生成式 AI 处理非英语文本耗能更多

由 Lisa Lock 编辑,Andrew Zinin 审校
本文已按照 Science X 的编辑流程与政策进行审核。编辑在确保内容可信度的同时,重点标出了以下属性:
添加到首选来源
一张图表,显示大语言模型在回应用户提示时,平均所需的能耗——涉及全球使用最广泛的八种语言,以及若干常用的印度语言。该图表基于以下电子表格中的数据:AI energy use by language(按语言划分的 AI 能耗)。图片来源:图表:Scott Hyde,密歇根大学工程学院。
密歇根大学工程学院的研究人员指出,在处理和生成非英语文本时,生成式 AI 会消耗更多能源,有时还会给出准确性更低的答案,这可能加剧现有的全球不平等。
这种能源差距在不使用拉丁字母、且可供训练的数字内容较少的语言中最为明显。AI 模型生成非英语文本的能耗可能高出 1.4 至 180 倍,而大多数西欧语言的能耗仅为两倍或更低。然而,回应印地语和孟加拉语的提示——这两种语言各有数亿使用者——所需的能耗分别是回应同等英语提示的 7.9 倍和 15.3 倍。
“通常能耗更高的语言,其模型输出的准确性也往往更低,”密歇根大学 AI Lab 计算机科学与工程专业博士生、该研究的第一作者 Naihao Deng 表示。该研究已在计算语言学协会(Association for Computational Linguistics)的 Empirical Methods in Natural Language Processing 会议上发布。
“大多数人会希望用母语提问来使用 AI 模型,而随着 AI 被融入社会更多领域,他们可能会感到不得不用它;但这种双重惩罚使得某些人更有可能因价格过高而被挡在 AI 之外,也让企业更不愿意为某些语言提供模型服务。”
目前,根据研究人员的测量结果以及 Ethnologue 中可用的人口统计数据,约有 52 亿人用母语生成文本所消耗的能源是英语使用者的两倍以上。不过,研究人员也提出了缩小这一差距的建议,以确保用户不会因为想用自己选择的语言使用 AI 而受到惩罚。
测量 AI 的语言能源差异
研究人员通过测量三个开源 AI 模型——阿里巴巴的 Qwen 3、Google 的 Gemma 3 和 Meta 的 Llama 3.1——所消耗的能源,发现了能源与准确率方面的差异。他们让每个模型回答 Belebele 中的多项选择题,该数据集用于评估 AI 模型在 122 种不同语言中的阅读理解能力。随后,他们使用密歇根工程学院开发的 ML.Energy 框架测量了回答每个问题所消耗的能源。
为了检验其他类型任务中是否存在差异,研究人员还让模型回答了一组小学数学应用题,以及其他基准数据集中包含的开放式用户提示。由于这些数据集并非多语言,研究人员使用 Meta 的翻译工具 No Language Left Behind 将英文题目翻译成了另外七种语言。
平均而言,所有模型处理英文文本时消耗的能源都更少。这种能源差异在更大的模型规模以及不同类型的计算机处理器上依然存在。
更多的 Token,更高的成本
在处理其他语言时,模型消耗的能源更多,因为它们必须生成更多的 Token——Token 是 AI 使用的基本数据单位。Token 可以代表整个短语和单词,也可以代表单个音节和字符。AI 模型会先把提示词转换为 Token 再进行处理,然后把输出的 Token 转换回人类语言,形成最终回复。
对于字符需要更多代码才能显示的书写系统,这一问题会进一步加剧。例如,印地语、尼泊尔语和马拉地语使用的天城文,以及孟加拉文,其基本字符会变形并与重音符号组合,从而产生多达数百万个不同的字符。
Token 数量越多,占用的计算机内存就越多,消耗的能源也越多。由于 Token 占用了更多内存,模型可用于处理信息的内存就更少。因此,它只能以更小的批次处理这些 Token,导致每个 Token 的能耗更高。与英文文本相比,这种效应会进一步放大能源差异。
更多的 Token 还可能带来更高的直接成本,因为企业通常按生成的 Token 数量来对模型调用定价。
缩小 AI 的语言能源差异
为帮助解决这种成本差异,研究人员建议企业针对更多语言优化其服务。这可以包括把用户查询路由到为特定语言定制的模型。企业也可以在把用户提示词送入更通用的模型之前,先将其翻译成英文——前提是翻译足够可靠。
“AI 开发者应当投入更多精力,为那些存在能源与成本差距的语言创建更多资源,”Rada Mihalcea 表示。她是 Janice M. Jenkins 计算机科学与工程讲席教授、密歇根大学 AI 实验室主任,也是该研究的通讯作者。“通过众包方式让母语者帮助翻译材料,或创建资源以获得足够的训练数据,也有助于提升效率。”
Naihao Deng 等,《语言—能源鸿沟:测量多语言 LLM 推理的能源成本》,2026 年自然语言处理经验方法会议(EMNLP 2026)论文集(2026)。DOI: 10.18653/2026.emnlp-main.2113
这篇报道的背后是谁?
艺术史学士,物质文化硕士。曾任博物馆编辑、急救人员、器官移植协调员。自 2021 年起为 Science X 从事编辑工作。完整简介 →
物理学硕士,具备科研经历。长期关注科学新闻。在 Science X 的编辑工作中发挥关键作用。完整简介 →
引用:生成式 AI 在处理非英语文本时消耗更多能源(2026 年 11 月 5 日),2026 年 11 月 6 日检索自 https://techxplore.com/news/2026-11-generative-ai-energy-english-text.html
本文受版权保护。除为私人学习或研究目的进行的任何合理使用外,未经书面许可,不得复制本文任何部分。内容仅供参考。
获取即时摘要文本(Gist)
在三个开源 AI 模型中,生成非英语文本的能耗是英语的 1.4–180 倍,部分语言还表现出更低的准确率。更高的 Token 数量增加了内存占用和能耗成本。针对特定语言的模型以及改进的多语言资源有望缩小这些差距。
本摘要由 LLM 自动生成。完整免责声明
译文由英文原文机器翻译,术语按行业惯例统一,可在文末展开原文核对。
查看英文原文(Generative AI uses more energy on non-English text)
Generative AI uses more energy on non-English text
Generative AI uses more energy on non-English text
by Derek Smith, University of Michigan
edited by Lisa Lock, reviewed by Andrew Zinin
This article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring the content's credibility:
Add to Preferred Sources
A chart showing the amount of energy required, on average, for large language models to respond to user prompts in eight languages ranked among the world’s most widely spoken, as well as several commonly spoken Indic languages. The chart is based on data available in this spreadsheet: AI energy use by language. Credit: GRAPH: Scott Hyde, University of Michigan Engineering.
When processing and producing non-English text, generative AI burns more energy and sometimes provides less accurate answers, potentially reinforcing existing global inequalities, according to University of Michigan Engineering researchers.
The energy disparity is greatest in languages that don't use Latin script and have less digital content available for training. AI models could use 1.4 to 180 times more energy to generate non-English text, with most Western European languages coming in at twice the energy expenditure or less. However, responding to prompts in Hindi and Bengali, each spoken by hundreds of millions of people, requires 7.9 and 15.3 times more energy, respectively, than responding to an equivalent English prompt.
"The languages that typically consume more energy also tend to have lower accuracy in the model output," said Naihao Deng, a doctoral student in computer science and engineering from the U-M AI Lab and the first author of the study, which was presented at the Empirical Methods in Natural Language Processing conference of the Association for Computational Linguistics.
"Most people will want to use AI models by asking questions in their native languages, and they may feel pressured to use it as it's incorporated into more parts of society, but the double penalty makes it more likely that certain people will be priced out of AI and that companies will be more reluctant to serve models in certain languages."
At present, around 5.2 billion people have to use more than twice as much energy to generate text in their first language as English speakers do, according to the researchers' measurements and demographic data available in Ethnologue. But the researchers also provide recommendations to bridge the gap and ensure that users aren't penalized for trying to use AI in their language of choice.
Measuring AI's language energy disparity
The researchers found the energy and accuracy disparities by measuring the amount of energy used by three open-source AI models: Alibaba's Qwen 3, Google's Gemma 3 and Meta's Llama 3.1. They asked each model to answer multiple-choice questions included in Belebele, a dataset for assessing the reading comprehension of AI models in 122 different languages. They then measured the amount of energy used to answer each question using the ML.Energy framework, developed at Michigan Engineering.
To check for disparities in other types of tasks, the researchers also asked the models to answer a collection of grade-school math word problems and open-ended user prompts included in other benchmarking datasets. Because these datasets aren't multilingual, the researchers translated the English questions into seven other languages using Meta's translation tool, No Language Left Behind.
All the models used less energy for English text, on average. The energy disparity persists at larger model sizes and across different types of computer processors.
Extra tokens, extra cost
When working in other languages, the models used more energy because they had to generate more tokens, the basic units of data used by AI. Tokens can represent entire phrases and words or single syllables and characters. AI models convert prompts into tokens before processing them, and then convert the output tokens into human language for the final response.
This problem escalates for scripts whose characters require more code to display. For example, the Devanagari script used in Hindi, Nepali and Marathi and the Bengali script have primary characters that change form and combine with accents to produce up to millions of unique characters.
A higher number of tokens takes up more computer memory and burns more energy. With more memory consumed by tokens, the model has less available for processing information. So, it processes the tokens in smaller batches, which uses more energy per token. This effect compounds the energy disparity compared with English text.
More tokens can also result in higher direct costs, since companies often price model access by the number of tokens generated.
Reducing AI's language energy disparity
To help fix the cost disparity, the researchers recommend that companies optimize their services for more languages. This could entail routing user queries to models tailored to specific languages. Companies could also translate user prompts into English before feeding them to a more generalized model, provided the translation is reliable.
"AI developers should be creating more resources for those languages with an energy and cost disparity," said Rada Mihalcea, the Janice M. Jenkins collegiate professor of computer science and engineering, director of the U-M AI Lab and corresponding author of the study. "Crowdsourcing native speakers to help translate material or create resources to get enough training data could also help improve efficiency."
Naihao Deng et al., The Language–Energy Divide: Measuring Energy Costs of Multilingual LLM Inference, Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing (2026). DOI: 10.18653/2026.emnlp-main.2113
Who's behind this story?
BA art history, MA material culture. Former museum editor, paramedic, and transplant coordinator. Editing for Science X since 2021. Full profile →
Master's in physics with research experience. Long-time science news enthusiast. Plays key role in Science X's editorial success. Full profile →
Citation: Generative AI uses more energy on non-English text (2026, November 5) retrieved 6 November 2026 from https://techxplore.com/news/2026-11-generative-ai-energy-english-text.html
This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.
Get Instant Summarized Text (Gist)
Across three open-source AI models, generating non-English text used 1.4–180 times as much energy as English, with some languages also showing lower accuracy. Higher token counts increased memory use and energy costs. Language-specific models and improved multilingual resources could reduce these disparities.
This summary was automatically generated using LLM. Full disclaimer
This site uses cookies to assist with navigation, analyse your use of our services, collect data for ads personalisation and provide content from third parties. By using our site, you acknowledge that you have read and understand our Privacy Policy and Terms of Use.
原文为英文,正文由自动抓取并保留原文。
原文发布于 TechXplore(Generative AI uses more energy on non-English text),标题、摘要与正文由机器翻译,术语按行业惯例统一,未经人工逐句校订。
