这是用户在 2024-3-19 15:48 为 https://deloitte.wsj.com/cfo/facing-the-power-hungry-side-of-generative-ai-2519e791 保存的双语快照页面,由 沉浸式翻译 提供双语支持。了解如何保存?
  • Most Popular News

Facing the Power-Hungry Side of Generative AI
面对生成式 AI 耗电的一面

The environmental impact of the disruptive technology may require a new push toward clean and efficient energy use

Facing the Power Hungry Side of Generative AI

Enthusiasm about generative AI investment has helped to fuel a surge in equity prices in the United States, with an expectation that investment will grow rapidly in the coming few years and potentially have a favorable impact on productivity. Yet another aspect of generative AI has drawn less attention, namely its high power usage that potentially could cause a sizable increase in carbon emissions.

Ira Kalish
Ira Kalish 艾拉·卡利什

Granted, generative AI has the potential to assist in minimizing carbon emissions, among other benefits. The technology can be used to detect pipeline leaks and contribute to the development of new materials and processes that are more energy efficient. Still, the use of billions of devices connected to the internet, and especially the use of data centers, with the air conditioning needed to cool servers, will likely lead to considerable consumption of electricity.

It is reported1 that connected devices will account for 3.5% of global carbon emissions in 2025, and possibly 14% of global emissions in 2040. The future increase would likely be due to generative AI as well as increased usage of devices in emerging countries. In addition, a study2 by scientists at Cornell University found that training large language models, which must be done continually, consumes a large amount of electricity, thereby pumping more carbon into the atmosphere. 
据悉 1 ,到2025年,联网设备将占全球碳排放量的3.5%,到2040年可能占全球排放量的14%。未来的增长可能是由于生成式人工智能以及新兴国家设备使用量的增加。此外,康奈尔大学科学家的一项研究 2 发现,训练大型语言模型(必须持续进行)会消耗大量电力,从而将更多的碳泵入大气中。

What can be done? First, the world will need to shift toward producing electricity using clean energy. This would include solar, wind, hydroelectric, nuclear, and hydrogen power. Second, producers of AI-related services will have to engage in more efficient usage of energy. A recent study3 offered suggestions on how to accomplish this. Yet unless governments compel companies to follow these suggestions, it will not be clear whether efficiencies are being adopted. 
可以做些什么?首先,世界需要转向使用清洁能源发电。这将包括太阳能、风能、水力发电、核能和氢能。其次,人工智能相关服务的生产者将不得不更有效地利用能源。最近的一项研究 3 就如何实现这一目标提出了建议。然而,除非政府迫使公司遵循这些建议,否则是否正在采用效率尚不清楚。

As a resident of California, I recall that, about a decade ago, we faced a temporary yet serious shortage of electricity. In part it was due to a rapid increase in demand, almost entirely from the rapid rollout of new servers and data centers in the state’s large technology industry. This was, I think, a warning about what could happen globally in the years to come. 

Emerging Countries and the Carbon Calculus

Meanwhile, despite all the efforts underway to reduce carbon emissions, the International Energy Agency (IEA) reports 4 that global emissions hit a record level in 2023, up 1.1% from the previous year. Moreover, the increase was entirely due to emerging countries. Emissions in advanced economies fell 4.5% from 2022 to 2023, hitting a 50-year low. Emissions fell 9% in Europe and 4.1% percent in the United States. In advanced economies, renewables and nuclear accounted for 50% of electricity generation. 
与此同时,尽管为减少碳排放做出了所有努力,但国际能源署 (IEA) 报告称 4 ,2023 年全球排放量创下历史新高,比上一年增长 1.1%。此外,这一增长完全是由于新兴国家。从2022年到2023年,发达经济体的排放量下降了4.5%,创下50年来的新低。欧洲的排放量下降了9%,美国的排放量下降了4.1%。在发达经济体,可再生能源和核能占发电量的50%。

Of greater concern is that emissions were up 5.2% in China and 7% in India. The IEA says that 40% of the increase in global emissions last year was due to burning fossil fuels as a replacement for hydropower. Climate-induced droughts led to a decline in the ability to produce hydropower. Going forward, this could be a serious issue. 

The good news is that, in the past 10 years, global emissions grew at an annual rate of just 0.5%, the lowest rate since the 1930s. However, that was due, in part, to the disruptive impact of the pandemic, which suppressed travel. Going forward, the IEA anticipates big increases in the use of clean energy to produce electricity, and in the use of electric vehicles.

—by Ira Kalish, chief global economist, Deloitte Touche Tohmatsu Limited 
——作者:Ira Kalish,德勤关黄有限公司首席全球经济学家

  1. The Guardian: “‘Tsunami of data’ could consume one fifth of global electricity by 2025,” December 2017
    《卫报》:“到 2025 年,'数据海啸'可能消耗全球五分之一的电力”,2017 年 12 月
  2. Cornell University: “Carbon Emissions and Large Neural Network Training,” April 2021
    康奈尔大学:“碳排放和大型神经网络训练”,2021 年 4 月
  3. Google Research: “Good News About the Carbon Footprint of Machine Learning Training,” December 2022
    Google Research:“关于机器学习训练碳足迹的好消息”,2022 年 12 月
  4. The Wall Street Journal: “Global CO2 Emissions Hit Record High in 2023, IEA Says,” March 2024
    《华尔街日报》:“国际能源署称,2023 年全球二氧化碳排放量创历史新高”,2024 年 3 月
Published on  Mar 16, 2024 at 3:00 AM
发表于 2024年3月16日 上午3:00

This publication contains general information only and Deloitte is not, by means of this publication, rendering accounting, business, financial, investment, legal, tax, or other professional advice or services. This publication is not a substitute for such professional advice or services, nor should it be used as a basis for any decision or action that may affect your business. Before making any decision or taking any action that may affect your business, you should consult a qualified professional advisor.

Deloitte shall not be responsible for any loss sustained by any person who relies on this publication.

About Deloitte 关于德勤

Deloitte refers to one or more of Deloitte Touche Tohmatsu Limited, a UK private company limited by guarantee (“DTTL”), its network of member firms, and their related entities. DTTL and each of its member firms are legally separate and independent entities. DTTL (also referred to as “Deloitte Global”) does not provide services to clients. In the United States, Deloitte refers to one or more of the US member firms of DTTL, their related entities that operate using the “Deloitte” name in the United States and their respective affiliates. Certain services may not be available to attest clients under the rules and regulations of public accounting. Please see www.deloitte.com/about to learn more about our global network of member firms.
德勤是指英国私人担保有限公司(“DTTL”)德勤有限公司(Deloitte Touche Tohmatsu Limited)的一家或多家、其成员所网络及其相关实体。DTTL及其每个成员公司在法律上都是独立的实体。DTTL(也称为“德勤全球”)不向客户提供服务。在美国,德勤是指DTTL的一家或多家美国成员所、在美国使用“德勤”名称运营的关联实体及其各自的关联公司。根据公共会计的规则和条例,某些服务可能无法提供给鉴证客户。请参阅 www.deloitte.com/about 以了解有关我们全球成员所网络的更多信息。

Copyright © 2024 Deloitte Development LLC. All rights reserved.
版权所有 © 2024 Deloitte Development LLC.保留所有权利。

What to read next...


Deloitte’s Executive Perspectives dives deeper into critical business issues to deliver timely and actionable content to help support decision-making and build careers. Through research, technology perspectives and analyses, interviews, and more, Executive Perspectives for CFOs provides finance executives a customized resource to help them address the strategic, operational, and regulatory issues they face in managing their organizations and careers.Learn more about Deloitte's executive programs.