Cornell Engineers Innovate AI Hardware with Mechanical Motion
Cornell engineers use tiny vibrating beams to rethink AI hardware
Cornell Chronicle
Image: Cornell Chronicle
Researchers at Cornell University have created a novel computing device that uses tiny vibrating beams to read information mechanically, enhancing energy efficiency in AI and scientific computing. This device integrates ferroelectric materials, allowing for precise analog value storage and computation, potentially transforming future hardware design.
- 01The new device, called a ferroelectric microelectromechanical system (FeMEMS), utilizes a 20-nanometer layer of hafnium zirconium oxide.
- 02It can represent approximately 200 distinguishable electromechanical states, providing fine control over analog values.
- 03The device aims to reduce energy consumption by integrating memory and computation, addressing inefficiencies in current computing systems.
- 04The research was led by doctoral student Shubham Jadhav and Professor Amit Lal at Cornell's Duffield College of Engineering.
- 05Supported by DARPA, the next steps include developing larger arrays for more complex operations.
Advertisement
In-Article Ad
Researchers at Cornell University have developed a groundbreaking computing device that utilizes tiny vibrating beams to read information through mechanical motion, rather than traditional electrical methods. This innovative approach, detailed in the journal Nano Letters, combines ferroelectric materials with a microelectromechanical system (FeMEMS), allowing for energy-efficient storage and computation of analog values. The device can represent around 200 distinguishable electromechanical states, significantly improving precision in analog computing. By integrating memory and computation, the researchers aim to reduce the energy expenditure associated with data transfer in artificial intelligence (AI) and scientific computing. The project, led by doctoral student Shubham Jadhav and Professor Amit Lal, seeks to revisit and explore alternative computing methods as traditional scaling of CMOS technology becomes more challenging. Future developments will focus on creating larger arrays to facilitate complex matrix operations, potentially revolutionizing AI hardware and beyond.
Advertisement
In-Article Ad
This development could lead to more energy-efficient AI hardware, impacting industries reliant on AI and scientific computing.
Advertisement
In-Article Ad
Reader Poll
What do you think about the potential of mechanical computing in AI?
Connecting to poll...
More about Cornell University
Read the original article
Visit the source for the complete story.