Chinese researchers have made strides in artificial intelligence with the launch of a fully flexible AI chip, detailed in a paper released on January 29 in Nature. This advancement, driven by teams from Tsinghua University, Peking University, and other institutions, targets sectors such as wearable health monitoring and flexible robotics.
The newly introduced FLEXI series of digital compute-in-memory AI chips addresses the growing need for lightweight and adaptable computing solutions, particularly as AI merges with the Internet of Things. Traditional silicon chips lack the flexibility required for integration into wearable devices, while current flexible processors suffer from inefficiencies.
Characterized by ultra-low power consumption and high energy efficiency, the FLEXI chip utilizes low-temperature polycrystalline silicon thin-film transistors, making it ultra-thin and bendable. The FLEXI-1 variant, which is only 31.12 square millimeters and contains 10,628 transistors, operates at a remarkably low power of just 55.94 microwatts. Notably, it can withstand over 40,000 cycles of 180-degree bending without performance loss and has shown stable performance in tests lasting up to six months, achieving a 99.2% accuracy in arrhythmia detection.