The distributed electrochemical random-access memory, ECRAM is a new tool to handle AI.

 

"Researchers have, for the first time, decoded how Electrochemical Random-Access Memory (ECRAM) works, using a special technique to observe internal electron behavior even at extreme temperatures. This hidden mechanism, where oxygen vacancies act like shortcuts for electrons, could unlock faster AI systems and longer-lasting smartphones, laptops, and tablets." (ScitechDaily, Ghost Highways in Memory Chips – The Secret Electron Shortcut to Lightning-Fast AI)

New things like AI require new equipment. The AI is a combination of complex algorithms, and that system requires a new, faster way to handle memory and its combination with the software. In traditional computing, the microchip calls the program from RAM (Random Access Memory) for driving. When the computer must always call millions or even billions of algorithms. 

That thing makes traditional computing slow. "In-memory" computing where data is handled in multiple memory blocks at the same time can be the answer to that problem. In that model, every single memory chip has its own processor. 

Whenever the computer drives a program the code travels through the microchip. That kind of way where all data travels through the microprocessors is quite slow. There are tested systems where each RAM component has its own microchip that gives the system a new multicore way to handle data. The data can be handled in a memory chip which decreases the need to transport it a long way. 

That decreases the temperature in the wires. However, the problem is that all microchips create temperature. And that system just transports the temperature problem to another location. 

The new systems called "in-memory computing" where calculations happen straight in memory are anyway tools that allow to make faster and more effective computers. If we think that the system uses a distributed way of data-handling processes, that system can mimic the quantum computer. 

The system their calculations happen non-centralized in multiple independently operating memory units that can operate as a virtual quantum computer. The system can share calculations or operations with those data handling units. Each of those data-handling units can act like a qubit state. So "in-memory computing" can allow the system that can operate as a virtual quantum computer to make it faster than a regular, centralized computer. 

"A schematic representation of in-memory computing using electrochemical memory devices (ECRAMs) arranged in a cross-point array structure, mimicking the way synapses in the brain process information. When voltage is applied to the device, ions move within the channel, enabling simultaneous computation and data storage. This study reveals how ions and electrons behave under applied voltage, uncovering the device’s internal operational dynamics. Credit: (Pohang University of Science and Technology,POSTECH) ((ScitechDaily, Ghost Highways in Memory Chips – The Secret Electron Shortcut to Lightning-Fast AI)

"In-memory computing" allows the system to handle multiple operations at the same time. The system can share the memory blocks into segments. And when the system accomplishes its job.

It can give output to the screen. Then the block waits for new orders. Or it can tell other blocks that it's free, and if those other blocks require help it can offer its resources to them. 

 But in the same way as human brains the "in-memory computer" can continue some other operations backward. "In-memory computers" acts like human brains. 

There is always space for a new mission. If the system must run multiple tasks at the same time, it can reserve a certain number of memory units for each job. When some job is done the computer can share those free resources with jobs that need them. 

"In-memory computer"mimics human brains. Because the system has multiple data-handling units that can share multiple problems the system doesn't need to stop for a new mission. The single unit must stop but the entirety can always take the mission. The reason why our brains use that model is simple. All neurons work as groups. There are always free neurons that can take a new job. And when the other neurons finish their mission. Those neurons can call others to assist them. If the problem is difficult brains will connect more and more neurons to operate with it. 

But that system requires new memory technology called Electrochemical Random-Access Memory, ECRAM. The system mimics human memory. This means there are two states in that memory. The electric state and chemical state. Basically, the idea of the ECRAM chemical memory is simple. When the molecule is in a certain position, it has a state of 0. And the other position is 1. The problem is how to make that thing in practical solution. Those positions can be physical or they can be the ion states. Or anion can be 1 and an ion can be 0. There are multiple ways to make that thing possible. ECRAM technology should produce less heat than a regular electric data-handling process. 

1) "ECRAM (Electrochemical Random-Access Memory): An electrochemical memory device whose channel conductivity varies according to the concentration of ions within the channel. This behavior allows for the expression of analog memory states. The device features a three-terminal structure consisting of a source, drain, and gate. By applying voltage to the gate, ion movement is controlled, and the channel conductivity is read through the source and drain."(ScitechDaily, Ghost Highways in Memory Chips – The Secret Electron Shortcut to Lightning-Fast AI)

2) "Parallel Dipole Line Hall System, PDL Hall System: A Hall measurement system composed of two cylindrical dipole magnets. When one magnet is rotated, the other rotates automatically, enabling the generation of a strong, superimposed magnetic field. This configuration allows for enhanced sensitivity in observing internal electron behaviors." (ScitechDaily, Ghost Highways in Memory Chips – The Secret Electron Shortcut to Lightning-Fast AI)

https://pubs.aip.org/aip/apl/article/106/6/062407/29245/A-parallel-dipole-line-system

https://www.quantamagazine.org/what-is-distributed-computing-20241125/

https://scitechdaily.com/ghost-highways-in-memory-chips-the-secret-electron-shortcut-to-lightning-fast-ai/