Researchers are one step closer to making “brainlike” computers even more powerful.
A team at the National Institute of Standards and Technology (NIST) has designed a sophisticated superconductor circuit that could allow artificial neural systems to operate “100,000 times faster than the human brain.”
Each of about 86 billion cells, or neurons, that make a human brain, interact through thousands of connections known as synapses. They use short electrical pulses called spikes to communicate with each other, forming a basis of cognition.
This is something that scientists have tried to emulate for decades, building artificial neural networks that mimic the brain by having electronic components act as neurons. The use of conventional digital electronics, however, limits the potential of these systems in terms of complexity and speed.
“As the chips become larger and more complex, the signals between their individual components become backed up like cars on a gridlocked highway and reduce computation to a crawl,” NIST said.
To overcome this challenge, researchers have designed a circuit board that uses photons, a tiny light that is the smallest possible optical signal, to represent spikes in neuron interaction. The circuit described in a paper published by Nature Electronics is the first one “that behaves much like a biological synapse yet uses just single photons to transmit and receive signals,” according to NIST.
It can be “especially energy-efficient,” as it uses a single-photon detector and “a Josephson junction” to achieve the result. The Josephson junction is a “sandwich” of superconducting materials separated by a thin insulating film that can be manipulated through an electric current to produce small voltage pulses, or fluxons.
As these electric pulses accumulate, they merge into their own current, forming a “superconducting loop.”
“This behavior is similar to that of biological synapses. The stored current serves as a form of short-term memory, as it provides a record of how many times the neuron produced a spike in the near past,” NIST said.
"Consequential" problem solver
The duration of this “memory” can last from hundreds of nanoseconds to milliseconds and beyond, depending on the time it takes for the electric current to decay in the superconducting loops.
“This means the hardware could be matched to problems occurring at many different time scales – from high-speed industrial control systems to more leisurely conversations with humans,” NIST said, describing the findings as an “important milestone.”
Neuron interactions, or synapses, are a crucial computational component of the brain, and the same is true for artificial systems that mimic it. Combining synapses with on-chip sources of light will solve “large, consequential” problems in the future, NIST said.
More from Cybernews:
Subscribe to our newsletter