Despite the fact that we often talk aboutartificial intelligence, in the structure of the human and machine brain, there is one key difference - this is their structure. Neural networks built on neurons and on chips and transistors are not the same thing. However, a fairly interesting discovery was made by scientists at the University of California James Gimzevsky and Adam Stig. They, as the editorial office of phys.org reports, have created a chip that in its structure is very similar to the human brain. And this may be the first step towards creating truly thinking machines.
How to create an artificial brain
In collaboration with his Japanese counterpartsThe National Institute of Materials Science, the team created an experimental device that demonstrated characteristics similar to certain types of “behavior” of the brain - learning, remembering, forgetting information, staying awake and even sleeping.
The way the device is constantly evolving is verysimilar to how the human brain behaves. It can “come up” with various types of behavior patterns that are not repeated. - say scientists in a press release on their work.
The device that the researchers createdmade of a “tangle” of silver nanowires with an average diameter of only 360 nanometers (approx. ed. nanometer is one billionth of a meter). The nanowires were coated with an insulating polymer about 1 nanometer thick. In general, the device itself turned out to be 10 square millimeters in size.
Randomly Collected on Siliconhighly organized structures that are surprisingly similar to those that form the neocortex, the part of the brain involved in higher functions such as language, perception and cognition, have formed a plate of nanowire. One of the features that distinguish a nanowire network from conventional electronic circuits is that the electrons passing through them cause a change in the physical configuration of the network. During the study, an electric current caused silver atoms to migrate from the polymer coating and form new compounds. The system had about 10 million such connections, which are similar to synapses, thanks to which brain cells connect and communicate.
Researchers attached two electrodes tothe resulting sample to test how the neural network works. After the current flowed through the network, the bonds between the nanowires were maintained for one minute, which was reminiscent of the process of learning and memorization taking place in the brain. In other cases, the connection abruptly terminated after the end of the charge, imitating the “forgetting process”.
This is interesting: Human mini-brains created in the laboratory can feel pain.
In other experiments, the research teamfound that with less energy consumption, the device shows a behavior that is similar to what neuroscientists see when they use a functional MRI scan to obtain images of the brain of a sleeping person. With more power, the nanowire network behaved in the same way as an awake brain. And what do you think, is this really a step towards the creation of smart cars? Write about it in our chat in Telegram.
Our approach can be useful for creating new types of equipment that are both energy efficient and capable of handling complex data sets.
Due to their resemblance to the inner workings of the brainfuture devices based on nanowire technology can be extremely energy efficient. The human brain operates at a power roughly equivalent to that of a 20-watt incandescent lamp. In contrast, computer servers that perform time-consuming tasks use hundreds of times more electricity.