Researchers have developed a smart material that could take AI devices to a new human-like level. It should work differently depending on the stimulation and be able to adapt like a synapse.
For decades, researchers around the world have been trying to find alternatives to silicon for certain applications. The idea behind it is the production of electronic components based on certain molecules. But they often failed because these molecules usually behave unpredictably and cannot be put together so easily.
So-called neuromorphic computing pursues a similar goal. The basis is hardware that is inspired by the brain. The goal is to produce a material that can store information, perform calculations and be customizable at the same time. Currently used approaches are often based on oxide materials and filamentary circuits. But these still only function like carefully constructed systems that imitate learning.
Neuromorphic computing: Device adapts based on stimulation
New research from India suggests that both problems can be solved using the same approach. A team at CeNSE created a small molecular device that can perform various tasks. The basis lies in how the researchers stimulate the device.
It can store information and can be a logic gate, a selector, an analog processor or an electronic synapse. Chemical design and computers should go hand in hand. The scientists produced 17 specially designed ruthenium complexes.
These are chemical compounds in which a central ruthenium atom is surrounded by ligands. Due to their stability and versatility, these are used primarily as catalysts, in cancer therapy and in photochemistry. The team then examined how small changes in the shape of the molecule and the surrounding ionic environment influence the behavior of the electrons.
Will even more efficient and intelligent AI hardware be coming soon?
By adjusting the ligands and ions, the device showed different functions. The researchers observed that switching from analog to digital and vice versa is possible without any problems. There is a lot of theoretical knowledge behind the research about physics and quantum theory.
The approach is intended to pave the way for neuromorphic hardware, where learning can be encoded directly into the material itself. The team is already working on placing the materials on silicon chips. The goal is to develop future AI hardware that is both energy efficient and inherently intelligent.
Also interesting:
