文本选自:The Economist(经济学人)
作者:Unknown
原文标题:What does a leaked Google memo reveal about the future of AI?
原文发布时间:13 May 2023
What does a leaked Google memo reveal about the future of AI?
They have changed the world by writing software. But techy types are also known for composing lengthy memos in prose, the most famous of which have marked turning points in computing. Think of Bill Gates's "Internet tidal wave" memo of 1995, which reoriented Microsoft towards the web; or Jeff Bezos's "API mandate" memo of 2002, which opened up Amazon's digital infrastructure, paving the way for modern cloud computing. Now techies are abuzz about another memo, this time leaked from within Google, titled "We have no moat". Its unknown author details the astonishing progress being made in artificial intelligence (AI)—and challenges some long-held assumptions about the balance of power in this fast-moving industry.
AI burst into the public consciousness with the launch in late 2022 of 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 to release their own LLM-powered chatbots. Such systems can generate text and hold realistic conversations 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.
But that assumption is wrong, says the Google memo. It notes that researchers in the open-source community, using free, online resources, are now achieving results comparable to the biggest proprietary models. It turns out that LLMS can be "fine-tuned" using a technique called low-rank adaptation, or LoRa.
一份泄露的谷歌备忘录揭示了人工智能的未来?
精听党背景导读
5月4日,一份遭到泄露的谷歌内部文件在科技圈里热传。一位谷歌高级软件工程师在这份文件中称,谷歌正在丧失人工智能(AI)优势,但赢家也不会是ChatGPT开发商OpenAI,而是开源社区打造的AI。
这位工程师认为,谷歌与OpenAI的竞争分散了公司对开源技术快速发展的注意力。在开源社区,许多独立研究人员利用AI技术取得了快速和意想不到的进展。
文本选自:The Economist(经济学人)作者:Unknown原文标题:What does a leaked Google memo reveal about the future of AI?原文发布时间:13 May 2023关键词:备忘录 AI 未来
精听党带着问题听
1.“护城河”用英语可以怎么表达?2. 如何理解第二段中的“deep-pocketed”?3. 培训一个大型语言模型需要多少成本?
精听党选段赏析
标题解读
What does a leaked Google memo reveal about the future of AI?
一份泄露的谷歌备忘录揭示了人工智能的未来?
leaked adj. 泄露的,曝光的;1. leak v. 渗入,漏出;漏,渗漏;泄露,透露;
The roof was leaking. 屋顶在漏雨。
memo n. 备忘录;
reveal vt. 揭示,透露;表明,证明;展示,显示;1. to reveal a secret 泄露一条秘密;
AI abbr. 人工智能(artificial intelligence);
段一
They have changed the world by writing software. But techy types are also known for composing lengthy memos in prose, the most famous of which have marked turning points in computing. Think of Bill Gates’s “Internet tidal wave” memo of 1995, which reoriented Microsoft towards the web; or Jeff Bezos’s “API mandate” memo of 2002, which opened up Amazon’s digital infrastructure, paving the way for modern cloud computing. Now techies are abuzz about another memo, this time leaked from within Google, titled “We have no moat”. Its unknown author details the astonishing progress being made in artificial intelligence (AI)—and challenges some long-held assumptions about the balance of power in this fast-moving industry.
techy n. 科技通;科技迷;(尤指)电脑通,电脑迷;
compose vt. 组成,构成;作曲;撰写(信、诗或演讲稿);1. He started at once to compose a reply to Anna. 他立刻开始给安娜写回信。
lengthy adj. 漫长的,长时间的;(讲话、文章等)冗长的,过于详尽的;1. The board members held a lengthy meeting to decide future policy. 董事会成员开了一次漫长的会议来决定未来的政策。
prose n. 散文,白话文;(尤指待翻译的)一段散文;平淡乏味的文章(或话语,表达法);
mark vt. 是…的迹象;成为…的征兆;表明;1. This speech may mark a change in government policy. 这篇演讲表明政府的政策可能会有变化。
API abbr. 应用程序接口(Application Programming Interface);
mandate n. 授权,委托;
infrastructure n. 基础设施,基础建设;
abuzz adj. 对…议论纷纷的;
moat n. 护城河;壕沟;
astonishing adj. 令人十分惊讶的;使人大为惊奇的;难以置信的;1. She ran 100m in an astonishing 10.9 seconds. 她以惊人的10.9秒速度跑完了100米。
参考译文
他们通过编写软件改变了世界。但技术人员也以撰写冗长的散文备忘录而闻名,其中最著名的备忘录标志着计算机领域的转折点。想想比尔·盖茨1995年的“互联网浪潮”备忘录,它将微软重新定位于网络;2002年杰夫·贝佐斯的“应用程序接口授权”备忘录打开了亚马逊的数字基础设施,为现代云计算铺平了道路。现在,技术人员正在讨论另一份备忘录,这次是从谷歌内部泄露出来的,标题是“我们没有护城河”。书中不知名的作者详细描述了人工智能(AI)领域正在取得的惊人进展,并挑战了人们长期以来对这个快速发展的行业中权力平衡的一些假设。
段二
AI burst into the public consciousness with the launch in late 2022 of 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 to release their own LLM-powered chatbots. Such systems can generate text and hold realistic conversations 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.
burst vi. 猛冲;突然出现;1. The sun burst through the clouds. 太阳破云而出。
consciousness n. 知觉,清醒;思想,意识;(群体的)观念体系;感知,察觉;
startup n. 新兴公司;初创公司;
prompt vt. 促使;导致;激起;1. The discovery of the bomb prompted an increase in security. 此次发现炸弹促使当局加强了安全工作。
dominate vt. 统治,支配;在…中占首要地位;
deep-pocketed adj. 有钱的;
参考译文
随着2022年底ChatGPT的推出,人工智能突然进入了公众的视野。ChatGPT是一款由与微软关系密切的初创公司OpenAI开发的“大型语言模型”(LLM)驱动的聊天机器人。它的成功促使谷歌和其他科技公司发布了他们自己的大型语言模型聊天机器人。这样的系统可以生成文本并进行真实的对话,因为它们已经使用了来自互联网的数万亿个单词进行了训练。培训一个大型语言模型需要几个月的时间,耗资数千万美元。这导致人们担心人工智能将被一些财力雄厚的公司所主导。
段三
But that assumption is wrong, says the Google memo. It notes that researchers in the open-source community, using free, online resources, are now achieving results comparable to the biggest proprietary models. It turns out that LLMS can be “fine-tuned” using a technique called low-rank adaptation, or LoRa.
open-source adj. (计)源代码开放的,提供程序源的;
comparable to 可比较的,比得上的;
proprietary adj. 专卖的;专营的;专利的;1. proprietary brands 专利品牌;
turn out 被发现是;1. The forecast turned out to be completely wrong. 这项预测被发现是完全错误的。
fine-tune vt. 调整;使有规则;对进行微调;1. We do not try to fine-tune the economy on the basis of short-term predictions. 我们不会根据短期预测而试图对经济作微调。
low-rank adj. 低质的;低档的;
参考译文
但谷歌的备忘录指出,这种假设是错误的。它指出,开源社区的研究人员使用免费的在线资源,现在正在取得与最大的专有模型相当的结果。事实证明,大型语言模型可以使用一种称为低档适应的技术进行“微调”。
精听党每日单词
leaked
/liːkt/ adj. 泄露的,曝光的;
memo
/ˈmemoʊ/ n. 备忘录;
reveal
/rɪˈviːl/ vt. 揭示,透露;表明,证明;展示,显示;
AI
abbr. 人工智能(artificial intelligence);
techy
/ˈteki/ n. 科技通;科技迷;(尤指)电脑通,电脑迷;
compose
/kəmˈpoʊz/ vt. 组成,构成;作曲;撰写(信、诗或演讲稿);
lengthy
/ˈleŋθi/ adj. 漫长的,长时间的;(讲话、文章等)冗长的,过于详尽的;
prose
/proʊz/ n. 散文,白话文;(尤指待翻译的)一段散文;平淡乏味的文章(或话语,表达法);
mark
/mɑːrk/ vt. 是…的迹象;成为…的征兆;表明;
API
abbr. 应用程序接口(Application Programming Interface);
mandate
/ˈmændeɪt/ n. 授权,委托;
infrastructure
/ˈɪnfrəstrʌktʃər/ n. 基础设施,基础建设;
abuzz
/əˈbʌz/ adj. 对…议论纷纷的;
moat
/moʊt/ n. 护城河;壕沟;
astonishing
/əˈstɑːnɪʃɪŋ/ adj. 令人十分惊讶的;使人大为惊奇的;难以置信的;
burst
/bɜːrst/ vi. 猛冲;突然出现;
consciousness
/ˈkɑːnʃəsnəs/ n. 知觉,清醒;思想,意识;(群体的)观念体系;感知,察觉;
startup
/ˈstɑːrtʌp/ n. 新兴公司;初创公司;
prompt
/prɑːmpt/ vt. 促使;导致;激起;
dominate
/ˈdɑːmɪneɪt/ vt. 统治,支配;在…中占首要地位;
deep-pocketed
/ˌdiːp ˈpɑːkɪtɪd/ adj. 有钱的;
open-source
/ˌoʊpən ˈsɔːrs/ adj. (计)源代码开放的,提供程序源的;
comparable to
可比较的,比得上的;
proprietary
/prəˈpraɪəteri/ adj. 专卖的;专营的;专利的;
turn out
被发现是;
fine-tune
/ˌfaɪn ˈtuːn/ vt. 调整;使有规则;对进行微调;
low-rank
/loʊ ræŋk/ adj. 低质的;低档的;
精听党文化拓展
谷歌是在其I/O开发者大会上发布了最新的通用大语言模型PaLM2。内部文件显示,该模型采用了3.6万亿个令牌(token)进行训练。令牌指的是单词串,这是训练大语言模型的重要基础,因为这可以教给模型如何预测字符串中可能出现的下一个单词。
之前版本的PaLM发布于2022年,当时使用的令牌为7800亿个。
虽然谷歌很希望展示其人工智能技术的实力,以及嵌入搜索、电子邮件、字处理和电子表格后的效果,但该公司却不愿发布其训练数据的规模和其他细节。微软支持的OpenAI也对其最新的GPT-4大语言模型的细节信息保密。
这些公司表示,之所以不披露这些信息,是出于商业竞争考虑。谷歌和OpenAI都在努力吸引想要用聊天机器人代替传统搜索引擎,从而直接获取答案的用户。
但随着人工智能军备竞赛日趋白热化,相关研究人员也呼吁企业加大透明度。
自从发布PaLM2之后,谷歌一直表示新的模型比之前的大语言模型更小,这就意味着该公司的技术效率得以提升,但却可以完成更加复杂的任务。内部文件显示,PaLM2基于3400亿个参数训练——这项指标可以说明该模型的复杂程度。最初的PaLM则基于5400亿个参数训练。
精听党每日美句
That man is the richest whose pleasure are the cheapest.
能处处寻求快乐的人才是最富有的人。