Clik here to view.
A new neuromorphic element called a “spin-memristor” mimics the energy-efficient operation of the human brain to reduce the power consumption of artificial intelligence (AI) applications to 1/100th of traditional devices. TDK developed this “spin-memristor” as the basic element of a neuromorphic device in collaboration with the French research outfit CEA.
It’s apparent by now that energy consumption utilizing big data and AI will boom, inevitably leading to complexity around the computational processing of vast amounts of data. So, TDK aims to develop a device that electrically simulates human brain’s synapses: the memristor.
Image may be NSFW.
Clik here to view.
Figure 1 The “spin-memristor” has been demonstrated to function as the basic element of a neuromorphic device. Source: TDK
Here, it’s important to note that conventional memory elements store data as either 0 or 1. On the other hand, a spin-memristor can store data in analog form, just as the brain does. That enables it to perform complex computations with ultra-low power consumption.
While memristors for neuromorphic devices already exist, they face critical challenges, including changes in resistance over time, difficulties in controlling the precise writing of data, and the need for control to ensure data is retained. TDK’s spin-memristor overcomes these issues and provides immunity to environmental influences and long-term data storage while reducing power consumption by cutting leakage current.
Practical applications
After jointly developing spin-memristor with CEA, TDK is partnering with the Center for Innovative Integrated Electronic Systems at Tohoku University to create practical applications for this device. While the tie-up between TDK and CEA has demonstrated that spin-memristors can serve as the basic element of a neuromorphic device, manufacturing them requires the integration of semiconductor and spintronic manufacturing processes.
Spintronics is a technology that utilizes both the charge and spin of electrons or the spin element alone. TDK’s AI semiconductor development program, in collaboration with Tohoku University, will work on fusing memristors with spintronics technology.
Image may be NSFW.
Clik here to view.
Figure 2 TDK is collaborating with Tohoku University to develop practical applications for spin-memristors. Source: TDK
It’s worth noting that the integration between semiconductor and spintronic manufacturing processes has already been accomplished in a similar product: MRAM. TDK chose Tohoku University as its partner mainly because it’s a leading academic institution in MRAM research and development.
Related Content
- Will memristors prove irresistible?
- Memristor emulates neural learning
- Knowm’s memristors alive and shipping
- How Memristors Could Help Drive AV Evolution
- Memristor Computer Emulates Brain Functions
The post Spin memristor mimics brain for energy efficiency in AI appeared first on EDN.
