Spaniard ‘copying’ brain to develop artificial intelligence in Korea

“The goal is not to copy nature as it is, but to learn from something that is already beautiful and try to develop new things. We know very little about how our brains work in general and even how we think, but it’s fascinating. With the vibration of the two membranes in our ears, we can find out where something we cannot see or touch is in three dimensions. instantly The young man from Barcelona told EL PERIÓDICO DE ESPAÑA from Daegu, South Korea, where he has worked for two years, in our heads and these are the result of much more advanced processing than existing computers.

For those like him who want to explore the secrets of the brain, the challenge is to understand exactly how procedures developed over millions of years of evolution are articulated and can be applied to improve current technology.

He explains that a variety of AI models will emerge over the next few years, from those who rely on a traditional computing system to those who aspire to replicate the model through so-called neuromorphic computation.

Miguel, one of the international experts at the Hermes Institute and a master’s degree in biomedicine from the University of Barcelona, ​​spent six years doing his doctorate in Japan to study the mechanisms and neural circuits of language learning. From there it jumped to Korea, where this study of the way auditory circuits interact with the motor brain could develop an artificial intelligence revealed in 2022 as one of the technological advances that will mark the next decade.

“As we see with the brain, the image it presents to us is not what our eyes see, so we listen with the brain. The brain ‘invents’ and ‘fills in’so to speak, most of what we perceive It is based on guesswork and previous experience and maintains a representation of the outside world that is constantly selectively ‘updated’ by sensory input. Therefore, there are phantom sound pathologies that are not perceived by the ear, but are in our head. This type of knowledge allows us to understand how sound travels through our neurons and therefore to know a little more about how that processing happens,” Miguel explains.

However, the human brain is not limited to ‘filling’, it has an extraordinary adaptability when it comes to sound. For example, if we take a sound at 60 decibels instead of 30, it sounds a little louder in our head, but as the young Barcelona man explained, it actually has to make a sound. 1000 times stronger. “The same thing happens in our own voice or when we exercise. You don’t hear the sounds you produce very loud because your brain drowns them out, it’s an involuntary survival mechanism,” he says.

For example, if these processes can be synthesized, the ability of autonomous vehicles to adapt to new situations and unexpected stimuli could be improved. This is just an example because they have already evolved, but previously a car could be prepared to diagnose a child, an adult, or a dog and not pass over it, but what if a boar or a scooter came along? AI had to be able to process this new element for which it was not ready, and judging relevance instantlyThere is no time for traditional processing.

“Until now, AI has worked with classifiers, so to speak. It gives more or less standard answers for A, B or C, and that’s what we thought was in the brain, but that’s not true, most things aren’t true. Don’t trust the stimulus and the response. 80% of brain activityfor example, it consists in maintaining an internal representation of what is going on around us”, he notes.

AI SUSTAINABILITY IS IN THE BRAIN

But even if it were possible to find a way to emulate all this via a conventional computer, that would be practically impossible and impractical. “Even supercomputers do step-by-step algorithmic operations that slow down processes and waste enormous amounts of energy. quantum computersOn the other hand, they don’t do algorithmic operations, and neither do our brains today, although they have other limitations,” explains the engineer.

In this way, he says, computing is currently done “bit by bit,” meaning that “even when we simulate neural circuits, we do it as if it were software on a CPU that follows algorithmic rules that are completely unrelated to the operation of the brain”. At this point, neuromorphic computinga field that aims to replicate the articulation of human thought in both chips and processes.

Currently, only in South Korea and Taiwan, where Miguel works, they can produce the most advanced chips in this field, and this has all the implications for the technological advancement race between America, Europe and Asia.

In addition, it is estimated that 3% of all electricity used in the world today is consumed in data centers and this rate may reach 13% by 2030. Optimizing computational models, as the brain has done in millions of years of evolution, will make artificial intelligence a much more sustainable technology.