The first living computer already exists: it works with neurons from living mice, incorporated into the computation, which process information better than the most sophisticated electronic chips. They learn like biological brains and will generate thinking robots in the future.
The average mouse, which weighs just 40 grams, has 71 million neurons stored in a brain that weighs just 0.4 grams: enough to master complex thinking and represent abstract information at the neural level, it was recently discovered.
Perhaps in the future this tiny privileged brain will go even further: a group of scientists from the University of Illinois at Urbana-Champaign, in the United States, extracted 80,000 neurons from the brain of a mouse and built with them what we could call the first living computer.
It is a very simple computer, as it only recognizes patterns of light and electricity, but it has a completely new component: the processing of this information from the patterns of light and electricity is not done by a computer chip, but directly by an extracted neuron or nerve cell. of the privileged brain of a mouse.
The results of this research, led by André Dou, were presented this month at the annual meeting of the American Physical Society in Las Vegas, Nevada, reports NewScientist magazine.
neurons and chips
Neurons and computer chips (miniaturized electronic circuits) basically work the same way, that is, both process information.
Based on this premise, a form of computation was developed, called neuromorphicthat mimics the functioning of neurons: it aims at an efficiency equivalent to that of a biological brain with less energy consumption.
Neuromorphic computing is still at an early stage of research, with modest attempts to build computer architectures modeled after the patterns of a biological brain.
This computation aims to improve the efficiency of the artificial neural networks on which Artificial Intelligence (AI) is built, but it could be on the verge of a qualitative leap if Dou’s proposal were technically possible. and the thunder box which opened the new investigation.
AI and organoids
Brain-inspired AI algorithms run on conventional computers, but what the new research proposes, where its originality lies, is to use real living neurons as an essential part of the configuration of a modern computer that mimics the biological brain.
The first thing this team of scientists did was cultivate around 80,000 neurons extracted from the brain of a mouse in a laboratory dish, with the help of a network of electrodes that send electrical signals to stimulate tissue culture.
This tissue culture technique, long considered a blend of science and art, has in the past produced neurons with which to organoidsa kind of tiny preserved brains that have been used for both medical research and the study of intelligence.
The most brilliant thing about the new research is that she connected these living neurons to a fiber optic cable (which transmits information via photons, or particles of light) as well as to the network of electrodes that contributed to their culture, so that the nerve cells could experience light and electricity outside the brain.
With all these elements they created an unusual computer with an architecture that we are not used to: a box the size of a hand, which works in an incubator to keep the neurons alive and as operational as any other component of the computer’s structure.
After several learning sessions, the researchers were able to make the living neurons correctly interpret the patterns of light and electricity created in that box (a kind of miniature frankenstein), as if they were inside a biological brain, that is, they processed information exactly as they would in their natural environment.
This neural reaction, obtained from the stimuli provided, both by the optical fiber (which does not transmit electricity) and by the electrodes of the experiment, was registered by an additional computer chip and the miracle happened: True neural (biological) computing could be operationally incorporated into a mechanical processor (chip) and, moreover, process information in record time for ordinary computing.
This interface between a biological neuron and an electronic circuit has worked for the first time with encouraging results, although it is a first step in the long-term goal of developing living computers and robots with cognition-like capabilities, such as learning or attention, achieved by integrated living neurons. to the computational process.
NewScientist highlights the interruption This proposal represents: although living neurons have been used in the past by robots to process information about their environment, a computer has always been needed to act as a bridge between biological and robotic systems. The new research achieves this connection without intermediariesdirectly between the neurons and the computer chip.
This leap amplifies the technological impact that neuromorphic computing can have, as the neurons embedded in a robot will be able to efficiently and quickly process much more information from the environment at once, and even ensure cognition should any smaller part of the robot fail. .
The robotics that can be developed from this technology will far surpass existing robots, but this will require it to continue to mature in size and complexity, which will still take time. After this investigation, however, the question that remains is no longer whether thinking robots will exist, but when.
Abstract: A11.00001: In vitro live neurons in a reservoir computing framework. Zhi Dou et al. APS March Meeting 2023. Session A11: Physics of Neural Systems I.