These were the ‘Adams and Eves’ of facial recognition technology

When we unlock our phones, pass through airport security, and — depending on where you live in the world — walk down the street, our faces are analyzed for the exact combination of nose width, eye spacing and jaw shape that allows an algorithm to see and recognize us. For billions of people, facial recognition technology has become part of daily life, allowing machines to assess our humanity and our physical identities in less than a second. In return for giving up this part of ourselves, we get a product or service or a sense of safety.

But to learn all of our faces, it began with an original group. These “Adams and Eves” of facial recognition were part of a US Department of Defense initiative in the 1990s. Close to a decade ago, the artist Trevor Paglen was researching government datasets — which, though not widely seen, are in the public domain — when he came across one featuring the faces of thousands of government employees that the Defense Advanced Research Projects Agency used to develop and train the nascent technology.

Since then, he has exhibited a selection of the images as a series called “It Began as a Military Experiment,” which will display again in a 20-year retrospective of his work, “ The Most Merciful Thing in the World ,” that is also part of the FotoFocus Biennial in Cincinnati this month. Titled after an H.P. Lovecraft quote, the wider collection of his work examines what is often unseen or hidden from us, including photographs of hundreds of unidentified objects in Earth’s orbit, and hazy images of military sites magnified beyond what the human eye can see. Over the course of Paglen’s career, the themes he has explored across our relationship to technology, the ways in which we are surveilled, and the invisible technological processes that shape our world have only become more urgent.

Courtesy the Artist/Fellowship/Jessica Silverman Gallery, San Francisco/Pace Gallery, New York

As an artist, Paglen explained by phone, “I honestly go back to the oldest trick in the book, which is take something familiar and make it strange, or take something strange and make it familiar.” He added: “But that what that means always changes because history is changing.”

In “It Began as a Military Experiment,” the photographs show subjects of different ages, gender and skin tone against a blank sky-blue backdrop. Paglen ran each of the images through an algorithm to identify the points on the face that are measured and analyzed through facial recognition programs, and he shows the markers to the viewer — small white letters hovering over the corners of eyes and cupid’s bows and nostril edges, which translate people into data.

The artist making unsettling AI images of the human body

The overall effect is uncanny. There is no way to place the subjects in time or space, beyond the occasional dated haircut. They mostly gaze off to the side, ignoring the camera. To the modern eye, a series of uniform, close-cropped portraits of a range of people might bring to mind the AI-generated portraits that at first shocked us with their realism . Now, we have to question every image put in front of us, looking for the increasingly more subtle signs that it might not be real at all. And, as AI has exploded, we have all quickly — and uncomfortably — had to adjust to the idea that we all have become the dataset on which these algorithms are trained, as they continue to gobble up our images and cultural output in the name of self-improvement.

When I spoke to Paglen, it was the day after the Anthropic researcher Jacob Coxon publicly announced on X that he had resigned from his job, and warned that the top labs building AI “earnestly believe that it could kill us all by the end of the decade.” In the weeks since, the mainstream discourse around AI has sharply turned for the worst. But it’s exactly the type of thing that Paglen has tried to capture in his work — the truest definition of the sublime, that heady mix of awe, beauty and horror in a world that’s beyond our cognitive understanding.

“Our whole relationship to technology to me feels very supernatural, feels very in this world of cosmic horror,” he said. “And that is what our politics feels like to me as well. This YOLO, let’s-summon-every-demon-we can-imagine-and-see-what-happens feeling.”

But when he created “It Began as a Military Experiment,” the conversations around the algorithms we were creating and the datasets we were feeding them were much different. In the mid-to-late 2010s, the conversations were largely around bias. Were algorithms used in policework prone to profiling? Was too much of our technology based on the perspective of privileged men?

While those were all important conversations to have, he also believes it’s a fallacy to believe it could be designed differently. Ultimately, he said, they are tools of capitalism, tools for extraction, which will never care about being fair.

“The bias conversation puts you in a position where it’s almost like you think it can be fixed and it is possible to have equitable AI systems, and that always felt wrong to me because once you introduce these things into a market logic, they are going to discriminate,” he explained. “That’s what they do, and you’re going to discriminate in order to figure out how to extract capital from different people in different ways.”

Now when he looks back at these early facial recognition portraits, he feels their vulnerability more than anything else. The subjects were being photographed solely because they were government employees, and he wonders just how much they understood about the technology at the time and its implications.

“If you’re being told to do this as part of your job, do you really consent?” he posed. “That’s one layer.”

The other is that they couldn’t have possibly foreseen how our biometric data would underpin so many of our invisible systems today.

“Is it possible to consent to have something that you did today used in the distant future in ways that you could not imagine?” he asked. “We retroactively become a part of systems that didn’t exist.” Yesterday it was them, today it’s all of us.

当我们解锁手机、通过机场安检，甚至——取决于你居住在世界何处——走在街上时，我们的面部都会被分析，精确捕捉鼻子宽度、眼距和下颌形状等特征，以便算法能够识别我们。对于数十亿人来说，面部识别技术已成为日常生活的一部分，使机器能够在不到一秒的时间内评估我们的人性特征和生理身份。作为放弃这部分个人信息的回报，我们获得了产品或服务，或者一种安全感。

但要掌握我们所有人的面孔，一切都始于一个最初的群体。这些面部识别领域的“亚当和夏娃”是美国国防部在20世纪90年代一项计划的一部分。大约十年前，艺术家特雷弗·帕格伦（Trevor Paglen）在研究政府数据集时——这些数据集虽然不为人知，但属于公共领域——他偶然发现了一个包含数千名政府雇员面孔的数据集，国防高级研究计划局（DARPA）曾利用这些数据开发和训练这项新兴技术。

此后，他以“一切始于军事实验”（It Began as a Military Experiment）为题，展出了部分作品。本月，他将在辛辛那提FotoFocus双年展上再次展出其20年回顾展“世间最仁慈之物”（The Most Merciful Thing in the World）。该展览的标题源自H.P.洛夫克拉夫特的一句名言，其更广泛的作品系列探讨了我们常常看不到或被隐藏的事物，包括地球轨道上数百个不明物体的照片，以及放大到肉眼无法辨识程度的军事设施的朦胧影像。在帕格伦的艺术生涯中，他探索的主题——我们与科技的关系、我们被监视的方式，以及塑造我们世界的无形技术进程——变得越来越紧迫。

图片由艺术家/奖学金/旧金山杰西卡·西尔弗曼画廊/纽约佩斯画廊提供

帕格伦在电话中解释说，作为一名艺术家，“我其实总是用最古老的技巧，那就是把熟悉的事物变得陌生，或者把陌生的事物变得熟悉。” 他补充道：“但这种技巧的意义总是在变化，因为历史在变化。”

在作品《一切始于一场军事实验》中，不同年龄、性别和肤色的人物以纯天蓝色的背景为衬托。帕格伦运用算法处理每张照片，识别出面部特征点，这些特征点会被面部识别程序测量和分析。他将这些标记展示给观众——一些白色的小字母悬浮在眼角、唇峰和鼻翼边缘，将人转化为数据。

这位艺术家创作了令人不安的人工智能人体图像

整体效果令人毛骨悚然。除了偶尔出现的过时发型，我们根本无法将照片中的人物置于特定的时间和空间。他们大多侧目而视，对镜头视而不见。在现代人看来，一系列风格统一、特写镜头拍摄的人物肖像，或许会让人联想起那些最初以其逼真度震惊世人的AI生成肖像。如今，我们不得不质疑呈现在眼前的每一幅图像，寻找那些越来越微妙的蛛丝马迹，以防它完全是虚构的。随着人工智能的爆炸式发展，我们都不得不迅速——且不自在地——适应这样一个事实：我们都成为了这些算法的训练数据集，它们以自我提升之名，不断吞噬着我们的图像和文化产出。

我与帕格伦交谈的第二天，人类学研究员雅各布·考克森在X杂志上公开宣布辞职，并警告说，顶尖的人工智能实验室“真心相信，人工智能可能会在本十年末毁灭我们所有人”。此后几周，关于人工智能的主流讨论急转直下，朝着更糟糕的方向发展。但这正是帕格伦试图在他的作品中捕捉到的那种感觉——崇高之美的真谛，那种令人陶醉的敬畏、美丽与恐惧交织在一起的感觉，存在于一个超越我们认知理解的世界中。

“在我看来，我们与科技的整个关系都充满了超自然色彩，感觉就像置身于一个宇宙恐怖的世界，”他说。“而我们的政治也给我带来了同样的感觉。一种‘人生苦短，及时行乐’、‘让我们召唤所有能想象到的恶魔，看看会发生什么’的感觉。”

但当他创作《一切始于军事实验》时，围绕我们创建的算法和输入算法的数据集的讨论与现在截然不同。在2010年代中后期，讨论主要集中在偏见问题上。警务工作中使用的算法是否容易导致种族定性？我们的技术是否过多地基于特权男性的视角？

尽管这些讨论都很重要，但他认为认为可以采用不同的设计方案是一种谬论。他最终表示，这些制度都是资本主义的工具，是榨取财富的工具，它们永远不会在意公平。

“关于偏见的讨论会让你觉得偏见是可以解决的，可以建立公平的人工智能系统，但我总觉得这种想法是错的，因为一旦把这些东西引入市场逻辑，它们就会进行歧视，”他解释说。“它们本来就是这样运作的，为了从不同的人身上以不同的方式榨取资本，你也会进行歧视。”

如今，当他回望这些早期的面部识别肖像时，他感受到的更多是它们的脆弱。照片中的人物仅仅因为是政府雇员而被拍摄，他不禁思考，当时的这些人究竟对这项技术及其影响了解多少。

他反问道：“如果这是你工作的一部分，你真的同意吗？这是其中一层含义。”

另一方面，他们不可能预见到，我们的生物识别数据如今会如何支撑我们如此多的无形系统。

他问道：“你是否能够同意让别人在你遥远的未来以你无法想象的方式使用你今天所做的事情？我们仿佛被追溯性地纳入了一个原本并不存在的系统。”昨天是他们，今天是我们所有人。