让光“理解”和“认识”语义

Core Insights - Generative artificial intelligence is advancing towards more complex real-world applications, with a focus on next-generation computing chips like optical computing due to increasing demands for computational power and energy consumption [1][2] Group 1: Technological Breakthroughs - Shanghai Jiao Tong University's research team has achieved a significant breakthrough in optical computing with the development of the LightGen chip, which supports large-scale semantic visual generation models [1] - LightGen integrates millions of optical neurons on a single chip and utilizes an optical generation model training algorithm that does not rely on true values, overcoming three key bottlenecks for end-to-end optical implementation of large-scale generation tasks [1] Group 2: Performance Evaluation - LightGen demonstrates a computational power and energy efficiency improvement of two orders of magnitude compared to top digital chips, and theoretically, it can achieve a seven-fold increase in computational power and an eight-fold increase in energy efficiency when using cutting-edge input devices [2] - The research has been highlighted as a key paper by the journal Science, emphasizing the necessity of developing chips capable of executing real-world tasks, particularly for high-latency and energy-intensive tasks like large-scale generative models [2] Group 3: Future Implications - LightGen paves the way for next-generation computing chips to support advanced artificial intelligence applications, providing new research directions for faster and more energy-efficient generative intelligent computing [2]