从基础研究浅析人工智能技术发展趋势

Analysis of the trend of artificial intelligence technology on basic research

近六十多年来,人工智能在算法、算力和数据的共同驱动下,获得了飞速发展,但仍处于弱人工智能阶段。重点分析了人工智能算法和算力方面的基础研究现状和发展趋势,弱人工智能迈向强人工智能亟待基础研究上的革命性突破。算法层面,深度学习算法模型缺乏可释性和可泛化性,在基础理论上遇到瓶颈,亟待基础理论上的突破;算力层面,因集成电路工艺制程逼近微观物理极限导致摩尔定律失效和电子芯片算力增长趋缓,通用计算芯片架构受制于冯诺依曼瓶颈,以神经形态芯片为代表的人工智能芯片方兴未艾;数据层面,细分领域的高质量数据集匮乏制约人工智能技术应用发展,未来高质量数据集将不断构建。总之,人工智能底层技术将在未来相当长时间内缓慢前进,但产业化应用正在蓬勃发展。

During the past sixty years, artificial intelligence( AI) has achieved rapid development jointly promoted by algorithms,computing power, and big data, but it is still in the stage of artificial narrow intelligence. The status and trends of basic research in AI algorithms and computing power are analyzed. The evolution of artificial narrow intelligence to artificial general intelligence will depend on breakthrough in AI basic theory research. On the aspect of AI algorithms, the deep learning algorithm model lacks interpretive reasoning and generalizability. AI encounters bottlenecks in basic theory and urgently needs a breakthrough. On the as-pect of computing power, due to the CMOS physical limits the Moore ′ s law is approaching failure and the growth of computing power is slowing down, the general computing chip architecture is limited by Feng Neumann ′ s bottleneck and AI chips represented by neuromorphic chips are in the ascendant. On the aspect of data, the lack of high-quality data sets in specific area restricts AI technology application and more high-quality data sets will be continuously constructed in the short future. In short, the basic AI technology will slowly advance for a long time in the future, but the AI applications are booming from right now.