Chips are everywhere: smartphones, supercomputers, remote sensing robots. Now, MIT engineers have created a microchip design that allows sensors and processors to be easily replaced or added, like LEGO bricks. A reconfigurable modular chip like this could be useful for upgrading smartphones, computers, or other devices without producing as much garbage. Also, it could be useful for artificial intelligence applications. Their article describing the technology was published this week in the journal Natural electronics.
Here is how the chip is configured. It has alternating layers for detection and processing. Instead of having copper wires, the layers of the chip communicate internally via optical signals, more specifically with light-emitting diodes (LEDs). These two features make it easy to interchange various elements on individual layers with other elements.
“As we enter the era of the sensor network-based Internet of Things, the demand for advanced multi-function computing devices will dramatically increase,” Jeehwan Kim, an associate professor of mechanical engineering at MIT, said in a statement. Press release. “Our proposed hardware architecture will provide great versatility for edge computing in the future.” (Edge computing refers to electronics that can independently process data without having to connect to a central server).
To test how the chip works on simple tasks, the team created a prototype with image sensors, LEDs and a processor containing “artificial brain synapses” – silicon, silver and copper components that mimic the how the brain transmits information (the team also calls these memristors). Instead of just transmitting information in binary (like 0 or 1), the strength of the output electric current from memristors depends on the strength of the incoming current. This allows it to have a range of values based on the strength of the signals. And it always remembers the value associated with the signal strength so that the calculations remain constant. A connected circuit, or network, of these neurons could directly process and classify the on-chip signals.
[Related: The trick to a more powerful computer chip? Going vertical.]
The researchers trained a version of the stacked chip to recognize the letters M, I, and T. (For MIT.) This chip had photosensors to receive the visual signal and pass it on to other layers that encoded the image as form of sequence of LED pixels. and classified the signal based on the strength of the incoming light. The researchers used laser light to shine different letters onto the chip, and the chip was generally able to recognize the letter given to it, although it did better with brighter and brighter images. At some point, the researchers added a “denoising” processor that helped the chip better understand blurry images.
The team imagines that this modular capability will allow them to add features such as image recognition to smartphone cameras or health monitoring sensors to electronic skins.
“We can build a general chip platform, and each layer could be sold separately as a video game,” Jeehwan Kim said. “We could create different types of neural networks, like for image or voice recognition, and let the customer choose what they want, and add to an existing chip like a LEGO.”