Chips that detect odors or how technology tries to mimic brain function

Chips that detect odors or how technology tries to mimic brain function
Neuromorphic engineering seeks to replicate the functioning of the human nervous system with the intention of solving complex problems in real time and with greater energy efficiency.

Disciplines like neurotechnology and neuroscience yearn to replicate the functioning of the human brain. To develop a technology capable of calling into question whether the nervous system is artificial. This challenge is almost endless without the collaboration of all kinds of areas. As happens within this vital organ, complexity reigns in each of its corners. Billions of neurons, electrical impulses, impulse coding, information processing ... Neuromorphic engineering, started in the late 1980s by Carver Mead and which combines sciences such as mathematics, computer science and physics, wants to contribute its bit. sand; especially in solving complex problems in real time and improving energy efficiency.

As Bernabé Linares-Barranco, a research professor at CSIC, explains, the brain is designed in such a way that it minimizes the amount of impulses necessary to undertake a function. It has very optimized how it encodes the information and the amount of signals it requires to send it. Every nerve fiber and each neuron has its individual metabolism and consumption. Even so, under normal conditions, the brain consumes between 10 and 20 watts, a figure still far from neuromorphic computing, "he says.

At a similar crossroads is information processing. Teresa Serrano, researcher at the CSIC, points out that a hardware is needed to process all changes in real time. The most basic example for her is that of an image. A camera captures a specific moment, static; but the brain composes on different images because it continuously decodes a changing reality through the retina. “In the case of a conventional computer, data processing is done in sequences, like the camera. The neuromorphic seeks that real time. Take a leap in interpreting impulses. Whether in robotics or any other discipline, you need a chip with these capabilities to go further than with classical computing, "he says.
Intel is one of the companies that has most strongly opted for this technology. At the end of 2017 he released Loihi, a neuromorphic chip capable of learning thanks to the 130,000 artificial neurons that make it up. By being able to communicate with each other, it creates a network similar to that woven by the human brain itself. The latest advance has come from researchers at Cornell University and the Intel Pohoiki Springs platform, which integrates 768 chips and some 100 million neurons. Last week they published a study demonstrating the ability of this hardware to detect dangerous chemicals through smell.

The multinational assures that it quickly learned the neural representation of each of the smells and recognized them all, even in a context of considerable blocking. Linares-Barranco comments that it is possible because the system is based on eight organic sensors –as it happens with the three primary colors that make up the entire palette–, where each one is sensitive to a basic odor unit. When a specific smell arrives, express in code it as impulses and this train of impulses is deal with by the Loihi chip to recognize the pattern of the combination .
A wide range of applications

According to Intel, this innovation would lead doctors to diagnose diseases, airport security personnel to detect weapons and explosives, police and border control to more easily locate and seize narcotic substances, and even serve to create smoke and monoxide detectors. carbon filters for households. 
The things of the olfactory structure share a great similarity with the properties of other brain regions, such as the hippocampus. That is to say, it is feasible that with the neuromorphic one, thanks to this advance, we can attend to general problems of associative memory of high dimension , predicts Mike Davies, director of the neuromorphic computing laboratory of Intel.

Electronic noses do not represent precisely unknown terrain for this engineering. In the perfume industry they are common to create new fragrances - even to remove certain smells. NASA invented its own in 2004 to detect when ammonia levels reached dangerous limits for astronauts or to predict a fire. The problem is that it is difficult to replicate the functioning of the nervous system. It has not made the leap to intelligence that is presupposed to the human being. At the moment, it only identifies the substances that it has preset, such as Loihi with the dangerous ones, but is unable to interpret the information it collects.

The future looks bright for the neuromorphic. In three decades he is getting closer to his long-awaited dream. Still a long way to go, but with promising results. If you follow the same evolution with respect to low consumption, both in the computational and hardware fields, the brain is not so far away. Linares-Barranco considers that it will be feasible to manufacture fully autonomous robots, without connection to the electricity grid and with greater speed in decision-making. There is a lot of possibility for those who are communicating, like those who learn tasks. For 100% autonomous driving, for example, it will be a great ally because it is a technology that needs a reflection time similar to that of the human being , concludes Serrano.

Next Post »