人工智能是指通过计算机程序或硬件，模拟人类的感知、认知、决策和执行能力[1]。AI的发展经历了三个关键阶段：符号逻辑时代（1950s-1970s）、机器学习时代（1980s-2000s）和深度学习时代（2012年至今）[1]。其中，2012年ImageNet挑战赛成为重要分水岭，深度学习错误率比上一年砍掉一半[1]。2017年Transformer架构的诞生引入了自注意力机制，成为大模型的基石[1]，随后2022年ChatGPT的问世标志着生成式AI的广泛应用，再到2025年DeepSeek以低成本、高性能、开源模式打破垄断[1]。
推动AI发展的三大要素分别是算法、算力和数据，分别担当大脑、发动机和燃料的角色[1]。AI的技术实现分为五层结构：基础设施层、框架层、模型层、应用层和Agent层[1]。从应用看，AI已广泛应用于内容生成、客服、知识管理等场景[1]。展望未来，到2026年，72%的企业预计已完成Agent试点[1]，中国大模型市场规模突破700亿元，企业级智能体市场增至449亿元[1]。
Artificial intelligence is defined as the simulation of human perception, cognition, decision-making, and execution capabilities through computer programs or hardware [1]. The field has evolved through three distinct waves: the symbolic logic era spanning the 1950s to 1970s, the machine learning period from the 1980s through 2000s, and the deep learning age that began in 2012 and continues today [1].
The year 2012 marked a pivotal turning point when the ImageNet challenge demonstrated deep learning's breakthrough potential, with error rates cut in half compared to the previous year [1]. This momentum accelerated with the emergence of the Transformer architecture in 2017, which introduced self-attention mechanisms that became foundational to large language models [1]. The landscape shifted dramatically in 2022 with ChatGPT's arrival, followed by DeepSeek's 2025 disruption of the market through its low-cost, high-performance, and open-source approach [1].
The development of AI rests on three fundamental elements: algorithms serving as the brain, computing power functioning as the engine, and data acting as fuel [1]. The technological infrastructure supporting AI comprises five layers—basic infrastructure, frameworks, models, applications, and agents—each playing a critical role in the ecosystem [1]. Current applications span content generation, customer service, and knowledge management across six major scenarios [1], with projections indicating that 72 percent of enterprises will have completed Agent pilot programs by 2026 [1]. China's large language model market has exceeded 7 billion yuan in scale, while the enterprise-level intelligent agent market has grown to 4.49 billion yuan [1].
