
图131428-1:
(20)
Reserachers are struggling to understand how artificial intelligence models
know things no one told them
(摘自 Scientific American - September 2023 涉及词汇:baffle, ace)
对整篇文章感兴趣的小伙伴可以私信我获取期刊电子版,欢迎探讨交流哦!
No one yet knows how chatGPT and its artificial-intelligence cousins will transform the world, and one reason is that no one really knows what goes on inside them. Some of these systems' abilities go far beyond what they were trained to do—and even their inventors are baffled as to why. A growing number of tests suggest these AI systems develop internal models of the real world, much as our own brain does, although the machines' technique is different.
没有人知道chatGPT和它的人工智能表亲将如何改变世界,其中一个原因是没有人真正知道它们内部发生了什么。其中一些系统的能力远远超出了它们被训练的能力——甚至它们的发明者也不知道为什么。越来越多的测试表明,这些人工智能系统开发了现实世界的内部模型,就像我们自己的大脑一样,尽管机器的技术不同。
"Everything we want to do with them in order to make them better or safer or anything like that seems to me like a ridiculous thing to ask ourselves to do if we don't understand how they work," says Ellie Pavlick of Brown University, one of the researchers working to fill that explanatory void.
布朗大学(Brown University)的埃莉·帕夫利克(Ellie Pavlick)是致力于填补这一解释空白的研究人员之一,她说:“在我看来,如果我们不了解它们的工作原理,我们做一些让它们变得更好、更安全或诸如此类事情的努力,是很荒谬的。”
At one level, she and her colleagues understand GPT (short for "generative pre-trained transformer") and other large language models, or LIMs, perfectly well. The models rely on a machine-learning system called a neural network. Such networks have a structure modeled loosely after the connected neurons of the human brain. The code for these programs is relatively simple and fills just a few screens. It sets up an autocorrection algorithm, which chooses the most likely word to complete a passage based on laborious statistical analysis of hundreds of gigabytes of Internet text. Additional training ensures the system will present its results in the form of dialogue.
In this sense, all it does is regurgitate what it learned—it is a "stochastic parrot," in the words of Emily Bender, a linguist at the University of Washing-ton. (Not to dishonor the late Alex, an African Grey Parrot who understood concepts such as color, shape and "bread" and used corresponding words intentionally.) But LLMs have also managed to ace the bar exam, write a sonnet about the Higgs boson and make an attempt to break up their users' marriage. Few had expected a fairly straightforward autocorrection algorithm to acquire such broad abilities.
在某种程度上,她和她的同事们非常了解GPT(“生成预训练转换器”的缩写)和其他大型语言模型(lim)。这些模型依赖于一种被称为神经网络的机器学习系统。这种网络的结构松散地模仿了人类大脑中相互连接的神经元。这些程序的代码相对简单,只占用几个屏幕。它设置了一个自动纠错算法,通过对数百GB的互联网文本进行费力的统计分析,选择最可能的单词来完成一篇文章。还有其他的训练以确保该系统将以对话的形式提出其结果。
从这个意义上说,它所做的一切就是反刍它所学到的东西——用华盛顿大学语言学家艾米丽·本德(Emily Bender)的话来说,它是一只“随机鹦鹉”。(不是为了羞辱已故的亚历克斯,这是一只非洲灰鹦鹉,它能理解颜色、形状和“面包”等概念,并有意使用相应的词)。但是这些大语言模型还可以通过律师资格考试,或者写一首关于希格斯玻色子的十四行诗,甚至试图破坏他们用户的婚姻。很少有人预料到一个相当简单的自动校正算法会获得如此广泛的能力。
That GPT and other Al systems perform tasks they were not trained to do, giving them "emergent abilities," has surprised even researchers who have been generally skeptical about the hype over LIMs. "I don't know how they're doing it or if they could do it more generally the way humans do—but they've challenged my views," says Melanie Mitchell, an Al researcher at the Santa Fe Institute.
GPT和其他人工智能系统能够执行它们没有接受过训练的任务,这使它们获得了“突发能力”,这甚至让那些对人工智能的炒作持怀疑态度的研究人员也感到惊讶。圣达菲研究所的人工智能研究员Melanie Mitchell说:“我不知道它们是如何做到的,也不知道它们是否能像人类那样更普遍地做到这一点——但它们挑战了我的观点。”
【1】Some of these systems' abilities go far beyond what they were trained to do—and even their inventors are baffled as to why.
1.baffle 及物动词,意为“使困惑、难住”,在句中常可直接加宾语,还可加ing变为形容词,还可用作被动语态当中。看几个句子:
1. The new security system baffled the thieves and prevented them from breaking in. 新的安全系统难住了小偷,使他们无法破门而入。 (摘自 The New York Times)
2. I was constantly ill, with a baffling array of symptoms. 我不断地生病,还伴有一大堆莫名其妙的症状。
3. I'm baffled why she hasn't called. 我不明白她为什么没打电话。
baffle 还可以用作名词,意为“(控制声、光、液体等流动的)隔板,挡板,反射板”。看一个来自于《科学美国人》的例句:
The scientist used a baffle to separate the gases in the reaction chamber. 科学家用挡板把反应室里的气体分开。
2.句子改写:
The capabilities of some of these systems exceed what they were originally designed to handle, leaving their creators perplexed as to the reasons behind this.
改写后的句子请注意perplex的用法,还有as to the reason behind this 对于 as to why的替换。
【2】But LLMs have also managed to ace the bar exam, write a sonnet about the Higgs boson and make an attempt to break up their users' marriage.
1. ace 名词,意为“(纸牌中的)A;一流选手;(网球比赛中)直接得分的发球”。可以活用为形容词,意为“一流的,极好的”,还可以活用为动词,意为“(在考试中)取得好成绩、表现极好”。如:
1. Despite the loss of their ace early in the game, Seattle beat the Brewers 6-5. 尽管在那场比赛开始不久就失去了一流队员,西雅图队仍以6比5击败了酿酒人队。
2. We had an ace time. 我们过得真痛快。
3. He aced every subject. 他每科都考得很好。
2. 句子改写:
However, LLMs have also managed to excel in the bar exam, composed a sonnet about the Higgs boson, and attempted to disrupt their users' marriage.
主要注意改写后句子中excel和disrupt的用法:
1. 原句中 ace 是及物动词,可以直接加宾语“the bar exam”;改写后句子中excel是不及物动词,不可以直接加宾语,需要加介词in后再接宾语(正确的理解方法为:介词短语做状语修饰excel)。
2. disrupt 基本上原意替换 break up,但disrupt是更为学术的用法,而break up通常用于非正式的口语表达中。
【3】回译练习:
暂停(税收抵免的)新申请突显了国税局内部的高度警觉,其认为税收抵免已经被滥用和误用。
原句请点击:考研英语:5分钟读外刊话题(词汇语法) 美国的退税欺诈
来源:今日头条
作者:英哥研语
点赞:1
评论:0
标题:考研英语 | 5分钟外刊阅读:ChatGPT做出了它爹都想不明白的事
原文:https://www.toutiao.com/article/7288683672826741307
侵权告知删除:yangzy187@126.com
转载请注明:网创网 www.netcyw.cn/b131428.html



