New computer memory tech could power the AI of the future

A research team, led by the University of Cambridge, has developed a novel computer memory design, which promises to significantly improve performance while reducing the energy demands of internet and communications technologies.

As per the university, AI, algorithms, internet usage, and other data-driven technologies are estimated to require over 30% of our global electricity consumption within the next decade.

“To a large extent, this explosion in energy demands is due to shortcomings of current computer memory technologies,” said first author Dr Markus Hellenbrand, from Cambridge’s Department of Materials Science and Metallurgy. “In conventional computing, there’s memory on one side and processing on the other, and data is shuffled back between the two, which takes both energy and time.”

The researchers experimented with a new type of technology known as resistive switching memory. Unlike conventional memory devices that can encode data in two states (one or zero), this novel type of memory can enable a continuous range of states.

This is done by applying an electrical current on specific materials, causing the electrical resistance to increase or decrease. The various changes in electrical resistance create different possible states to store data.

“A typical USB stick based on continuous range would be able to hold between ten and 100 times more information, for example,” explained Hellenbrand.

The team developed a prototype device based on hafnium oxide, which had so far proven to be challenging for resistive switching memory applications. That’s because the material has no structure at the atomic level. Hellenbrand and his co-scientists, however, found a solution: throwing barium into the mix.

“These materials can work like a synapse in the brain.

When barium was added, it formed highly-structured barium “bridges” between thick films of hafnium oxide. At the point where these bridges meet the device contacts, an energy barrier was created, allowing the electrons to cross. The energy barrier can be raised or lowered, which changes the resistance of the hafnium oxide composite, and in turn allows multiple states to exist in the material.

“What’s really exciting about these materials is they can work like a synapse in the brain: they can store and process information in the same place, like our brains can,” Hellebrand said.

The researchers believe that this could lead to the development of computer memory devices with far greater density and performance but lower energy consumption, making the technology especially promising in the field of AI and machine learning.

A patent of the technology has been filed by Cambridge Enterprise, the university’s commercialisation arm, and the scientists are now working with the industry to run larger feasibility studies. They claim that integrating hafnium oxide into existing manufacturing processes won’t prove challenging, as the material is already being used in semiconductor production.