Home Sciences They create a chip that can be configured like the human brain

They create a chip that can be configured like the human brain

40
0

Researchers at Purdue University in the United States have developed an electronic chip that updates its electronic circuits by itself. It is capable of learning continuously and serving as the basis for a computer that would operate as the biological brain.

Researchers at Purdue University in the United States have developed a chip that can be constantly reprogrammed and is capable of learning just like the human brain.

When the human brain learns something new, it adapts. But when AI learns something new, it tends to forget the learned information.

As companies increasingly use data to improve the way AI recognizes images, learns languages ​​and performs other complex tasks, new research, published in the journal Science, explains how a computer’s chips can be reconfigured. computer to continue learning about what has already been learned, just as the biological brain does.

“The brains of living things can learn continuously throughout their lives. Now we have created an artificial platform for machines to learn throughout their useful life, “explains Shriram Ramanathan, architect of the new development, in a statement.

AI embedded in hardware

Unlike the brain, which constantly forms new connections between neurons to enable learning, the circuitry of a computer chip does not change: a circuit that a machine has been using for years is no different from the circuit that was originally built in a factory. .

This limitation represents a problem in making AI more portable, such as when it is integrated into autonomous vehicles or robots in space, since at many times they would have to make decisions on their own in isolated environments.

Read:  A sick planet: 13 million deaths a year are due to preventable environmental causes

If AI could be built directly into hardware, rather than simply running in software, as AI often does, these machines could operate more efficiently, the researchers say.

In the new study, Ramanathan and his team built a new piece of hardware that can be reprogrammed on demand via electrical pulses.

Ramanathan believes that this adaptability would allow the device to take on all the functions needed to build a brain-inspired computer. “If we want to build a brain-inspired computer or machine, then consequently we want to have the ability to continually program, reprogram, and change the chip,” says Ramanathan.

new material

The chip’s hardware is a small rectangular device made of a material called perovskite nickelate, which is very sensitive to hydrogen.

Applying electrical pulses at different voltages allows the device to mix a concentration of hydrogen ions in a matter of nanoseconds, creating states that the researchers observed could be mapped to corresponding functions in the biological brain.

When the device has more hydrogen near its center, for example, it can act like a neuron, a single nerve cell. With less hydrogen in place, the device serves as a synapse, a connection between neurons, which is what the brain uses to store memory in complex neural circuits.

Read:  Pollution killed more people in 2019 than Covid in two years

Operational Neural Micronetwork

Through simulations of the experimental data, the researchers demonstrated that the internal physics of this device create a dynamic structure for an artificial neural network that is able to more efficiently recognize the patterns and digits of the electrocardiogram, compared to the static network. .

This neural network built into the chip’s hardware uses “reservoir computing,” which spontaneously replicates how different parts of the brain communicate and transfer information.

The researchers also showed that as new problems arise, this dynamic network can “pick and choose” which circuits are best suited to address those problems.

Adaptable to semiconductors

Since the team was able to build the device using standard semiconductor-compatible fabrication techniques and operate the device at room temperature, Ramanathan believes this technique can be easily adopted by the semiconductor industry.

“We show that this device is very robust,” explains Michael Park, another of the researchers.

“After programming the device for a million cycles, the reconfiguration of all functions is remarkably reproducible,” he adds.

The researchers are working to demonstrate these concepts on large-scale test chips, which would be used to build a brain-inspired computer.

Reference

Reconfigurable perovskite nickelate electronics for artificial intelligence. Hai Tian Zhang et al. Science, 3 Feb 2022, Vol 375, Issue 6580, p. 533-539. DOI: 10.1126/science.abj7943

Previous articleKanye West documentary: The rapper’s demands unsettle the director just before its premiere
Next articleWhat are mosquitoes for and what would happen if they disappeared from the Earth?