本文来自《麦肯锡中国金融业CEO季刊》，版权归麦肯锡所有。该季刊主要围绕生成式AI（以下简称“GenAI”）主题，通过4大章节共8篇文章，全面深入分析了GenAI对各主要行业的影响、价值链投资机会、中国GenAI市场现状和未来趋势以及企业如何布局GenAI，从而真正挖掘其价值。
随着ChatGPT的火爆出圈，GenAI成为各行各业关注和热议的话题。全球科技巨头和AI厂商纷纷下场，唯恐错过此番科技盛宴。行业翘楚和媒体将GenAI浪潮类比昔日的移动互联网机遇，认为它将对全球经济和各个行业带来深远影响，企业也将迎来重大变革机遇。
GenAI蓬勃发展产业规模也在高速增长，投资者纷纷入局。2022年GenAI市场收入为400亿美元，预计2027年及2032年将分别达到3,990亿美元和13,040亿美元，2022~2032年复合增长率达42%。
中国市场，2022年规模约人民币660亿元，2020~2025年复合增速将达到84% ，2025年中国GenAI市场将占全球市场规模（2,170亿美元）的14%。
由于内容较多一共178页，下面主要为大家介绍中国GenAI市场现状和未来趋势。
一、市场规模
2022年全球GenAI市场整体收入为400亿美元，预计2027年及2032年将分别达到3990亿美元和1.3万亿美元，2022~2032年复合增长率高达42%。 数字化转型网www.szhzxw.cn
到2025年，GenAI产生的数据将占所有生成数据的10%（2021年仅为1% 2）。
预计2022~2035年全球市场规模增量主要来源于训练侧硬件、广告应用和软件。其中，在训练侧硬件方面的增量约4,440亿美元，仅GenAI基础设施服务（GenAI as a service）一项，增量就有2,448亿美元，年复合增长率将达到60%；
在广告应用方面，相关年复合增长率将达到125%，增量达1924亿美元；而软件方面，在约2,800亿美元的增量中，GenAI助理软件增量可观，达890亿美元，年复合增长率有望达到70%。
2022年中国GenAI市场规模约人民币660亿元，预计2020~2025年复合增速将达84%，到2025年将占全球市场规模（2,170亿美元）的13%。 数字化转型网www.szhzxw.cn
与此同时，近两年国内GenAI融资市场热度持续上升，前十大融资事件总规模达8.7亿美元。
其中，某领先的端到端AI制药公司、光某大模型初创公司等头部企业获得逾2亿美元的单笔融资，多轮融资累计规模分别超3亿美元、2.5亿美元。
二、技术栈
（一）专用硬件：基础建设资源受限，但国产化进程不断加速。
由于无法获得高端GPU（图形处理器）芯片A100与H100，国内大模型研发机构在算力上遭遇瓶颈。为应对上述困境，中国本土芯片研发厂商加强科研，在算力技术上实现重要突破。
目前，中国市场可以获得的A800 GPU芯片，传输速度仅为顶尖产品A100的70%。由于AI技术发展高度依赖先进GPU，需借助芯片进行大量模型训练和扩展，部分国内科研机构和科技企业因此受到较大影响，尤其在目前大模型参数量快速膨胀、算力需求显著增加的情况下。
为了应对算力瓶颈，国内领先算力芯片公司力求在产品端实现突破。例如，国内某量产商业人工智能芯片厂商，在头部大型云服务提供商旗下一款知识增强大语言模型产品中，实现了对部分业务场景的部署。
某专注于通用GPU芯片及解决方案研发及销售的高科技公司和另一计算机图形芯片设计公司则希望开发出接近高端芯片性能的产品。
全球领先的ICT（信息与通信）基础设施和智能终端提供商旗下的人工智能处理器则被用于内部大模型开发训练。 数字化转型网www.szhzxw.cn
目前，国产GPU的算力规模大多在1000TOPS以内，上述某计算机图形芯片设计公司已能达到2000TOPS，但距离国际领先的H100芯片高达4000TOPS的算力仍有一定差距。
（二）云平台：全球竞争格局趋同，未来国内GenAI厂商将依赖云平台完成模型训练与调整。
在云平台领域，国内和海外市场均呈现头部厂商占据绝对主导地位的垄断格局。2022年，国内四家头部厂商占据国内79% 4的市场份额；在海外市场，截至2023年一季度，前三大厂商占据全球65%的市场份额。
通用大模型需要海量数据作训练用途，但中文网站数量从全球来看占比仅为1.4% （英文网站占比达54% ），可用作训练的公开中文语料库（含文本、图片、视频等素材）往往数量有限、质量不均。
与此同时，国内用户在网站和移动端App产生的海量数据，基于保护用户隐私的原因，大部分也未能应用在大模型训练上，对模型训练的效率和精度都造成不利影响。
尽管在训练数据上受到诸多限制，中国通用大模型技术仍然不断追赶国际领先水平，参数规模紧跟国际领先水准并实现快速提升。
相较于适用于多领域、多任务的通用大模型，行业大模型更依赖于开发者对垂直场景的理解和海量行业数据支持，在算力受限的背景下，行业大模型更有可能成为中国GenAI厂商的制胜赛道。
就行业大模型而言，国内市场正呈现差异化竞争格局：某头部互联网大厂的行业大模型根植于游戏、广告、内容创作等多媒体领域，成为提升广告和游戏场景创作效率的生产力工具；
国内前三的超大规模云服务商开发的产业级知识增强大模型则凭借其对中文语言的强大理解能力，赋能汽车行业售后服务与知识库构建、医疗行业单据识别、社科行业版权保护与词条管理等；
某全球领先的ICT（信息与通信）基础设施和智能终端提供商旗下的包含NLP、CV、多模态和科学计算的中文大模型则在工业物流、新药研发、气象预测等领域得到广泛应用。
（三）基础模型：通用大模型数量可观，行业大模型可能是国产GenAI的出路。
在通用大模型方面，截至2023年5月，中国已发布79个10亿参数级以上的大模型，在大模型发布数量上位居全球第二，仅次于美国。在全球模型发布数量前十的GenAI大模型厂商中，中国研发机构及厂商占据4席。 数字化转型网www.szhzxw.cn
通用大模型需要海量数据作训练用途，但中文网站数量从全球来看占比仅为1.4% （英文网站占比达54% ），可用作训练的公开中文语料库（含文本、图片、视频等素材）往往数量有限、质量不均。
与此同时，国内用户在网站和移动端App产生的海量数据，基于保护用户隐私的原因，大部分也未能应用在大模型训练上，对模型训练的效率和精度都造成不利影响。
尽管在训练数据上受到诸多限制，中国通用大模型技术仍然不断追赶国际领先水平，参数规模紧跟国际领先水准并实现快速提升。 数字化转型网www.szhzxw.cn
（四）模型库和工具：国内外围绕开源模型，涌现出大模型工具独立厂商。
海外市场围绕可自由调整代码的开源模型，已涌现出一批独立厂商，可实现模型训练和微调、模型部署、模型应用开发等细分技术功能。随着国内GenAI市场的不断发展，各种开发和维护工具等也将逐渐成熟。
（五）应用：中国创业公司尚处于早期阶段，聚焦领域相对集中。
中国GenAI领域的创业公司融资轮次靠前，集中在天使轮到Pre-A轮之间，融资金额大多小于1亿元人民币，正处于方兴未艾的发展初期，行业潜力巨大。
其中，某全球领先的跨国科技公司旗下的人工智能聊天机器人作为准独角兽企业，于2022年11月完成10亿元A+轮融资，投后估值约10亿美元。
反观海外GenAI行业，科创企业规模体量相对较大，涌现出多家细分应用领域独角兽企业，其中某总部位于伦敦的全球化AI研究型企业估值高达38亿美元。
中国GenAI垂直应用领域主要聚焦文本、图像和音视频生成，海外市场则在上述领域以外出现了大量基于GenAI的开发平台、数据分析平台及代码编写平台，其原因在于海外以早期程序语言（如COBOL）编写的系统较多，许多企业面临较高编程人力成本，因此对编程辅助软件有较高需求。
同时，目前中国GenAI企业目标客群多为终端用户，而海外市场存在大规模企业客户群体，随着中国SaaS市场成熟以及企业付费意愿提升，预计国内文本生成和图像生成初创企业将快速拓展至企业客户市场。
三、商业化应用
中国GenAI行业的商业化应用呈现出两大特点：一是行业分布集中，主要集中在商业化发展较成熟的优势产业；二是大部分企业尚处于积极探索自身商业模式的早期阶段。
（一）中国GenAI应用集中于商业化发展成熟的优势产业
中国GenAI企业利用国内垂直场景优势，将GenAI优先应用于商业化基础设施较完善的新兴行业中；而欧美同行则利用当地成熟的SaaS市场，将GenAI应用到高科技、通信及各种传统行业（如医疗、教育等），应用范围更加广泛。 数字化转型网www.szhzxw.cn
在国内，GenAI应用发展最迅速的领域包括电商、传媒、娱乐和游戏，尤其是数字虚拟人和电商视频营销，而大部分传统行业（如金融、能源、教育等）仍处于小规模试点阶段。
GenAI应用得以在互联网电商平台蓬勃发展，原因在于中国拥有优质的电商和供应链生态以及庞大的消费者群体，为GenAI应用落地提供了契机。
典型的行业应用案例包括视频行业的某生成式AI的3D短视频内容厂商、电商行业的某领先GenAI视频大模型研发企业以及游戏行业的某互联网平台旗下的游戏与AI研究和应用机构等。
上述企业中，生成式AI的3D短视频内容厂商能够基于文本生成3D视频内容，并通过抖音、快手、B站等短视频平台进行推广，极大提高了短视频内容生产者的创作效率；
GenAI视频大模型研发企业则深耕电商行业，以AI生成虚拟人主播，赋能各行业、各品牌电商；游戏与AI研究和应用机构则利用人工智能打造AI陪玩、AI竞技机器人，并通过自然语言技术赋予非玩家角色人设，使其能够在游戏中与玩家互动对话。
（二）中国市场GenAI商业模式尚在探索中
目前，大部分中国市场的GenAI创业公司刚刚完成标准化产品输出，开始进入初级商业化探索阶段，市场上主流的商业模式包括云资源售卖、模型API调用、SaaS收费、素材收费等。
由于国内企业对软件的付费意愿不高、市场有待进一步培育，以及企业对SaaS部署方式存在数据安全方面的顾虑，大规模应用GenAI的商业模式仍有待探索。 数字化转型网www.szhzxw.cn
在欧美市场，SaaS付费订阅模式已基本成熟，一众海外GenAI企业纷纷构建起可持续的SaaS商业模式。
四、风险管理
GenAI赋能各行各业的同时，也带来了诸如公平性受损、侵犯知识产权、信息泄露、恶意使用、安全威胁、模型幻觉及第三方风险等负面影响和挑战。其中，模型幻觉、恶意使用、信息泄露三大风险值得重点关注。
恶意使用：不当使用模型输出可能会带来欺诈、虚假信息传播等负面影响，中国GenAI应用丰富、用户群体庞大，所面临的恶意使用风险较为突出，但由于国内在实名认证，内容合规，和打击虚假信息等方面落实到位，有利于追溯恶意使用源头，从而防范和控制相关风险。
信息泄露：用于模型训练的隐私或敏感信息可能会被GenAI应用作为输出内容，同时人工智能模型存储系统内的个人信息也可能面临网络攻击和泄漏，因此需要在内部部署数据、又处于强监管下的重点行业和企业，需要主动防范信息泄露风险。
安全威胁：不仅负责任的AI（RAI-Responsible AI）受到监管和企业关注，安全的AI（SAI-Securing AI)也逐渐进入企业和公众视野，特别是AI系统全生命周期安全保护。有效的AI威胁管理直接影响到防范信息泄露和恶意使用。近期全国信息安全标准化技术委员会起草了相关评估规范和标准，为算法提供者给与了机器学习算法生命周期安全指导意见7，也可为监管评估提供参考。
模型幻觉：模型输出结果不符合事实。模型幻觉普遍存在于国内外所有的GenAI模型和应用，但因为国内大模型的研发经验和技术实力尚有提升空间，且中文语料相比英文更为复杂，让模型幻觉问题变得更加棘手。 数字化转型网www.szhzxw.cn
为规范GenAI的应用，中国早在2022年已经开始布局AI相关监管，例如2022年3月发布了互联网信息服务算法推荐管理规定，2022年11月发布了互联网信息服务深度合成管理规定。
2023年7月，国家网信办联合国家发展改革委、教育部、科技部等7部门发布《生成式人工智能服务管理暂行办法》，对模型训练过程中可能涉及的数据歧视、知识产权保护、个人隐私及肖像权保护等作出明确规定，行业监管框架进一步完善。
放眼海外，美国国家标准与技术研究院于2023年1月发布了AI风险管理框架，着力为AI开发，治理，和运营提供风险控制最佳实践。欧盟则于2023年6月通过《人工智能法案》授权草案，对不同等级的人工智能所承担的责任和义务进行划分，通过风险定级来实现分层治理，并着力推动跨产业的人工智能立法，以进一步防范人工智能可能带来的多重风险。
五、结语
当下全球人工智能市场正处于高速增长阶段，本文从市场规模、技术栈、商业化应用和风险管理四大维度，对国内外GenAI行业的发展现状和趋势进行了深入浅出的分析。
尽管目前国内外市场在诸如大模型开发、应用布局、商业模式等多个领域都存在较大差异，我们看到中国GenAI行业正在不断追赶国际领先水平，预计未来中国GenAI相关技术和应用将逐渐成熟，并进一步探索出适合自身发展的商业模式。 数字化转型网www.szhzxw.cn
翻译：
McKinsey: Current situation and future development trend of China’s generative AI market
This article is from McKinsey China Financial CEO Quarterly. Copyright 2015 McKinsey. The quarterly magazine mainly focuses on the theme of generative AI (hereinafter referred to as “GenAI”), through 4 chapters, a total of 8 articles, a comprehensive and in-depth analysis of GenAI’s impact on major industries, value chain investment opportunities, the current situation and future trend of China’s GenAI market, and how enterprises layout GenAI, so as to truly tap its value.
With the popularity of ChatGPT, GenAI has become a topic of concern and hot discussion in all walks of life. Global technology giants and AI manufacturers have come to the end, lest they miss this technological feast. Industry leaders and media have likened the GenAI wave to the mobile Internet opportunities of the past, believing that it will have a profound impact on the global economy and various industries, and enterprises will also usher in major opportunities for change.
GenAI is booming and the scale of the industry is also growing at a high speed, and investors have entered the game. In 2022, the GenAI market revenue is $40 billion, and is expected to reach $399 billion and $1,304 billion in 2027 and 2032, respectively, with a compound growth rate of 42% from 2022 to 2032. 数字化转型网www.szhzxw.cn
The Chinese market, the size of about 66 billion yuan in 2022, 2020 to 2025 compound growth rate will reach 84%, 2025 China GenAI market will account for 14% of the global market size (217 billion US dollars).
Due to the large content, a total of 178 pages, the following is mainly to introduce the current situation and future trend of China’s GenAI market.
Market size
The overall revenue of the global GenAI market in 2022 is $40 billion, and it is expected to reach $390 billion and $1.3 trillion in 2027 and 2032, respectively, with a compound growth rate of 42% from 2022 to 2032.
By 2025, GenAI will generate 10% of all data generated (only 1% in 2021).
It is expected that the growth of the global market size from 2022 to 2035 will mainly come from training hardware, advertising applications and software. Among them, the increase in training side hardware of about $444 billion, GenAI infrastructure services alone (GenAI as a service), the increase of $244.8 billion, the compound annual growth rate will reach 60%;
In advertising applications, the CAGR will reach 125%, an increase of $192.4 billion; In terms of software, in the approximately $280 billion increment, GenAI assistant software increment is considerable, reaching $89 billion, and the compound annual growth rate is expected to reach 70%.
In 2022, the size of China’s GenAI market is about 66 billion yuan, and it is expected that the compound growth rate will reach 84% from 2020 to 2025, and it will account for 13% of the global market size (US $217 billion) by 2025. 数字化转型网www.szhzxw.cn
At the same time, the heat of the domestic GenAI financing market continued to rise in the past two years, and the total scale of the top ten financing events reached $870 million.
Among them, head enterprises such as a leading end-to-end AI pharmaceutical company and a large model startup company obtained a single financing of more than $200 million, and the cumulative scale of multiple rounds of financing exceeded $300 million and $250 million, respectively.
Technology stack
(1) Dedicated hardware: infrastructure resources are limited, but the localization process is accelerating.
Due to the inability to obtain high-end GPU (graphics processor) chips A100 and H100, domestic large model research and development institutions have encountered bottlenecks in computing power. In response to the above difficulties, China’s local chip R&D manufacturers have strengthened scientific research and achieved important breakthroughs in computing technology.
At present, the A800 GPU chip available in the Chinese market has a transmission speed of only 70% of the top product A100. Because the development of AI technology is highly dependent on advanced Gpus, a large number of model training and expansion need to be carried out by chips, and some domestic scientific research institutions and technology enterprises are greatly affected, especially in the case of the rapid expansion of the number of large model parameters and the significant increase in computing power demand.
In order to deal with the bottleneck of computing power, the leading domestic computing chip companies strive to achieve breakthroughs in the product side. For example, a domestic mass-produced commercial artificial intelligence chip manufacturer has implemented the deployment of some business scenarios in a knowledge-enhanced large language model product owned by the first large cloud service provider.
One high-tech company focused on the development and sale of general-purpose GPU chips and solutions and another computer graphics chip designer wanted to develop products with near-high-end chip performance. 数字化转型网www.szhzxw.cn
Ai processors from the world’s leading ICT (information and communication) infrastructure and smart terminal providers are used for in-house large-scale model development training.
At present, the computing power scale of domestic Gpus is mostly within 1000TOPS, and the above mentioned computer graphics chip design company has reached 2000TOPS, but there is still a certain gap between the international leading H100 chip and the computing power of 4000TOPS.
(2) Cloud platform: Global competition pattern convergence, the future domestic GenAI manufacturers will rely on the cloud platform to complete model training and adjustment.
In the field of cloud platform, both domestic and overseas markets show a monopoly pattern in which the leading manufacturers occupy an absolute dominant position. In 2022, the four leading domestic manufacturers occupy 79% of the domestic market share; In the overseas market, as of the first quarter of 2023, the top three manufacturers accounted for 65% of the global market share.
The general large model requires massive data for training purposes, but the number of Chinese websites accounts for only 1.4% globally (English websites account for 54%), and the open Chinese corpus (including texts, pictures, videos and other materials) that can be used for training is often limited in quantity and uneven in quality.
At the same time, due to the reason of protecting user privacy, most of the massive data generated by domestic users on websites and mobile apps cannot be applied to large-scale model training, which adversely affects the efficiency and accuracy of model training.
Despite many limitations on training data, China’s general large model technology is still catching up with the international leading level, and the parameter scale is keeping up with the international leading level and achieving rapid improvement.
Compared with the general large model applicable to multi-fields and multi-tasks, the industry large model is more dependent on the developer’s understanding of the vertical scene and massive industry data support, in the context of limited computing power. The industry large model is more likely to become the winning track for Chinese GenAI manufacturers.
In terms of the industry model, the domestic market is showing a differentiated competitive pattern:. The industry model of a leading Internet manufacturer is rooted in the multimedia field of games, advertising, content creation. And has become a productivity tool to improve the efficiency of advertising and game scene creation;
The industrial-level knowledge enhancement model developed by the top three hyperscale cloud service providers in China, with its strong understanding of Chinese language, enables the automotive industry after-sales service and knowledge base construction, medical industry document identification, social science industry copyright protection and term management, etc.
A global leading ICT (information and communication) infrastructure and intelligent terminal provider’s Chinese large models including NLP, CV, multimodal and scientific computing are widely used in industrial logistics, new drug research and development, weather forecasting and other fields. 数字化转型网www.szhzxw.cn
(3) Basic model: The number of general large models is considerable, and the industry large model may be the way out of domestic GenAI.
In terms of general large models, as of May 2023, China has released 79 large models above the 1 billion parameter level, ranking second in the world in the number of large models released, second only to the United States. Among the top ten GenAI large model manufacturers in the world, Chinese R & D institutions and manufacturers occupy 4 seats.
The general large model requires massive data for training purposes. But the number of Chinese websites accounts for only 1.4% globally (English websites account for 54%). And the open Chinese corpus (including texts, pictures, videos and other materials) that can be used for training is often limited in quantity and uneven in quality.
At the same time, due to the reason of protecting user privacy, most of the massive data generated by domestic users on websites and mobile apps cannot be applied to large-scale model training, which adversely affects the efficiency and accuracy of model training.
Despite many limitations on training data, China’s general large model technology is still catching up with the international leading level. And the parameter scale is keeping up with the international leading level and achieving rapid improvement.
(4) Model libraries and tools: Independent manufacturers of large model tools have emerged around open source models at home and abroad.
In the overseas market, a number of independent vendors have emerged around the open source model that can be freely adjusted code, which can realize the subdivision technical functions such as model training and fine tuning, model deployment, and model application development. With the continuous development of the domestic GenAI market, various development and maintenance tools will gradually mature. 数字化转型网www.szhzxw.cn
(5) Application: Chinese startups are still in their early stages, and their focus areas are relatively concentrated.
The financing rounds of startups in the field of GenAI in China are in the forefront, concentrated between angel and Pre-A rounds. And the financing amount is mostly less than 100 million yuan. Which is in the early stage of development, and the industry has great potential.
Among them, the artificial intelligence chatbot of A global leading multinational technology company. As a quasi-unicorn enterprise, completed a billion yuan A+ round of financing in November 2022. With a post-investment valuation of about $1 billion.
In contrast, the overseas GenAI industry, the scale of scientific and creative enterprises is relatively large. And a number of unicorn enterprises have emerged in the field of subdivision application. Of which a global AI research enterprise headquartered in London is valued at up to $3.8 billion.
China’s GenAI vertical application field mainly focuses on text, image and audio and video generation, and overseas markets have a large number of GenAI-based development platforms, data analysis platforms and code writing platforms outside of the above fields. The reason is that overseas systems written in early programming languages (such as COBOL) are more. And many enterprises face higher programming labor costs. Therefore, there is a high demand for programming aid software.
At the same time, the current target customer group of China’s GenAI enterprise is mostly end users. And there are large-scale enterprise customer groups in overseas markets. With the maturity of China’s SaaS market and the increase of enterprises’ willingness to pay. It is expected that domestic text generation and image generation start-ups will rapidly expand to the enterprise customer market.
Commercial application
The commercial application of China’s GenAI industry shows two characteristics:. First, the industry distribution is concentrated, mainly concentrated in the commercial development of more mature advantages of the industry. Second, most companies are still in the early stages of actively exploring their own business models. 数字化转型网www.szhzxw.cn
(1) China’s GenAI application is concentrated in the commercial development of mature advantageous industries
Chinese GenAI enterprises make use of the advantages of domestic vertical scenarios to prioritize GenAI application in emerging industries with complete commercial infrastructure; European and American counterparts take advantage of the mature local SaaS market to apply GenAI to high-tech, communications and a variety of traditional industries (such as medical care, education, etc.), and a wider range of applications.
In China, the most rapidly developing areas of GenAI applications include e-commerce, media, entertainment and games. Especially digital virtual people and e-commerce video marketing. While most traditional industries (such as finance, energy, education, etc.) are still in the small-scale pilot stage.
GenAI applications can flourish in the Internet e-commerce platform. Because China has a high-quality e-commerce and supply chain ecology and a large consumer group. Which provides an opportunity for GenAI applications to land.
Typical industry application cases include a generative AI 3D short video content manufacturer in the video industry. A leading GenAI video large model research and development enterprise in the e-commerce industry. And a game and AI research and application institution under an Internet platform in the game industry. 数字化转型网www.szhzxw.cn
Among the above enterprises, 3D short video content manufacturers of generative AI can generate 3D video content based on text, and promote it through short video platforms such as Douyin, Kuaishou, and B Station, greatly improving the creation efficiency of short video content producers.
GenAI video large model research and development enterprises are deeply cultivating the e-commerce industry, generating virtual human anchors with AI, empowering various industries and brands of e-commerce; The Game and AI research and Application Institute uses artificial intelligence to build AI-accompanied, AI-competitive robots, and uses natural language technology to give non-player characters, so that they can interact with players in the game.
(2) The GenAI business model in the Chinese market is still being explored
At present, most of the GenAI startups in the Chinese market have just completed standardized product output and begun to enter the stage of primary commercialization exploration, and the mainstream business models in the market include cloud resource sales, model API calls, SaaS charges, material charges, etc.
Because domestic enterprises are not willing to pay for software, the market needs to be further cultivated. And enterprises have data security concerns about SaaS deployment. The business model of large-scale application of GenAI still needs to be explored.
In the European and American markets, the SaaS paid subscription model has basically matured. And a number of overseas GenAI companies have built a sustainable SaaS business model.
Risk management
While GenAI empowers various industries, it also brings negative impacts and challenges such as fairness impairment, intellectual property infringement, information leakage, malicious use, security threats, model illusion, and third-party risks. Among them, model illusion, malicious use, information leakage three major risks deserve attention. 数字化转型网www.szhzxw.cn
Malicious use, information leakage, security threats, model hallucinations
Malicious use: Improper use of model output may bring fraud, false information spread and other negative effects. China’s GenAI applications are rich, large user groups, the risk of malicious use is more prominent. But due to the domestic real-name authentication, content compliance. And fight against false information and other aspects of the implementation in place, is conducive to tracing the source of malicious use, so as to prevent and control related risks.
Information leakage: Private or sensitive information used for model training may be applied by GenAI as output content. While personal information in the artificial intelligence model storage system may also face cyber attacks and leaks. So key industries and enterprises that need to deploy data internally and are under strong supervision need to proactively prevent the risk of information leakage.
Security threats: Not only Responsible AI (RAI-Responsible AI) is receiving regulatory and corporate attention. But also SAI-Securing AI (Sai-securing AI) is gradually entering the corporate and public view. Especially the full life cycle security protection of AI systems. Effective AI threat management has a direct impact on preventing information leakage and malicious use. Recently, the National Information Security Standardization Technical Committee drafted relevant evaluation norms and standards, which gave algorithm providers guidance on the life cycle security of machine learning algorithms, 7 and can also provide reference for regulatory evaluation.
Model illusion: The model output does not conform to the facts. Model illusion is common in all GenAI models and applications at home and abroad. But because of the domestic research and development experience and technical strength of large models. There is still room for improvement. And Chinese corpus is more complex than English, so the problem of model illusion becomes more difficult.
In order to regulate the application of GenAI, China has begun to lay out AI-related regulations as early as 2022
In order to regulate the application of GenAI. China has begun to lay out AI-related regulations as early as 2022. Such as the Internet information service algorithm recommendation management regulations released in March 2022. And the Internet information service depth synthesis management regulations released in November 2022. 数字化转型网www.szhzxw.cn
In July 2023, the National Cyberspace Administration, together with the National Development and Reform Commission, the Ministry of Education, the Ministry of Science and Technology and other seven departments issued the Interim Measures for the Management of Generative Artificial Intelligence Services, which made clear provisions on data discrimination, intellectual property protection, personal privacy and portrait rights protection that may be involved in the model training process, and further improved the industry regulatory framework.
Looking overseas, the US National Institute of Standards and Technology released an AI Risk Management framework in January 2023, focusing on providing best practices for risk control in AI development, governance, and operations. The European Union passed the draft authorization of the Artificial Intelligence Act in June 2023, dividing the responsibilities and obligations of different levels of artificial intelligence, achieving hierarchical governance through risk grading, and promoting cross-industry artificial intelligence legislation to further prevent multiple risks that may be brought about by artificial intelligence. 数字化转型网www.szhzxw.cn
Conclusion
At present, the global artificial intelligence market is in the stage of rapid growth. This paper from the market size, technology stack, commercial application and risk management four dimensions. The development status and trend of GenAI industry at home and abroad were analyzed in a simple way.
Although there are big differences in domestic and foreign markets such as large model development, application layout, business model and other fields. We see that China’s GenAI industry is constantly catching up with the international leading level. It is expected that China’s GenAI related technologies and applications will gradually mature in the future. And further explore a business model suitable for its own development.
本文由数字化转型网（www.szhzxw.cn）转载而成，来源于AIGC开放社区；编辑/翻译：数字化转型网宁檬树。
免责声明: 本网站(https://www.szhzxw.cn/)内容主要来自原创、合作媒体供稿和第三方投稿，凡在本网站出现的信息，均仅供参考。本网站将尽力确保所提供信息的准确性及可靠性，但不保证有关资料的准确性及可靠性，读者在使用前请进一步核实，并对任何自主决定的行为负责。本网站对有关资料所引致的错误、不确或遗漏，概不负任何法律责任。
本网站刊载的所有内容(包括但不仅限文字、图片、LOGO、音频、视频、软件、程序等) 版权归原作者所有。任何单位或个人认为本网站中的内容可能涉嫌侵犯其知识产权或存在不实内容时，请及时通知本站，予以删除。
