There is a fundamental difference between the brain and artificial intelligence in the way they learn things. If the brain tends to adapt by considering its previous knowledge, artificial intelligence is inclined to forget the information it has already learned.
Recently, scientists seeking to fill this gap in AI have designed a new microchip that can better mimic the ability of the human brain to learn continuously.
A microchip mimicking the human brain
Shriram Ramanathan, a professor at Purdue University’s School of Materials Engineering, specializes in studying how materials could mimic the brain to improve computing.
“The brain of living beings can continuously learn throughout their lives. We have now created an artificial platform for machines to learn throughout their lives. »
The circuits of a computer chip remain staticwhich is the opposite of brain. The latter constantly forms new connections between neurons allowing learning.
The challenge is to make more portable AI. Indeed, machines would be able to operate more efficiently if AI could be integrated directly in the material instead of run on software.
With the aim of build a computer inspired by the brain, Ramanathan and his team built a prototype that could be reprogrammed on demand by electrical impulses. This adaptability is necessary for the device to create a machine endowed with the didactic capacity of the brain.
“If we want to build a brain-inspired computer or machine, we need to have the ability to program, reprogram, and change the chip all the time. »
Like the brain, the chip stores memory
The chip consists of perovskite nickelates very sensitive to hydrogen. By applying electrical pulses at different voltages, the device can change the concentration of hydrogen ions within nanoseconds. According to the researchers, the states resulting from this process are comparable to the corresponding functions of the brain.
When the chip has more hydrogen near its centershe plays the role of a neuron, i.e. transmit information to other nerve cells. If, on the contrary, it has less on this location, it acts as a synapse, a connection between neurons, used by the brain to store memory in complex neural circuits.