【观点】恒生电子董事长刘曙峰：从大模型视角重新看金融科技

10月21日下午，恒生电子董事长刘曙峰在全球财富管理论坛•2023上海苏河湾大会的主题论坛“金融业未来趋势与发展展望”环节发表主题演讲。

刘曙峰表示，大模型实现了从垂直、专用向通用、深层次的重大突破，推动人类进入科技替代脑力劳动的数智化阶段，为金融科技发展带来了新机遇。但模型评测、模型适配、模型部署问题造成的模型选择难，建设成本过高和周期过长、自主算力支持不足、专业人才短缺造成的算力供应不足，专业准确性、合规及监管问题、数据安全、业务流程处理能力造成的应用成熟度不足，是大模型在金融行业落地面临的主要挑战。从技术角度展望，一是要推动金融行业大模型的第一个原生应用的研发，二是要面向大模型实现科技体系向数智化方向进行体系重构。未来，要围绕如何实质性改善用户体验、提高业务效率、实现利润增长等目标，推进大模型技术在金融科技领域的落地应用。

我主要从技术的角度谈谈金融业未来趋势与发展展望。 CXO UNION-CXO联盟（cxounion.cn）

讲到技术问题，毫无疑问当前最受关注的是大模型技术的最新突破。近一年来，大模型技术实现了从垂直、专用向通用、深层次的重大突破，这超出很多人预料，带有很大偶然性，称为“涌现”现象。从ChatGPT发布到现在不到一年的时间里，在不断刷新新的上限、刷新AI达到的新高度。同时，在投资方面也是空前的，从OPEN AI发布到GPT4.0，GPT5.0还在过程中，从几亿到几十亿、几百亿、上千亿，接下来新的发布有望突破万亿的参数级别。如果说400亿是涌现现象的一个起点，那么我们不知道到万亿参数的时候机器智能会给我们带来什么新的惊喜。

一、大模型推动第三次信息化浪潮

大模型带热的是算力，计算机算力需求基本每两个月翻一番，今年依然是一卡难求的状况，这是大模型不断刷新的结果。大模型是信息技术的最新突破，信息技术的进步是这几十年重大的事情，很多人认为等同工业革命。

在我看来，这背后基本经过了三个阶段：一是信息化阶段，用电脑处理一些流程性工作；二是网络化阶段，PC到移动互联网，交互方式发生了根本性变革，时间空间被压缩；三是数智化阶段，即今天大模型带来的突破。数智化阶段实际上是基于真实决策能力的替代，这就使得整个社会包括对金融行业的进步可能会产生巨大的影响。

信息革命的三次浪潮，每次都催生了伟大的企业，至少从市值上看起来达到万亿美元。企业市值达到几万亿美元速度越来越快，算力企业在这一轮最先达到。从金融科技的角度来看，这三个阶段技术的发展在中国同样出现类似状况，第一个阶段——信息化阶段，我们通过科技手段实现了突破；第二个阶段——互联网阶段，出现金融应用，改变了支付方式；第三个阶段——数智化阶段还将会有什么新突破，需要观察，我们拭目以待。

二、金融行业对大模型应用的关注和落地概况

回到具体现实情况，我们今年做了较大范围的调研。调研结果显示，金融行业对大模型行业的应用保持了极大关注，基本上所有金融机构都在尝试、调研，但是真正落地的应用还相对较少。我列了一些从投顾助手到代码开发处在不同应用场景阶段的情况，总体来看接近10%的金融机构已经在落地中或者落地了大模型应用，无论是内部还是外部。同时，大家对未来大模型应用方面的投入做了一定的规划。 CXO UNION-CXO联盟（cxounion.cn）

三、金融行业大模型应用面临的挑战与落地途径

金融大模型行业面临的挑战是，选择什么样的模型？需要考虑到模型的技术能力、合规性、适配性以及如何评测等问题。算力供应不足如何解决？以及应用层如何发展？应对系列挑战，我们提出以下三方面落地途径：

第一，选模型。我们和信通院一起发布了一个评测大模型在金融应用领域的标准，根据标准可以评测不同的模型在金融应用行业的合规性、决策质量表现情况如何等，同时发布了GPT金融行业大模型的第一个版本。

第二，混合部署。建议算力不足通过混合部署模式解决。私有化数据、核心部位需要私有化部署。公共数据和算力是可以以行业或者以公有云的方式部署，解决当前数据不足和算力不足的问题。有了算力和数据，尤其是实时的数据，可以有效支撑决策体系。 CXO UNION-CXO联盟（cxounion.cn）

第三，建场景。基于场景驱动开发大模型在金融行业的应用，如投顾助手、资产管理中的投研场景。从助手到投研报告的生成，这些技术的实际效果都已经被看到了。长期来看，借助于大模型能力金融机构的很多运营环节可能可以达到50%以上甚至达到70%以上的效率提升，垂直应用的领域会越来越广泛。 CXO UNION-CXO联盟（cxounion.cn）

此外，如何在应用过程中不断建立客户的黏性，建立数据的闭环，形成数据飞轮，模型才可以变得越用越好，越用越有力量，构成一个成长的闭环？从商业角度来看，这是竞争过程中非常关键的问题。

四、展望

以上是当前现状，我从技术角度再做一些未来展望。

第一，我们看到过去一些APP等应用的用户数达到1亿级别的时间越来越短，最新的是ChatGPT仅用了2个月时间。我们不知道金融行业基于大模型原生应用会以什么方式出现，什么时候出现。我相信，如果金融行业出现大模型原生应用的话，它仍将会是一个强大的创新并吸引众多用户使用，也会带来金融行业的业态新变化。

第二，面向大模型，需要重新建构科技体系，乃至组织架构，甚至整个行业结构，所以称之为数智化体系重构。随着数字化应用、大模型应用的深入，未来逐步会有数智化体系的重构。 CXO UNION-CXO联盟（cxounion.cn）

我想从最新、最受关注的大模型技术突破视角看金融科技和对金融业未来的展望。科技的发展最终是要落实到实质性的改善用户体验、数量级提高业务效率以及带来有质量的增长，而非仅仅只是流量。过去十年出现了大量的新技术突破：比如今天已经看到AI、大模型、虚拟现实最新的技术、脑机接口技术等等。接下来，我们需要致力于和专注于新技术如何在行业真正得到应用，改善客户体验，提高业务效率以及带来有质量的成长。这个改变也许可以称为金融科技发展的下半程。对于金融机构而言，这样的环境变化应受到每个人的关注，并尽快应用技术带来新的成长。

翻译：

Liu Shufeng, chairman of Hundsun Electronics: Re-look at financial technology from the perspective of large models

On the afternoon of October 21, Liu Shufeng, chairman of Hundsun Electronics, delivered a keynote speech at the theme forum “Future Trends and Development Prospects of the Financial Industry” at the Global Wealth Management Forum 2023 Shanghai Suhewan Conference. CXO UNION-CXO联盟（cxounion.cn）

Liu Shufeng said that the large model has achieved a major breakthrough from vertical, special to general, deep, and promoted human beings to enter the stage of digital intelligence where science and technology replace mental labor, and brought new opportunities for the development of financial technology. However, model selection is difficult due to model evaluation, model adaptation and model deployment problems, high construction cost and long cycle, insufficient support for autonomous computing power, insufficient supply of computing power caused by shortage of professional talents, and insufficient application maturity caused by professional accuracy, compliance and supervision issues, data security and business process processing capabilities are the main challenges facing the implementation of large models in the financial industry. Looking forward from the technical point of view, one is to promote the research and development of the first native application of the large model of the financial industry, and the second is to restructure the scientific and technological system in the direction of digital intelligence for the large model. In the future, it is necessary to promote the application of large model technology in the field of financial technology around how to substantially improve user experience, improve business efficiency, and achieve profit growth.

I will mainly talk about the future trend and development prospect of the financial industry from the perspective of technology.

Speaking of technical issues, there is no doubt that the latest breakthrough in large model technology is the most concerned at present. In the past year, large model technology has achieved a major breakthrough from vertical, special to general, deep, which exceeds many people’s expectations, with a great deal of chance, known as the “emergence” phenomenon. In the less than a year since ChatGPT was released, new ceilings and new heights of AI have been constantly refreshed. At the same time, it is also unprecedented in terms of investment, from the release of OPEN AI to GPT4.0, GPT5.0 is still in the process, from hundreds of millions to billions, tens of billions, hundreds of billions, and the next new release is expected to break through the parameter level of trillions. If 40 billion is a starting point for emergence, then we don’t know what new surprises machine intelligence will bring us when we reach trillion parameters.

First, the big model promotes the third wave of informatization

The big model with hot computing power, computer computing power demand basically doubles every two months, this year is still a card difficult to find the situation, which is the result of the big model refresh. The Big model is the latest breakthrough in information technology, the progress of information technology is the major event of these decades, many people think of it as equivalent to the industrial revolution. CXO UNION-CXO联盟（cxounion.cn）

In my opinion, there are basically three stages behind this: First, the information stage, using computers to deal with some process work; The second is the networking stage, PC to mobile Internet, the way of interaction has undergone fundamental changes, time and space are compressed; The third is the stage of number intelligence, that is, the breakthrough brought by today’s big model. The stage of digital intelligence is actually based on the substitution of real decision-making ability, which makes the progress of the whole society, including the financial industry, potentially have a huge impact.

Each of the three waves of the information revolution has produced great companies that, at least in terms of market capitalization, appear to be worth trillions of dollars. The market value of enterprises reaching several trillion dollars is getting faster and faster, and the computing power enterprises are the first to reach it in this round. From the perspective of financial technology, the development of technology in these three stages also appears similar situation in China, the first stage – the information stage, we have achieved breakthroughs through scientific and technological means; The second stage – the Internet stage, the emergence of financial applications, change the way of payment; The third stage – the number of intelligent stage will have any new breakthroughs, we need to observe, we will wait and see. CXO UNION-CXO联盟（cxounion.cn）

Second, the financial industry’s attention to the application of large models and landing overview

Going back to the actual situation, we did a wide range of research this year. The survey results show that the financial industry has maintained great attention to the application of the large model industry, and basically all financial institutions are trying and researching, but the real application is still relatively few. I listed a number of different application scenarios from investment counsel to code development, and overall, close to 10% of financial institutions have already implemented or implemented large model applications, whether internal or external. At the same time, we have made certain plans for the future investment in the application of large models.

Iii. Challenges and landing approaches of large model application in the financial industry

The challenge facing the financial big model industry is, what model to choose? Issues such as the technical capability, compliance, suitability, and how to measure the model need to be considered. How to solve the shortage of computing power supply? And how does the application layer evolve? In response to a series of challenges, we propose the following three approaches:

First, choose the model. Together with the ICT Center, we released a standard to evaluate the financial application of large models, according to which different models can be evaluated in the financial application industry compliance, decision quality performance, etc., and released the first version of the GPT financial industry large model. CXO UNION-CXO联盟（cxounion.cn）

Second, hybrid deployment. Hybrid deployment mode is recommended to solve the computing power shortage. Privatized data and core areas need to be privatized. Public data and computing power can be deployed in the industry or in the public cloud to solve the current problem of insufficient data and computing power. With computing power and data, especially real-time data, it can effectively support the decision-making system.

Third, build the scene. Based on scenario-driven development of large model applications in the financial industry, such as investment assistant, investment research scenarios in asset management. From assistants to the generation of research reports, the practical effects of these technologies are already being seen. In the long run, with the help of large model capacity, many operational links of financial institutions may be able to achieve more than 50% or even more than 70% efficiency improvement, and the fields of vertical application will be more and more extensive.

In addition, how to continuously establish customer stickiness in the application process, establish a closed loop of data, form a data flywheel, so that the model can become better and better with more and more power, and form a growth of the closed loop? From a business point of view, this is a very critical issue in the competitive process. CXO UNION-CXO联盟（cxounion.cn）

Iv. Outlook

Above is the current situation, I will make some future outlook from the technical point of view.

First, we have seen that in the past, some apps and other applications have reached the 100 million user level in a shorter and shorter time, the latest is ChatGPT only took 2 months. We don’t know how and when big model-based native applications will emerge in the financial industry. I believe that if there is a large model native application in the financial industry, it will still be a powerful innovation and attract many users, and will also bring new changes in the business format of the financial industry.

Second, facing the large model, it is necessary to rebuild the scientific and technological system, even the organizational structure, and even the entire industry structure, so it is called the reconstruction of the digital intelligent system. With the deepening of digital application and large model application, there will be some reconstruction of intelligent system in the future.

I want to look at fintech and the future of finance from the perspective of the latest and most talked about big model technology breakthroughs. Technology is ultimately about substantial improvements in user experience, orders of magnitude more business efficiency, and quality growth, not just traffic. The last decade has seen a number of new technological breakthroughs: today we see AI, large models, the latest technologies in virtual reality, brain-computer interface technology, and so on. Next, we need to be committed to and focused on how the new technology can really be applied in the industry, improve the customer experience, increase business efficiency and lead to quality growth. This change might be called the second half of fintech’s development. For financial institutions, such changes in the environment should be of concern to everyone, and the application of technology as soon as possible to bring new growth. CXO UNION-CXO联盟（cxounion.cn）

由CXO UNION-CXO联盟（cxounion.cn）转载而成，来源于中国财富管理50人论坛；编辑/翻译：CXO UNIONCXO联盟小U。

如需加入CXO UNION（CXO联盟）高管社群，请联系社群小伙伴哦~

免责声明: 本网站(http://www.cxounion.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。