纠缠,而后日以作夜

外滩大会AI概念艺术展 · Inclusion 2026 — AI Concept Art Exhibition

《天书》

“A Book from the Sky”

Gene Kogan · 吉恩·科根
作品页面

作品 · Artwork

年份 · Year
2015
类型 · Medium
平面喷绘 + 数字影像 · Graphic Print Work + Digital Video

《来自天空的书》是一篇开创性的作品,将人工智能(AI)与艺术领域相连接。该作品创作于2015年,运用深度卷积生成对抗网络(DCGAN)探索中文手写体的潜在空间,网络在约一百万个手写简体汉字上训练后,能够生成语料库中不存在的、以假乱真的伪汉字,并在字与字之间的潜空间中连续变换。这一项目标志着人工智能发展史上的一个关键转折点,自2015年以来,AI已从识别任务转向生成能力,凸显了其从学习数据中创造新颖且有意义内容的潜力。

作品得名于徐冰1988年的《天书》——徐冰以宋明版刻风格刻制了数千个自创的伪汉字。

“A Book from the Sky” by Gene Kogan is a groundbreaking piece that bridges the domains of artificial intelligence (AI) and art. Created in 2015, “A Book from the Sky” utilizes Deep Convolutional Generative Adversarial Networks (DCGAN) to explore the latent space of Chinese handwriting, trained on a labeled subset of roughly one million handwritten simplified Chinese characters, the generator produces convincing fake characters found nowhere in the original dataset, and traverses the space between them in continuous interpolations. The project signifies a pivotal moment in the evolution of AI, transitioning from recognition tasks to generative capabilities since 2015, highlighting AI’s potential to create novel, meaningful content from learned data.

The work takes its name from Xu Bing’s 1988 “Book from the Sky” (天书), in which Xu composed thousands of fictitious glyphs in the style of traditional prints of the Song and Ming dynasties.

背景 · Background

2015年12月,DCGAN 论文与开源代码发布数周后,作者以 HIT-OR3C 语料库中约一百万个手写简体汉字训练网络创作了本作品,是最早有意识运用生成对抗网络创作的艺术作品之一。作品呼应徐冰1988年的装置《天书》——由数千个不可读的伪汉字构成。

Created in December 2015, weeks after the DCGAN paper and code were released, by training the network on about one million handwritten simplified Chinese characters from the HIT-OR3C corpus — among the first intentional artworks made with generative adversarial networks. The piece references Xu Bing's 1988 installation “Book from the Sky,” composed of thousands of unreadable pseudo-Chinese characters.

艺术家 · Artist

Gene Kogan

Gene Kogan

吉恩·科根

吉恩·科根(Gene Kogan)是一位运用人工智能与生成系统进行创作的艺术家和工程师。他最初是音乐信息检索方向的机器学习工程师,后通过 Processing、openFrameworks 等开源工具转向互动与生成艺术,以早期深度学习模型建立起创意编程实践,并创建了免费课程 ml4a,向一代技术人员传授创意与生成式 AI。他举办过200多场推广生成式 AI 的工作坊,是最早讲授去中心化 AI 与自主智能体的讲者之一,并共同创立了 Eden 与 Mars College。其作品曾在佳士得、林茨电子艺术节、首尔媒体城市双年展、上海明当代美术馆、温特图尔摄影博物馆与多伦多国际电影节展出,并于2019年与2025年登上 NeurIPS、2026年登上 CVPR。他2019年发起的 Abraham 项目提出“自主人工艺术家”的构想,如今是一件在十三年之约下每日创作的活的作品。

Gene Kogan is an artist and engineer working with AI and generative systems. Originally trained as a machine learning engineer in music information retrieval, he turned toward interactive and generative art through open-source kits like Processing and openFrameworks, building a creative coding practice with early deep learning models and sharing it through ml4a, a free curriculum that taught creative and generative AI to a generation of technologists. He has given over 200 workshops popularizing generative AI, was among the earliest to lecture on decentralized AI and autonomous agents, and co-founded Eden and Mars College. His work has been exhibited at Christie's, Ars Electronica, the Seoul Mediacity Biennale, the Ming Contemporary Art Museum in Shanghai, Fotomuseum Winterthur, and the Toronto International Film Festival, and presented at NeurIPS in 2019 and 2025 and CVPR in 2026. His 2019 project Abraham proposed an autonomous artificial artist, now a living work producing art daily under a thirteen-year covenant.

网站 / 社交媒体

关注议题 · AI Topics

自主人工艺术家与智能体生态(Abraham、Eden);去中心化 AI 与智能体社会;生成模型作为创作协作者;创意 AI 教育与开源实践。

Autonomous artificial artists and agent ecosystems (Abraham, Eden); decentralized AI and agent societies; generative models as creative collaborators; creative-AI education and open source.