- Published on
人工智能使制造虚假信息更容易(Disinformation and AI: Generation confusion)
- Authors
对于虚假信息，社交媒体将传播成本降至零，而生成式人工智能则将生成成本降至零。
双语精读
Disinformation and AI: Generation confusion
虚假信息与人工智能：生成式混乱
When it comes to disinformation, "social media took the cost of distribution to zero, and generative AI takes the cost of generation to zero," says Renee DiResta of the Stanford Internet Observatory.
斯坦福大学网络观察室的蕾妮·迪雷斯塔表示，对于虚假信息，“社交媒体将传播成本降至零，而生成式人工智能则将生成成本降至零”。
Large language models such as GPT-4 make it easy to produce misleading news articles or social-media posts in huge quantities. And AI can produce more than text. Cloning a voice using AI used to require minutes, or even hours, of sample audio.
GPT-4 等大语言模型很容易生成大量误导性新闻文章或社交媒体帖子。人工智能可以生成的不仅仅是文本。过去使用人工智能克隆声音需要几分钟甚至几小时的样本音频。
Last year, however, researchers at Microsoft unveiled VALL-E, an AI model that is able to clone a person's voice from just a three-second clip of them speaking, and make it say any given text.
然而，去年，微软的研究人员推出了VALL-E，这是一种人工智能模型，能够用一个人说话的三秒片段克隆出他的声音，并让它说出任何给定的文本。
OpenAl, the American company behind GPT-4, has developed a similar tool, Voice Engine, which can convincingly clone any voice from a 15-second clip. It has not yet released it, recognising "serious risks, which are especially top of mind in an election year".
GPT-4背后的美国公司OpenAl开发了一款类似的工具“声音引擎”，它可以用15秒的音频片段克隆任何声音，以假乱真。该工具尚未发布，因为OpenAl认识到“其存在严重风险，在选举年尤其受到关注”。
Similarly, Sora, from OpenAl, can produce surprisingly realistic synthetic videos, in response to text prompts, of up to a minute in length. OpenAI has yet to release Sora to the public, partly on the ground that it could be used to create disinformation.
同样，来自OpenAl的Sora可以根据文本提示词生成逼真程度令人惊讶的合成视频，长度可达一分钟。OpenAI尚未公开发布Sora，部分原因是它可能被用来制造虚假信息。
As well as providing new ways to discredit or misrepresent politicians, AI tools also raise the spectre of personalised disinformation, generated to appeal to small groups (think soccer moms in a specific town).
除了提供抹黑或歪曲政客的新方法外，人工智能工具还引发了对个性化虚假信息的恐慌，这些虚假信息是为了吸引小群体而产生的（想想某个城镇的足球妈妈）。
It may even be possible to "microtarget" individuals with disinformation, based on knowledge of their preferences, biases and concerns.
人工智能甚至可以根据个人的偏好、偏向和关心的问题，用虚假信息“微观定位”个人。
Though all of this is worrying, it is worth remembering that not all aspects of the technology are negative. Al, it turns out, can be used for fighting disinformation as well as producing it.
虽然所有这些都令人担忧，但值得记住的是，并非人工智能的所有方面都是负面的。事实证明，人工智能既可以用来制造虚假信息，也可以用来打击虚假信息。
词汇预习
response [高考]
美[rɪˈspɑːns] | 英[rɪˈspɒns]
n. 反应；回答；响应；答复
release [高考]
美[rɪˈliːs] | 英[rɪˈliːs]
n. 释放；让渡；发行 vt. 释放；让与；准予发表；发射
sample [高考]
美[ˈsæmpl] | 英[ˈsɑːmpl]
vt. 取样；尝试；抽样检查 n. 样品；样本；例子 adj. 试样的，样品的；作为例子的
distribution [高考]
美[ˌdɪstrɪˈbjuːʃn] | 英[ˌdɪstrɪˈbjuːʃn]
n. 分发；分销，配送；（电影在各院线的）发行，上映；分配，分布
audio [高考]
美[ˈɔːdioʊ] | 英[ˈɔːdiəʊ]
adj. 声音的；成音频率的 n. 音频；音响设备
election [高考]
美[ɪˈlekʃn] | 英[ɪˈlekʃn]
n. 选举；当选；选举权
media [高考]
美[ˈmiːdiə] | 英[ˈmiːdiə]
n. 传播媒介；(medium的复数)媒体；新闻媒介
generation [高考]
美[ˌdʒenəˈreɪʃn] | 英[ˌdʒenəˈreɪʃn]
n. 一代（人）；一代人的时间；（产品发展的）代；产生
appeal [高考]
美[əˈpiːl] | 英[əˈpiːl]
n. 恳求；呼吁；上诉；吸引力 n. (体育比赛中)诉请裁决 v. 呼吁；有吸引力；求助；诉请
length [高考]
美[leŋθ] | 英[leŋθ]
n. 长度；时间长短；距离；一段；距离的测量单位
clone [高考]
美[kloʊn] | 英[kləʊn]
n. 克隆；无性繁殖；复制品 v. 克隆；复制
appeal to [高考]
美[əˈpiːl tu] | 英[əˈpiːl tu]
对 ... 产生吸引力； 呼吁； 恳求； 申诉
in response [高考]
美[ɪn rɪˈspɑns] | 英[in riˈspɔns]
作为回答
clip [高考]
美[klɪp] | 英[klɪp]
v.剪辑；修剪；削减；固定 n.夹；夹子；速度；钳
confusion [四级]
美[kənˈfjuːʒn] | 英[kənˈfjuːʒn]
n. 惶惑；混淆；混乱；不确定状态
surprisingly [四级]
美[sərˈpraɪzɪŋli] | 英[sə(r)ˈpraɪzɪŋli]
adv. 惊人地；意外地
synthetic [四级]
美[sɪnˈθetɪk] | 英[sɪnˈθetɪk]
adj. 合成的；人造的；综合的；虚伪的 n. 合成物；人工制品
make it [四级]
美[mek ɪt] | 英[meik it]
成功 达到预定目标,性交易,幸存
up to [四级]
美[ʌp tu] | 英[ʌp tu]
胜任；直到；多达；该由 ... 决定
based on [六级]
美[best ɑn] | 英[beɪst ɔn]
以 ... 为基础； 基于
discredit [专八]
美[dɪsˈkrɛdɪt] | 英[dɪsˈkredɪt]
v. 败坏…的名声；怀疑；不信 n. 丧失名誉；怀疑；不信任；丢脸
convincingly [考研]
美[kən'vɪnsɪŋlɪ] | 英[kən'vɪnsɪŋlɪ]
adv. 令人信服地；有说服力地
generative [雅思]
美[ˈdʒenərətɪv] | 英[ˈdʒenərətɪv]
adj. 生殖的；生产的；有生殖力的
重点讲解
Disinformation and AI: Generation confusion
虚假信息与人工智能：生成式混乱
disinformation n. （故意发布的）虚假信息
【例】Social media platforms are trying to combat misinformation.
社交媒体平台在打击虚假信息。
【辨析】misinformation n. 错误信息；错误情报
【拓展】false/fake/untrue adj. 虚假的；不真实的
misleading/deceptive adj. 误导性的；欺骗性的
hallucination n. 幻觉；人工智能生成虚假信息
distribution n.
1. 分发；散布
【例】the distribution of food to the flood victims
给洪灾难民分发食物
【近义词】allocation n. 分发；分配
2. 分布
【例】the distribution of panda in China
熊猫在中国的分布
When it comes to disinformation, "social media took the cost of distribution to zero, and generative AI takes the cost of generation to zero," says Renee DiResta of the Stanford Internet Observatory.
斯坦福大学网络观察室的蕾妮·迪雷斯塔表示，对于虚假信息，“社交媒体将传播成本降至零，而生成式人工智能则将生成成本降至零”。
disinformation n. （故意发布的）虚假信息
【例】Social media platforms are trying to combat misinformation.
社交媒体平台在打击虚假信息。
【辨析】misinformation n. 错误信息；错误情报
【拓展】false/fake/untrue adj. 虚假的；不真实的
misleading/deceptive adj. 误导性的；欺骗性的
hallucination n. 幻觉；人工智能生成虚假信息
distribution n.
1. 分发；散布
【例】the distribution of food to the flood victims
给洪灾难民分发食物
【近义词】allocation n. 分发；分配
2. 分布
【例】the distribution of panda in China
熊猫在中国的分布
Large language models such as GPT-4 make it easy to produce misleading news articles or social-media posts in huge quantities. And AI can produce more than text. Cloning a voice using AI used to require minutes, or even hours, of sample audio.
GPT-4 等大语言模型很容易生成大量误导性新闻文章或社交媒体帖子。人工智能可以生成的不仅仅是文本。过去使用人工智能克隆声音需要几分钟甚至几小时的样本音频。
unveil v. 揭幕；展示；公布
【例】to unveil a new product line/a plan
推出新产品线/公布计划
【近义词】release/launch v. 推出
reveal/disclose v. 公布；揭露
Last year, however, researchers at Microsoft unveiled VALL-E, an AI model that is able to clone a person's voice from just a three-second clip of them speaking, and make it say any given text.
然而，去年，微软的研究人员推出了VALL-E，这是一种人工智能模型，能够用一个人说话的三秒片段克隆出他的声音，并让它说出任何给定的文本。
unveil v. 揭幕；展示；公布
【例】to unveil a new product line/a plan
推出新产品线/公布计划
【近义词】release/launch v. 推出
reveal/disclose v. 公布；揭露
OpenAl, the American company behind GPT-4, has developed a similar tool, Voice Engine, which can convincingly clone any voice from a 15-second clip. It has not yet released it, recognising "serious risks, which are especially top of mind in an election year".
GPT-4背后的美国公司OpenAl开发了一款类似的工具“声音引擎”，它可以用15秒的音频片段克隆任何声音，以假乱真。该工具尚未发布，因为OpenAl认识到“其存在严重风险，在选举年尤其受到关注”。
convincingly adv. 令人信服地
【例】All of the actors in Dream of the Red Chamber (1987) played their roles convincingly.
87版《红楼梦》中所有演员的表演都很入戏。
【拓展】convincing adj. 令人信服的
convince v. 说服；使信服
【近义词】persuasively adv. 有说服力地
top of mind 最受关注的；最重要的
【例】For many people, safety is top of mind when purchasing a new car.
对许多人来说，买新车的时候最看重的就是安全性。
【近义词】top of one’s list/agenda 某人最重要的事
synthetic adj. 合成的；人造的
【例】synthetic fabrics 合成织物
【拓展】synthesis n. 合成物；结合体
【前缀】synonym n. 同义词
sync n. 同步；一致
on the ground that 理由是
【例】The proposal was rejected on the ground that it was too costly.
The proposal was rejected on grounds of costs.
这个提案被拒绝了，理由是成本太高。
【近义词】for the reason that 理由是
Similarly, Sora, from OpenAl, can produce surprisingly realistic synthetic videos, in response to text prompts, of up to a minute in length. OpenAI has yet to release Sora to the public, partly on the ground that it could be used to create disinformation.
同样，来自OpenAl的Sora可以根据文本提示词生成逼真程度令人惊讶的合成视频，长度可达一分钟。OpenAI尚未公开发布Sora，部分原因是它可能被用来制造虚假信息。
convincingly adv. 令人信服地
【例】All of the actors in Dream of the Red Chamber (1987) played their roles convincingly.
87版《红楼梦》中所有演员的表演都很入戏。
【拓展】convincing adj. 令人信服的
convince v. 说服；使信服
【近义词】persuasively adv. 有说服力地
top of mind 最受关注的；最重要的
【例】For many people, safety is top of mind when purchasing a new car.
对许多人来说，买新车的时候最看重的就是安全性。
【近义词】top of one’s list/agenda 某人最重要的事
synthetic adj. 合成的；人造的
【例】synthetic fabrics 合成织物
【拓展】synthesis n. 合成物；结合体
【前缀】synonym n. 同义词
sync n. 同步；一致
on the ground that 理由是
【例】The proposal was rejected on the ground that it was too costly.
The proposal was rejected on grounds of costs.
这个提案被拒绝了，理由是成本太高。
【近义词】for the reason that 理由是
As well as providing new ways to discredit or misrepresent politicians, AI tools also raise the spectre of personalised disinformation, generated to appeal to small groups (think soccer moms in a specific town).
除了提供抹黑或歪曲政客的新方法外，人工智能工具还引发了对个性化虚假信息的恐慌，这些虚假信息是为了吸引小群体而产生的（想想某个城镇的足球妈妈）。
discredit v. 败坏名声；使丢脸
【例】They discredited her good name with ugly gossip.
他们散布恶毒的流言蜚语，破坏她的好名声。
【拓展】credit n. 信誉；信用
【近义词】defame v. 污蔑；诽谤
bias n. 偏见；偏向
【例】To avoid bias, the research must be conducted in a controlled and impartial manner.
为了避免偏见的影响，研究必须以受控的、中立的方式进行。
【拓展】prejudice n. （带有敌意的）偏见；歧视
It may even be possible to "microtarget" individuals with disinformation, based on knowledge of their preferences, biases and concerns.
人工智能甚至可以根据个人的偏好、偏向和关心的问题，用虚假信息“微观定位”个人。
discredit v. 败坏名声；使丢脸
【例】They discredited her good name with ugly gossip.
他们散布恶毒的流言蜚语，破坏她的好名声。
【拓展】credit n. 信誉；信用
【近义词】defame v. 污蔑；诽谤
bias n. 偏见；偏向
【例】To avoid bias, the research must be conducted in a controlled and impartial manner.
为了避免偏见的影响，研究必须以受控的、中立的方式进行。
【拓展】prejudice n. （带有敌意的）偏见；歧视
Though**** all of**** this**** is**** worrying****,**** it**** is**** worth remembering that**** not all aspects of**** the technology**** are negative****.**** Al****,**** it**** turns out,**** can be**** used**** for fighting**** disinformation**** as**** well**** as**** producing it****.****
虽然所有这些都令人担忧，但值得记住的是，并非人工智能的所有方面都是负面的。事实证明，人工智能既可以用来制造虚假信息，也可以用来打击虚假信息。
课后练习
What concern does Renee DiResta express about the impact of social media and generative AI?
A. Generative AI will make voice cloning a lengthy process.
B. Social media will change the way people consume information.
C. Social media and generative AI will create an explosion of information.
D. Generative AI will significantly lower the cost of creating disinformation.
Why has OpenAI chosen not to release Voice Engine and Sora to the public?
A. There are concerns about their potential use in creating disinformation.
B. They require further testing to improve voice and video quality.
C. They are still under development and not fully functional.
D. They are too costly to be made publicly available.
What potential threat does AI pose according to the passage?
A. It makes it easier to expose politicians’ misbehavior.
B. It allows for disinformation targeting individuals.
C. It may foster people’s bias towards social issues.
D. It can divide people into small isolating groups.
What is the author's attitude towards generative AI tools?
A. Skeptical, with a belief that they should not be used.
B. Unreservedly positive, focusing only on their benefits.
C. Balanced, acknowledging both positive and negative aspects.
D. Cautious, calling for proper regulation of AI tools and social media.