- Published on
机器知道我们知道它知道什么吗(下)
- Authors
乌尔曼博士指出，过去几十年里，机器学习领域的研究人员一直在计算机模型中竭力捕捉人类知识的灵活性。
双语精读
Dr. Ullman noted that machine learning researchers have struggled over the past couple of decades to capture the flexibility of human knowledge in computer models. This difficulty has been a "shadow finding," he said, hanging behind every exciting innovation.
乌尔曼博士指出，过去几十年里，机器学习领域的研究人员一直在计算机模型中竭力捕捉人类知识的灵活性。他说，这一难题就像是在"找影子"，存在于每一项令人兴奋的创新背后。
Researchers have shown that language models will often give wrong or irrelevant answers when primed with unnecessary information before a question is posed; some chatbots were so thrown off by hypothetical discussions about talking birds that they eventually claimed that birds could speak.
研究人员已经表明，在提出问题之前，如果事先提示一些不必要的信息，语言模型往往会给出错误或不相关的答案；一些聊天机器人被关于会说话的鸟儿的假设性讨论给搞晕了，以至于它们最终声称鸟儿是会说话的。
Because their reasoning is sensitive to small changes in their inputs, scientists have called the knowledge of these machines "brittle."
因为它们在推理时对输入信息的微小变化很敏感，所以科学家们称这些机器的知识是"脆性易碎的"。
Dr. Gopnik compared the theory of mind of large language models to her own understanding of general relativity.
戈普尼克博士将大型语言模型的心智理论比作她自己对广义相对论的理解。
"I have read enough to know what the words are," she said. "But if you asked me to make a new prediction or to say what Einstein's theory tells us about a new phenomenon, I'd be stumped because I don't really have the theory in my head."
"我读了足够多的东西，知道这几个词是什么意思，"她说，"但如果你让我做一个新的预测，或者让我用爱因斯坦的理论去解释一个新现象，我就会被难住，因为我脑子里其实没有这个理论。"
By contrast, she said, human theory of mind is linked with other common-sense reasoning mechanisms; it stands strong in the face of scrutiny.
她说，与之形成对比，人类的心智理论与其他常识推理机制是联系在一起的，它经得起审视。
In general, Dr. Kosinski's work and the responses to it fit into the debate about whether the capacities of these machines can be compared to the capacities of humans -- a debate that divides researchers who work on natural language processing.
总体而言，科辛斯基博士的研究和对这一研究的各种回应属于一种辩论的范畴，即这些机器的能力是否可以与人类的能力相提并论，研究自然语言处理的人就这一问题分成了两派。
Are these machines stochastic parrots, or alien intelligences, or fraudulent tricksters?
这些机器是具有随机性的学舌鹦鹉，还是异于人类的智能，还是具有欺诈性的骗子？
A 2022 survey of the field found that, of the 480 researchers who responded, 51 percent believed that large language models could eventually "understand natural language in some nontrivial sense," and 49 percent believed that they could not.
2022年对该领域进行的一项调查发现，在做出回应的480名研究人员中，51%的人认为大型语言模型最终可以"在某种重大意义上理解自然语言"，49%的人认为它们不能做到这一点。
Dr. Ullman doesn't discount the possibility of machine understanding or machine theory of mind, but he is wary of attributing human capacities to nonhuman things.
乌尔曼博士并不否认机器有理解能力或机器有心智理论的可能性，但他对认为非人类事物有人类能力的这种想法持谨慎态度。
He noted a famous 1944 study by Fritz Heider and Marianne Simmel, in which participants were shown an animated movie of two triangles and a circle interacting. When the subjects were asked to write down what transpired in the movie, nearly all described the shapes as people.
他提到了弗里茨·海德和玛丽安· 西梅尔在1944年进行的一项著名研究，在这项研究中，参与者观看了一部两个三角形和一个圆形进行互动的动画影片。当受试者被要求写下电影中发生了什么时，几乎所有人都把这些形状当作人来描述。
"Lovers in the two-dimensional world, no doubt; little triangle number-two and sweet circle," one participant wrote. "Triangle-one (hereafter known as the villain) spies the young love. Ah!"
"毫无疑问，这是二维世界里的恋人，二号小三角和甜美的圆形。"一位参与者写道，"一号三角(以下简称反派)在暗中窥视这对年轻的恋人。啊!"
It's natural and often socially required to explain human behavior by talking about beliefs, desires, intentions and thoughts. This tendency is central to who we are -- so central that we sometimes try to read the minds of things that don't have minds, at least not minds like our own.
用信念、欲望、意图和想法来解释人类行为是很自然的，也经常是为社会所要求的。这种倾向是何以为人的核心所在，正因为它处于如此核心的地位，以至于我们有时候试图读懂那些无心智之物的心智，至少它们的心智和我们的并不相同。
词汇预习
participant [高考]
美[pɑːrˈtɪsɪpənt] | 英[pɑːˈtɪsɪpənt]
n. 参与者
contrast [高考]
美[ˈkɑːntræst , kənˈtræst] | 英[ˈkɒntrɑːst , kənˈtrɑːst]
n. 对比；差别；对照物 v. 对比；成对照 n. [计算机] 反差
capture [高考]
美[ˈkæptʃər] | 英[ˈkæptʃə(r)]
vt. 捕获；占领；夺取；吸引；(用照片等)留存 n. 捕获；战利品
innovation [高考]
美[ˌɪnəˈveɪʃn] | 英[ˌɪnəˈveɪʃn]
n. 创新；革新
flexibility [高考]
美[ˌfleksə'bɪləti] | 英[ˌfleksə'bɪləti]
n. 灵活性；弹性；适应性；柔韧性
eventually [高考]
美[ɪˈventʃuəli] | 英[ɪˈventʃuəli]
adv. 最后；终于
possibility [高考]
美[ˌpɑːsəˈbɪləti] | 英[ˌpɒsəˈbɪləti]
n. 可能；可能性；可能的事情；潜在的价值
theory [高考]
美[ˈθiːəri] | 英[ˈθɪəri]
n. 学说；理论；原理；意见
debate [高考]
美[dɪˈbeɪt] | 英[dɪˈbeɪt]
n. 辩论；讨论 vt. 争论；思考 vi. 辩论
sensitive [高考]
美[ˈsensətɪv] | 英[ˈsensətɪv]
adj. 灵敏的；敏感的；感光的；易受伤害的；善解人意的 n. 敏感的人
phenomenon [高考]
美[fəˈnɑːmɪnən] | 英[fəˈnɒmɪnən]
n. 现象；奇人；奇事
discount [高考]
美[ˈdɪskaʊnt , dɪsˈkaʊnt] | 英[ˈdɪskaʊnt , dɪsˈkaʊnt]
n. 折扣；贴现率 vt. 打折扣；贴现；不全信；不重视 vi. 贴现；减息贷款
tendency [高考]
美[ˈtendənsi] | 英[ˈtendənsi]
n. 趋势；倾向；秉性
shadow [高考]
美[ˈʃædoʊ] | 英[ˈʃædəʊ]
n. 影子；阴影；阴暗；荫；暗处 adj. 影子内阁的；阴影的 vt. 投阴影于；跟踪；预示 vi. 渐变；变阴暗
triangle [高考]
美[ˈtraɪæŋɡl] | 英[ˈtraɪæŋɡl]
n. 三角(形)
no doubt [高考]
美[noʊ daʊt] | 英[nəʊ daʊt]
无疑； 必定
sensitive to [高考]
美[ˈsɛnsɪtɪv tu] | 英[ˈsensitiv tu:]
对 ... 敏感
debate about [高考]
美 | 英
就……辩论
in the face of [高考]
美[ɪn ðə feɪs əv] | 英[ɪn ðə feɪs əv]
面对； 尽管
finding [高考]
美[ˈfaɪndɪŋ] | 英[ˈfaɪndɪŋ]
n. 调查(或研究)的结果；[律] 裁决
relativity [高考]
美[ˌreləˈtɪvəti] | 英[ˌreləˈtɪvəti]
n. 相对性；相对论
brittle [四级]
美[ˈbrɪtl] | 英[ˈbrɪtl]
adj. 易碎的；尖利的；脆弱的 n. 果仁薄脆糖
socially [四级]
美[ˈsoʃəlɪ] | 英['səʊʃəlɪ]
adv. 在社交上；在社会上；与礼貌行为相关
irrelevant [四级]
美[ɪˈreləvənt] | 英[ɪˈreləvənt]
adj. 不恰当的；无关系的；不相干的
scrutiny [六级]
美[ˈskruːtəni] | 英[ˈskruːtəni]
n. 监视；细看；周密的调查
wary [六级]
美[ˈweri] | 英[ˈweəri]
adj. 小心的；机警的
learning [专八]
美[ˈlɜːrnɪŋ] | 英[ˈlɜːnɪŋ]
n. 学问；学识；学习
dimensional [专八]
美[dɪˈmɛnʃənl] | 英[dɪ'menʃənəl]
adj. 维度的；次元的；空间的；尺寸的
hereafter [专八]
美[ˌhɪrˈæftə(r)] | 英[ˌhɪərˈɑ:ftə(r)]
adv. 此后；将来；来世 n. 将来；来世
fraudulent [专八]
美[ˈfrɔːdʒələnt] | 英[ˈfrɔːdjələnt]
adj. 欺诈的；不正的；不诚实的
nonhuman [考研]
美[ˌnɒn'hjumən] | 英[ˌnɒn'hju:mən]
adj. 非人类的
hypothetical [考研]
美[ˌhaɪpəˈθɛtɪkəl] | 英[ˌhaɪpəˈθetɪkl]
adj. 假设的；假定的
don [托福]
美[dɑn] | 英[dɒn]
n. 先生（西班牙用语）；(牛津、剑桥大学的)指导教师 v. 穿上
noted [托福]
美[ˈnoʊtɪd] | 英[ˈnəʊtɪd]
adj. 著名的；有声望的
animated [雅思]
美[ˈænəˌmetɪd] | 英[ˈænɪmeɪtɪd]
adj. 生气勃勃的；栩栩如生的；动画片的
重点讲解
Dr. Ullman noted that machine learning researchers have struggled over the past couple of decades to capture the flexibility of human knowledge in computer models. This difficulty has been a "shadow finding," he said, hanging behind every exciting innovation.
乌尔曼博士指出，过去几十年里，机器学习领域的研究人员一直在计算机模型中竭力捕捉人类知识的灵活性。他说，这一难题就像是在"找影子"，存在于每一项令人兴奋的创新背后。
capture v.
1. 捕获；夺得
【例】to capture a monkey/a large market share 捕获一只猴子/夺得一大块市场份额
2. 引起（注意、想象、兴趣等）
【例】His brilliant performance captured the audience’s imagination.
他的精彩表演使观众为之神往.
3. 准确地描绘；刻画
【例】The picture captured her unique charm.
这张照片捕捉到了/表现出了她的独特魅力。
【近义词】catch
Researchers have shown that language models will often give wrong or irrelevant answers when primed with unnecessary information before a question is posed; some chatbots were so thrown off by hypothetical discussions about talking birds that they eventually claimed that birds could speak.
研究人员已经表明，在提出问题之前，如果事先提示一些不必要的信息，语言模型往往会给出错误或不相关的答案；一些聊天机器人被关于会说话的鸟儿的假设性讨论给搞晕了，以至于它们最终声称鸟儿是会说话的。
prime v. 事先指点；使做好准备
【例】I'd been primed so I knew not to mention her son.
有人跟我透过风，所以我知道不要提起她儿子。
【搭配】
1. prime sb with sth 事先为某人提供某物
【例】She primed herself with useful information before attending the meeting.
开会前，她事先收集了一些有用的信息，让自己做好准备。
2. prime sb to do sth 事先让某人做某事
【例】He had primed his friends to give the journalists as little information as possible.
他已经事先知会他的朋友，尽量少向记者透露消息。
be thrown off/throw...off
1. 使摆脱
【例】to throw off your worries 摆脱你的忧虑
2. 使犯错误
【例】Any small interruption is likely to throw off my calculations.
一点点的打扰就可能会使我的计算出错。
【拓展】throw...out 打乱；使不知所措
【例】The noise threw his concentration out.
嘈杂声使他难以集中思想/扰乱了他的注意力。
Because their reasoning is**** sensitive to**** small changes in**** their inputs****,**** scientists**** have**** called**** the knowledge of**** these machines**** "brittle****."
因为它们在推理时对输入信息的微小变化很敏感，所以科学家们称这些机器的知识是"脆性易碎的"。
Dr****.**** Gopnik**** compared**** the theory**** of**** mind**** of**** large language**** models**** to**** her own understanding of**** general relativity****.****
戈普尼克博士将大型语言模型的心智理论比作她自己对广义相对论的理解。
"I have read enough to know what the words are," she said. "But if you asked me to make a new prediction or to say what Einstein's theory tells us about a new phenomenon, I'd be stumped because I don't really have the theory in my head."
"我读了足够多的东西，知道这几个词是什么意思，"她说，"但如果你让我做一个新的预测，或者让我用爱因斯坦的理论去解释一个新现象，我就会被难住，因为我脑子里其实没有这个理论。"
stump v. 难住；把……难倒
【例】The police were stumped about the motive behind the murder.
警方给难住了, 无法查明这次谋杀的动机。
By contrast, she said, human theory of mind is linked with other common-sense reasoning mechanisms; it stands strong in the face of scrutiny.
她说，与之形成对比，人类的心智理论与其他常识推理机制是联系在一起的，它经得起审视。
stump v. 难住；把……难倒
【例】The police were stumped about the motive behind the murder.
警方给难住了, 无法查明这次谋杀的动机。
In general, Dr. Kosinski's work and the responses to it fit into the debate about whether the capacities of these machines can be compared to the capacities of humans -- a debate that divides researchers who work on natural language processing.
总体而言，科辛斯基博士的研究和对这一研究的各种回应属于一种辩论的范畴，即这些机器的能力是否可以与人类的能力相提并论，研究自然语言处理的人就这一问题分成了两派。
divide v. 使产生分歧；使意见不一
【例】The issue has divided the government.
这个问题在政府中引起了意见分歧。
【拓展】divided adj. 有分歧的；抵触的
【例】divided loyalty 互相抵触的忠诚
Divorce is an agony of divided loyalties for children.
离婚对孩子来说是很痛苦的，他们不知道该向着谁。
Are these machines stochastic parrots, or alien intelligences, or fraudulent tricksters?
这些机器是具有随机性的学舌鹦鹉，还是异于人类的智能，还是具有欺诈性的骗子？
alien adj. 陌生的；异己的；不相容的
【例】The idea is alien to our religion.
这种思想与我们的宗教不相容。
fraudulent adj. 欺诈性的
【例】fraudulent advertising 虚假不实的广告
【拓展】fraud n. 欺诈行为；骗子
A 2022 survey of the field found that, of the 480 researchers who responded, 51 percent believed that large language models could eventually "understand natural language in some nontrivial sense," and 49 percent believed that they could not.
2022年对该领域进行的一项调查发现，在做出回应的480名研究人员中，51%的人认为大型语言模型最终可以"在某种重大意义上理解自然语言"，49%的人认为它们不能做到这一点。
alien adj. 陌生的；异己的；不相容的
【例】The idea is alien to our religion.
这种思想与我们的宗教不相容。
fraudulent adj. 欺诈性的
【例】fraudulent advertising 虚假不实的广告
【拓展】fraud n. 欺诈行为；骗子
Dr. Ullman doesn't discount the possibility of machine understanding or machine theory of mind, but he is wary of attributing human capacities to nonhuman things.
乌尔曼博士并不否认机器有理解能力或机器有心智理论的可能性，但他对认为非人类事物有人类能力的这种想法持谨慎态度。
discount v. （认为不重要而）不相信，不考虑
【例】They have discounted the rumors.
他们不相信那些谣传。
【近义词】dismiss v. 不予理会；不予考虑
wary adj. 谨慎的；小心翼翼的
【例】Be wary of strangers who offer you a ride.
提防那些主动让你搭车的陌生人。
He noted a famous 1944 study by Fritz Heider and Marianne Simmel, in which participants were shown an animated movie of two triangles and a circle interacting. When the subjects were asked to write down what transpired in the movie, nearly all described the shapes as people.
他提到了弗里茨·海德和玛丽安· 西梅尔在1944年进行的一项著名研究，在这项研究中，参与者观看了一部两个三角形和一个圆形进行互动的动画影片。当受试者被要求写下电影中发生了什么时，几乎所有人都把这些形状当作人来描述。
discount v. （认为不重要而）不相信，不考虑
【例】They have discounted the rumors.
他们不相信那些谣传。
【近义词】dismiss v. 不予理会；不予考虑
wary adj. 谨慎的；小心翼翼的
【例】Be wary of strangers who offer you a ride.
提防那些主动让你搭车的陌生人。
"Lovers in the two-dimensional world, no doubt; little triangle number-two and sweet circle," one participant wrote. "Triangle-one (hereafter known as the villain) spies the young love. Ah!"
"毫无疑问，这是二维世界里的恋人，二号小三角和甜美的圆形。"一位参与者写道，"一号三角(以下简称反派)在暗中窥视这对年轻的恋人。啊!"
spy v. 做间谍；暗中监视
【例】to spy (on) the enemy 暗中监视敌人的活动
【搭配】spy out 查明情况
【例】to spy out the land 查明地形
It****'s natural and often socially**** required**** to**** explain human behavior**** by**** talking about beliefs,**** desires,**** intentions**** and thoughts****.**** This**** tendency**** is**** central to**** who we**** are --**** so**** central that**** we**** sometimes try to**** read**** the minds of**** things**** that**** don't have**** minds,**** at**** least not minds like**** our own.****
用信念、欲望、意图和想法来解释人类行为是很自然的，也经常是为社会所要求的。这种倾向是何以为人的核心所在，正因为它处于如此核心的地位，以至于我们有时候试图读懂那些无心智之物的心智，至少它们的心智和我们的并不相同。
课后练习
What can be learned about the knowledge of language models?
A. Its vastness can complement the flexibility of human knowledge.
B. The models need background information to use their knowledge.
C. Its accuracy is easily influenced by irrelevant information.
D. The models can correct the mistakes in their knowledge.
Dr. Gopnik’s example of general relativity illustrates that ____.
A. physical laws are beyond the comprehension of machines
B. machines acquire knowledge by reading extensively
C. machines and humans share the same learning mechanism
D. the knowledge of machines is superficial and isolated
The study by Fritz Heider and Marianne Simmel shows that ____.
A. humans have the advantage of imagination over machines
B. people tend to attach qualities of humans to nonhuman things
C. people are sensitive to emotions and interpersonal relationships
D. machines can imitate the pattern of the interaction between people
It is implied in the last paragraph that ____.
A. theory of mind often leads people to form false opinions
B. some of the social requirements are not actually reasonable
C. the difference between machine and human minds should be noted
D. the intelligence of machines can be incomprehensible to humans