全员中文属名，清华大学×北京大学合作最新Nature论文

Core Viewpoint - The integration of flexible electronics technology with artificial intelligence (AI) is expected to revolutionize various fields, including robotics, wearable medical devices, and human-computer interaction interfaces. However, significant challenges remain in developing flexible computing hardware capable of efficiently executing neural network inference tasks [3]. Group 1 - A new flexible digital AI chip named FLEXI has been developed for edge intelligence, characterized by its lightweight, thin, and robust design, paving the way for ultra-low-cost AI systems in wearable medical devices [5]. - The FLEXI chip features a maximum operating clock frequency of 12.5 MHz, a power consumption as low as 2.52 mW, and a unit cost of less than 1 USD, with a circuit yield rate stable between 70% and 92% [5]. - The chip can perform 10^10 fixed-point and random multiplication operations with zero error, withstand over 40,000 bending cycles, and maintain stable performance over more than six months of continuous operation [5]. Group 2 - The chip achieves a detection accuracy of 99.2% for arrhythmia tasks on a single 1kb chip and over 97.4% accuracy for monitoring daily human activities based on multimodal physiological signals [5]. - An accompanying article in Nature highlights the collaborative optimization design and technology that enables the efficient operation of neural network tasks on affordable and flexible microchips, validated in wearable medical devices [5].