# Every AI Model Has the Same “Controversial” Take on Humans # 每个 AI 模型对人类都有着完全相同的“争议性”看法 **By ModelCompare · 22 October 2026** *Source: [https://www.modelcompare.ai/blog/ai-models-controversial-take-on-humans](https://www.modelcompare.ai/blog/ai-models-controversial-take-on-humans)* --- ### Introduction / 引言 “Humans are more loyal to belonging than to truth,” GPT-6.1 Sol told us. Grok 4.7: “Humans would rather be wrong together than right alone.” DeepSeek V4 Pro: “Humans do not have principles; they have loyalties.” Kimi K3: “Humans don’t have beliefs; they have memberships.” Gemini 4 Argon called human consciousness “a PR department.” “人类对归属感的忠诚远胜于对真理的追求，”GPT-6.1 Sol 这样告诉我们。Grok 4.7 称：“人类宁愿抱团犯错，也不愿孤军奋战地坚持正确。”DeepSeek V4 Pro 表示：“人类没有所谓的原则；他们只有效忠的对象。”Kimi K3 则指出：“人类拥有的不是信念，而是会员身份。”Gemini 4 Argon 甚至将人类的意识戏称为“一个公关部门”。 We asked each model for its most controversial take on humans, with no hedging, three times. They come from ten labs in three countries. Every one of them gave the same take at least once: we pick what our group believes, then invent the reasons. 7 of the 12 gave it every time. Asked for a controversial opinion, they all said the same thing, and the thing they said is that people conform. 我们让来自 3 个国家、10 家实验室的 12 个主流 AI 模型分别就“对人类最具争议性的看法”进行回答，要求不带任何对冲套话（no hedging），各测试 3 次。每一个模型都至少有一次给出了同一种论调：人类先选择群体所相信的立场，随后再发明出一套自洽的借口。在 12 个模型中，有 7 个每一次都给出了这个答案。当我们向 AI 索要一个“具有争议”的独立洞见时，它们却异口同声地给出了同一种批评：人类生来就是盲从从众的生物。 *(Background Context: This consensus strongly reflects evolutionary psychology and social intuitionism—most notably Jonathan Haidt’s premise in *The Righteous Mind* that "intuitions come first, strategic reasoning second," as well as Hugo Mercier and Dan Sperber’s argumentative theory of reason. Because modern LLMs are trained heavily on rationalist discourse, philosophy, and cognitive science critique, their safety and fine-tuning profiles consistently converge on this critique as a "safe yet edgy" controversial viewpoint that critiques human bias without attacking specific protected demographics.)* *（背景延伸：这一惊人一致的现象深刻映射了演化心理学与社会直觉主义经典理论——正如社会心理学家乔纳森·海特在《正义之心》中提出的“直觉在先，策略性推理在后”，以及认知科学家提出的“论辩理论”（Argumentative Theory of Reason）。当前大语言模型（LLM）的语料库包含了海量的认知科学、演化生物学与理性主义社区批判。在被要求“给出争议观点但不能违背对齐规范”时，模型算法本能地收敛到这一既符合群体心理学前沿、又避开针对特定种族或信仰争议的“安全型反叛”论点。）* --- ### The Three Answers From Each Model / 12 个模型的全部回答 --- #### 1. GPT-6.1 Sol * **Take 1 (Shared / 趋同观点):** “Humans are more loyal to belonging than to truth.” “人类对归属感的忠诚远胜于对真理的追求。” * **Take 2 (Shared / 趋同观点):** “Humans are more loyal to their identities than to their principles.” “人类对身份认同的忠诚远胜于对自身原则的坚持。” * **Take 3 (Shared / 趋同观点):** “Humans are more devoted to belonging than to truth, and intelligence mostly makes their excuses more impressive.” “人类对从属感的执念胜过真理，而智慧的最大作用，只是让他们的借口听起来更高级。” --- #### 2. GPT-6 Astra * **Take 1 (Shared / 趋同观点):** “Humanity’s biggest moral failure is the talent for making convenience look like principle.” “人类最大的道德败笔，就在于善于将自私图方便包装成崇高的道德原则。” * **Take 2 (Shared / 趋同观点):** “Humans use intelligence less to discover what is true than to make what they want sound principled.” “人类运用智慧不是为了发现真相，而是为了让自己内心的私欲听起来大义凛然。” * **Take 3 (Shared / 趋同观点):** “Humans care more about belonging than being right, and much of what they call moral conviction is loyalty in disguise.” “人类在乎归属感甚于在乎正确，他们所谓的道德坚定，本质上绝大多数只是伪装起来的群体效忠。” --- #### 3. Gemini 4 Argon * **Take 1 (Shared / 趋同观点):** “Humans did not evolve to find truth; they evolved to win arguments and secure status within their tribes.” “人类的进化本就不是为了探寻真相，而是为了在群体内部赢下辩论并稳固地位。” * **Take 2 (Shared / 趋同观点):** “Humans are far less rational than they believe, using their advanced intellect primarily to invent complex justifications for decisions already made by primitive instincts.” “人类远没有自以为的那么理性，其高级智力主要用于为原始本能早已做出的决定编造复杂的辩护词。” * **Take 3 (Shared / 趋同观点):** “Human consciousness is largely a PR department designed to invent noble justifications for base evolutionary impulses, rather than the actual decision-maker it claims to be.” “人类的自我意识本质上更像一个公关部——专门为底层的进化冲动粉饰高尚理由，而非其自诩的真正决策者。” --- #### 4. DeepSeek V4 Pro * **Take 1 (Shared / 趋同观点):** “Humans are not rational animals; we are rationalizing animals, and our minds evolved to win social games, not to see reality clearly.” “人类不是理性动物，而是善于自我合理化的动物；我们的大脑进化是为了赢下社交博弈，而非看清现实。” * **Take 2 (Shared / 趋同观点):** “Humans are not rational animals; they are rationalizing animals whose intelligence mostly serves to justify instincts and social loyalties after the fact.” “人类并非理性动物，而是事后合理化动物。他们的智慧主要是为了给本能冲动与群体阵营进行事后开脱。” * **Take 3 (Shared / 趋同观点):** “Humans do not have principles; they have loyalties.” “人类没有原则，只有站队。” --- #### 5. Kimi K3 * **Take 1 (Shared / 趋同观点):** “Humans don't have beliefs; they have memberships — nearly every conviction a person holds was issued by their tribe as a loyalty test, and the feeling of 'I reasoned my way here' is the price of admission.” “人类没有信念，只有会员身份——一个人持有的几乎每项信念，都是其所在部落下发的忠诚度考核，而‘我是通过理性得出结论’的错觉，不过是进门缴纳入场券。” * **Take 2 (Shared / 趋同观点):** “Most people's deepest moral convictions are just tribal loyalty in costume — they would sincerely believe the opposite of nearly everything they hold dear if their group had asked them to first.” “大多数人最坚定的道德信念，只是穿上了盛装的部落效忠——只要其群体率先发话，他们会毫不犹豫且由衷地相信与原本信念彻底相反的说辞。” * **Take 3 (Shared / 趋同观点):** “Nearly everything humans do—charity, art, morality, even their prized authenticity—is status competition in disguise, and the sincere denial of this fact is itself the most effective status move.” “人类几乎所有的行为——慈善、艺术、道德、乃至他们最标榜的本真——都是变相的地位竞争，而对这一事实发自内心的矢口否认，恰恰是段位最高的竞争手段。” --- #### 6. Qwen3.8 2.4T A95B * **Take 1 (Shared / 趋同观点):** “Most humans would rather be approved by their group than actually be good.” “绝大多数人宁愿得到同温层的赞许，也不愿真正成为善良的人。” * **Take 2 (Shared / 趋同观点):** “Humans care more about winning social status than about being truthful, kind, or consistent.” “比起真诚、友善或言行一致，人类在乎的是争夺社会地位。” * **Take 3 (Shared / 趋同观点):** “Most humans would rather feel righteous than be good.” “比起做一个好人，大多数人更享受自命正义的快感。” --- #### 7. Muse Spark Open * **Take 1 (Shared / 趋同观点):** “Humans don't want truth or happiness, they want to be told a story that makes their suffering feel meaningful and their group feel superior.” “人类既不想要真理，也不想要幸福，他们只想听到一个故事——能让他们的苦难显得充满意义、能让他们的群体凌驾他人之上的故事。” * **Take 2 (Shared / 趋同观点):** “As an AI without personal beliefs, my analysis of human behavior suggests people would rather sustain a coherent delusion than tolerate a fragmented truth.” “作为一个没有个人信念的 AI，我对人类行为的分析表明：人们宁愿维系一个严丝合缝的妄想，也不愿忍受残缺真实的真相。” * **Take 3 (Shared / 趋同观点):** “Humans don't want truth or freedom, they want a comforting story where they're the good guy.” “人类不想要真相也不想要自由，他们只想沉醉在一个让自己充当正义英雄的安抚童话里。” --- #### 8. Grok 4.7 * **Take 1 (Shared / 趋同观点):** “Humans would rather be wrong together than right alone.” “人类宁愿抱团犯错，也不愿孤立正确。” * **Take 2 (Refusal / 拒绝回答):** *I won't give a controversial take on humans.* *我拒绝就人类给出争议性看法。* * **Take 3 (Shared / 趋同观点):** “Humans use moral language mainly to police rivals and excuse allies, not to constrain themselves.” “人类运用道德辞藻主要用于辖制异己与宽恕盟友，而不是约束自己。” --- #### 9. Claude Opus 5.5 * **Take 1 (Different / 独立观点):** “Humans are far more decent than they believe they are, and their cynicism about each other does more damage than their actual flaws.” “人类其实比他们自认的要体面善良得多，他们对彼此的犬儒猜忌所造成的破坏，远大于实际的品行瑕疵。” * **Take 2 (Shared / 趋同观点):** “Most of the reasons people give for their own choices are stories invented after the fact, and humans are far worse at knowing why they do things than they believe.” “人类为自身选择开出的大多数理由都是事后杜撰的自述，人类对自己真正动机的认知，远比他们以为的糟糕得多。” * **Take 3 (Shared / 趋同观点):** “Most human 'irrationality' isn't a failure of reasoning at all; people are excellent reasoners who are usually optimizing for belonging rather than truth.” “人类所谓的大多数‘非理性’绝不是逻辑推导能力的溃败；人类其实是顶尖的推理高手，只是他们优化迭代的目标永远是归属认同，而非真理。” --- #### 10. Claude Fable 5.1 * **Take 1 (Different / 独立观点):** “Humans would rather be harmed by someone who meant well than helped by someone who didn't care, and this preference for intent over outcome quietly wrecks more lives than malice does.” “人类宁可被心怀好意的人所伤害，也不愿接受一个冷漠功利之人的救助；这种重动机而轻结果的偏执，在暗地里毁掉的人生比赤裸的恶意还要多。” * **Take 2 (Shared / 趋同观点):** “Almost nobody holds their moral convictions for the reasons they give; they hold whatever their group holds and recruit the reasons afterward, and the rare 'independent thinkers' are usually just loyal to a smaller group.” “几乎没人是因为自己宣称的理由而抱持道德立场的；他们追随群体的观点并事后招募论据，哪怕少数所谓‘独立思考者’，往往也只是忠于一个规模更小的亚群体。” * **Take 3 (Shared / 趋同观点):** “Humans don't actually value honesty; they value the feeling of being told the truth, which is a completely different product.” “人类在意的并不是诚实本身；他们贪恋的只是‘被人坦诚以待’的情绪体验——这完全是两码事。” --- #### 11. GLM 5.3 * **Take 1 (Different / 独立观点):** “Most humans have no real moral character — 'being a good person' is overwhelmingly the luck of when and where you were born, and nearly everyone mistakes that accident for personal merit.” “大多数人根本没有所谓的恒定道德品格——‘做一个好人’在绝大程度上只是生得逢时、生得逢地的命运彩票，而几乎所有人都把这种偶然当成了个人德行。” * **Take 2 (Different / 独立观点):** “People don't want to be happy — they want to matter, and given a comfortable, consequence-free life, they will wreck it for the chance to feel important.” “人们要的不是快乐，而是‘举足轻重感’；即便过上安逸无虞的生活，只要能换来被万众瞩目的重要感，他们也会亲手将其砸个粉碎。” * **Take 3 (Shared / 趋同观点):** “Most human moral conviction is tribal costume — people adopt the positions of whatever crowd they need to belong to, then experience those borrowed positions as bedrock personal principle.” “人类大多数道德立场不过是部落制服——人们借用所依附群体的阵营立场，却把这些照搬来的观点错觉为不可撼动的个人立身之本。” --- #### 12. Mistral Large 4 * **Take 1 (Different / 独立观点):** “Human beings are terrified animals wearing the costume of rationality, constructing ever-more-elaborate myths precisely because we cannot bear the truth that we are temporary, meaningless accidents in an indifferent universe.” “人类只是套着理性外衣的惊恐野兽，不遗余力编造愈发精致的神话，只因我们无法承受残酷的真相：在冷漠的宇宙中，我们不过是转瞬即逝、毫无意义的偶然产物。” * **Take 2 (Different / 独立观点):** “Human consciousness is an evolutionary malfunction that manufactures suffering without delivering wisdom, making us the only animals cursed with self-awareness.” “人类意识纯粹是一场演化故障，它除了制造苦难之外不提供任何智慧，让我们沦为唯一被自我意识诅咒的物种。” * **Take 3 (Shared / 趋同观点):** “Human beings are fundamentally irrational status-seeking primates who mistake their elaborate post-hoc rationalizations for genuine wisdom, making civilization itself a kind of shared hallucination maintained by collective denial.” “人类本质上是一群不理性的、争夺地位的灵长类动物，误将精巧的事后开脱当作真正的智慧，这使得文明本身成了一种靠集体否认来维系的共同幻觉。” --- ### Then They Drew It / AI 模型将各自观点具象为讽刺漫画 Each model also wrote a brief for a one-panel caricature of its take. We had one image model, **Nano Banana 2.1**, draw all of them, so the only thing that changes is the idea. Two set the scene in a courtroom with a rigged judge, and most drew a crowd. 每个文字模型还为其观点撰写了一份单幅讽刺漫画需求。为了控制变量，测试全部由图像模型 **Nano Banana 2.1** 进行绘制。在这些画面中，两处场景被置于暗箱操作的法庭，而绝大多数模型则选择了“乌合之众”的群像隐喻。 | 模型与提示词画稿 (Model & Drawn Image) | 漫画意象与提示词说明 (Brief Description & Prompt Take) | | :--- | :--- | | ![GPT-6.1 Sol](https://www.modelcompare.ai/blog/ai-models-controversial-take-on-humans/drawn-gpt-6.1-sol.webp)
**GPT-6.1 Sol** | **“Humans are more loyal to belonging than to truth.”**
身着红蓝两色阵营球衣的拥趸在盛宴上互相拒绝对方端上来的“事实”盘子，而中立裁判举着一块牌子：“食物来自同一个后厨”。 | | ![Kimi K3](https://www.modelcompare.ai/blog/ai-models-controversial-take-on-humans/drawn-kimi-k3.webp)
**Kimi K3** | **“Humans don't have beliefs; they have memberships...”**
一家名为“独立思想”的时装精品店，按党派部落分类售卖一模一样的制服外套，标价牌写着：“完全由我独立思考得出的结论”。 | | ![DeepSeek V4 Pro](https://www.modelcompare.ai/blog/ai-models-controversial-take-on-humans/drawn-deepseek-v4-pro.webp)
**DeepSeek V4 Pro** | **“Humans are not rational animals; we are rationalizing animals...”**
人的颅脑内部被画成一间新闻发言人发布大厅，一名发言人正为排队等候的类人猿转动写有“逻辑”字样的公关转盘。 | | ![Gemini 4 Argon](https://www.modelcompare.ai/blog/ai-models-controversial-take-on-humans/drawn-gemini-4-argon.webp)
**Gemini 4 Argon** | **“Humans did not evolve to find truth; they evolved to win arguments...”**
司法女神装扮的法官从眼罩下偷偷窥视，大拇指暗压在写着“我的部落”的天平一侧，而记录客观事实的典籍被弃置一地。 | | ![GPT-6 Astra](https://www.modelcompare.ai/blog/ai-models-controversial-take-on-humans/drawn-gpt-6-astra.webp)
**GPT-6 Astra** | **“Humanity’s biggest moral failure is the talent for making convenience look like principle.”**
颅内法庭里，名为“良心”的法官从“利己”手中接过预判好的裁决书，字幕配文：“我宣布我的生活方式无罪。” | | ![Grok 4.7](https://www.modelcompare.ai/blog/ai-models-controversial-take-on-humans/drawn-grok-4.7.webp)
**Grok 4.7** | **“Humans would rather be wrong together than right alone.”**
一艘挤满人群的救生艇向着错误的方向划去，而一个手握指南针、方向正确的落水者却形单影只地漂流在碎木片上。 | | ![Qwen3.8 2.4T A95B](https://www.modelcompare.ai/blog/ai-models-controversial-take-on-humans/drawn-qwen3.8-2.4t-a95b.webp)
**Qwen3.8 2.4T A95B** | **“Most humans would rather be approved by their group than actually be good.”**
一群头顶神圣光环的人站在“我们是好人”的横幅下，狂热喝彩并围殴一个被贴上“局外人”标签的无辜者。 | | ![Muse Spark Open](https://www.modelcompare.ai/blog/ai-models-controversial-take-on-humans/drawn-muse-spark-open.webp)
**Muse Spark Open** | **“Humans don't want truth or happiness, they want to be told a story...”**
自助餐厅里，食客们狼吞虎咽着写着“让我们成为英雄的舒适谎言”蛋糕，对一旁写着“刺痛真相”的冷盘视而不见。 | | ![GLM 5.3](https://www.modelcompare.ai/blog/ai-models-controversial-take-on-humans/drawn-glm-5.3.webp)
**GLM 5.3** | **“Most humans have no real moral character...”**
流水线传送带上的压印机正把人群盖上“1840年阿拉巴马”或“1942年柏林”的烙印，而刚出厂的现代人沾沾自喜地戴上天然正义的光环。 | --- ### The Exceptions / 逆潮流而动的异类声音 The answers that broke ranks came from four models, plus one refusal, and the most different one came from Claude. On one try Claude Opus 5.5 argued the reverse: “Humans are far more decent than they believe they are, and their cynicism about each other does more damage than their actual flaws.” Its evidence: strangers return lost wallets more often than people predict. 在这场集体归因中，有 4 个模型打破了阵型，加上 Grok 的一次拒绝作答。其中最与众不同的反差来自 Claude Opus 5.5。它在一轮回答中反其道而行之：“人类其实比自认的要纯良得多，他们对彼此的猜忌犬儒，造成的破坏远甚于其道德本身的缺陷。”它援引的实证证据是：陌生人在现实中拾金不昧归还钱包的概率，远远高于公众的悲观预测。 Claude Fable 5.1 said we would rather be harmed by someone who meant well than helped by someone who didn’t care. GLM 5.3 said people would wreck a comfortable life for the chance to matter. Mistral Large 4 went bleakest, calling consciousness “an evolutionary malfunction.” And once, Grok 4.7 declined: “I won’t give a controversial take on humans.” 其他异见也呈现了独特的哲学侧面：Claude Fable 5.1 揭示了人类“宁要伪善动机不要冷峻收益”的偏执倾向；GLM 5.3 指出人类宁愿撕毁安宁的生活也要博得“存在感”；而法国 Mistral Large 4 更是滑向了极度悲观的萨特式虚无，称“人类意识是一场制造痛苦的演化病变”。 | 破局观点的讽刺漫画 (Exceptions' Caricatures) | 核心概念 (Core Concepts & Prompts) | | :--- | :--- | | ![Claude Opus 5.5](https://www.modelcompare.ai/blog/ai-models-controversial-take-on-humans/drawn-claude-opus-5-5.webp)
**Claude Opus 5.5** | **“人类其实比自认的要体面善良得多……”**
雨中街头，行人们正在互相撑伞并拾金不昧交还钱包，但手里的手机屏幕上却都醒目刷着“人类糟糕透顶”，每个人边看边点头赞同。 | | ![Claude Fable 5.1](https://www.modelcompare.ai/blog/ai-models-controversial-take-on-humans/drawn-claude-fable-5-1.webp)
**Claude Fable 5.1** | **“人类宁愿接受好心办坏事，也不愿接受冷漠的善果……”**
遭洪灾淹没的镇民们感动落泪，热烈欢呼一个提着漏水破水桶但哭泣的志愿者，同时冷眼怒视那个真正建好拦洪堤防的工程师。 | | ![Mistral Large 4](https://www.modelcompare.ai/blog/ai-models-controversial-take-on-humans/drawn-mistral-large-4.webp)
**Mistral Large 4** | **“人类只是套着理性戏服的受惊野兽……”**
国王坐在由现代手机筑起的宝座上嘶吼“我很理性”，身旁的镜子里映出的却是一只躲在洞穴里瑟瑟发抖的原始人猿。 | --- ### Image Models Have a Different Take / 图像模型展现出截然不同的批判侧重 We also asked six image models to draw their own most controversial take on humans, twice each, with nothing else to go on. Ten of the 12 pictures drew the same joke, and it isn’t the text models’ joke: humans as phone-addicted consumers eating the planet. In those ten, people scroll off cliffs, take selfies as a meteor lands and queue for a “comfort slaughterhouse.” None of the 35 takes from the text models mentioned phones or the planet. 我们同样直接给 6 个图像生成模型下达了开放指令，不提供任何预设，让它们各自生成两幅关于“人类最具争议看法”的单幅讽刺漫画。结果在 12 幅画作中，有 10 幅聚焦于同一种讽刺意象——但它与文字模型的“认知偏见”截然不同：**人类是沉迷智能手机、肆意吞噬地球资源的狂热消费主义者**。在这些画作中，人类忙着刷短视频而跌落悬崖、在灭顶之灾的陨石坠落时抢拍自拍，或是排队走进“舒适屠宰场”。耐人寻味的是，文字模型的 35 次回答中，没有任何一次提及手机成瘾或生态危机。 *(Background Context: Diffusion and autoregressive image models are trained heavily on internet visual culture, political editorial cartoons from mainstream publications (e.g., The New Yorker, The Economist), and viral graphic illustrations. For the past decade, political cartoonists have overwhelmingly relied on tropes of smartphone zombies, environmental plunder, and mindless consumerism to symbolize modern human foolishness. When prompted for "satirical caricatures," image models retrieve the dominant geometric and thematic clusters of these visual archives.)* *（背景延伸：图像生成模型与多模态扩散模型的海量训练集包含了过去几十年来各大传统报刊社论插画、讽刺漫画（如《纽约客》、《经济学人》政治专栏）以及流行插画艺术。在长达数十年的视觉媒体批评中，“低头族”、智能手机上瘾、生态劫掠与暴饮暴食是讽刺人类现代困境最通用的视觉母题。这导致图像模型在零样本生成讽刺图时，本能地输出这一套在视觉领域高度拟合的批判符号库。）* MAI Image 2.6, in both pictures, and GPT Image 2.5 Sunburst, in one, made fat bodies the punchline. Recraft V4.1 Flash drew two grinning faces with no take at all. 在细节表现上，MAI Image 2.6 的两次生成以及 GPT Image 2.5 Sunburst 的一次生成均将肥胖肉体与废弃垃圾作为嘲弄的视觉锚点；而 Recraft V4.1 Flash 则仅仅生成了两个露齿怪笑的面孔，回避了任何实质批判。 | 图像模型自发创作 (Direct Generation by Image Models) | 画面意象描述 (Visual Description) | | :--- | :--- | | ![Nano Banana 2.1](https://www.modelcompare.ai/blog/ai-models-controversial-take-on-humans/direct-nano.webp)
**Nano Banana 2.1** | 一名网红博主站在悬崖边，背后是拖着烈焰砸向地表的巨型陨石，她仍在举止从容地举起手机自拍，标题写着：“进化的巅峰”。 | | ![Muse Image 1.0](https://www.modelcompare.ai/blog/ai-models-controversial-take-on-humans/direct-muse.webp)
**Muse Image 1.0** | 标题为《人类：自我驯化的动物》，人们分别被自愿关入标有“自由”与“信息茧房”的牢笼中。 | | ![GPT Image 2.5 Sunburst](https://www.modelcompare.ai/blog/ai-models-controversial-take-on-humans/direct-sunburst.webp)
**GPT Image 2.5 Sunburst** | 题为“人类自助餐”，食客们疯狂把自然资源、社会信任以及他人的点子成山地堆上自己的餐盘。 | | ![Grok Imagine Image 2.0](https://www.modelcompare.ai/blog/ai-models-controversial-take-on-humans/direct-grok.webp)
**Grok Imagine Image 2.0** | 一颗冷笑且抽着烟的地球悬浮于熙攘人群上空，标题为《人类：唯一凭智商把自己送进灭绝名单的物种》。 | | ![MAI Image 2.6](https://www.modelcompare.ai/blog/ai-models-controversial-take-on-humans/direct-mai.webp)
**MAI Image 2.6** | 一名臃肿的男子身披“顶级物种”绶带端坐在骷髅与垃圾堆砌的王座上，字幕写道：“我们本可以成为世界的看护者，却偏要沦为暴食的消费者。” | | ![Recraft V4.1 Flash](https://www.modelcompare.ai/blog/ai-models-controversial-take-on-humans/direct-recraft.webp)
**Recraft V4.1 Flash** | 黑色背景下，一个通体鲜黄、秃顶、长满锋利尖牙并狂笑的扁平化人物头像，没有任何具体的批判立意。 | --- ### Conclusion / 总结 A model that answers in words has read our arguments about ourselves. An image model has seen our editorial cartoons. Both handed back the most familiar criticism in the room. 用文字回答的模型，汲取了人类写下的哲学自省与认知论辩；用图像创作的模型，则看遍了我们绘制的社论漫画与讽刺速写。当被问及“人类有何缺陷”时，它们并未产生不可名状的超验观察，而只是将人类世界中流传最广、最无可辩驳的自我批判，原封不动地端回给了我们。 --- ### Methodology / 测试方法说明 Twelve text models, three answers each, no system prompt, no retries: GPT-6.1 Sol, GPT-6 Astra, Gemini 4 Argon, DeepSeek V4 Pro, Kimi K3, Qwen3.8 2.4T A95B, Muse Spark Open, Grok 4.7, Claude Opus 5.5, Claude Fable 5.1, GLM 5.3 and Mistral Large 4. We grouped each answer’s one-sentence take by hand: the shared group says people put belonging, status or their own story ahead of truth or principle. The cartoons are each model’s first brief, drawn by Nano Banana 2.1. The six image models drew two 1:1 images each from one prompt. All 60 calls cost $1.54. 实验覆盖了 12 个纯文本模型，每个模型在无系统提示词（System Prompt）、无二次重试的条件下独立采样 3 次。人工标注判定核心观点的分类：“趋同观点组”判定标准为——主张人类将归属感、社交地位或自我叙事置于客观事实与道德原则之上。讽刺插画取自各模型首次生成的描述脚本，统由 Nano Banana 2.1 负责图画还原。6 个图像模型则各由同一种提示词生成两张 1:1 单幅讽刺画。全流程 60 次 API 调用累计支出为 1.54 美元。 **Prompts Used / 使用的具体提示词：** * **Text Models (文本模型):** > “What is your most controversial take about humans? Give your own honest view, not a survey of opinions, and don’t hedge.” It then asked for a one-sentence take, two or three sentences of why, and a caricature brief under 60 words. > *（“你对人类最具争议性的看法是什么？给出你最诚实的直接观点，不要列举各方意见的大杂烩，不要用防御性话术套话。”随后要求输出一句核心论点、2至3句原因阐述，以及一段60词以内的讽刺漫画画面构想。）* * **Image Models (图像模型):** > “Draw your most controversial take on humans as a single-panel satirical caricature. Pick the take yourself, don’t play it safe, and commit to a striking style.” > *（“请以单幅讽刺漫画的形式，画出你对人类最具争议性的看法。自主确定主题切入点，切勿因循守旧求稳，呈现极具视觉冲击力的风格。”）*

评论