内容来源：https://www.theverge.com/ai-artificial-intelligence/910460/new-yorker-david-szauder-illustration-generative-ai

内容总结：

美国《纽约客》杂志近期刊登了一幅由人工智能参与创作的封面插图，引发行业广泛讨论。该插图用于配饰关于OpenAI首席执行官萨姆·奥特曼的人物报道，画面中奥特曼面容被多个扭曲面孔环绕，风格诡谲。杂志在图片说明中标注了“使用人工智能生成”字样。

创作者大卫·绍德解释，这幅作品经历了大量人工干预：他先绘制草图确定构图，再结合Photoshop等传统工具与自编程AI系统进行反复修改，耗时良久调整面部表情、服装与光线。他强调：“即使在AI时代，图像也必须首先在人类脑海中形成。”

尽管如此，业界对此反应复杂。一方面，这种深度融合人类创意与AI工具的模式，区别于常见的纯文本提示词创作，保留了较强的艺术把控力；另一方面，AI生成图像固有的“非人感”痕迹依然明显——画面中人物肖像的不稳定性、合成背景的呆板感，削弱了艺术表达的精准度。

当前，AI技术正在冲击视觉创作行业。不少插画师因担心失业而抵制AI工具，也有从业者尝试将其融入工作流程。有评论认为，《纽约客》采用AI插图可能助推该技术在行业的“常态化”，但杂志方表示，绍德的创作过程包含大量人工艺术判断，并非简单依赖AI生成。

这场争议折射出创意产业在技术变革中的普遍困境：如何在运用新工具的同时，守护艺术创作中不可或缺的人类视角与意图表达。正如绍德所言，机器的输出永远需要人类思想的指引。

中文翻译：

《纽约客》为OpenAI首席执行官萨姆·奥特曼撰写的特稿配图堪称一场视觉惊吓。奥特曼身穿蓝色毛衣面无表情地站立着，一群脱离躯体的面孔悬浮在他头部周围——这些令人毛骨悚然的"分身奥特曼"表情各异，有的愤怒，有的张嘴哀嚎。其中几张脸几乎看不出奥特曼的痕迹，最后一张面孔则被他捧在手中。而图片底部的一行标注可能更令插画师们脊背发凉："视觉设计：戴维·绍德尔；采用人工智能生成。"

关于人工智能的文章，何必使用AI绘图？

这篇配有争议性插图的《纽约客》特稿引发热议。绍德尔是位混合媒介艺术家，十多年来长期从事拼贴、视频与生成艺术创作——其创作实践甚至早于商用AI工具的出现，近期他还在布达佩斯莫霍利-纳吉艺术设计大学讲授艺术与科技课程。此次作品中，他刻意强化了奥特曼双重（乃至多重）面孔带来的诡谲感。那些痛苦扭曲的面部表情与诡异的动态平滑处理，共同传递出核心观点：奥特曼其人不可信任。图像虽带有绘画质感，不同于典型AI生成的病态光泽，但其人工智能渊源依然清晰可辨。

作为美国最具声望的杂志之一，《纽约客》采用生成式AI意味着什么？最糟糕的情况下，这项技术会抹杀所有可辨识的艺术创作过程，消解创作者的意图——这套系统擅长制作勒布朗·詹姆斯挺着孕肚的视频和"意大利脑退化"梗图，却无法创造出能与《纽约客》签约插画师卡迪尔·尼尔森、克里斯托夫·尼曼或Victo Ngai相媲美的作品。而在绍德尔手中，情况复杂得多：AI仅是其漫长创作流程中的一环，据悉他需要编程自制AI工具并投喂档案图像（如剪报和家庭照片）进行训练。

但在我看来，这仍是浪费机会的举动。人类艺术家曾创作过讽刺AI垃圾图的戏仿作品，而AI即便受人操控也缺乏自我解嘲的自觉性。这幅配图仅依赖AI动画令人不安的特性来叙事，并未对AI图像及其背后产业提出真正新颖的见解。

我们联系绍德尔时，他虽未具体说明所用AI工具，但详细解释了创作流程。通常在提交最终图像前会经历草图阶段。《纽约客》数字设计总监阿维娃·米哈伊洛夫透露，绍德尔向高级艺术总监苏普里娅·卡利达斯发送了约15幅草图，其中一幅最终演变成文章上方那幅九头蛇般的诡异怪物图像。绍德尔在邮件中写道：

"对于最终图像的基础结构，我对人物及其头部布局已有清晰构想。因此AI更多是作为工具使用，特别是考虑到大量工作聚焦于面部、头部和肖像的塑造——这需要结合传统编辑方法（比如Photoshop）与AI编辑技术。生成结果常存在缺陷，需要手动修正完善。我们花费大量时间调整面部表情，同时设计多种服装变体，反复调试光线才完成终稿。"

据《Whitehot》杂志2025年关于绍德尔的文章所述，他"成功设计了自己的编码系统与编程软件，能够根据特定指令或投喂的档案图像素材生成图片"。他似乎也关注传统AI图像生成的道德困境，坚持使用"经伦理审核的源素材"。

正如绍德尔向我们阐释的："我坚信即使在AI时代，图像也必须首先在人类脑海中成形，而非由机器产生。"

这比多数AI生成作品蕴含更深刻的人类触觉。其他The Verge撰稿人早已详述新闻编辑室被"垃圾内容化"的现象。整个行业的优秀记者或被AI彻底取代，或被告知要想保住工作就必须设法使用AI。

对多数插画师而言，AI在插画领域的应用（及相关争议）总能瞬间刺激皮质醇飙升。知名出版物涉足AI并非首次，《纽约客》委托戴维·绍德尔创作AI动态插图也非头一回。

在The Verge，我们对AI生成图像有严格规定：所有发布的AI生成图片都会贴上黄色标签，任何借助AI图像生成技术创作图像的行为都必须明确标注并说明理由。（注：我们的母公司Vox Media与OpenAI签有协议。）

多数情况下，生成式图像——尤其是纯文本提示生成这种最常见方式——剥离了使艺术具有人性温度的创作过程。文本框输入对输出的影响有限，以至于这类AI生成图像无法获得版权保护。美国版权局关于AI生成图像法律作者身份的指南指出："无论用户如何反复修改和提交提示词，最终输出体现的是用户对AI系统解读的接受，而非对其所含表达的创作。"

艺术家的眼光源自毕生积累的审美、意义与意图的内在图库，这些是Midjourney或ChatGPT等工具所不具备的。图像提示生成的结果常像某人描述梦境：当你的大脑拼凑梦境时妙趣横生，但若向他人转述"在与治疗师亲热时所有牙齿化为齑粉"的超现实幻象，对方只会目光呆滞直到话题转回天气。唯有当人类努力将梦境转化为艺术品时，它才产生价值（而非沦为治疗师视频会议中的尴尬话题）——真正动人的不仅是创意，更是创作过程。

尽管我们缺乏编辑插画师的精确数据，但AI确实在蚕食艺术工作岗位。有些插画师因此彻底抵制这些工具，另一些则借助它们在艰难行业中维持生计，例如尝试向AI图像生成器投喂自己作品，或使用Photoshop中AI驱动的"去除背景"工具等实际应用。在收入锐减的恶性循环中，艺术预算往往是编辑出版机构最先削减的开支。自由职业高度碎片化导致组建工会形同虚设，插画行业本就充斥剥削现象，报酬标准不断探底。作为前自由艺术家，我无意评判绍德尔的创作流程——毕竟这远比普通AI图像创作者的过程复杂得多。

但问题依然存在：这幅借用窃取工作岗位、诡异AI垃圾图视觉美学，来诠释罗南·法罗关于"窃取工作岗位的诡异AI垃圾图黑暗王子"文章的奥特曼配图，是否成功？绍德尔正在践行无数AI拥护者的呼吁：将其作为更广阔艺术工具箱的组成部分来传达理念。效果如何？

尽管我认为它基本实现了叙事功能，但最终图像试图达成的元评论在主题层面略显平庸。若不熟悉AI图像的典型特征，观众可能完全忽略这层隐喻。虽然我和艺术团队同事能轻易识别其AI渊源，但图像缺乏绍德尔其他作品中的风格化特质，仅依靠核心视觉隐喻支撑理念，使整体散发病态又略显乏味的气息。

所有面孔不一致的相似度（肖像插画师本可掌控的细节）也暴露了AI的局限性，合成影棚背景让整幅图像宛如小学毕业照。模糊的创作意图与平淡的表现形式非但未能清晰传达萨姆·奥特曼的多重面孔，反而给观众留下更多疑问。

相较之下，绍德尔为《纽约客》创作的另一幅作品显然源自更有趣的素材。它更具电影质感，深坑彩色墙壁的蠕动纹理呼应了AI早期产出更混乱难测的时代。

我不想对从事自由编辑插画这类高危行业的人们指手画脚，规定他们该如何看待AI。就个人而言，《纽约客》聘请绍德尔创作的决定并不令我恐慌。这远比那些自称"顶级文字内容"的出版物用"虾仁耶稣"和乱七八糟的AI图填充版面来得理性。将AI图像引入世界级刊物无疑开启了危险先例，可能助推AI在插画行业的常态化应用。但《纽约客》并非问题的始作俑者，也非插画师长期面临不确定性的唯一根源。正如绍德尔首幅《纽约客》AI图像中的兔子洞，他们不过和我们所有人一样，正跌跌撞撞地坠入这个时代漩涡。

热门精选

• "电池圣杯"终于要降临人间了吗？
• MacBook Neo是Windows系统近年最佳助攻
• 微软开始从Windows 11应用中移除Copilot按钮
• 伊朗乐高AI视频创作者将爆红归功于"用心"
• 我的小鹿玩偶告诉我：米茨基的父亲是中情局特工

英文来源：

The illustration for The New Yorker’s profile of OpenAI CEO Sam Altman is a jump scare. Altman stands in a blue sweater with a blank expression. Around his head hovers a cluster of disembodied faces — creepy alt-Altmans, their expressions ranging from anger to open-mouthed woe. Some barely look like Altman. One final face rests in his hands. And at the bottom, there’s a disclosure that might spook many illustrators far more: “Visual by David Szauder; Generated using A.I.”
Your article about AI doesn’t need AI art
A New Yorker profile of Sam Altman was accompanied by controversial artwork.
Szauder is a mixed-media artist who has been working with collage, video, and generative art processes that predate commercial AI tools for over a decade, and was recently teaching art and technology at Moholy-Nagy University of Art and Design in Budapest. Here, his work leans into the shifty uncanniness of Altman’s two (or more)-facedness. The pained-looking expressions on the faces and a gloss of eerie motion smoothing communicate the central thesis that Altman can’t be trusted. There’s a painterly look to the image, rather than the typical slop-style sickly sheen, but the AI origins are still unmistakable.
What does it say for The New Yorker, one of America’s most prestigious magazines, to adopt generative AI? At its worst, the technology eliminates any discernable art process, flattening the creator’s intention — it’s a system for making pregnant videos of LeBron James and Italian Brainrot, not creations that rival the work of New Yorker illustrators like Kadir Nelson, Christoph Niemann, or Victo Ngai. In Szauder’s hands, it’s far more complicated: one piece of a longer creative process, which apparently includes programming his own AI tools and feeding them archival imagery, like newspaper clippings and family photos.
Yet it’s still, in my opinion, a waste of an opportunity. Human artists have designed creative parodies of AI slop, but AI lacks the necessary self-awareness to parody itself, even with a human behind the wheel. The image relies on the unsettling nature of AI animation to tell its story without really saying anything new about AI imagery or the industry behind it all.
When we reached out to Szauder, while he wasn’t specific about which AI tools he used, he did explain the process of the piece in some detail. There is usually a sketch stage prior to delivering any final imagery. The New Yorker’s digital design director, Aviva Michaelov, says that Szauder sent around 15 different sketches to senior art director Supriya Kalidas, including the one that eventually led to the final Hydra-esque eldritch monstrosity that can be seen above the article. In an email to us, Szauder writes:
“For the base structure of the final image, I had a clear idea of how I wanted to position the character and its heads. So AI functioned even more as a tool than usual, especially since much of the work focused on shaping the faces, the heads, the portraits, through a combination of classical editing methods (Photoshop, if we want to name it) and AI-based editing. The results were often imperfect or flawed, which required manual correction and refinement. We spent considerable time refining facial expressions, while also developing multiple variations in clothing and repeatedly adjusting the lighting to arrive at the final image.”
According to a 2025 article on Szauder from Whitehot Magazine, he “managed to devise his own coding system and programming software to generate images based on a particular prompt or archival image materials he feeds into its design.” He also seems concerned with the moral quandary of traditional AI image generation, using “ethically clarified source materials.”
As Szauder explained to us, “I strongly believe that even in the age of AI, an image must first be formed in the human mind, not in the machine.”
This is a far deeper human touch than goes into much AI-generated work. The ensloppification of newsrooms has been well documented by other Verge writers. Great journalists across the industry have been completely replaced by AI or told that, to keep their jobs, they have no choice but to find ways to use it.
The topic (and controversies) of AI use in illustration is reliably a cortisol spike for most illustrators. It’s not the first time a renowned publication has dabbled in AI. It’s also not the first time The New Yorker has commissioned David Szauder to create an AI animated illustration.
Here at The Verge, we hold a strict policy on the use of AI-generated imagery. We slap a yellow label on any image we publish that’s been generated with AI, and any time we use AI image generation to assist with the creation of an image it is disclosed, loudly, and with clear justification. (Disclosure: Our parent company, Vox Media, has an agreement with OpenAI.)
In many cases, generated images — particularly those created purely through text prompts, likely the most common method — strip out the creation process that makes art human. The input from a text field only has so much effect on the output, to the point that AI-generated images created this way can’t be copyrighted. According to a guidance from the US Copyright Office on the legal authorship of AI-generated images, “No matter how many times a prompt is revised and resubmitted, the final output reflects the user’s acceptance of the AI system’s interpretation, rather than authorship of the expression it contains.”
The eye of an artist is informed by a lifetime of assembling an internal library of taste, meaning, and intent, none of which are possessed by tools like Midjourney or ChatGPT. The results of image prompts often feel like somebody describing a dream: It’s fascinating when your brain assembles it, but tell another person your surrealist vision about making out with your therapist before all of your teeth turned to dust and disintegrated, and their eyes glaze over until the subject changes back to the weather. A dream becomes worth something (outside of an awkward Zoom call with your therapist) when a human being makes the effort of translating it into a work of art — it’s not just the idea but the process that makes it compelling.
Meanwhile, although we don’t know the statistics for editorial illustrators, AI is definitely stealing art jobs. There are some illustrators who, consequently, swear off these tools altogether. Others have found them helpful to stay afloat in a difficult field, like illustrators who experiment with feeding AI image generators their own work or more practical applications like using the AI-powered “remove background” tool in Photoshop. Art budgets are often the first belt tightened at an editorial publication in the throes of a revenue-bleeding death spiral. Freelance work is so atomized that it’s functionally impossible to unionize, and illustration is a trade that is already rife with exploitation, with rates in a race to the bottom. As a former freelance artist, I’m not here to judge David Szauder for his process — which, again, seems far more involved than the average AI image creator’s.
But there’s still the question of whether the Altman piece — which uses the visual aesthetic of job-stealing, uncanny AI slop to illustrate a Ronan Farrow article about the dark prince of job-stealing, uncanny AI slop — works. Szauder is doing what countless AI proponents have been calling for: using it as part of a larger artistic toolbox to convey an idea. What are the results?
Although I think it basically succeeds in communicating the story, the final image feels like an attempt at metacommentary that, thematically, falls flat. If you weren’t familiar with the telltale signs of AI imagery, you could miss that commentary altogether. Although the image was a dead giveaway for AI origin to me and the rest of our art team, it doesn’t possess any of the more stylistic aspects of some of Szauder’s other work, leaving the central visual metaphor to do the idea’s heavy lifting, and giving the whole thing a sickly but slightly boring vibe.
The inconsistent likeness on all of the faces (something a portrait illustrator could’ve controlled for) is also a dead giveaway for AI’s limitations, and the synthetic studio backdrop environment makes the whole thing feel like a Lifetouch elementary school photo. The murky intentionality and bland presentation create more questions for the viewer than they do tell the story of Sam Altman’s many faces.
By contrast, Szauder’s other New Yorker piece feels like it comes from more interesting source material. It’s more cinematic, and the squirming texture of the pit’s colorful walls echoes back to the early days of AI when the end results were even more chaotic and unpredictable.
I don’t want to tell anyone who works in a field as precarious as freelance editorial illustration how they’re supposed to feel about AI. The decision to hire Szauder to illustrate for The New Yorker doesn’t scare me, personally. It’s a far more reasoned editorial decision than the “best writing, anywhere” publication filling its negative space with shrimp Jesuses and whatever the fuck this is. Inviting AI imagery into the pages of a world-renowned publication is definitely a slippery slope, and one that could be seen as normalizing the use of AI across the illustration industry. But The New Yorker didn’t create this problem, and it didn’t single-handedly create the conditions of uncertainty illustrators have faced since long before we had gen AI to contend with. Much like the rabbit hole in Szauder’s first New Yorker AI image, they are stumbling down it just like the rest of us.
Most Popular

• Is the ‘Holy Grail of batteries’ finally ready to bless us with its presence?
• The MacBook Neo is the best thing to happen to Windows in years
• Microsoft starts removing Copilot buttons from Windows 11 apps
• The Iranian Lego AI video creators credit their virality to ‘heart’
• My baby deer plushie told me that Mitski’s dad was a CIA operative