经济学人双语精读TE-2023-06-08期考研英语|人工智能经济学：每个锅里都有一只随机鹦鹉？（PDF版+Word版+音频）
The economics of AI：A stochasticparrotin every pot?
What aleakedmemofrom Googlerevealsabout the future ofartificialintelligence
【1】They have changed the world by writing software. But techy types are also known forcomposinglengthymemos in prose, the most famous of which have marked turning points in computing. Think of Bill Gates’s “Internettidalwave”memoof1995, which reoriented Microsoft towards the web; or Jeff Bezos’s “APImandate”memoof2002, which opened up Amazon’s digital infrastructure, paving the way for modern cloud computing. Nowtechiesareabuzzabout anothermemo, this timeleakedfrom within Google,titled“We have nomoat”. Itsunknownauthor details theastonishingprogress being made inartificialintelligence (AI)—and challenges some long-heldassumptionsabout the balance of power in this fast-moving industry.
【2】AIburstinto the publicconsciousnesswith the launch in late2022of ChatGPT, a chatbot powered by a “large language model” (LLM) made by OpenAI, a startup closely linked to Microsoft. Its success prompted Google and other tech firms toreleasetheir own LLM-powered chatbots. Such systems cangeneratetext and holdrealisticconversations because they have been trained using trillions of words taken from the internet. Training a large LLM takes months and costs tens of millions of dollars. This led to concerns that AI would be dominated by a few deep-pocketed firms.
【3】But thatassumptionis wrong, says the Googlememo. It notes that researchers in the open-sourcecommunity, using free, onlineresources, are nowachievingresultscomparableto the biggestproprietarymodels. It turns out that LLMs can be “fine-tuned” using atechniquecalled low-rankadaptation, or LoRa. This allows an existing LLM to be optimised for aparticulartask far more quickly and cheaply than training an LLM from scratch.
【4】Activity in open-source AIexplodedin March, when LLaMA, a model created by Meta, Facebook’s parent, wasleakedonline. Although it is smaller than the largest LLMs (its smallest version has7bnparameters, compared with540bn for Google’s PaLM) it was quickly fine-tunedto produce resultscomparableto the original version of ChatGPT on some tasks. As open-source researchers built on each other’s work with LLaMA, “atremendousoutpouringofinnovationfollowed,” thememo’s author writes.
【5】This could haveseismicimplicationsfor the industry’s future. “Thebarrierto entry for training andexperimentationhas dropped from the total output of a major research organisation to one person, an evening, and abeefylaptop,” the Googlememoclaims. An LLM can now be fine-tunedfor $100in a few hours. With its fast-moving,collaborativeand low-cost model, “open-source has somesignificantadvantages that we cannotreplicate.” Hence thememo’s title: this may mean Google has nodefensive“moat” against open-sourcecompetitors. Nor, for that matter, does OpenAI.
【6】Not everyone agrees with thisthesis. It is true that the internet runs on open-source software. But people use paid-for,proprietarysoftware, from Adobe Photoshop to Microsoft Windows, as well. AI may find a similar balance. Moreover,benchmarkingAI systems isnotoriouslyhard. Yet even if thememois partly right, the implication is that access to AI technology will be far more democratised than seemed possible even a year ago. Powerful LLMs can be run on alaptop; anyone who wants to can now fine-tune their own AI.
【7】This has both positive and negativeimplications. On the plus side, it makesmonopolisticcontrol of AI by ahandfulof companies far less likely. It will make access to AI much cheaper,accelerateinnovationacross the field and make it easier for researchers to analyse the behaviour of AI systems (their access toproprietarymodels waslimited),boostingtransparencyand safety. But easier access to AI also means bad actors will be able to fine-tune systems fornefariouspurposes, such as generatingdisinformation. It means Westernattemptstopreventhostileregimesfromgainingaccess to powerful AI technology will fail. And it makes AI harder to regulate, because thegenieis out of thebottle.
【8】Whether Google and its ilk really have lost theirmoatin AI will soon becomeapparent. But as with those previous memos, this feels like another turning point for computing.
①短语：
1.原文：What a leaked memo from Google reveals aboutthe future of artificial intelligence
词典：reveal about揭示
例句：Here"s what people are saying about the deal and what it reveals about Newsweek and the media"s future.
下面是人们对这次收购以及这次收购预示的《新闻周刊》乃至整个传媒的未来的看法。
2.原文：But techy types are also known forcomposing lengthy memos in prose,
词典：beknown for因...而众所周知；知道
例句：The result of the ballot will not be known for two weeks.
本次无记名投票的结果将在两周后才会揭晓。
3.原文：Think ofBill Gates’s “Internet tidal wave” memo of 1995,
词典：thinkof想到；考虑
例句：By any means,we should tell him frankly what we think of his proposal.
无论如何,我们应该直率地告诉他我们对他建议的看法。
4.原文：which opened up Amazon’s digital infrastructure, paving the way for modern cloud computing.
词典：pavethe way 为...铺路；为....作准备
例句：I believe this treaty will pave the way to peace in Europe.
我相信这个条约将为欧洲的和平铺平道路。
5.原文：AI burst intothe public consciousness with the launch in late 2022 of ChatGPT,
词典：burst into 闯入；突然发作
例句：Between astonishment and joy, she couldn"t help bursting into tears.
她惊喜交集,控制不住放声大哭起来。
6.原文：It turns outthat LLMs can be “fine-tuned” using a technique called low-rank adaptation, .
词典：turn out 发现（结果）
例句：But those reasons turn out to be red herrings.
但这些动机最后证明都是混淆视听的。
7.原文：The barrier to entry for training and experimentation has dropped from the total output of a major research organisation
词典：drop from 从...落下；从...除去
例句：Alexander has been dropped from his multimillion-dollar-a-year job as spokesman for the company.
亚历山大已经被拿掉了年薪数百万的发言人的工作。。
8.原文：It will make access to AI much cheaper,
词典：access to 接近；有权使用
例句：Access to justice is a basic right.
享有公正是一种基本权利。
9.原文：But as with those previous memos, this feels like anotherturning pointfor computing.
词典：turning point转折点
例句：The promotion proved to be a turning point in his career.
这次提升最后证明是他职业生涯的一个转折点。
②长难句
1.原文：It notesthatresearchers in the open-source community, usingfree, online resources,are now achievingresultscomparable to the biggest proprietary models.
2.分析：本句的结构为主谓宾，主语是形式主语it，看回前文会发现是指Googlememo，谓语为notes，宾语为that引导的宾语从句；从句的主语为researchers ，in the open-source community为介词短语作后置定语，谓语为areachieving，宾语为results，using为现在分词作状语，comparable to修饰results作定语。
3.译文：它指出，开源社区的研究人员使用免费的在线资源，他们取得的成果现在可与最大的专有模型相媲美。
1.原文：Itwill makeaccess to AImuch cheaper, accelerateinnovation across the field and make it easierfor researchersto analyse the behaviour of AI systems(their access to proprietary models was limited), boostingtransparency and safety.
2.分析：本句是并列句，出现了好几个并列谓语动词。首先句子为主谓宾结构，主语为It，并列的谓语分别为willmake，accelerate和make，注意句子并列三个谓语动词的时候是可以逗号将前两个动词隔开，最后一个用and连接；每个动词后面接的是对应的宾语，比如第一小句谓语为willmake，宾语为accesstoAI，muchcheaper为宾补；第二小句谓语为accelerate，宾语为innovation；across the field为状语；第三小句动词为make，形式宾语为it，to analyse the behaviour of AI systems是真正的宾语；最后的boosting为整个句子的非谓语，现在分词作结果状语。
3.译文：这也大大降低了人工智能的成本，加速整个领域的创新，并使研究人员更容易分析人工智能系统的行为（他们对专有模型的访问是有限的），从而提高透明度和安全性。
③写作技巧:
On the plus side, it makes monopolistic control of AI by a handful of companies far less likely.
从好的方面来说，这大大降低了少数公司垄断控制人工智能的可能性。
表达： On the plus side “好的方面”，写作时在分析利弊的时候，可以使用该表达。
例句：On the plus side, scientists say social distancing has been effective in slowing down the spread of coronavirus.
在有利方面，科学家称增大社交距离确实有效地减缓了冠状病毒的传播。。
④背景知识：
Stochastic parrot（随机鹦鹉）：随机鹦鹉是由华盛顿大学语言学家 Emily M. Bender 创造的词汇，2021 年 3 月，Bender 等人发表了一篇名为《On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? ���》的论文 —— （随机鹦鹉的危险：语言模型是否太大？）在论文发表后，两位女性合著者，失去了谷歌道德人工智能团队的联合领导职位。之后随机鹦鹉逐渐走红，至少在学术界，它成为了一个流行词汇。最近，随着 ChatGPT 等大模型的走红，很多人又用「随机鹦鹉」来形容大模型没有理解能力、出现幻觉（hallucinations）等行为，诟病这些模型只会制造语法上合理的字串或语句，并没有做到真正的理解。
⑤段落大意：
【1】谷歌备忘录闻名世界
【2】人们担心大型语言模型为少数公司垄断
【3】&【4】“低秩适应”的微调技术
【5】微调技术带来的影响
【6】&【7】微调技术的利弊
【8】计算机领域将迎来新起点
PS:各位研友下载请文末点击阅读原文.考研英语杂志经济学人英文外刊|2022年经济学人英文杂志（2022年经济学人周刊英杂志已更新完毕）366外刊社每日分享英文杂志,考研英语杂志,考研英语外刊双语精读,经济学人杂志,英文杂志下载。
文件格式:True PDF ，①.下载方式:网盘下载②.是否支持编辑:支持③.文字内容:支持选取复制★能否打印:支持打印★
戳“阅读原文”下载外刊精读PDF Doc Mp3
