Page 103 - WYSS2022 总结册-1
P. 103

潘云鹤                                潘爱民                                 Mike Davies
                                                                                                                                  Yunhe Pan                          Aimin Pan

                                                                                                                                  中国工程院院士、之江实验室                      之江实验室数字反应堆首席架构师                     英特尔神经形态计算实验室主任
                                                                                                                                  人工智能领域首席科学家                        Chief Reactor Architect,            Director of Intel' Neuromorphic
                                                                                                                                  Academician of CAE,                Zhejiang Lab                        Computing Lab
                                                                                                                                  Chief Scientist in Artificial
                                                                                                                                  Intelligence of Zhejiang Lab



                                                                                                                                  智能计算在计算机科学领域中正逐步形                  之江实验室最新研发的“之江瑶光”智                   神经形态计算的目标是弥合生物大脑和
                                                                                                                                  成自身体系,开始走向纵深发展。我相信, 能计算操作系统,可以进行算力聚合和                                  当今计算技术在效率、适应性和速度
                                                                                                                                  通过开展持续和深入的国际合作交流, 任务调度。该系统能够面向不同的计算                                    方面的数量级差距。过去几年,神经形
                                                                                                                                  打造智能计算创新生态,将会加速推动                  任务提供相应的算法、函数或框架,具                   态计算研究取得了重大突破,英特尔的
                                                                                                                                  智能计算在科学研究、产业发展、社会                  备算力聚合和数据编排的能力,可以让                   Loihi 等芯片首次证明,从感官感知到数
                                                                                                                                  治理等领域的应用。                          数据高效地到达计算集群和节点,实现                   据高效学习再到组合优化,神经形态计

                                                                                                                                  Intelligent computing is gradually   规模化算力的高效使用,支撑各领域的                 算都取得了令人瞩目的成效。
                                                                                                                                  forming its own system in computing   计算需求。                            The goal of neuromorphic computing
                                                                                                                                  science, and it’s beginning to develop   Zhejiang Alkaid OS, an intelligence   is to bridge the order of magnitude
                                                                                                                                  in depth. I believe that continuous and   computing system newly developed by   difference between biological brains
                                                                                                                                  in-depth international cooperation   Zhejiang Lab, is capable of aggregating   and c ompu ting te c hno lo gie s in
                                                                                                                                  and exchanges and the building of   computing power and scheduling     terms of efficiency, adaptability and
                                                                                                                                  an intelligent computing innovation   tasks. This system can provide the   speed. Over the past few years, major
                                                                                                                                  ecology will accelerate the application   corresponding algorithms, functions   breakthroughs have been made in
                                                                                                                                  of intelligent computing in scientific   or frameworks for different computing   research on neuromorphic computing.
                                                                                                                                  research, industry development, social   tasks, has the ability of computing   Intel’s Loihi and other neuromorphic
                                                                                                                                  governance, etc.                   power aggregation and data sequence,   chips have demonstrated for the first
                                                                                                                                                                     and can enable data to efficiently reach   time that impressive results have been
                                                                                                                                                                     computing clusters and nodes, thus   achieved in neuromorphic computing,
                                                                                                                                                                     achieving efficient use of large-scale   from sensory perception to efficient data
                                                                                                                                                                     computing power and supporting the   learning and combinatorial optimization.
                                                                                                                                                                     computing needs in various fields.





                                                                                                                                                                                                                                     101
   98   99   100   101   102   103   104   105   106   107   108