.png){kind=logo}
We’ve already invented mind reading technology, now it’s time to apply what we’ve learned to commercialize it. Advancements in artificial intelligence have made pattern prediction more efficient and accurate than ever. Each day Brain-Computer Interfaces (BCI), like Muse or Neurable, improve in usability and precision. So why do I still have to get the remote out of my couch cushion when I want to change a channel? Why do I have to use my hands to steer? Why can’t any of my “smart” and expensive devices just predict what I’m thinking?
Simply Put, They Are Capable of Reading Your Mind
However, the relatively young field of neurotechnology has come with barriers in software, hardware, and user experience. Countless researchers work every day to document new methods and applications for harnessing neural (mental) activity. At MindsApplied, we’ve implemented their research from a software development perspective in an effort to push the boundaries of neurotechnology and healthcare by commercializing telepathy.
Like the previously mentioned BCIs, we use electroencephalography via an EEG headset that is easily connectable to any device via Bluetooth. A few of the other ways scientists gather data from brain activity are through magnetism, blood flow, and dissection. However, the information network of the brain is similar to a computer: not by using wires, but by running on electricity. The reliability of these signals, and their similarity to every other device we use, are why EEGs have become the standard for non-invasive neurotechnology. Yet, harnessing this electricity is about as difficult as it was for Benjamin Franklin to capture lightning. Nevertheless, his perseverance to overcome the challenges of their time was fueled by his focus on the magnitude of its opportunities.
The Possibilities of the Internal Frontier
They’re in the back of your car’s headrest so you can turn off the AC about as quickly as you even realized you were cold. You wear them in your beanie, beret, or ball cap to tell your mom how her grandson’s playing without needing to take your eyes off of him. Soldiers cognitively convey enemy movement on a silent or thunderous battlefield before deciding to ask chatGPT about the requirements for a secretarial desk job.
The applications in control and communication are endless.
Tech tycoons like Elon Musk, Bill Gates, and Jeff Bezos are investing time and resources into a way to output words directly from our psyche. It’s assumed that without needing to take the time to type, we’ll post or search faster and more often.
Speech conveyed directly from the brain will unlock new paths towards brain health treatments that were either previously unavailable or impractical. Those suffering from neuromuscular disabilities, considered “locked in” or paralyzed, will be able to not just converse but even interact with their surrounding environment. When trained to distinguish layers of the psyche, mind reading software will advance our psychoanalytic understanding of the unconscious to a degree unrivaled since the time of Sigmund Freud. However, this technology cannot truly advance if its development is centered solely towards medicinal augmentation. The foremost researchers of telepathy have been and will continue to be constrained by limited institutional resources and pedantic methodological requirements.
However, the quickest integration of mind reading software is with technology already revolving around headwear. Facebook and Snapchat have invested heavily into sunglasses that can record content, and hold expectations for integration of augmented reality. Though functioning similarly to the non-invasive EEGs mentioned earlier, projects like Galea and Looxid have taken this a step further, as they are capable of supporting both electrodes and virtual reality. With such a platform, telepathic technology won’t just be a feature but a necessity.
So, What Exactly is Telepathy?
It’s a very abstract and subjective
concept. Does it consist of communicating words like “Dinner’s at six” directly into the thought stream of my son who’s in the middle of a video game? And if I am outputting my thoughts, how do we deal with the subconscious? I don’t want to accidentally inform the pizza guy of my social security number instead of my credit card.
Let’s narrow it down to an understandable and more tangible question:
“How can patterns in brain activity, corresponding to thoughts, be translated into meaningful sentences?”
The general hypothesis and most achievable approach to the question above is: we can train a deep learning neural network on snippets of brain activity captured from any BCI while thinking various words. From this, we can produce a model which uses online classification to predict, in real-time, the thoughts you are most likely thinking. Telepathy is not so much an achievement, but rather a direction, as translation can only be improved upon. This is the focus of MindsApplied in creating the neural communication software we call Cognichat, and here are the barriers we are overcoming:
Software Can only Advance as Quickly as its Hardware
Like all software, we are constricted by the quality and abilities of its hardware. The clearest signals can be obtained using invasive (in the brain) BCIs such as Neuralink’s N1. However, surgically implanted devices have yet to build a positive and trustworthy repertoire with the general public. They’re expensive to implement and maintain, all while in the back of our heads we fear potential short circuiting, complete mind takeover, and overreach from hackers or the government.
Several companies have made advancements at ‘everyday’ BCIs like glasses from the aforementioned social media companies or headphones from Neurable. Though to reliably encompass the entire cortex of the brain, we require devices with more comprehensive electrode placements, in turn, making them unwieldy and unbecoming. As I place one on my head, I feel more like leading the X-Men than watching a baseball game.
Regardless, software is a complementary good to hardware so at the least, we must prepare to keep up alongside its advancements. The prevailing attempts at telepathic technology require inconvenient data intake methods, after-the-fact prediction, and/or penetration of the skull. Other researchers have found success at translating neuro-muscular activity corresponding to the jaw, though we believe this approach limits future integration with other neurosoftware, like outputting images, and potential users, such as those with a vocal tract or facial disorder.
Though it can be accommodated to intracortical needs, for commercial purposes, we’ve focused on developing our software for non-invasive BCIs. They are more affordable, integrate more easily with other technologies, and are far less experimental. However, being non-invasive, EEG signals can have significant interference due to bodily movements, and hair and skull thickness. These can present artifacts which skew the training data and in turn adversely affect the prediction. We focus on building advanced algorithms to filter out noise and isolate the desired electrical activity. However, these will need to be greatly improved before biosensors can be spread to places like a car headrest.
Intricacies of the Brain
We often describe the brain like a computer. It has different areas for various functions like output, input, and memory. However, in a computer these can be individually assessed and understood; the brain is more like if all the wires and boards were fused together. Along with electrical signals, the brain has a variety of chemical components. Assessing these chemicals provides more accurate models of prediction, however the pricing and wearability of such devices greatly reduce commerciality. Contrary to commerciality, the strongest forms of analysis will compare data gathered from a variety of intake methods.
Furthermore, these sections are not 'connected' per say, but instead work as one singular but dynamic network. This makes it almost impossible to focus on any specific area of activity without neglecting another. As well, one section of neuron may not correlate with the same thought over time. On top of these, folds of the brain are layered, so when activity is captured from a certain section, there's a need to distinguish the depth at which the signals were fired.
These problems are faced by intracortical and external BCI's alike, across all aspects of neurotechnology. As the industry is still in its scientific development, most advancements are restricted by their aim at research publication instead of marketability. However, acting as an independent software development team, Minds Applied has been able to apply the research uncovered but with an approach focused on the user.
The Importance of the User’s Experience
The overall development of the neurotech industry, such as improvements in hardware and software and more specifically artificial intelligence (AI), requires an increase in the amount of consumers using the technology. Companies like OpenBCI and Neurosity continue to improve the user’s physical experience when interacting with their BCI hardware, but when it comes to creating new software, each mind is like a fingerprint, so we must make predictions based on an individual user’s patterns in brain activity.
Training a cognitive neural network to make these predictions based on learned thoughts requires a substantial amount of data taken from a variety of users. Transfer learning (training a network on one user’s thoughts and then using it to predict another’s) has yet to yield optimistic results thus far, signaling that for the most accurate predictions, each user will need to partake in their own pre-calibration and such a calibration will need to record copious amounts of specific data.
We have been addressing the obstacles faced in the commercialization of telepathic technology by focusing on developing software to optimize user experience (UX), data collection, and pattern prediction. Garnering public interest and usage of neurotechnology requires a focus on UX and perception. Therefore, it is paramount that a user interface be intuitive, visually appealing, and possibly entertaining.
Every user is required to train on a baseline of words for minimal functionality, such as those related to directions, time, emotions, etc. However, for apt and segmented word training, they can select from categories for which they are interested in discussing, utilizing, or improving upon. Calibration consists of seeing the word followed by a visual or auditory stimulation like a cross or ping. The user has been told that while seeing or hearing this stimulus, they should think of the word previously shown either once or in repetition. After each applicable word thought has been recorded, they can either repeat the calibration to enhance prediction accuracy, select a larger or different set of words, or move on to the online classification and real time conversational aspect. To maintain precision with consistently changing neural activity, we also record information during the online portion.
However, predicting thoughts has the potential to go beyond just the projection of words onto a screen. Therefore, MindsApplied is creating more than just a chat application but complete Telepathy as a Service (TAAS): a favorable data intake and training program that can be applied to a suite of applications such as controlling a prosthetic or visualizing images.
Developing Trustworthy Technology
Beyond experience, there needs to be trust in the technology being used. We’ve mentioned the fears that come along with hardware, but a database of thoughts brings uncertainties with it just as well. We’ve worked in correspondence with researchers who are ensuring ethical neuroscientific practices. MindsApplied has addressed this by including descriptions of the safety and risks that come with storing any user’s sensitive information. As well, we are developing a blockchain type of storage which encrypts and gives users control over their own neural data.
Furthermore, thoughts can only be output from a model trained on words the user selected for themselves. Meaning sensitive information and topics aren’t translatable unless a user wants them to be. Though, for mental health purposes, it is possible to consensually record an all-encompassing database of language aimed at uncovering the secrets of the unconscious. However, in terms of communication, users need not fear any such ‘Freudian Blips’.
Distress over potential electrical shocks should be aimed at alternative technologies that provide electrostimulation in an attempt to connect neurons related to motor skills or memory. Commercial EEG headsets are built to take in the data of electrical activity already occurring in your brain.
Thought Prediction Using the Crypt Algorithm
Here at MindsApplied, we’ve developed a process which we’ve named the Crypt Algorithm for its use in translating the data provided by the brain and the brains’ own similarities to a stronghold of information. After data collection, its application is focused on analyzing and outputting thoughts by comparing:
Manners of thinking: When speaking, the different ways we think about the same words or phrases can evoke different degrees and connections of neural activity. Such as words related to emotion versus those for directional needs.
Cerebral locations: Through intake channels scattered over the brain, we’re distinguishing which regions provide the strongest and most replicable signals related to imagined and spoken words.
Neural Network (NN) Configurations: NN architectures, experimental designs, and brain data recorded across subjects is compared to output the most accurate model for prediction.
The Crypt Algorithm analyzes areas, thoughts, and ways of thinking that can provide replicable patterns of data corresponding to thought words. We are additionally working with linguists to formulate a type of Cognispeak: A manner or subject of speech that provides the best translatability such as through rhythm or by emphasizing phonemes. After all, we may not just be looking for a way to read the mind but, as well, a new way to think so that our minds can be read. Changes to the way we communicate may seem cumbersome, but the nuances would only be similar to how one talks to their dog or even more conveniently like acronyms and emojis for texting.
It is evident from the referenced research into imagined word translation that models attempting to make predictions from a larger set of words lose their precision, causing close to all technology to be impractical and inapplicable. Using AI, the Crypt Algorithm ‘thinks’ similarly to the compartmentalization of a human brain. Instead of choosing from an array containing as many words as possible, it improves prediction accuracy in Cognichat by reducing the set of probable words such as through questioned prompts and circumstantial context. At the least, simpler sentences with high accuracy ensure telekinetic reliability which is needed in steering or environmental interaction (such as with IoT). This we call, fittingly, Cognitrol.
The Crypt Algorithm can improve upon its own output by cross referencing multiples of trained models. To create longer and more fluid sentences, we entwine thoughts produced from Cognispeak with Natural Language Processing as is used in a chatbot.
Thinking Towards the Future
With the ability to converse solely from a BCI, we’re left with identifying the best place to give and respond to feedback. Designing our TAAS applications in the popular video game software, Unity, allows for seamless integration into content creation and Extended Reality. However, to first address a more prevalent and immediately accessible technology market, our conversations will initially be in the form of SMS to allow Cognichat users communication with anyone else via their mobile device.
The results we, and others in the fields of neuroscience, technology, and linguistics, have achieved thus far propel us to continue experimentation with optimism. They fuel our confidence and fill our imaginations with a fully-fledged future of telepathic technology. By pinpointing the roadblocks to such a future, we can deconstruct them piece by piece. The first chapter in the book of reading the mind has already been written. The last one will come directly from our brains.
Written by JM Wesierski and Dr. Justin Jarovi, edited by Emily Dinh and Lars Olsen, with original artwork by Lina Cortéz.
JM Wesierski has a background in business, technology, linguistics, and neuroscience. He’s developed a variety of cutting edge software focused on improving communication and mental health. Along with multiple publications, he details his work in the journal A Mind Applied. MindsApplied grew out of his collaboratory research on telepathic neurotechnology.
Dr. Justin Jarovi is a neuroscientist and brain-computer interface enthusiast. He received his Ph.D. from the University of Toronto, where he investigated the brain’s memory circuits and their interplay with decision-making. He is passionate about the AI and Machine-Learning applications of neurotechnology and wants to make the field accessible to all.
Emily Dinh is a data specialist who works in the medical device industry and is part of a computational cognitive neuroscience lab. She is currently obtaining her MS in Artificial Intelligence.
Lars Olsen is a regulatory medical writer. He works in the pharmaceutical industry writing submission-level documents, and has additional experience with medical devices and pharmacovigilance.
Lina Cortéz is an electronics engineer and neurotechnology enthusiast who is highly interested in the application of brain-computer interface in robotics control.