AI and Central Banking
Thank you for the opportunity to speak to you today.
感谢您今天有机会与您交谈。
1 It is an honor and a pleasure to be here with you in Singapore, a crossroads for global trade and finance, to discuss the transformational nature of artificial intelligence (AI).
1 我很荣幸也很高兴来到新加坡这个全球贸易和金融的十字路口，与大家一起讨论人工智能 (AI) 的变革本质。
Like other central banks around the world, including the Monetary Authority of Singapore, we at the Federal Reserve have been exploring the use of AI in our operations for quite some time as well as considering the implications of AI's adoption for the financial sector and the broader economy.
与包括新加坡金融管理局在内的世界其他中央银行一样，美联储长期以来一直在探索在业务中使用人工智能，并考虑人工智能的采用对金融部门和更广泛的经济的影响。
At the Fed, we've been interested in the effects of AI on the economy and its role in the financial system for decades.
在美联储，几十年来我们一直对人工智能对经济的影响及其在金融体系中的作用感兴趣。
Remarkably, in 1997, Governor Susan Phillips delivered a speech noting the use of AI in consumer loan underwriting.
值得注意的是，1997 年，州长苏珊·菲利普斯 (Susan Phillips) 发表演讲，指出人工智能在消费贷款承销中的应用。
2 In the past year, no fewer than seven speeches by Fed Governors have had "AI" in the title.
2 过去一年，美联储理事的演讲不少于七次，标题中包含“人工智能”。
Today I'll discuss the opportunities presented by AI relevant for central bankers as well as the risks it may pose that policymakers should consider.
今天，我将讨论人工智能为央行行长带来的机遇，以及政策制定者应考虑的可能带来的风险。
I will leave you with three main takeaways.
我将给您留下三个主要收获。
The State of AI Innovation and Deployment My first main point is that while AI is a big deal that will transform economies, there are a range of outcomes for how it could do so.
人工智能创新和部署的现状我的第一个要点是，虽然人工智能是一件将改变经济的大事，但它如何实现这一目标会产生一系列结果。
AI—algorithms that mimic human thought, communication, and choices—has been with us for decades, but AI entered a new era with the launch of ChatGPT in late 2022.
人工智能（模仿人类思维、沟通和选择的算法）已经陪伴我们数十年，但随着 2022 年底 ChatGPT 的推出，人工智能进入了一个新时代。
Generative AI (GenAI) captured our imagination with convincing conversations in which it is possible to go deep on a wide range of topics.
生成式人工智能（GenAI）通过令人信服的对话激发了我们的想象力，在对话中可以深入探讨广泛的主题。
Earlier forms of AI were often the bailiwick of digitally native companies, but GenAI is spreading rapidly through the economy.
早期形式的人工智能通常是数字原生公司的管辖范围，但 GenAI 正在经济中迅速传播。
As of 2024, three in four large companies were using GenAI, though some report it has yet to improve their bottom line.
截至 2024 年，四分之三的大公司正在使用 GenAI，尽管有些人报告称它尚未改善他们的利润。
3 Smaller companies have been slower to adopt GenAI, with adoption rates reported in the high single digits, albeit with a high degree of heterogeneity among sectors.
3 规模较小的公司采用 GenAI 的速度较慢，尽管各行业之间存在高度异质性，但采用率仍处于较高个位数。
Also, one could surmise, based on other surveys of individuals, that work-related use is more widespread among employees than their CEOs realize.
此外，根据其他对个人的调查，人们可以推测，员工中与工作相关的使用比他们的首席执行官意识到的更为广泛。
4
4
A recent survey by the Federal Reserve Bank of New York showed that firms plan to retrain their workforces to take advantage of AI to enhance productivity, with widespread layoffs limited.
纽约联邦储备银行最近的一项调查显示，企业计划对员工进行再培训，以利用人工智能来提高生产力，但大范围裁员有限。
5 But survey respondents also report that AI has led firms to scale back hiring, a development that may be contributing to the recent low levels of job creation in the U.S. economy, a concern for many workers and particularly for newer entrants to the job market.
5 但调查受访者还报告说，人工智能已导致企业缩减招聘规模，这一发展可能导致美国经济近期创造的就业机会水平较低，这是许多工人，特别是新进入就业市场的人的担忧。
Taking a step back, as I have noted in the past, I see two basic scenarios for how AI can transform the economy.
退后一步，正如我过去所指出的，我看到了人工智能如何改变经济的两个基本场景。
6 In the first scenario, there is incremental adoption of GenAI that augments existing tasks and jobs.
6 在第一个场景中，GenAI 的逐步采用增强了现有的任务和工作。
In the second scenario, a revolution occurs.
在第二种情况下，一场革命发生了。
GenAI transforms the nature of work and leisure, boosting the efficiency of research and development, remaking industries, and creating firms with new—perhaps radically new—business models.
GenAI 改变了工作和休闲的性质，提高了研发效率，重塑了行业，并创建了具有新的（也许是全新的）商业模式的公司。
Right now, it is difficult to predict which scenario (or perhaps one or more intermediate scenarios) will come to pass.
目前，很难预测哪种情况（或者可能是一种或多种中间情况）将会发生。
We can already see incremental change as GenAI is increasingly integrated with standard workplace software.
随着 GenAI 越来越多地与标准工作场所软件集成，我们已经可以看到渐进式的变化。
With its natural language interface, GenAI is inherently user friendly, so few workers need special skills or unusually onerous training to use it.
凭借其自然语言界面，GenAI 本质上是用户友好的，因此很少有工人需要特殊技能或异常繁重的培训来使用它。
At the same time, some start-ups have a more revolutionary flavor because they are centered on AI from the outset.
与此同时，一些初创企业由于从一开始就以人工智能为中心，因此更具革命性。
One indicator of how the labor market is evolving toward deeper integration with AI is the skills mentioned in job postings.
劳动力市场如何与人工智能更深入融合的指标之一是职位发布中提到的技能。
While overall the share of job listings that mention AI-related skills is small—about 5 percent—in the information sector, it is about 20 percent.
虽然总体而言，在信息行业，提及人工智能相关技能的职位列表所占比例很小（约 5%），但这一比例约为 20%。
The financial sector, where firms are always looking for a technological edge, is not far behind, with 1 in 10 job postings mentioning AI.
企业一直在寻求技术优势的金融行业也不甘落后，十分之一的职位发布中提到了人工智能。
7 So we can see that the skills needed in some key sectors are already changing.
7 因此我们可以看到，一些关键领域所需的技能已经发生了变化。
The speed of that change is likely to increase.
这种变化的速度可能会加快。
If the AI changes happen gradually, workers and firms will have time to adjust, but if they happen rapidly, there may be significant dislocations in the short term.
如果人工智能的变化是逐渐发生的，工人和企业将有时间进行调整，但如果发生得很快，短期内可能会出现严重的混乱。
A massive wave of data center investment has begun, pointing to signs of confidence among leading AI companies that the use of AI at scale throughout the economy is just around the corner.
大规模的数据中心投资浪潮已经开始，这表明领先的人工智能公司有信心在整个经济中大规模使用人工智能指日可待。
If they're right and AI is useful enough to keep what is currently projected to be $3 trillion of new data center capacity utilized effectively, we can expect significant changes in economies.
如果他们是对的，并且人工智能足够有用，可以有效利用目前预计的 3 万亿美元的新数据中心容量，我们可以预期经济将发生重大变化。
Investment in capital generally raises labor productivity and offers the potential for higher output growth without pressure on inflation over the longer term.
资本投资通常会提高劳动生产率，并提供更高产出增长的潜力，而不会造成长期通胀压力。
As I have discussed in previous remarks, if these changes are significant, they can also affect the conduct of monetary policy.
正如我在之前的讲话中所讨论的，如果这些变化很大，它们也会影响货币政策的实施。
8 Of course, it may be the case instead that investment exceeds short-term demand, in which case there may be losses and adjustments to the AI sector.
8 当然，也可能出现投资超过短期需求的情况，这种情况下，人工智能行业可能会出现亏损和调整。
AI and the Financial Sector The second key point I would like to make is that the financial sector is adopting AI quickly, and while there are many benefits to this adoption, the risks will need to be managed carefully.
人工智能和金融部门我想说的第二个要点是，金融部门正在迅速采用人工智能，虽然这种采用有很多好处，但需要谨慎管理风险。
So far, AI adoption in the financial sector appears to be most concentrated in areas that can enhance operational efficiency, including applications that involve text analysis, classification, and information search inside the firm, as well as customer-facing functions.
到目前为止，人工智能在金融领域的采用似乎最集中在可以提高运营效率的领域，包括涉及公司内部文本分析、分类和信息搜索的应用程序，以及面向客户的功能。
These incremental improvements to common business functions are a key reason to be hopeful about AI raising labor productivity in that sector.
这些对常见业务功能的逐步改进是人们对人工智能提高该领域劳动生产率抱有希望的一个关键原因。
At the same time, there is significant investment in experimentation with AI for core functions for financial services.
与此同时，金融服务核心功能的人工智能实验也投入了大量资金。
Data-driven financial-sector-specific tasks, including credit decision support, fraud detection, and trading are using AI-specific tools.
数据驱动的金融部门特定任务，包括信贷决策支持、欺诈检测和交易，正在使用人工智能特定工具。
Ensuring that AI is used appropriately for these functions faces appreciable challenges.
确保人工智能适当地用于这些功能面临着相当大的挑战。
First, the amount of organizational change needed by financial services firms to utilize GenAI may be substantial.
首先，金融服务公司为了利用 GenAI 需要进行大量的组织变革。
History suggests progress may be slow.
历史表明进展可能会很缓慢。
Adoption of machine learning, an AI technology that preceded GenAI, was concentrated in firms that were highly digitized from their founding—and even in those cases, adoption was a long process.
机器学习是一种先于 GenAI 的人工智能技术，其采用主要集中在自成立之日起就高度数字化的公司，即使在这些情况下，采用也是一个漫长的过程。
9 Fintech firms organized to exploit AI from their founding can play a key role in driving efficiency forward in the sector, providing services to the incumbent firms.
9 从成立之初就致力于利用人工智能的金融科技公司可以在推动该行业效率、为现有公司提供服务方面发挥关键作用。
10 But productivity may even decrease in the short term, as heavy investments in business-process improvements take time to play out to productivity gains.
10 但生产率甚至可能在短期内下降，因为对业务流程改进的大量投资需要时间才能发挥生产率的提高。
A second challenge is the practical constraints of rushing into AI for core business activities in the financial sector, as firms need to ensure that the resulting processes and outcomes are consistent with relevant laws and appropriate risk management.
第二个挑战是金融领域核心业务活动采用人工智能的实际限制，因为企业需要确保由此产生的流程和结果符合相关法律和适当的风险管理。
Large institutions are exploring the use of GenAI, including agentic AI, in their financial models—but doing so requires care.
大型机构正在探索在其财务模型中使用 GenAI，包括代理人工智能，但这样做需要谨慎。
To successfully leverage the potential of GenAI on a sustainable basis, decisions based on those models must be well controlled, numerically and legally precise, explainable, and replicable.
为了在可持续的基础上成功地利用 GenAI 的潜力，基于这些模型的决策必须得到良好的控制、在数字和法律上精确、可解释和可复制。
AI developers still struggle to some extent with all of those criteria.
人工智能开发人员在某种程度上仍然难以满足所有这些标准。
We need to reduce the risk that AI reinforces biases in consumer lending.
我们需要降低人工智能加剧消费贷款偏见的风险。
And we also need to guard against the risks that could result from the use of AI in financial markets.
我们还需要防范人工智能在金融市场中使用可能带来的风险。
For example, profit maximization by AI-powered trading algorithms may result in tacit collusion, market manipulation, or trading strategies that result in significant market volatility or even systemic risk.
例如，人工智能交易算法的利润最大化可能会导致默契合谋、市场操纵或导致市场大幅波动甚至系统性风险的交易策略。
11
11
We will need innovation that is responsive to these risks to see additional advances in the use of AI for a broad array of core financial services functions.
我们需要应对这些风险的创新，以看到人工智能在广泛的核心金融服务功能中的使用取得更多进展。
AI and Central Banking A third and final point I would like to leave you with is that central banks, including the Fed, need to keep up with AI by increasing our speed of adoption for our own operations.
人工智能和中央银行 我想留给你们的第三点也是最后一点是，包括美联储在内的中央银行需要通过提高我们在自己的业务中采用人工智能的速度来跟上人工智能的步伐。
The nature of central banking work is inherently careful, considered, and measured when evaluating anything new.
央行工作的本质是在评估任何新事物时都非常谨慎、深思熟虑和衡量。
This is particularly true for any new technologies.
对于任何新技术来说尤其如此。
12 But it seems clear already that the many advantages offered by AI could assist central banks in at least some of their operations, and the speed at which this technology is moving makes it appropriate to proactively engage in using AI for our own operations.
12 但似乎已经很清楚的是，人工智能提供的许多优势至少可以协助中央银行的部分业务，而且这项技术的发展速度使得我们有必要积极主动地将人工智能用于我们自己的业务。
That is why the Federal Reserve is using AI, where appropriate, to increase staff efficiency and effectiveness.
这就是美联储在适当情况下使用人工智能来提高员工效率和效力的原因。
My view is that GenAI is a transformative technology that central banks need to remain engaged with to ensure an ability to execute as technology evolves.
我的观点是，GenAI 是一项变革性技术，央行需要继续参与其中，以确保随着技术的发展而有执行能力。
As we push ahead on efficiency gains, it is important that we leverage the right tools for the task at hand, recognizing that GenAI is not always the best choice.
当我们推进效率提升时，重要的是我们要利用正确的工具来完成手头的任务，并认识到 GenAI 并不总是最佳选择。
Some of the challenges that we face can be addressed by robotic process automation or traditional AI methods.
我们面临的一些挑战可以通过机器人流程自动化或传统的人工智能方法来解决。
These are the same kinds of questions that every business and organization considering AI should be weighing.
这些问题是每个考虑人工智能的企业和组织都应该权衡的。
At the Federal Reserve, we have focused on ensuring we can leverage AI capabilities by establishing an AI program and governance framework for the use of AI technologies.
在美联储，我们致力于通过建立人工智能计划和使用人工智能技术的治理框架来确保我们能够利用人工智能能力。
13 We are taking an enterprise-wide approach of learning-by-doing and broadly adopting high-value uses of GenAI, such as writing, coding, and research activities.
13 我们正在全企业范围内采取边干边学的方法，并广泛采用 GenAI 的高价值用途，例如写作、编码和研究活动。
We have taken a "hands on keys" approach to having staff engage with it.
我们采取了“动手操作”的方法来让员工参与其中。
We are identifying the business processes that can be improved and transformed with the technology.
我们正在确定可以通过该技术改进和转变的业务流程。
One internal application of GenAI I am particularly excited about that helps us achieve all these goals is technology modernization.
让我特别兴奋的 GenAI 内部应用之一是技术现代化，它可以帮助我们实现所有这些目标。
We are applying GenAI-enabled tools within clear guardrails to translate legacy code, generate unit tests, and accelerate cloud migration.
我们正在清晰的护栏内应用支持 GenAI 的工具来翻译遗留代码、生成单元测试并加速云迁移。
So far, the result of this usage is faster delivery, improved quality, and enhanced developer experience.
到
And it will likely mean better outcomes in support of the American people.
这可能意味着支持美国人民会取得更好的结果。
Given AI's current and prospective role in economic activity, we are devoting the necessary resources to understanding it, including by analyzing not only AI's economic and financial implications, but also exploring how AI can enhance our financial stability work, strengthen supervisory and regulatory capabilities, and ensure the smooth functioning of our payment systems.
鉴于人工智能当前和未来在经济活动中的作用，我们正在投入必要的资源来理解它，包括不仅分析人工智能的经济和金融影响，还探索人工智能如何加强我们的金融稳定工作，加强监督和监管能力，并确保我们的支付系统的顺利运行。
These are just some of the ways that the Fed, like other organizations, is using AI to make us more productive and capable.
这些只是美联储与其他组织一样利用人工智能提高我们的生产力和能力的一些方式。
These efforts may also help us understand the effect of AI on the economy, the banking system, and the payment system.
这些努力也可能有助于我们了解人工智能对经济、银行系统和支付系统的影响。
That task will be a major job for central banks in the years ahead.
这项任务将是未来几年央行的一项主要工作。
AI has the potential to fundamentally change the economy and society.
人工智能有潜力从根本上改变经济和社会。
And as central bankers, we need to keep up.
作为央行行长，我们需要跟上。
1.
1.
The views expressed here are my own and are not necessarily those of my colleagues on the Federal Reserve Board or the Federal Open Market Committee.
这里表达的观点是我自己的观点，不一定是我在美联储委员会或联邦公开市场委员会的同事的观点。
Return to text
返回正文
2.
2.
See Susan M.
参见苏珊·M.
Phillips (1997), " Risk Management ," speech delivered at the Asset/Liability and Treasury Management Conference of the Bank Administration Institute, Chicago, November 4.
Phillips (1997)，“风险管理”，在银行管理学院资产/负债和资金管理会议上发表的演讲，芝加哥，11 月 4 日。
Return to text
返回正文
3.
3.
See Aditya Challapally, Chris Pease, Ramesh Raskar, and Pradyumna Chari (2025), "The GenAI Divide: State of AI in Business 2025," MIT NANDA Report, July, https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Business_2025_Report.pdf. There are a range additional surveys with various results; see, for example, Richard Horton, Jan Michalski, Stacey Winters, Douglas Gunn, and Jennifer Holland (2025), "AI ROI: The Paradox of Rising Investment and Elusive Returns," October 22, https://www.deloitte.com/uk/en/issues/generative-ai/ai-roi-the-paradox-of-rising-investment-and-elusive-returns.html. Return to text
请参阅 Aditya Challapally、Chris Pease、Ramesh Raskar 和 Pradyumna Chari (2025)，“GenAI 鸿沟：2025 年人工智能商业状况”，麻省理工学院 NANDA 报告，7 月，https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Business_2025_Report.pdf。还有一系列其他调查，结果各不相同；例如，参见 Richard Horton、Jan Michalski、Stacey Winters、Douglas Gunn 和 Jennifer Holland (2025)，“人工智能投资回报率：不断增长的投资与难以捉摸的回报的悖论”，10 月 22 日， https://www.deloitte.com/uk/en/issues/generative-ai/ai-roi-the-paradox-of-rising-investment-and-elusive-returns.html。返回正文
4.
4.
On adoption by large firms, see McKinsey & Company (2025), "The State of AI: Agents, Innovation, and Transformation," November 5, https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai. On adoption by smaller firms, see the Business Trends and Outlook Survey from the U.S. Census Bureau, which is available on its website at https://www.census.gov/programs-surveys/btos.html .
关于大公司的采用，请参阅 McKinsey & Company (2025)，“人工智能的现状：代理、创新和转型”，11 月 5 日，https://www.mckinsey.com/capability/quantumblack/our-insights/the-state-of-ai。关于小型公司的采用情况，请参阅美国人口普查局的商业趋势和前景调查，该调查可在其网站 https://www.census.gov/programs-surveys/btos.html 上获取。
On adoption by individuals, see Alexander Bick, Adam Blandin, and David J.
关于个人收养，请参阅 Alexander Bick、Adam Blandin 和 David J.
Deming (2024), " The Rapid Adoption of Generative AI ," Working Paper Series 32966 (Cambridge, Mass.: National Bureau of Economic Research, September; revised February 2025).
Deming (2024)，“生成式 AI 的快速采用”，工作论文系列 32966（马萨诸塞州剑桥：国家经济研究局，9 月；2025 年 2 月修订）。
Return to text
返回正文
5.
5.
See Ben Hyman, Jaison R.
参见本·海曼、杰森·R.
Abel, Natalia Emanuel, Nick Montalbano, and Richard Deitz (2025), " Are Businesses Scaling Back Hiring Due to AI?
Abel、Natalia Emanuel、Nick Montalbano 和 Richard Deitz (2025)，“企业是否因人工智能而缩减招聘规模？
" Federal Reserve Bank of New York, Liberty Street Economics (blog), September 4.
“纽约联邦储备银行，Liberty Street Economics（博客），9 月 4 日。
Other surveys have shown similar results; see, for example, Jeremy Korst, Stefano Puntoni, and Prasanna Tambe (2025), " Accountable Acceleration: Gen AI Fast-Tracks Into the Enterprise (PDF) ," 2025 Report, October 29.
其他调查也显示了类似的结果；例如，请参阅 Jeremy Korst、Stefano Puntoni 和 Prasanna Tambe (2025)，“负责任的加速：Gen AI 快速进入企业 (PDF)”，2025 年报告，10 月 29 日。
Return to text
返回正文
6.
6.
See Michael S.
参见迈克尔·S.
Barr (2025), " Artificial Intelligence: Hypothetical Scenarios for the Future ," speech delivered at the Council on Foreign Relations, New York, February 18.
巴尔 (2025)，“人工智能：未来的假设情景”，2 月 18 日在纽约外交关系委员会发表的演讲。
Return to text
返回正文
7.
7.
Job-posting statistics are based on the classification by Lightcast and are calculated using the methodology developed in Daron Acemoglu, David Autor, Jonathon Hazell, and Pascual Restrepo, "Artificial Intelligence and Jobs: Evidence from Online Vacancies," Journal of Labor Economics, vol. 40 (April), pp.
职位发布统计数据基于 Lightcast 的分类，并使用 Daron Acemoglu、David Autor、Jonathon Hazell 和 Pascual Restrepo 所开发的方法进行计算，“人工智能与工作：来自在线职位空缺的证据”，《劳动经济学杂志》，第 1 卷。 40（4 月），第 15 页。
S293–340.
S293-340。
Data are available by subscription from Lightcast at https://lightcast.io. Return to text
数据可通过 Lightcast 订阅获取，网址为：https://lightcast.io。返回正文
8.
8.
See Michael S.
参见迈克尔·S.
Barr (2025), " Artificial Intelligence and the Labor Market: A Scenario-Based Approach ," speech delivered at the Reykjavík Economic Conference 2025, Central Bank of Iceland, Reykjavík, Iceland, May 9.
Barr (2025)，“人工智能和劳动力市场：基于场景的方法”，在 2025 年雷克雅未克经济会议上发表的演讲，冰岛中央银行，冰岛雷克雅未克，5 月 9 日。
Return to text
返回正文
9.
9.
See Timothy Bresnahan (2024), "What Innovation Paths for AI to Become a GPT?" Journal of Economics & Management Strategy, vol. 33 (Summer), pp.
请参阅 Timothy Bresnahan (2024)，“人工智能成为 GPT 的创新路径是什么？”经济与管理战略杂志，卷。 33（夏季），第 33 页。
305–16.
305–16。
Return to text
返回正文
10.
10.
See Michael S.
参见迈克尔·S.
Barr (2025), " AI, Fintechs, and Banks ," speech delivered at the Federal Reserve Bank of San Francisco, San Francisco, April 4.
Barr (2025)，“人工智能、金融科技和银行”，4 月 4 日在旧金山联邦储备银行发表的演讲。
Return to text
返回正文
11.
11.
The most recent Financial Stability Report is available on the Federal Reserve Board's website at https://www.federalreserve.gov/publications/files/financial-stability-report-20251107.pdf .
最新的金融稳定报告可在联邦储备委员会网站上获取：https://www.federalreserve.gov/publications/files/financial-stability-report-20251107.pdf。
Return to text
返回正文
12.
12.
A report from the Bank of International Settlements notes that central bankers have the prospect of improving data quality, enhancing operations, and improving decisionmaking with the use of AI and provides a framework for considering questions of governance and risk management when doing so; see Bank of International Settlements (2025), " Governance of AI Adoption at Central Banks (PDF) ," January.
国际清算银行的一份报告指出，央行行长有望利用人工智能来提高数据质量、加强运营和改进决策，并为在这样做时考虑治理和风险管理问题提供了一个框架；请参阅国际清算银行 (2025)，“中央银行采用人工智能的治理 (PDF)”，1 月。
Return to text
返回正文
13.
13.
See the "Board of Governors of the Federal Reserve System Compliance Plan for OMB Memorandum M-25-21," which is available on the Federal Reserve's website at https://www.federalreserve.gov/publications/compliance-plan-for-OMB-memorandum-m-25-21.htm .
请参阅“联邦储备系统理事会 OMB 备忘录 M-25-21 合规计划”，该计划可在美联储网站 https://www.federalreserve.gov/publications/compliance-plan-for-OMB-memorandum-m-25-21.htm 上获取。
Return to text
返回正文