Translated data: Tsinghua University has unveiled the world's first integrated memory-computing chip, supporting efficient on-chip learning, which can quickly complete model training for different tasks on the chip itself. The chip's energy consumption is only 1/35 of that of ASICs under advanced processes, potentially achieving a 75-fold increase in energy efficiency while protecting privacy. This breakthrough will propel the development of the AI field, enabling edge devices to better adapt to user habits and new scenarios. Researchers have experimentally verified the capabilities of on-chip learning, including tracking light spots with a small car and image recognition. This technology is expected to overcome the computational bottleneck posed by the von Neumann architecture.