New light-emitting artificial neurons could run AI systems more reliably

Over the past decades, computer scientists have developed increasingly advanced artificial intelligence (AI) systems that perform well on various tasks, including the analysis or generation of images, videos, audio recordings and texts. These systems power various highly performing software, including automated transcription apps, large language model (LLM)-powered conversational agents like ChatGPT, and various other platforms. Despite their remarkable performance, most AI systems are computationally intensive and can rapidly drain the energy of existing devices. Electronics engineers have thus been trying to develop brain-inspired hardware systems that are better suited for running these systems.

Recipe for neuromorphic processing systems?

Memristors achieve stable resistance values tied to fundamental constants of nature

Researchers at Forschungszentrum Jülich, together with international collaborators, have demonstrated for the first time that memristors—novel nanoscale switching devices—can provide stable resistance values directly linked to fundamental constants of nature. This paves the way for electrical units such as electrical resistance to be traced back far more simply and directly than it has been possible to date. By contrast, conventional, quantum-based measurement technology is so demanding that it can only be carried out in a few specialized laboratories worldwide. The paper is published in the journal Nature Nanotechnology. Since 2019, all base units of the International System of Units (SI)—including the meter, second, and kilogram—have been based on fundamental natural constants. For example, the kilogram, which was once based on the "prototype kilogram," is now linked to Planck's constant h. A meter is defined by the speed of light, and a second by the oscillation of the cesium atom.

Atom-thin crystals provide new way to power the future of computer memory

Picture the smartphone in your pocket, the data centers powering artificial intelligence, or the wearable health monitors that track your heartbeat. All of them rely on energy-hungry memory chips to store and process information. As demand for computing resources continues to soar, so does the need for memory devices that are smaller, faster, and far more efficient. A new study by Auburn physicists has taken an important step toward meeting this challenge. The study, "Electrode-Assisted Switching in Memristors Based on Single-Crystal Transition Metal Dichalcogenides," published in ACS Applied Materials & Interfaces, shows how memristors—ultra-thin memory devices that "remember" past electrical signals—switch their state with the help of electrodes and subtle atomic changes inside the material.

Iontronic Circuits: Building Intelligence in Brine

Imagine a future in which computers process information not with streams of electrons but with hydrated ions flowing through salt water, a system that mimics how the brain itself computes. This emerging field—known as iontronics, a portmanteau of ions and electronics—is rapidly growing as researchers design neuromorphic computing devices, inspired by animal nervous systems and powered by electrolyte solutions at the nanoscale [1–3]. Since Leon Chua introduced the memory resistor or “memristor” in the 1970s [4], these components have been considered revolutionary building blocks for neuromorphic computing. A memristor’s electrical resistance depends on the current that flowed through it before it was powered off, offering a way to store information. Unlike solid-state memristors, fluidic ones still face challenges in terms of scalability and integration with a circuit. A study by Patricio Ramirez of the University of Valencia in Spain and colleagues recently presented a preliminary step toward a scalable version [5]. The researchers demonstrated that nanofluidic memristors, based on membranes containing conical nanopores, can be organized into functional circuits. By arranging membranes in series and in parallel configurations, they showed how synaptic weight—that is, conductance—can be manipulated to represent brain-like properties that could be used in future logic and fluidic neuromorphic systems [6].

AI chips could get a sense of time with memristor that can be tuned

Artificial neural networks may soon be able to process time-dependent information, such as audio and video data, more efficiently. The first memristor with a "relaxation time" that can be tuned is reported today in Nature Electronics, in a study led by the University of Michigan. Memristors, electrical components that store information in their electrical resistance, could reduce AI's energy needs by about a factor of 90 compared to today's graphical processing units. Already, AI is projected to account for about half a percent of the world's total electricity consumption in 2027, and that has the potential to balloon as more companies sell and use AI tools. "Right now, there's a lot of interest in AI, but to process bigger and more interesting data, the approach is to increase the network size. That's not very efficient," said Wei Lu, the James R. Mellor Professor of Engineering at U-M and co-corresponding author of the study with John Heron, U-M associate professor of materials science and engineering.

Researchers develop thin film as resistance-switching material for next-generation memristive devices

Memristive devices are capable of retaining their internal resistance, thus offering superior performance compared to conventional devices that use integrated circuits. Several materials have been explored for the manufacture of these devices. In recent years, transition metal oxides have gradually become widely popular for this purpose. Due to their increasing application in diverse domains like artificial intelligence systems, memristive devices must now overcome several issues related to data retention, endurance, and a large number of conductance states. Moreover, the individual fabrication of these devices is time-consuming. As a result, several challenges need to be addressed to improve their performance and reliability. In a recent study led by Professor Min Kyu Yang from Sahmyook University, Korea, researchers have developed a silver (Ag)-dispersive chalcogenide thin film for use as a resistance-switching material in memristive devices. Their paper is published in the journal Applied Surface Science.

*astonished emoji face* this is amazing, 700c non-volatile memory! this is huge for venus missions, deep earth, holy geez! actually amazing work for real! this allows deep gas planet exploration, extreme plasma physics calculations done better through better observations, this allows us to literally use high temperature high pressure plasma to be used for calculations which opens up huge new computational methods involving light and sound based rotating virtual charge vortex fields that can be stable at room temperature or, for this, high pressure high control oscillating fields that make good superposition qbits that take advantage of collapsing virtual fields as transistors. basically quasi particles that get stuck but only at high temperatures, they can create virtual fields, they're typically way too hot and high temp for memory to be used.