electroencephalography
Measuring the electrical activity in the brain is clinical in nature, and performed mostly by medical doctors. This is done using the EEG test. This is used in order to determine certain neurological disorders like the following:
- Epilepsy
- Strokes
- Sleep Disorders
- Head Trauma
- Encephalitis
- Brain Tumors
- Encephalopathy
- Memory Loss
- Dementia
The EEG indicates whether the patient has any of the following disorders associated with their symptoms.
During the test, wires called electrodes are attached to the scalp of the patient. The electrical activity in the brain is measured as impulses by the electrodes and processed by a device which create graphical charts for analysis. The lines that appear on the chart are what the device recorded during the EEG test. Doctors will then use this data to provide the patient diagnostic information regarding their condition. By nature, these electrical signals are extremely weak, unlike the electricity that flows through our outlets at home. This requires amplification of the impulses in the brain to gather stronger signals. Once amplified, it is then converted using an ADC (Analog-to-Digital) process for discrete data signals that computer’s can understand.
The same technique of using electrodes can be applied to academic research in H2M, but using different methodologies. The application in this case is to see how machines can read the electrical signals in the human brain and use that data to provide assistance. Perhaps the main difference is that EEG are non-invasive, while BMI are invasive (depending on its application to probe deeper into the brain).
Brain-Machine Interfaces (BMI)
This provides a way for the nervous system to interface with digital electronic devices. Applications include restoration of sensory and motor function and the treatment of neurological disorders. In AI, the research in this field is focused on algorithms which can process the data from the brain for the purpose of helping patients who suffer from certain disorders that affect the nervous system. This includes epilepsy, nerve damage and traumas that affect neuromotor functions (e.g. physical disabilities).
Neuralink is a good example of BMI. They are developing scalable high-bandwidth BMI system, providing a wider data bus to the brain. This enables capturing more signals or data which are then processed by software. This makes use of arrays of small and flexible electrode “threads”, consisting of 3,072 electrodes per array distributed across 96 threads. Each thread contains 32 electrodes (3,072 electrodes / 96 threads). They also built a neurosurgical robot that inserts 6 threads (192 electrodes) per minute. The total time it should take to insert the threads is 16 minutes. Neuralink’s threads are connected to an N1 sensor chip, measuring just 4 x 5 mm (initial incarnation), which detect spikes of neural activity and send signals back to the brain.
Neuralink uses an invasive procedure. The challenge is the precise insertion of the threads to the brain. If not properly done, it can cause harm to the individual. This requires micron precision for avoidance of blood vessels and targeting specific brain regions. The number of electrodes inserted into the brain does matter. This is because the more electrodes you have, the more data you can gather. It is like having more bandwidth to access the brain.
What Neuralink is improving upon in BMI is the thread count which can gather more data from the brain and the use of more flexible materials that won’t injure human tissue. More threads means more channels, allowing larger bandwidth to the brain. Replacing stiff electrodes with more flexible ones also improves safety. Stiff electrodes don’t shift when the brain moves, thus it can cause damage to tissue. The flexible threads use polymers that are more elastic and shift with the movement of the brain.
Invasive vs. Non-Invasive
So which is better for H2M using BMI, invasive or non-invasive procedures?
According to Neuralink’s white paper:
“Noninvasive approaches can record the average of millions of neurons through the skull, but this signal is distorted and nonspecific.”
Invasive approaches on the other hand:
“Invasive electrodes placed on the surface of the cortex can record useful signals, but they are limited in that they average the activity of thousands of neurons and cannot record signals deep in the brain.”
In that regard, Neuralink still used the invasive approach:
“Most BMI’s have used invasive techniques because the most precise readout of neural representations requires recording single action potentials from neurons in distributed, functionally-linked ensembles.”
It has to do with the accuracy of data they can gather from the brain. The process is not simple, so it requires precise insertion so it will not harm the individual who will be attached to the electrodes.
The threads used are 4 to 6 μm (micrometers) in width, which is much thinner than human hair. These are also referred to as “flexible” threads, as they are less likely to cause damage to the human brain when compared to existing materials that are used in BMI. Therefore it should be safer for invasive procedures. There are criticisms about the risks of invasive procedures and that is going to need to be addressed not just by Neuralink, but all developers of advanced BMI systems.
Applications
Going back to the earlier questions, will these devices be used to record our most vivid dreams, read minds and communicate with thoughts? That is moving a little too far, but first things first. It will start with helping patients who have neurological disorders. Rather than develop something close to sci-fi that have many thinking of the singularity, perhaps a better way to think of this is a device that can help a patient who has lost sensation in their hand due to an injury start to regain those feelings again. This is for prosthetic implants which are due to loss of limbs.
According to Neuralink scientist Philip Sabes (In CNET article):
“You can use this technology in the brain to restore a sense of touch or vision.”
Facebook is also developing their own H2M, which can take interfaces to a whole new level. FB has a BCI (Brain-Computer Interface) unit that is developing applications that allow for more handsfree operation. Like Neuralink, they are starting with medical applications to help those with neurological disorders. One such application is working to help patients with neurological damage speak again by detecting intended speech from brain activity in real time. These applications, when proven successful, eventually become mainstream features that can be integral to social media apps.
According to Facebook’s Mark Chevillet:
“There’s no other way to do it today, but our team’s long-term goal is to make these things possible in a non-invasive, wearable device.”
Google is not to be left out of this conversation. They are also using their expertise in AI to develop H2M systems. This patent is just one example of that. Microsoft also has its own division for Human Computer Interaction (HCI). Other tech giants are getting in the game as this becomes a more lucrative business and the technology can be proven reality.
Final Thoughts
The more immediate need in healthcare is providing ways to treat neurological disorders. Patients will probably not object to implants, but that could be different when it comes to consumers. People are rational beings after all and you have to prove to them that these devices will be practical and safe to use outside of medical applications. FDA will have to regulate them since it requires implanting devices into the human body. Cybersecurity will definitely be another serious consideration for devices that are implanted. There are hacks that can affect the proper functioning of these devices, and that can lead to damage. It must be explored what attack vectors need to be checked for public safety.
For retail the obvious things are price and demand. It should be affordable like most consumer electronic devices. That requires scale and mass adoption, but this is not like developing the latest smartphone. A more ideal form is a wearable device rather than to have an implant which can have unexpected consequences. At the moment retail consumers are about instant and immediate use when it comes to products. People expect to be using their gadgets out of the box. Scheduling for an appointment with a doctor at the hospital to have a device implanted just doesn’t sound too exciting for many (could even be risky).
For other H2M systems, integration with BMI can provide mental control. Imagine the HDTV of the future. Instead of using a remote, the interface will be through a chip implant in your brain. It is attached to a VR/AR headset, and all you have to do is think the channel’s number and work like a remote control. You can use an interactive guide using your thoughts to answer the questions. This is an example of how the brain can work with computers through H2M interfaces. That is the direction where it expects to be going, so it can be considered progress when “A monkey has been able to control a computer with his brain” (Elon Musk).
