AI会"吃掉"自己？科学家找到解药

【议论文·合集6】| 高考英语阅读备考（30分）
来源：LiveScience | 难度：⭐⭐⭐ | 词数：约798词
涉及领域：人工智能 / 科技伦理
提分要点：观点推断题、作者态度题、议论文论证结构分析

📖 深度解析：当AI开始"近亲繁殖"，人类数据就是那一剂解药

本文探讨AI领域一个前沿议题：模型崩溃（model collapse）。文章采用"问题—发现—方案"的经典议论文结构，先指出大语言模型面临的数据枯竭危机，再引出伦敦国王学院Roudi教授团队的突破性发现——仅需一个人类数据点即可避免崩溃。作者态度明确但克制，既引用了"年底耗尽数据"的紧迫预言，也承认模型崩溃尚未在真实部署中发生，体现了科技议论文应有的客观平衡立场。

⚡ 语法考点：①文中频繁出现让步状语从句（while/although），如 "While the evolution of AI systems has shown no sign of slowing, there's a growing concern that..."——这类"让步+转折"结构是完形和阅读理解的高频逻辑考点。②虚拟语气出现在专家引用中："if they experienced model collapse, these machines could misdiagnose people"，语法填空中if条件句的时态搭配是必考点。

✍️ 写作可抄："There's a growing concern that..."（人们越来越担心……）是议论文开头引出争议话题的万能句式，可替换that从句内容用于任何社会议题。

🔑 核心词汇精讲

📰 中英对照正文

While the evolution of artificial intelligence (AI) systems has shown no sign of slowing, there's a growing concern that large language models (LLMs) will soon run out of human-made data to ingest and learn from.

虽然人工智能（AI）系统的发展没有放缓的迹象，但人们越来越担心大型语言模型（LLMs）很快将耗尽可供摄取和学习的人造数据。

Once this happens, scientists say, AI models will increasingly rely on synthetic AI-made information, which will lead to an effect called "model collapse." This is where LLMs spout gibberish and the AI systems they underpin deliver inaccurate answers and hallucinate information to queries far more commonly than they do today.

科学家们表示，一旦这种情况发生，人工智能模型将越来越依赖人工智能合成的信息，这将导致一种被称为"模型崩溃"的效应。在这种情况下，大语言模型会胡言乱语，它们所支撑的系统会给出不准确答案，对查询信息进行虚构的情况会比现在更加常见。

"That's especially worrying considering some experts think that we will run out of high-quality human-generated data by the end of the year," Yasser Roudi, a professor at King's College London, told Live Science. "If you had LLMs that were used in hospitals to analyze brain scans and find cancers, if while training another model they experienced model collapse, these machines could misdiagnose people."

伦敦国王学院教授亚西尔·鲁迪告诉《生活科学》杂志："考虑到一些专家认为我们到今年年底将耗尽高质量的人工生成数据，这尤其令人担忧。如果你有用于医院分析脑部扫描和检测癌症的大语言模型，在训练另一个模型时如果出现模型崩溃，这些机器可能会误诊患者。"

However, Roudi recently found that model collapse can be bypassed by adding a single human-made data point to an AI's training data, even if all the other data is AI-generated. The study was published May 14 in the journal Physical Review Letters.

然而，鲁迪最近发现，即使其他所有数据都是由人工智能生成的，通过向训练数据中添加一个人为的数据点，也可以避免模型崩溃。这项研究于5月14日发表在《物理评论快报》杂志上。

The researchers discovered that by introducing a single external human-made data point to a pool of synthetic data used in closed-loop training, they avoided model collapse. "This data point would be linked to a 'ground truth,' something we know undeniably to be true and independently verifiable," Roudi said.

研究人员发现，通过将一个单一的外部人造数据点引入闭环训练的合成数据池中，他们避免了模型崩溃。鲁迪说："这个数据点将与一个'基本事实'相关联，即我们毫无疑问知道是真实且可独立验证的事物。"

"This research is the first step in setting out some ground rules for preventing this from happening in the future," Roudi concluded. "While more work needs to be done, AI engineers developing the next ChatGPT-like product can use our findings to develop models that won't collapse."

鲁迪总结道："这项研究是制定基本规则以防止此类事情在未来发生的第一步。虽然还需要做更多工作，但开发下一个类似ChatGPT产品的人工智能工程师可以利用我们的发现来开发不会崩溃的模型。"

⚖️ 观点交锋

🔴 主流观点：AI数据即将耗尽，合成数据是唯一的出路——但合成数据本身会导致模型崩溃，形成死循环。
🟢 作者/受访者立场：Roudi团队发现"一个人类数据点"即可打破死循环，为AI可持续发展提供理论基础。
📌 作者态度：客观审慎——既呈现了"年底数据耗尽"的紧迫感，也指出模型崩溃尚未在真实系统中发生，研究仍处于早期阶段。

🧩 阅读理解

阅读下列各题，选出最佳答案。

1. What is "model collapse" as described in the article?
- A. A situation where AI models become too large to run efficiently.
- B. A phenomenon where AI systems produce increasingly inaccurate and meaningless outputs.
- C. A process where AI models learn from human data too slowly.
- D. A condition where AI hardware physically breaks down.

2. According to Roudi's research, what is the key to preventing model collapse?
- A. Training AI models only on human-generated data.
- B. Making AI models smaller and simpler.
- C. Adding at least one human-made "ground truth" data point to synthetic training data.
- D. Stopping AI models from using synthetic data entirely.

3. What can be inferred about the author's attitude toward Roudi's findings?
- A. Completely dismissive of the research's practical value.
- B. Cautiously optimistic while acknowledging limitations.
- C. Strongly convinced that the problem is already solved.
- D. Indifferent to the implications for AI development.

📌 参考答案：1-B　2-C　3-B

✍️ 信息提取

根据文章内容，用不超过三个词填写空白处。

① AI models may run out of __ data to learn from.
② Model collapse causes AI to _ information to queries.
③ Roudi's team published their findings in ___ journal.
④ The human-made data point must be linked to a verifiable ______.

📌 参考答案： ① human-made/human-generated ② hallucinate ③ Physical Review Letters ④ ground truth

💡 长难句手术室

原句：
"[That's especially worrying] (considering some experts think) [that we will run out of high-quality human-generated data by the end of the year] — so [if you're relying on this synthetic data], (but there's an almost existential threat it will sink your AI), you're in trouble."

符号解析：
- 主干：That's especially worrying... you're in trouble（这尤其令人担忧……你就麻烦了）
- [ ] = that引导的宾语从句（think的宾语）和if条件状语从句
- ( ) = considering引导的状语（考虑因素）和but引导的转折补充

⚡ 手术结论： "That's especially worrying considering that..."是议论文中表达担忧的高级句型。高考写作套用：That's especially concerning considering that + 令人担忧的事实。

💬 你站哪边？ 🔵 AI应该继续依赖合成数据扩大规模 🟢 AI必须保留人类数据作为"锚点"。评论区告诉我，下期重点讲！

读懂世界，拿捏高考。点击"在看"，领取本篇外刊 PDF 精读版。