Artificial synapse can run a million times faster than the human brain

(ORDO NEWS) — Although we are not yet close to replicating the complexity and intricacy of the human brain with anything artificial, scientists have made progress in creating certain specialized devices, such as the newly developed programmable resistor.

Resistors can be used to create analog neural networks in artificial intelligence systems based on a structure that mimics the human brain.

This latest device can process information about a million times faster than the brain synapses that tie neurons together.

In particular, the artificial synapse is intended for use in analog deep learning, an approach to AI development that increases speed while reducing power consumption, which is important for affordability as well as demands on the planet’s natural natural resources. resources.

The key to the significant improvements in this latest resistor is the use of a specially selected and efficient inorganic material. The team behind the project says the increase in the learning rate of AI neural networks promises to be significant.

“Once you get an analog processor, you no longer train networks that everyone else is working on,” says the computer. scientist Murat Onen from the Massachusetts Institute of Technology (MIT).

“You will be training networks with unprecedented complexity that no one else can afford, and therefore vastly outperform them all. In other words, it’s not a faster car, it’s a spaceship.”

The basis of the considered inorganic material is phosphosilicate glass (PSG) - silicon dioxide with the addition of phosphorus. Used as a solid electrolyte in a resistor, its nanosized pores allow protons to travel through it at previously unseen speeds when pulses of 10 volts are applied to the setup.

What’s more, PSG can be fabricated using the same fabrication techniques that are used to fabricate silicon circuits. This should facilitate integration into existing manufacturing processes without significant cost increases.

In the brain, synapses strengthen or weaken to control the flow of signals and other information. Here, controlling the movement of protons to influence the electrical conductivity has the same effect. It is fast, reliable and can operate at room temperature, which also makes it more practical.

“The speed certainly surprised me,” Onen says.

“Normally, we would not apply such extreme fields to devices, so as not to turn them into ashes. But instead, the protons ended up moving at tremendous speeds through the stack of devices, in particular, a million times faster than before.

“And this movement does not damage anything, thanks to the small size and low mass of protons. It’s almost like teleportation.”

The huge potential here lies in much faster AI training using less energy. work in parallel to increase speed.

As for the next step, the researchers will need to use what they have learned about the development of this resistor and adapt it so that it can be produced on a larger scale. it’s not easy, but the team is confident it’s possible.

The end result will be seen in AI systems that take on tasks such as determining what is shown in images or processing natural voice commands.

Everything that artificial intelligence has to learn by analyzing huge amounts of data has the potential to be improved. This extends to areas such as self-driving cars and medical image analysis.

Further research will enable these resistors to be built into real systems and overcome the potential performance bottlenecks that currently limit the voltage that can be applied.

“The road ahead will be very challenging, but also very exciting,” says computer scientist and study author Jesús del Alamo of the Massachusetts Institute of Technology.

—

Online:

Contact us: [email protected]

Our Standards, Terms of Use: Standard Terms And Conditions.