Our brains are incredibly complex organs, so much so that they are far from being fully understood by modern science.
One thing we do know is that the human brain is an incredibly efficient computing device — it operates at a much slower clock speed than modern microprocessors while still making billions of calculations per second.
Now, MIT computer scientists are trying to copy the computing efficiency of the human brain using neural networks made of superconducting nanowires.
RELATED: KEY TAKEAWAYS FROM ELON MUSK'S NEURALINK PRESENTATION: SOLVING BRAIN DISEASES AND MITIGATING AI THREAT
The "computer" in our heads
As per MIT Tech Review, our brains are powered by little more energy than "a bowl of porridge," while the world's most powerful supercomputers "use more power than large towns."
Despite this, the brain's incredible computing efficiency allows us to walk, converse, think, etc.
For a while now, computer scientists have been looking to build artificial neurons that are connected in brain-like networks. This, in principle, would lead to significantly more energy efficiency.
Superconducting neural networks
Emily Toomey and colleagues at MIT have designed a superconducting neuron made from nanowires. They claim it is a breakthrough — as it shows many similar behaviors to neural networks in the human brain.
In theory, the researcher's device matches the energy efficiency of the brain and could be "a building block for a new generation of superconducting neural networks that will be vastly more efficient than conventional computing machinery," the MIT press statement says.
Toomey and her colleagues say that superconducting nanowires have a nonlinear property that enables them to act like real biological neurons.
The nanowire’s superconductivity breaks down when the current flowing through it exceeds a threshold value — this mimics the way biological neurons do not fire unless the input signal exceeds a threshold level.
Moreover, when the nanowire's superconductivity breaks down, the resistance suddenly increases, creating a voltage pulse that significantly resembles the action potential, or electrical impulses, in a neuron — action potential effectively creates brain signals.
Connecting the nanowire to other wires and creating a network makes the simulation even more realistic.
Matching the human brain
The group of computer scientists claims that their superconducting neural network, in theory, can match the human brain in managing approximately 1014 synaptic operations per second per watt.
“The nanowire neuron can be a highly competitive technology from a power and speed perspective,” the group said in MIT's press release.
“The result would be a large-scale neuromorphic processor which could be trained as a spiking neural network to perform tasks like pattern recognition or used to simulate the spiking dynamics of a large, biologically-realistic network,” they say.
As always, patience is key: nanowire neural network supercomputers are nowhere near becoming a reality, but the findings show a lot of promise.
For more on superconducting neurons, have a look at the MIT team's research paper here.