Home > MarketWatch > Industry News
Jingtai Research|Artificial Intelligence Industry Chain Combing: Industry Background
Time:2024-06-23

25512366-68f65W.jpg?auth_key=1719158399-

Since 2018, pre-trained language model (PLM) and its "pre-training-fine-tuning" method have become the mainstream paradigm of natural language processing (NLP) tasks. In the AI 1.0 era, there are problems such as obvious fragmentation of models and insufficient AI generalization capabilities.


The "pre-training + fine-tuning" large model can significantly reduce the threshold for AI engineering, and the pre-trained large model has good versatility and generalization after learning and training of massive data, and the application manufacturers of subdivided scenarios can obtain significant results based on the large model through zero-sample and small-sample learning, making artificial intelligence is expected to build a unified intelligent base, and AI+ empowers all walks of life.


This round of generative AI is expected to gradually move from simple content generation to higher cognitive intelligence such as prediction, decision-making, and exploration.


01


The development history of artificial intelligence

25512366-a3fTgJ.jpg?auth_key=1719158399-



The rapid explosion of large models has redefined the artificial intelligence industry

Since 2018, pre-trained language model (PLM) and its "pre-training-fine-tuning" method have become the mainstream paradigm of natural language processing (NLP) tasks.


Larger models not only perform better on known tasks, but also demonstrate a strong generalization ability to complete more complex unknown tasks.


25512366-HL4943.jpg?auth_key=1719158399-

25512366-vTzstS.jpg?auth_key=1719158399-



Artificial intelligence industry chain

The AI industry chain mainly includes three layers: the basic layer, the technology layer, and the application layer.


25512366-rvh4vR.jpg?auth_key=1719158399-



|人工智能产业分类

AI模型大致可以分为决策式AI和生成式AI两种。决策式AI指学习数据中的条件概率分布,根据已有数据进行分析、判断、预测,主要应用模型有用于推荐系统和风控系统的辅助决策、用于自动驾驶和机器人的决策智能体。


生成式AI指学习数据中的联合概率分布,并非简单分析已有数据而是学习归纳已有数据后进行演技创造,基于历史进行模仿式、缝合式创作,生成了全新的内容,也能解决判别问题。


人工智能在经历前期技术积累和迭代后,逐渐突破传统分析型AI领域,迎来生成式AI的爆发期。从2012年至今,生成式AI急速发展,其源头就是DNN算法的升级,实现了语音和图像识别等功能。


25512366-3tL7d5.jpg?auth_key=1719158399-



|生成式人工智能应用领域

生成式人工智能的重大意义就在于能够为当前人力和AI 技术无法突破的难题提供可行的解决方案,从而颠覆了整个商业规则和运营模式。以下案例均为生成式AI 工具在前期发展阶段已经可以赋能的应用场景。


25512366-p3rqDz.jpg?auth_key=1719158399-



|全球市场规模预测

据彭博研究数据预测,全球生成式AI 市场 2023年收入规模约 670亿美元,到 2032年该规模将接近13,040亿美元,占总信息技术支出比例预计十年内从现阶段的小于1%增至10~12%。


从投资角度看,IDC预计 2023 年全球AIGC 投资规模为160 亿美元,2027年将达到约 1430亿美元,且其投资额占AI 整体支出比例将增至28%(2023年 9%)。


未来,AIGC应用将日渐普及,IDC最新报告显示,到2027年,全球约45%的企业将采用生成式人工智能工具进行产品和服务的开发和运营。


25512366-cndyvq.jpg?auth_key=1719158399-



|中国市场规模预测

中国将引领亚太区AIGC 投资,实现相关技术产业规模的爆发式增长。IDC最新预测数据显示,中国将继续成为亚太地区人工智能市场发展的主要增量,相关投资在亚太地区AI 总支出占比中超50%,约占全球AI 技术投资总规模的9%。作为向自动化和通用生产力迈进的关键技术,国内生成式AI 技术投资占比预计将在2027 年达到33.0%(2022年4.6%),投资规模超130亿美元,五年复合增长率为86.2%。


艾瑞咨询数据显示,目前中国仍处于大模型生态培育期,2023年 AIGC产业规模约为143 亿元,基本完成大模型最底层算力和应用平台等新型基础设施的搭建。预计于2028 年前后完成重点领域、关键场景的技术价值兑现,实现7,202 亿元的市场规模。


预计2030 年中国AIGC 产业规模有望突破万亿元,达到11,441 亿元,年复合增长率为115%。


25512366-QA74Sq.jpg?auth_key=1719158399-



TEL:
18117862238
Email:yumiao@jt-capital.com.cn
Address:20th floor, Taihe · international financial center, high tech Zone, Chengdu

Copyright © 2021 jt-capital.com.cn All Rights Reserved 

Copyright: JamThame capital 粤ICP备2022003949号-1  

LINKS

Copyright © 2021 jt-capital.com.cn All Rights Reserved 

Copyright: JamThame capital 粤ICP备2022003949号-1