THE INNOVATIVE SPIRAL

If we think about it, disruptive leaps in Artificial Intelligence have always had the same origin: the study of how our brain behaves at a physical and chemical level, and the modeling of this in digital systems. What a few years ago was very hard to implement (for example, computer vision models to identify objects in images, or natural language processing capable of extracting the intrinsic semantics from contexts) is currently within anyone’s reach – thanks, again, to the research by other Nobel Prize laureates.

When they investigated and identified how our visual cortex behaves, that was replicated in the form of convolutional networks and embeddings. Until then, all approaches to these problems were inefficient and unproductive. But this process is an innovative spiral. Thanks to thesebreakthroughs in Artificial Intelligence, advances in pattern-oriented modeling in complex systems allow neurocomputing research to be accelerated in such a way that Deep Learning models are making neurocomputing applications possible that had been unimaginable up until now.

MOVING THINGS WITHOUT MOVING

One of the areas that will generate the most business in the coming years is “Brain Computer Interface (BCI);” that is, interaction with digital systems (computers, machines, robots, drones, devices) through our thoughts. We already have an endless number of non-invasive devices (headbands similar to headphones) that are capable of recording the brain’s electromagnetic pulses (signals based on electroencephalography or EEG). From these signals (and with Artificial Intelligence) we are able to move a robot, manage a process, define gross motor movement in exoskeletons, and forget about keyboards, joysticks, and mice in simulation/virtual environments.

It is clear that immersive technologies like Virtual Reality have the potential to improve the remote operation of robots. Likewise, managing equipment and systems in the traditional way (either by keyboard or commands) is inefficient, since it distracts the technician from the problem to be solved, especially in critical environments. BCI technology, along with Virtual Reality, can take the operator to a virtual workspace where they are able to command the robots intuitively and directly (brain to machine) so that said learning can later be replicated in real environments.

EEG HEADSETS AND BCI SYSTEMS: IT’S THE THOUGHT THAT COUNTS

Recently, Ibermática in collaboration with the Universidad de Deusto and Virtualware has developed a VR system which allows for the training of industrial technicians together with robots in a VR environment through the use of EEG headsets. This, in turn, makes it possible for operators to have mental control of production equipment without the need for external control; only thoughts are needed. Meanwhile, operators can visually monitor the rest of the parts of the process. Clearly, in this context, major manufacturers like Meta, Intel, and Microsoft are already adding BCI systems to their video game platforms, in a revolution similar to the one pioneered in the past by the famous “Kinect.”

But the use of neurocomputing headsets does not only stop at controlling systems and virtual environments. Also (and this is a very mature and widespread application), using EEG signals from the brain, we can determine whether a user is liking what they’re seeing or not, if they’re comfortable in a virtual environment, if they’re stressed, apathetic, etc. We can even predict, based on different models, what the user’s reaction to specific environments will be.

Some of the most cutting-edge research is even making it possible to “translate” thoughts into images so that we can “see” what a user is thinking in real time. Clearly, this feature is worth its weight in gold in the metaverse. Knowing instantly if a user is attracted to a specific product, a specific logo, a specific environment, and/or a specific atmosphere, is the basis of a not-so-new advertising technique known as “neuromarketing.”

As an example, we recently published a study in which we can – through brain signals – know whether consumers really like new nutritional products in which sugar is replaced by other healthier components, simply by measuring their brain responses while they are trying them. The mind never lies. Customer satisfaction surveys do.

CUSTOM SYSTEMS TRAINING

There is still a problem to be solved, however: making models universal. Just like in the early days of speech-to-text transcription, when it was necessary to dedicate 20 minutes to training the Automatic Speech Recognition or ASR software in order for the machine to learn how we spoke (something totally outdated today with the current convolutional neural networks used by Siri, Alexa, and the transcription feature of our mobile phones), the same thing is happening with BCI systems. These systems, in order to be accurate, need a short bit of customized training for each user (15 minutes of adjustments). Like people’s voices, people’s brain patterns are unique and vary between individuals, with the difference that neural channels have a much more complex spectrum of frequencies, and this is a system that works in parallel on different channels (in different areas of the brain) at the same time and in an interconnected way.

There is an international race between companies to find a solution to this problem, given that the first to patent a working idea will have the exclusive rights over universal EEG modeling for virtual environments like the metaverse. The Ibermática Institute for Innovation (i3B) is working on two very promising lines of action. The first one, in which most of the current scientific and business community is involved, is to use highly advanced Artificial Intelligence techniques to come up with a mathematical model capable of modeling this complex network of signals, regardless of the user. That is, a model that allows for the generation of “embeddings” as if they were basic Lego pieces which, when put together, are able to reconstruct any EEG signal accurately, ignoring the particularities but maintaining the main idea of said signals (intentionality, thought, mobility, degree of attention, feelings, taste, etc..).

But there is another, much more tempting solution. It has been known for years now that neuronal behavior is not based only on classic mechanical behavior, but also on molecular behavior, depending on how the cell membranes of the neurons are affected by the sodium–potassium pump. There is a whole new discipline in Artificial Intelligence that is very promising in this regard, called neuromorphic computing, and the famous Spiking Neural Networks (SNNs).

But, if behavior is not purely classic, but based on the interaction of different particles, it seems sensible to think that a system that is able to accurately model the molecular behavior of said interactions, and extrapolate said behavior to the rest of the neurons in the different brain groups, would give us a solution not only to the problem of universality, but also to the challenge of building a very precise digital twin of the brain.

APPLICATIONS AND POTENTIAL

The applications of this probable solution would be multiple: we could simulate brain pathologies, like stroke and Alzheimer’s, and the evolution of brain tumors. And, of course, we could also simulate the behavior of our users in virtual environments like the metaverse. The technology that allows for this type of simulation at the quantum level in a natural, effective, and direct way does exist: quantum computing.

Until now it was not understood why the brain and its natural structures were able to process so much information in such a short time, despite having speeds much lower than that of current computer processors. The brain is completely parallel in nature, with a large number of processing elements (around 1,012 neurons) that are highly interconnected (up to 10,000 connections per neuron). Nevertheless, its processing time is slow, in the milliseconds. In addition, there are unicellular organisms (like the Protozoan Paramecium) that swim, find food, interact, learn, and remember without the need for any synapses.  So, then, how is it possible that reasoning arises in such a short period of time? It has recently been discovered that, in addition to synapses and neuronal structures, there are structures called “microtubes,” which, apparently, regulate the cellular behavior of Protozoa, and even that of neuronal synapses. That is, each neuron contains a “microtubular” structure composed of “self-assembled” polymers based on the tubulin protein, forming cylinders with hexagonal lattices in which the paired filaments intersect in accordance with the Fibonacci sequence, in helical symmetry.

These structures turn into “molecular automata,” allowing signal oscillations in a range of 1,015 operations per second for each neuron, similar to the entire brain synapse – as well as a 15-qubit quantum computer.

Ibermática currently has access to quantum computers with more than 15 qubits, so they’re in a position to take on this strategy. In fact, Ibermática is fortunate to be collaborating with Enrique Solano, a pioneer in quantum computing, with whom the entity is beginning to connect EEG headsets to quantum computers with the aim of finding universal models and simulating brain behavior in a quantum environment at the molecular simulation level, and doing so in a direct way (it is always easier to simulate a quantum oscillation with a quantum particle than with a classical particle), also taking advantage of superposition (the volume of information to process) and quantum collapse (applying subjectivity to an environment with massive amounts of information). Fascinating.

A REALISTIC/FUTURISTIC PRESENT

In conclusion, there are cutting-edge and very powerful projects in neurocomputing that I recommend following, such as the European Human Brain Project, the Blue Project, and the European Brain Research Area. It is these initiatives that allow for advances in neuroscience that help us to understand the human brain, allowing these discoveries to be extrapolated towards new Artificial Intelligence models, which, in turn, in a constant spiral of innovation, are making possible the construction of, for the first time in history, more effective virtual patient systems (a clinical metaverse), as well as a deeper understanding of brain models.

This trio (neurocomputing, AI , and quantum science), together with new technologies, such as quantum computing, is accelerating clinical personalization processes and new computational interfaces that will be key in our daily lives in the coming years. The emulation of the brain in 3D simulation systems (both overall and at the molecular level), as well as simulation at the neuronal level, and iterations of it, are making it possible to diagnose and improve the treatment of certain pathologies, such as strokes, tumors, ALS, and Alzheimer’s, allowing us to evaluate and predict the evolution of said diseases at a personal level.

We might think that this is a futuristic technology, yet the current state of the art ranges from basic research to the sale of hardware systems with AI processes already built in. There is a real chance to lead a new era of advanced neurocomputing, which, in turn, will give rise to new Artificial Intelligence models based on neurocomputational study of the brain. Thus, we will have an opportunity to generate a new industry around quantum BCI systems. The global market for BCI systems was valued at $1.5 billion in 2021, and is expected to increase at a compound annual growth rate of 17.16% over the next several years. This is a unique national opportunity to lead a new industry that is promising worldwide, while not missing the chance to promote this field’s strong presence in industry and on the market.

For this, different multidisciplinary talents must come together, such as data scientists, scientists in engineering and quantum computing, and experts in neuroscience. Moreover, governments, companies, universities, and technology centers should foster collaborative environments for projects and joint work at the national level. We are in time to lead this new digital revolution.