It has long been the fodder of science fiction entertainment, the idea of implanting integrated circuits (ICs) into human and animal brains to control organism behavior. This also includes the concept of creating silicon chips that literally think and perform like humans. Some may say the concept of creating devices and systems that think and behave like humans falls under the umbrella of artificial intelligence. Assuming humans have original, non-artificial thoughts, that assumption would be correct. However, we all know what happens when we assume.
On September 12, 2016 in Lausanne, Switzerland at ESSDERC/ESSCIRC 2016, an international project to develop technology and architectures for mimicking human neural behavior in ICs will review the state of the art in creating neuromorphic circuits and bring together device and design communities in this emerging field. The project, called the NeuRAM3 project, takes its name from neural-computing architectures in advanced monolithic 3D VLSI technologies. A workshop will feature speakers from different EU and international programs and groups involved in development of neuromorphic electronics to present for the purposes of discussing recent advances in the field.
According to Carlo Reita, director of technical marketing and strategy, nanoelectronics at Leti, the company coordinating the NeuRAM3 project, “Neuromorphic computing is based on mimicking the processes of the brain in a very simplified manner. In the brain, connections between neurons get reinforced – meaning better synapse connections – when there is a temporal correlation between signals coming into the neurons.”
He further emphasizes that project members are not trying to mimic this behavior using software programming in conventional computers, similar to in state-of-the-art deep-learning machines. Instead, they are employing time-domain electrical spikes in a/d circuits that are designed to learn behaviors using the properties of some materials and components in the circuits.
Some of the goals of the project are not much different from those of semiconductor makers in general. These include developing an ultra-low power, scalable and highly reconfigurable architecture, achieving a 50x improvement in power consumption compared to digital designs, and fabricating a monolithic 3D technology in FDSOI at 28 nm with integrated RRAM synaptic elements (keeping it all compact).
The three-year NeuRAM3 project is made up of teams from CEA Tech institutes Leti, STMicroelectronics, IBM Zurich, University of Zurich, CNR-IMM, imec, Jacobs University, and IMSE-CISC. Presenters at the workshop will offer diverse strategies for implementing synaptic devices, neurons, and different circuit architectures. Should you find yourself in Lausanne, Switzerland on September 12 and you want to check out the workshop, you can register by CLICKING HERE. For more details, visit http://esscirc-essderc2016.epfl.ch and http://www.leti.fr
