The electronic brain has learned to “forget” information

The research team has created an electronic simulation of neurons in the human brain, which can forget information if it is not in demand for a long time.

The brain is a perfect computing machine, so it’s not surprising that engineers tend to imitate its systems. The new study allowed us to take one more step towards full-fledged imitation. Scientists have created a device that "forgets" information in the same way as the human brain.

The new system is called “memristor” (memristor, from a combination of English. “Memory” and “resistor”) of the second-order, popmech.ru. It imitates the structure of the synapses of the human brain, and in the same way remembers and then “forgets” information if it is not accessed for a long time.

What is the practical use of this device? At present, no. However, it is precisely such systems that help to develop a new type of neurocomputer, the basis for artificial intelligence. Say what you like, but the developers voluntarily involuntarily have to imitate brain activity.

In the so-called In an “analog” neurocomputer, individual electronic components (such as a memristor) play the role of individual neurons and synapses. Engineers assure that this will not only speed up the calculations, but also reduce energy consumption. Today it is a hypothetical concept embodied only in the format of individual components. That's because for starters, scientists need to figure out how to teach electronics to mimic synaptic plasticity. This property, due to which the active synapses of the brain increase over time, and inactive synapses become weaker. Doctors believe that it is thanks to him that some memories are preserved for decades, while others disappear from memory.

For the new prototype, engineers used a material called hafnium oxide. It is already used to create microchips by companies like Intel. In fact, the “forgetfulness” of neurons is simulated by a defect at the interface between the silicon and hafnium parts of the processor. Usually it serves as a source of many problems, but in this case, the researchers turned the defect in favor - the memristor conductivity weakens over time and unclaimed information is erased.

Of course, the project will have a long way to go. It is possible that a perfect imitation of the human brain can lose in efficiency to the "classical" computing systems - but this can only be understood after testing in practice.