Developers create a Lego-style, artificial intelligence integrated circuit
A team of researchers from MIT has developed a groundbreaking Lego-like AI chip designed for edge computing and wearable electronics. This modular chip, inspired by the popular building toy, can be assembled and reconfigured like Lego pieces, offering flexible, custom hardware design optimized for specific AI tasks.
The chip consists of alternating layers of sensing and processing elements, along with light-emitting diodes (LED) for communication between layers. An optical system between each sensor and artificial synapse array enables communication without requiring a physical connection.
The design is currently configured to carry out basic image-recognition tasks, recognizing three specific letters, M, I, and T. The team trained each layer to recognize its corresponding letter, and the chip correctly classified clear images of each letter. However, it struggled with blurry images.
The chip's modularity offers several advantages, particularly for edge computing devices and wearable electronics. These devices often operate with limited processing resources and power budgets. By reconfiguring the AI hardware for specific tasks on demand, the chip promises improved performance and energy efficiency. It also supports versatility, enabling a single chip to handle multiple functions without needing replacement or redesign.
The research was supported by the Ministry of Trade, Industry, and Energy (MOTIE) from South Korea, the Korea Institute of Science and Technology (KIST), and the Samsung Global Research Outreach Program. The team's results are published in Nature Electronics.
The team previously developed artificial synapse arrays that function as a physical neural network, or "brain-on-a-chip." They envision applications such as adding layers to a cellphone's camera for more complex image recognition, creating healthcare monitors for wearable electronic skin, and developing modular chips for consumers to customize with the latest sensor and processor "bricks."
The researchers are eager to apply the design to edge computing devices and plan to add more sensing and processing capabilities to the chip. With its adaptable, efficient design, this Lego-like AI chip could revolutionize the field of edge computing, promising improved performance, lower power consumption, and enhanced functional flexibility tailored to diverse, resource-constrained environments.
- The MIT team's Lego-like AI chip, designed for edge computing and wearable electronics, boasts a modular structure, similar to the popular building toy, allowing for flexible, custom hardware design optimized for specific AI tasks in fields like science, engineering, and technology.
- One of the significant advantages of the chip's modularity is its potential for energy efficiency, particularly in edge computing devices and wearable electronics where processing resources and power budgets are often limited.
- The chip's developers intend to explore applications beyond image recognition, such as adding layers to a cellphone's camera for more complex data-and-cloud-computing tasks, creating healthcare monitors employing engineering and health technologies, and developing modular chips for consumers to customize with the latest sensor and processor components.
- In the realm of research, the team's results have been published in Nature Electronics, with support from the Ministry of Trade, Industry, and Energy (MOTIE) from South Korea, the Korea Institute of Science and Technology (KIST), and the Samsung Global Research Outreach Program.
- The chip's technology features an optical system that enables communication between layers without requiring a physical connection, and the use of artificial intelligence in conjunction with computing, engineering, and science to perform tasks such as image recognition.
- The chip is equipped with light-emitting diodes (LED) for communication between layers, and the current configuration can recognize three specific letters, M, I, and T, although it faces challenges with blurry images, indicating potential improvements in the field of artificial-intelligence technology.
