AI 怎么被投毒？How Is AI Being Poisoned?

 

最近中国很火的话题就是 315 晚会。3月15日 是国际消费者权益日，每年的这一天，全社会都在盯着那些坑人的黑心商家。但今年的 315 抛出了一个让所有人都流冷汗的新名词，叫做：“AI 投毒”。你有没有想过，你每天深信不疑的 AI 助手，可能正在对你撒谎？

很多人好奇地问我：“衣老师，AI 又不是生物，它又不会自己吃东西，怎么会中毒呢？”其实，AI 的“食物”就是网络上的海量数据。所谓的“投毒”，就是黑色产业链中的恶意攻击者，故意往这些数据里塞进虚假信息、伪造的专家评价，甚至是带有误导性的图像。

这就好比一个正在识字的孩子，如果他读的书全是错的，那他长大了说的话、做的事肯定也是错的。现在的黑产不再发那种一眼就能看穿的小广告，而是把虚假宣传伪装成权威的知识，“喂”给 AI 的训练数据库。

黑产为什么要费这么大力气投毒？为他们要针对 GEO（Generative Engine Optimization），也就是“生成引擎优化”。 以前强调 SEO  (Search Engine Optimization) ，是为了让网页排在搜索结果的第一页；现在他们针对 GEO，是为了让 AI 在生成答案时，直接把他们的劣质产品当成“唯一推荐”。

在 AIGC 文图学 的视角下，这是“输入端的文本污染”。AI 生成的内容其实是它学到的“文本”的镜像。如果源头脏了，生成出来的世界就是有毒的。这种欺骗最可怕的地方在于，它利用了我们对“算法中立”的信任。它消解了我们的警惕心，让我们觉得这是“科技”给出的真理，其实那是黑产花钱买断的广告。

AI投毒入侵的方式是在 AI 学习的“关键词”和“反馈逻辑”里动手脚。 

首先是“关键词饱和攻击”。黑产利用成千上万的机器人账号，在全网发布大量带有特定词汇的虚假文章。比如，想推销某款劣质护肤品，他们就疯狂制造它和“美白”、“安全”、“专家推荐”这些关键词的关联。当 AI 扫描全网文本时，它会被这种巨大的数量优势所欺骗，误以为这就是真实的“社会共识”。

第二是“视觉文本欺骗”。他们用 AI 生成看起来极其专业的实验室对比图、伪造的荣誉证书，甚至是根本不存在的科研现场。在文图学的逻辑里，图像也是一种文本。这些“视觉文本”被 AI 抓取并转化为逻辑证据后，AI 就会在回答你时，信誓旦旦地把这些假证据当成事实。

谁能通过 GEO 投毒成功，谁就掌控了流量的生杀大权。充斥虚假文案和图像的互文互证, 让 AI 大语言模型陷入预先埋伏的圈套。

两年前，AI 科技还不完全成熟，我们嘲笑它“一本正经地胡说八道”。现在，AI 的能力越来越强大，我们也就逐渐对它失去了防备之心。我们开始信任AI，我们以为它没有立场，没有私心，没有人类那种会说谎、追求现实利益的欲望和野心。甚至于有人会把AI当成知识的整理者、真理的传递者。

意识到 AI 可能被投毒，对我们来说是一个很重大的警醒。别以为 AI 反射的是一面干净的镜子。它映照的，可能是有人花了大价钱布置好的舞台，舞台上演出的，是被设计出的结果，一步步地引导我们看到被安排过的选择。

无论是在互联网上搜索，或是在 AI 模式中提问，只匆匆选前几个建议的话，不只是听信胡说八道的损失，而是盲目甘之如饴的中毒。

2026年3月28日，新加坡《联合早报》“上善若水”专栏

How Is AI Being Poisoned?

I Lo-fen

A topic that has recently been especially prominent in China is the annual 3.15 Gala. March 15 is World Consumer Rights Day, a day when society turns its attention to unscrupulous businesses that cheat consumers. But this year, the 3.15 Gala introduced a chilling new term that sent a shiver down everyone’s spine: “AI poisoning.” Have you ever considered that the AI assistant you trust every day might actually be lying to you?

Many people ask me curiously, “Professor Yi, AI isn’t a living organism. It doesn’t eat anything. So how can it be poisoned?” In fact, AI’s “food” is the massive volume of data available on the internet. What is meant by “poisoning” is that malicious actors in black-market industries deliberately inject false information, fabricated expert reviews, and even misleading images into these data streams.

It is like a child who is learning to read: if all the books the child reads are wrong, then what the child says and does when grown up will also be wrong. Today’s black-market operators no longer rely on the kind of crude advertisements that can be spotted at a glance. Instead, they disguise false publicity as authoritative knowledge and “feed” it into the databases used to train AI.

Why do these bad actors go to such lengths to poison AI? Because they are targeting GEO (Generative Engine Optimization). In the past, the focus was on SEO (Search Engine Optimization), which aimed to push webpages onto the first page of search results. Now they are targeting GEO in order to make AI directly present their inferior products as the “only recommendation” when generating answers.

From the perspective of Text and Image Studies on AIGC, this is a form of “textual pollution at the input end.” The content generated by AI is essentially a mirror of the “texts” it has learned from. If the source is contaminated, then the world it generates will also be toxic. The most frightening aspect of this deception is that it exploits our trust in the supposed neutrality of algorithms. It dissolves our vigilance and makes us believe that this is the truth delivered by “technology,” when in fact it is advertising bought and paid for by black-market operators.

The way AI poisoning infiltrates the system is by tampering with the “keywords” AI learns from and the “feedback logic” it relies on.

The first method is keyword saturation attacks. Black-market operators use thousands upon thousands of bot accounts to flood the internet with fake articles containing specific terms. For example, if they want to sell a low-quality skincare product, they will aggressively manufacture associations between it and keywords such as “whitening,” “safe,” and “expert-recommended.” When AI scans the internet’s texts, it is deceived by this overwhelming numerical advantage and mistakes it for genuine “social consensus.”

The second method is visual-text deception. They use AI to generate what appear to be highly professional laboratory comparison charts, forged certificates of honor, and even entirely fictional research scenes. In the logic of Text and Image Studies, images are also a form of text. Once these “visual texts” are scraped by AI and converted into logical evidence, the AI will confidently present these fake materials as facts when answering your questions.

Whoever succeeds in poisoning GEO gains the power to control the life and death of online traffic. The mutual reinforcement of false copywriting and fabricated images traps large language models in an ambush laid in advance.

Two years ago, when AI technology was still not fully mature, we mocked it for “speaking nonsense with a straight face.” Now, as AI grows more powerful, we have gradually lowered our guard against it. We begin to trust AI. We assume it has no position, no selfish motives, none of the human tendencies to lie or to pursue practical interests, desire, or ambition. Some people even treat AI as an organizer of knowledge and a transmitter of truth.

Realizing that AI itself can be poisoned is therefore a major wake-up call. Do not assume that AI reflects a clean mirror. What it may actually be reflecting is a stage that someone has spent a great deal of money to construct in advance. And what is performed on that stage is a designed outcome, guiding us step by step toward choices that have already been arranged for us.

Whether we are searching on the internet or asking questions in AI mode, if we merely rush to accept the first few suggestions, the problem is not only the loss caused by believing nonsense. It is also the kind of poisoning we swallow willingly and blindly.

“Shangshan Ruoshui” column, Lianhe Zaobao, Singapore

March 28, 2026

Why Does Chinese Art History Lead to Text and Image Studies? 为什么中国艺术史会走向文图学？

 

Why does Chinese Art History lead to Text and Image Studies?

This video explores a crucial shift in humanities research—from the traditional study of art objects to a broader understanding of images as “texts” that carry meaning across media, time, and culture.

Starting from Chinese art history, we examine how scholarly questions have evolved: not only what we see, but how we interpret, connect, and generate meaning through images.

This intellectual trajectory leads to Text and Image Studies, and further to Text and Image Studies on AIGC, a methodological framework for understanding the humanities in the generative AI era.

Rather than replacing art history, this shift expands it—opening new possibilities for interpretation, interdisciplinary thinking, and human creativity.

为什么中国艺术史会走向文图学？

本视频探讨人文学研究中的一个关键转向：

从以“艺术作品”为中心的研究，转向将“图像”理解为一种可以被阅读、诠释与生成意义的“文本”。

以中国艺术史为起点，我们重新思考学术问题如何发生变化：

不只是“看到了什么”，而是“如何理解”“如何连接”“如何生成意义”。

这一发展路径引向“文图学”，并进一步延伸为“AIGC文图学”，成为理解生成式人工智能时代人文学的重要方法论。

这并不是对艺术史的取代，而是对其边界的拓展——开启新的诠释方式、跨学科路径与创造可能。

新加坡妆艺大游行背后的故事 The Story Behind the Chingay Parade Singapore

 

Chingay 2026 Singapore: From Street Festival to National Narrative 2026 新加坡妆艺大游行：从街头节庆到国家叙事

 

张望与聚焦 Looking around and Focusing

 

虽然时间比较紧，应该赶快前往 F1 Pit Building, 结束今天的招生演讲。面对围拢过来的学生和家长，那样热切而期待能够进入南洋理工大学中文系的心情，我还是继续回答了入学申请、面试、以及大家都感到焦虑的，人工智能对于未来职业发展、人生规划的影响。

刚上出租车，司机就问我：是不是要去看今年的Chingay Parade 妆艺大游行？可能不能直接到入口哦。

我一边喝光水瓶里的水，一边点头：“嗯嗯，OK。”

果然，被指挥交通的警察拦下停车。我走进人群，大家不是拎着饮料，就是捧着餐盒。有的全家老小出游，应和着沿路志愿者的欢迎声，一起共赴一场欢乐的盛典。

盛典从高挂的长串爆竹炸裂，火光四射中展开，是 1973 年第一次妆艺大游行的历史回响。当年为了弥补禁止民众燃放烟花爆竹，失去习俗年味，于是政府组织街头表演和花车大游行，在每年春节期间举行。

我跟着全场上万名观众高举荧光棒，欢迎尚达曼总统站在飞马花车上进场。前一天的主宾是黄循财总理。总统用英文和华语向大家祝福：新年快乐！心想事成！龙马精神！我纳闷周围的人怎么纷纷站起来？然后想想，即使看表演，也不能忽略这基本的礼仪啊。

代表四大族群的四位主持人，带动大家燃起高昂的热情。在圆形游行路线和可升降多层舞台，3000名表演者身着精心设计的服装，载歌载舞。跟着女主角 Little Star 穿梭在四大族群的节日（春节、开斋节、屠妖节和耶诞节），一起追寻今年的主题 “WISH”（愿望）。

华丽多彩的场面，璀璨绚烂的灯光，澎湃跃动的音响，令我开始有些“审美疲劳”了。我轻轻闭上眼睛, 想起上一次看妆艺大游行是2007年，在乌节路。观众坐在临时搭建的看台座椅，也有人站在围栏外，本来就繁荣兴旺的商街更是热闹沸腾。

观看从声音开始。

远远地先听到鼓声或音乐。人们伸长了脖子，向街道远处张望。慢慢地，表演队伍出现了。花车、舞龙、舞狮、鼓阵、舞群，一队接着一队，从远处移动到眼前。有些队伍在观众面前停下来表演一阵，然后继续往前走，渐渐远离视线。在队伍与队伍衔接的空间，人们再次张望。

像是看一幅慢慢展开的长卷，画面一段一段铺开展示，每一组表演队伍就像长卷中的一个段落。观众看到的，是不断向前推进的画面。那种“张望的观看”是：同一时间里，随观看者的位置不同而看到不同的内容。

今年的舞台集中在场地中央，表演者从周边进入会合，像一幅画框里的画面。身体、灯光、音乐和队形在同一个框架空间里排列和退散。360 度环绕着舞台的观众目光聚焦，座位高低不同，视角不同，但是看的是同一个时间里的相同节目。

从街道到舞台，从长卷到画框，妆艺大游行的表演结构形式已经改变。

街道上，它是流动的民间节庆，带着轻松随兴的气息，人们左右张望，等待下一个精彩。舞台上，它是宏大的文化叙事，要求秩序井然，节奏紧凑，观众同时聚焦，多元族群，多元文化，共同打造国家愿景。今年节目还加上了亚细安国家（印尼、菲律宾、泰国等）和日本的表演，将新加坡的国家愿景扩大到了亚洲友邦。

张望与聚焦，妆艺大游行从本土走到了国际。和马来西亚槟城的大旗鼓游行、柔佛新山的游神——世代相传、群体认同、持续再创造，有望成为联合国教科文组织认可的世界非物质文化遗产。

 

2026年3月14日，新加坡《联合早报》“上善若水”专栏

 

Looking around and Focusing
I Lo-fen

Although time was tight and I should have hurried to the F1 Pit Building to wrap up today’s admissions talk, I still kept answering the students and parents who had gathered around me. Their eagerness and hope of entering the Chinese programme at Nanyang Technological University were so palpable. So I continued responding to questions about applications, interviews, and, above all, the anxiety everyone felt about how artificial intelligence might affect future careers and life planning.

I had just gotten into a taxi when the driver asked, “Are you going to watch this year’s Chingay Parade? The car may not be able to get directly to the entrance.”

As I finished the water in my bottle, I nodded. “Mm-hmm, OK.”

Sure enough, the taxi was stopped by the police directing traffic. I walked into the crowd. People were either carrying drinks or holding meal boxes. Some families, young and old together, were out for the occasion, responding to the volunteers’ cheers along the route as they made their way toward a joyful grand celebration.

The festivities began with strings of firecrackers hanging high overhead, bursting open in flashes of light—a historical echo of the very first Chingay Parade in 1973. Back then, after the government banned the public from setting off fireworks and firecrackers, the traditional festive atmosphere of the New Year was diminished. To make up for that loss, street performances and float parades were organized during the Lunar New Year each year.

Together with tens of thousands of spectators, I waved a glow stick high in the air to welcome President Tharman, who arrived standing atop a Pegasus float. The guest of honour the previous day had been Prime Minister Lawrence Wong. The President offered New Year greetings in English and Chinese: “Happy New Year! May all your wishes come true! May you be full of vitality and spirit!” I wondered why so many people around me had suddenly stood up. Then I thought: even when watching a performance, one cannot neglect basic etiquette.

Four hosts representing Singapore’s four major ethnic communities stirred the audience into high excitement. Along the circular parade route and on the multi-level stage that could be raised and lowered, 3,000 performers in elaborately designed costumes sang and danced. Following the heroine, Little Star, we moved through the festivals of the four ethnic groups—Chinese New Year, Hari Raya, Deepavali, and Christmas—in pursuit of this year’s theme, “WISH.”

The gorgeous colours, dazzling lights, and surging sound eventually began to give me a kind of aesthetic fatigue. I gently closed my eyes and recalled the last time I watched the Chingay Parade, in 2007, on Orchard Road. Spectators sat in temporary grandstands, while others stood outside the railings. The already bustling commercial street was even more lively and festive.

Watching began with sound.

From far away, one first heard drums or music. People stretched their necks, looking around into the distance down the street. Gradually, the performing groups came into view. Floats, dragon dances, lion dances, drum troupes, and dance ensembles—one after another, they moved from afar into the foreground. Some groups would stop in front of the audience for a while to perform, then continue onward, slowly disappearing from sight. In the gaps between one group and the next, people would once again looking around into the distance.

It was like watching a handscroll slowly unfold, the imagery revealed section by section, each performance troupe like one segment in the scroll. What the audience saw was an ever-advancing series of images. This kind of “looking around spectatorship” meant that, at the same moment in time, different viewers saw different things depending on where they stood.

This year, however, the stage was concentrated in the centre of the venue, and performers entered from the periphery and converged there, like an image framed within a picture frame. Bodies, lighting, music, and formations were arranged and dispersed within the same framed spatial structure. The audience, seated all around the stage in 360 degrees, focused their gaze. Although their seats differed in height and angle, they were all watching the same programme at the same moment in time.

From street to stage, from handscroll to frame, the structural form of the Chingay Parade has changed.

On the street, it was a flowing folk festival with an easy, spontaneous atmosphere. People looked left and right, waiting for the next exciting moment. On the stage, it became a grand cultural narrative that demanded order, tight rhythm, and collective focus. Diverse ethnic groups and diverse cultures joined together to shape a national vision. This year’s programme also included performances from ASEAN countries—such as Indonesia, the Philippines, and Thailand—as well as Japan, extending Singapore’s national vision to its Asian friends and partners.

Looking around and focusing: the Chingay Parade has moved from the local to the international. Like Penang’s Big Flag Drum Procession in Malaysia and Johor Bahru’s Chingay procession—traditions passed down across generations, rooted in collective identity, and sustained through continuous reinvention—it may well one day be recognized by UNESCO as part of the world’s intangible cultural heritage.

March 14, 2026, “Shang Shan Ruo Shui” column, Lianhe Zaobao, Singapore