高考英语 D 篇精读|AI 能不能说真话(2025届高三一轮复习联考(一) D 篇)

哈喽～ 宝子们！今日份高考英语精读来啦🥰 专门挑选了Z20名校联盟2025届高三第二次联考 D 篇 ，原版考场真题原文，中英逐段对照，生词短语都整理好了，附带高考级例句，边读边记轻松提分。

Can artificial intelligence be made to tell the truth? Probably not, but the developers of large language model (LLM) chatbots should be legally required to reduce the risk of errors, says a team of ethicists. "What we're just trying to do is to create a structure to get the companies to put a greater emphasis on truth or accuracy when they are creating the systems," says Brent Mittelstadt at the University of Oxford.
人工智能能被训练得只说真话吗？恐怕不能，但一个伦理学家团队认为，应当从法律上要求大型语言模型（LLM）聊天机器人的开发者降低出错风险。牛津大学的 Brent Mittelstadt 说："我们想做的，是建立一套机制，让这些公司在开发系统时更加重视真实性和准确性。"

LLM chatbots, such as ChatGPT, generate human-like responses to users' questions, based on statistical analysis of vast amounts of text. But although their answers usually appear convincing, they also tend to have "hallucination". "We have these really, really impressive generative AI systems, but they get things wrong very frequently, and as far as we can understand the basic functioning of the systems, there's no fundamental way to revamp that," says Mittelstadt.
LLM 聊天机器人（如 ChatGPT）基于对海量文本的统计分析，生成类似人类的回答。但尽管它们的答案通常显得很有说服力，却也往往会产生"幻觉"。Mittelstadt 说："这些生成式 AI 系统确实非常惊艳，但它们出错的频率很高；就我们对系统基本原理的理解而言，目前没有从根本上解决这一问题的办法。"

This is a "very big problem" for LLM systems, given they are being rolled out to be used in a variety of contexts, such as government decisions, where it is important they produce factually correct, truthful answers, and are honest about the limitations of their knowledge, he says.
他说，这对 LLM 系统来说是个"很大的问题"，因为它们正被推广到各种场景中使用，比如政府决策——在这些场景里，系统给出事实正确、真实可信的答案，并坦诚说明自身知识的局限，这一点至关重要。

To address the problem, Mittelstadt and his colleagues propose a range of measures. They say large language models should react in a similar way to how people would when asked factual questions. That means being honest about what you do and don't know. "It's about doing the necessary steps to actually be careful in what you are claiming," says Mittelstadt. This seems like a praiseworthy aim, but Eerke Boiten at De Montfort University, UK, questions whether the ethicists' demand is technically practical. "I don't understand how they expect legal requirements to mandate what I see as fundamentally technologically impossible," he says.
为解决这一问题，Mittelstadt 和同事提出了一系列措施。他们认为，大型语言模型在被问到事实性问题时，反应方式应与人类相似，也就是坦诚自己知道什么、不知道什么。Mittelstadt 说："关键是要采取必要步骤，在作出论断时真正保持审慎。"这一目标看似值得称赞，但英国德蒙福特大学的 Eerke Boiten 质疑，伦理学家们的要求在技术上是否可行。他说："我不明白，他们怎么能指望用法律规定去强制执行一件在我看来从根本上就技术上不可能的事。"

Mittelstadt and his colleagues do suggest some more straightforward steps that could make LLMs more truthful. The models should link to sources, he says — something that many of them now do to evidence their claims, while the wider use of a technique known as "retrieval-enhanced generation" to come up with answers could limit the likelihood of "hallucinations".
Mittelstadt 和同事确实提出了一些更直接、能让 LLM 更诚实的做法。他说，模型应当链接到信息来源——现在许多模型已经这样做来佐证自己的论断；而更广泛地使用一种被称为"检索增强生成"的技术来生成答案，也能降低产生"幻觉"的可能性。

生词 & 短语

ethicist /ˈeθɪsɪst/ n. 伦理学家
高考例句：The ethicist argued that the experiment was morally wrong.
文中语境：says a team of ethicists

truthfulness /ˈtruːθflnəs/ n. 真实性，诚实
高考例句：We should question the truthfulness of the report.
文中语境：put a greater emphasis on truth or accuracy

hallucination /həˌluːsɪˈneɪʃn/ n. 幻觉，（AI）虚构
高考例句：The fever caused him terrifying hallucinations.
文中语境：they also tend to have hallucination

revamp /ˌriːˈvæmp/ v. 彻底改造，修复
高考例句：The school plans to revamp its old library.
文中语境：there's no fundamental way to revamp that

roll out推出，投放使用
高考例句：The company will roll out the new app next month.
文中语境：they are being rolled out to be used

context /ˈkɒntekst/ n. 语境，场景
高考例句：You need to understand the word in context.
文中语境：a variety of contexts

propose /prəˈpəʊz/ v. 提议，提出
高考例句：They proposed a new plan at the meeting.
文中语境：Mittelstadt and his colleagues propose a range of measures

mandate /ˈmændeɪt/ v. 强制要求，授权
高考例句：The new law mandates that all children wear seat belts.
文中语境：expect legal requirements to mandate what is impossible

retrieval /rɪˈtriːvl/ n. 检索，取回
高考例句：The system speeds up information retrieval.
文中语境：retrieval-enhanced generation

原文重读

Can artificial intelligence be made to tell the truth? Probably not, but the developers of large language model (LLM) chatbots should be legally required to reduce the risk of errors, says a team of ethicists. "What we're just trying to do is to create a structure to get the companies to put a greater emphasis on truth or accuracy when they are creating the systems," says Brent Mittelstadt at the University of Oxford.

LLM chatbots, such as ChatGPT, generate human-like responses to users' questions, based on statistical analysis of vast amounts of text. But although their answers usually appear convincing, they also tend to have "hallucination". "We have these really, really impressive generative AI systems, but they get things wrong very frequently, and as far as we can understand the basic functioning of the systems, there's no fundamental way to revamp that," says Mittelstadt.

This is a "very big problem" for LLM systems, given they are being rolled out to be used in a variety of contexts, such as government decisions, where it is important they produce factually correct, truthful answers, and are honest about the limitations of their knowledge, he says.

To address the problem, Mittelstadt and his colleagues propose a range of measures. They say large language models should react in a similar way to how people would when asked factual questions. That means being honest about what you do and don't know. "It's about doing the necessary steps to actually be careful in what you are claiming," says Mittelstadt. This seems like a praiseworthy aim, but Eerke Boiten at De Montfort University, UK, questions whether the ethicists' demand is technically practical. "I don't understand how they expect legal requirements to mandate what I see as fundamentally technologically impossible," he says.

Mittelstadt and his colleagues do suggest some more straightforward steps that could make LLMs more truthful. The models should link to sources, he says — something that many of them now do to evidence their claims, while the wider use of a technique known as "retrieval-enhanced generation" to come up with answers could limit the likelihood of "hallucinations".

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 

本篇精读结束，日积月累坚持打卡，英语阅读和词汇能力稳步提升，我们下期再见～ 今天的文章就学到这里，坚持精读，阅读稳步提分！