- 主办单位
- 中国美术学院（CAA）
跨媒体艺术学院（SIMA）
网络社会研究所（INS） - 总召集人
- 黄孙权
- 学术委员会
- 李凯生、闵罕、黄孙权、陈界仁、刘怿斯、周蓬岸
引言
第十届年会“互联网之终？”——并非末世广告，而是对秩序更替的审问：当超连结的公共技艺（引用、互操作、透明可查）成为提示的自然语言界面（大型数据库、系统预设、代理决定）取代，我们是否正在跨越一条无形却决定性的界线？“AI之内／之外”亦非是末日派与荣景派（doomer/bloomer）的历史终结意识形态，这是我们已经太熟悉的资本主义运作的模式的广告词，如今，各社交平台拥有私数据（由个人数据构成的公司资产），使之成为不联网；现在AI治理的问题更隐讳，历史上所有人类的一切，人文艺术智力的结晶都将含括在内，将在未经公共授权、未标明出处与回馈机制的条件下，被编纂为一种可租用、可封装、可套利的“模型资本”。此外，谁以及什么样的资料得以被纳入模型的语言、资料与评测；谁与什么样的数据被排除在算力、标准与分润之外。这就是我们十年来反复追问的“协议”问题，亦即网络社会学式的“公共性”的关键。早期的网际世界以开放协议维系——HTTP、SMTP、RSS、ActivityPub……大家约定栏位、格式与互通性，称之为“互联”：我可以接得上你，你也能接得上我，规则公开、可审、可替换。如今的节点悄悄移位：我们面对的是模型界面——API条款、系统提示、对齐策略、内容筛检与评测束缚——它们同样扮演“协议”的角色，但多半是黑箱、可变、且由少数供应者单方推进。于是，此转变就从开放协议的“互联”（inter-operability），过渡到模型界面的“互导”（inter-guidance）。
由巨型AI公司引领的互导，实是“一般人工智能”（马克思意义上的）的剥削与极大利润化利用，由此，当工作就是提示（when work is prompt），这可能是人类文明发展中最危险的转折：人类共同的知识存量被清理过滤成“大语言模型”，语言这一社会关系的织机被改造成提词器的界面，工艺、判断与美学训练被压缩为一行行可计费通证费用的指令。危险不在于机器更聪明，而在于一般智能（general intellect）被公司化地圈占为产权，隐藏了掘矿─资本─劳力的广泛社会过程，并以黑箱评测与基建垄断获取梯度地租。当“工作就是提示”，我们若不重写规则，就等于把作者性、公共性与未来的分配权，一并交付给系统预设值：浏览分析判断被替换为“召唤”、手艺与编码被简化为“提示”、作者性被稀释为“参数”。
在网络社会年会迈向十周年之际，我们面对的不仅只是资本─技术的复合权力，与文化─技术的社会构造，我们更面临算法主权、语料主权、各民族国家在全球范围上的基础设施上资源斗争。也许没有比这个时刻接近马克思在《大纲》中论述的了，马克思说：“固定资本的发展表明，普遍的社会知识已成为直接的生产力。”他又说：“固定资本乃资本之为资本的最相称形态。”这不就是人工智能？人工智能难道不是资本主义的梦中情人？
十年之际，不是让我们忘记之前的问题，去更新解决方案，相反的是在延续的长战场中，找到互联网后期丧失的理想与实践可能，这恰恰好能够使我们面对看似不一样型态，确有着更为精细省力的剥削型态。
回顾网络社会研究所将把十年累积的系列研究（如 SparkLink／Archverse／合作社式模型与算力、城市论坛、青年学者论坛与hackathon），本届年会除了主题演讲外，另外展出文化与技术三部曲的研究资料与影像播放，以及青年学者论坛。
- 时间
- 2025年10月17-19日
- 地点
- UFO未知终端上海西岸油罐艺术公园一号罐
（上海市徐汇区龙腾大道2398号） - 直播地址
- 👉 B站直播链接👇
线下报名
请点击链接或扫码填写信息完成注册报名，可优先于现场签到获取同传设备与座位保留
（现场同传设备领取需抵押身份证件）
活动议程
10月17日
时间 (UTC+8) 活动内容
上午
青年学者论坛
9:00
入场
9:20-9:30
青年学者论坛 开幕致词：刘怿斯
9:30-10:30
自组论坛｜情感提词器
主持评论人：刘怿斯
发表人：
朱韵｜我的“爱人”：中国乙女游戏研究的学术图景（2019—2025）
陈苏元｜与AI谈恋爱：AI产品“爱”的商业构建、文本驯化与亲密关系实践——以“Replika”为例
郑叶颖｜无形/限之爱：通用对话型AI重织人类关系之网
叶云鹤｜从“肉身相遇”至“数据匹配”：论Blued的技术中介对中国Z世代男同性恋亲密关系的建构
（中国美术学院 网络社会研究所）
10:30-11:10
自组论坛｜被挟持的未来与替代想象的（不）可能：科技右翼、风险资本与反民主的幽灵
主持评论人：蒋斐然
发表人：
金博文｜风险资本的政治工程：科技右翼如何驱动美国国家公司化进程
杨启晨｜黑暗启蒙、敌基督者与技术弥赛亚：美国科技右翼如何构建当代神学叙事？
秦川｜人文艺术实验室在技术治理背景下的谱系观察，艺术学校的批判位置
（中国美术学院 当代艺术与社会思想研究所）
11:20-12:20
- 创意讲谈
- 主持评论：周蓬岸
- 发言嘉宾
- 罗霄
- 演讲主题
- 当生成成为日常：AI 时代的流变、协议与遗忘
10月18日
时间 (UTC+8) 活动内容
上午
主旨论坛一
9:00-9:30
第十届网络社会年会 开幕致辞
跨媒体艺术学院副院长 闵罕
网络社会研究所所长 会议总召 黄孙权教授
9:30-10:30
板块一 静默编辑
发言嘉宾：迪伦·雷比林（Dylan Reibling）
演讲主题： 追问基础设施：平台、预测与提示词
10:30-10:45
茶歇
10:45-11:45
- 板块三 AI 基建志
- 发言嘉宾
- 乔安娜·莫尔（Joana Moll）
- 演讲主题
- 内在的界面
11:45-12:45
板块一 静默编辑
对谈嘉宾：斯科特·摩尔（Scott Moore）、博格娜·科尼尔（ Bogna Konior）
对谈主题： 机器决策并非最终：质疑“静默的编辑者”
下午
主旨论坛二
14:00-15:00
- 板块四 当作品就是提示
- 发言嘉宾
- 尼克·蒙特福特（Nick Montfort）
- 演讲主题
- 编码、点击与提示：生成艺术的方式
15:00-16:00
- 板块四 当作品就是提示
- 发言嘉宾
- 费利克斯·斯塔尔德（Felix Stalder）
- 演讲主题
- 数字文化、生成式人工智能与艺术
16:00-16:15
茶歇
16:15-17:15
板块二 梯度地租
演讲主题： 杰克·斯蒂尔戈（Jack Stilgoe）
演讲主题： 让人工智能回归现实
晚上
18:30-20:30
作品放映讨论会：
《互联网的终结》（The End of the the Internet, 2025），迪伦·雷比林（Dylan Reibling）
20:30-10:00
Live Coding 开放式声音现场
Boyang GUAN｜崔涵｜周蓬岸
10月19日
上午
主旨论坛三
9:30-10:30
板块二 梯度地租
演讲主题：詹姆斯·斯坦霍夫（James Steinhoff）
演讲主题： 提示工程与（非）决定论的政治经济学
10:30-11:30
- 板块三 AI 基建志
- 发言嘉宾
- 路淼
- 演讲主题
- 热带地区的数据中心：西非的热量管理、中国科技与能源政治
下午
青年学者论坛
1:00-2:30
青年学者论坛｜Panel 1 静默编辑
主持评论人：黄孙权
发表人：
丁凡、袁金玉、张婧妮（同济大学 艺术与传媒学院）｜静默编辑的公共理性再建构：交往理性视角下的隐性权力机制批判研究
黄越敏（华东师范大学）、郭都末（英国皇家艺术学院）｜无限深的镜子：语言符号在后互联网时代的内循环
赵佳菁（中国美术学院 网络社会研究所）｜异端批注：越狱提示的斗争
李洋梓（南开大学 社会学系）｜从集体共鸣到算法响应：生成式AI对同人酷儿叙事与可见性政治的重塑
何乐其、Max Loeffler（上海纽大） ｜对齐的暴力：如何停止担忧并爱上灵异软件
卢曼曼（山东大学 新闻传播学院）｜从“超级傀儡”到“赛博超级傀儡”：虚拟影响者在智能广告中的权力操控与伦理审视（线上）
2:30-3:30
青年学者论坛｜Panel 2 梯度地租
主持评论人：黄孙权
发表人：
仝昭祥（中国美术学院 网络社会研究所）｜人工/智能的地理不均衡：数据标注员的劳动过程
许可（中国美术学院 网络社会研究所）｜数字流水线上的母亲：中国县域数据标注女工的母职困境
滑惜子（上海纽约大学 交互媒体艺术）｜寄生于系统之间：模型间的协议美学作为创作方法
王少玮（奥斯陆大学 人文学院媒介与传播系）｜提示即投注：短剧投流中的语义金融与流量的第四价值形式（线上）
3:30-3:45
茶歇
3:45-4:15
青年学者论坛｜Panel 3 Al基建志
主持评论人：潘霁
发表人：
牟浩杰（复旦大学 新闻学院）｜“绿色”反对“全景”：城市基础设施的摩擦与政治
张徐姗（清华大学 社会学系）｜循线而行：数据中心的替代性解剖学
4:15-6:30
青年学者论坛｜Panel 4 当作品就是提示
主持评论人：黄孙权 刘怿斯 潘霁
发表人：
姜奕竹（中国美术学院 网络社会研究所）｜想象的机器：AI生成艺术的美学与算法权力批判
孙景赫（中国美术学院 跨媒体艺术研究）｜界面、反馈与决策暴力：电子游戏与战争的同构逻辑
严慧（独立研究员、算法工程师）｜算法重塑的作者性：AI艺术创作中主体身份的技术建构与权力转移
任玄奇、吴于枫（天津美术学院 跨媒体艺术）｜从弱影像到提示词：AIGC时代梦核美学的算法收编与作者性危机
林檎（宁波诺丁汉大学 音乐传播与音乐戏剧）｜算法之外：AI歌姬从生产到表演的文化表达
姚丽金（中国美术学院 美丽中国研究院）｜判断作为基础设施：档案展劳动中的人机协作
唐兰、王芷子（西安美术学院 跨媒体艺术系）、石艺佳（西安交通大学 艺术教育中心）｜从行动者视角出发：重构AI生成艺术中的作者性、在场性与灵韵
金秋雨（東京藝術大学大学院 国際芸術創造研究科）｜提示机制与观众记忆：当代语境下个体经验的转变（线上）
刘磊（罗德岛设计学院 全球艺术与文化）｜助燃剂：“设计师之选：诺曼·提格 — 即兴合奏”与人工智能艺术与设计的机构化（线上）
19:30-21:00
黄孙权：“文化与技术一部曲：旧金山 | 个人计算机与反文化”影片放映讨论会
主题演讲嘉宾
（按演讲顺序排列）
迪伦·雷比林
Dylan Reibling
讲者简介迪伦·雷布林（Dylan Reibling）是一位加拿大电影导演，他的作品关注技术、政治与文化的交汇。他的首部长片纪录片《互联网的终结》（The End of the Internet, 2025）在哥本哈根国际纪录片节（CPH:DOX）参赛首映，并在BAFICI、慕尼黑纪录片节（DokFest Munich）及其他国际影展放映。影片探讨了去中心化运动以及数字基础设施如何塑造社会、政治和经济生活。雷布林曾与维斯媒体（VICE） 合作制作备受好评的系列纪录片网络战争（Cyberwar），并在北美和欧洲的各大影展与机构展示作品。他目前正在开发关于数字未来、技术前沿与政治思想的新项目。
演讲主题追问基础设施：平台、预测与提示词
摘要基础设施让什么成为可能——又让什么变得不可能？
本次主旨演讲探讨基础设施——无论是网络协议、平台，还是预测算法——如何塑造政治、社会与文化生活。导演迪伦·雷布林（Dylan Reibling）在制作纪录长片《互联网的终结》（The End of the Internet, 2025）期间，与技术专家、社会行动者及不同社群的交流中，梳理了分布式系统从冷战时期的起源到当下平台主导的历史轨迹，揭示了自由理想与商业现实之间的张力。
早期的去中心化网络愿景曾许诺韧性与非等级化的交流，这些理想与上世纪七十年代的反主流文化运动互相呼应。然而，万维网的商业化进程重新集中了控制权，使平台成为数字体验的主要中介，并在算法筛选、监控与商业化过程中生成了新的限制形式。
人工智能则带来了进一步的转变：早期的文本性基础设施强调溯源、处理与规训，而当代系统则依赖“提示”（prompt）与“预测”（prediction）。大型语言模型并非在推理，而是在根据过往数据生成概率性的延续，这引发了关于知识、想象力与规模极限的诸多疑问。
主旨演讲提出，对基础设施的批判性审视意味着揭示技术系统内部被构筑的“能供性”（affordances），因为这些能供性正塑造着我们的世界。若我们不加审视地接受技术所提供的无缝体验，就有可能放弃那些需要关怀、讨论与集体投入的制度与价值。因此，基础设施必须被视为不仅仅是技术的底层结构，更是决定我们所能想象与构建的未来的关键因素。
乔安娜·莫尔
Joana Moll
讲者简介她是一位常驻巴塞罗那和柏林的艺术家与研究者。她的作品批判性地审视技术资本主义叙事如何塑造机器、人类及生态系统的“字母化”过程，聚焦于数据的物质性、监控，以及通过数字媒体对公民社会的军事化。她的项目曾在全球多个重要场域展出，包括威尼斯双年展（Venice Biennale）、奥地利电子艺术节（Ars Electronica）、卡尔斯鲁厄艺术与媒体中心（ZKM），并被《纽约时报》（The New York Times）、《金融时报》（The Financial Times）等媒体报道。她共同创立了位于巴塞罗那的批判性界面政治研究小组（Critical Interface Politics Research Group），并曾与莫兹拉基金会（Mozilla Foundation）合作。目前，她担任科隆媒体艺术学院（KHM）网络学教授，同时也是巴塞罗那设计与工程学院的访问讲师。
演讲主题内在的界面
摘要2012年，一段广为流传的视频中，一个一岁大的孩子试图在纸质杂志上进行“滑动”“捏合”“放大”的手势操作——当图像没有任何反应时，她显得十分沮丧。视频的结尾，家长说了一句意味深长的话：“史蒂夫·乔布斯已经在她的操作系统里写入了代码。”本次演讲以这一瞬间为切入点，探讨科技如何将自身铭刻进我们的身体、姿态与想象之中，并揭示在数据提取时代，艺术实践如何揭露那些潜藏在技术、意识形态、物质性与资本之间的无形纠缠。
斯科特·摩尔
Scott Moore
讲者简介Scott Moore 是 Public Works 的创始人，该机构是一家开源生态基金，专注于支持艺术家与创意技术实践者。除了个体创作者之外，Public Works 还支持了多个项目，包括 Fuser、Gitcoin、Metalabel 与 Spline。Moore 同时担任旧金山非营利文化机构 Gray Area 的活跃顾问。业余时间，他亦参与多个“外制度性”（extitutional）学术团体的合作研究，其中包括 Metagov，主要探讨网络化的人类协作问题。
博格娜·科尼尔
Bogna Konior
讲者简介Bogna Konior 是一位学者与作家，其研究领域聚焦于新兴技术。她现任纽约大学上海分校（NYU Shanghai）媒体理论助理教授，同时在人工智能与文化研究中心（Artificial Intelligence & Culture Research Center）以及互动媒体艺术系（Interactive Media Arts Department）工作。
她著有《互联网的黑暗森林理论》（Dark Forest Theory of the Internet），该书将于 2025 年由 Polity Press 的 “Theory Redux” 丛书系列出版。
她与 Benjamin Bratton 及 Anna Greenspan 共同担任《机器决策并非最终：人工智能的中国史与未来》（Machine Decision is Not Final: China and the History and Future of Artificial Intelligence）一书的联合编辑，该书将于 2025 年由 Urbanomic 出版，并由麻省理工学院出版社（MIT Press）发行。
对谈主题机器决策并非最终：质疑“静默的编辑者”（Fireside Chat）
摘要我们今日所阅读的内容，越来越多地由那些不可见的编辑所塑造——即所谓的“静默编辑者”（quiet editor）：隐藏在提示式界面背后的系统提示（system prompts）、安全过滤器（safety filters）与检索管线（retrieval pipelines）。它们决定了谁能被看见、哪些语气被允许、哪些语境被抹除。它们以“可信度”的外观取代可验证的来源，以“合成的权威”取代脚注。
“静默的编辑者”在超越人类监督的规模与速度中运作，每秒做出成千上万次编辑性决策，判断何者可被视为合理、可信，甚至可言说。然而，在用户看来，这些决策却呈现为自然的、中立的、乃至不可避免的——一种无需上诉的机器裁决。
本场与 Bogna Konior 的对谈将探讨在此条件下，质疑与争辩的可能性与意义，尤其聚焦于她的“互联网黑暗森林理论”（Dark Forest Theory of the Internet）。讨论将延伸至当“公共理性”的基础设施变得专有且不透明时的后果，以及当人工智能系统能生成无限、乃至相互矛盾的证据时，会出现何种新的现实。
在一个沟通本身充满风险的世界里，编辑究竟是暴力，还是关怀？
尼克·蒙特福特
Nick Montfort
讲者简介尼克·蒙特福特（Nick Montfort）的创作成果包括十部由计算机生成的著作（由七家出版社出版）、合作项目《删除主义者》与《海与雀之间》，以及五十余项独立数字艺术项目。其最新诗集《全为胜利》全部由三个字母的单词构成。他在MIT出版社出版的著作包括《未来》及两部合编文集：《新媒体读本》与《输出：计算机生成文本选集（1953-2023）》。蒙特福特现任麻省理工学院教授，同时担任卑尔根大学数字叙事中心首席研究员，并主导名为“修辞实验室”的创新工作室。他常年居住在纽约市。
演讲主题编码、点击与提示：生成艺术的方式
摘要自可编程的通用计算机诞生之初，艺术家们就开始尝试用它来创作艺术，并发展出多种方法。其中，编写计算机程序（coding）无疑是最早让计算机“生成艺术”的方式。起初，这种编程直接基于特定处理器的指令集进行；不久之后，更高层次的编程语言出现，使艺术家与程序员能够以更易于理解的方式进行创作。计算机发明后仅二十年间，创新者们就已开发出专门用于艺术创作的平台与工具。例如，在创意文本生成领域，约瑟夫·魏岑鲍姆（Joseph Weizenbaum）因开发出计算机角色“DOCTOR”而闻名——这是历史上第一个聊天机器人。然而，他在20世纪60年代的作品 ELIZA 不仅仅是创造了一个角色，而是一个可用于构建不同对话系统的通用平台。同一时期，马克斯·马修斯（Max Matthews） 开发了通用合成器系统 MUSIC，而艾文·萨瑟兰（Ivan Sutherland） 的 Sketchpad 则首次实现了矢量绘图功能，成为图形用户界面的先驱。
如今，除了晦涩的底层编程（汇编语言与低级编码）、更易上手的高级语言编程，以及从超文本系统到照片编辑、三维建模等各类艺术创作工具之外，一种全新的计算机交互方式——生成式人工智能（Generative AI） 正在迅速兴起。艺术家不再需要编写针对处理器指令集的代码，也无需构建复杂的面向对象程序，甚至不必通过鼠标点击完成创作，而是可以通过输入提示词（prompt）来观察系统的生成结果。尽管我会对此次最新发展的具体细节发表一些看法，但我论述的核心观点是：新的计算艺术方式并不会取代旧的方式。如今，欧洲的演示场景（demoscene） 社群中仍有创作者使用极低层的编程语言来制作实时动画；一些艺术家依然亲自编写程序来处理图像，而非依赖Photoshop的现成功能。尽管创作平台与工具层出不穷，现场编程（live coding） 的艺术家往往会自己搭建平台——就像音乐家发明自己的乐器一样。在现场编程表演中，代码的编写与修改通常会实时投影并公开展示，使得透明性甚至某种意义上的编程教育成为表演的一部分。因此，真正值得思考的问题不是：当提示词写作取代其他方式时我们该怎么办？而是：当生成式人工智能成为计算机艺术生态的一员时，这一新的艺术生态将呈现出怎样的面貌？
费利克斯·斯塔尔德
Felix Stalder
讲者简介费利克斯·斯塔尔德（Felix Stalder）现任苏黎世艺术大学教授，执教于美术学院“艺术与人工研究”专业。他的研究聚焦于文化、政治与技术动态的交汇点，尤其关注基于共同体的新型生产模式、版权制度、数据化进程、人工智能以及主体性转型等领域。
除了学术工作外，他还是一位活跃的文化实践者，长期担任批判性网络文化重要平台“网时代”的协调人——从早期的邮件列表到如今在联邦宇宙中的节点。他是维也纳世界信息研究所与技术政治工作组的核心成员。其学术成果丰硕，重要著作包括：《数字团结》（2014年）、《数字状况》（2016年、2018年、2023年）、《共同体的美学》（2021年）、《数字无意识》（2021年）、《从共同体到NFT》（2022年）以及《具身数据实践中的当代性》（2025年）。更多信息可访问：http://felix.openflows.com/
演讲主题数字文化、生成式人工智能与艺术
摘要如果我们把“文化”理解为社会用以组织意义、并据此引导行动的所有活动之总和，那么生成式人工智能正在对数字语境下的文化带来一种深刻的转变。
最初，数字文化由三种过程的互动所塑造：指涉性（referentiality）——意义源于事物之间的关联；共同性（communality）——意义在群体中被协商形成；以及算法性（algorithmicity）——意义通过自动化的选择过程而生成。
然而，大型语言模型（LLMs）及其他类型的生成式人工智能并不依循这些路径。它们所生成的内容看似具有文化意义，实则只是通过对大量变量的相关性计算得出的结果，并不依赖外部的指涉、逻辑或语义。由此，这类系统无法生成用以“指导行动的意义”，而是生成“行动”来引导意义。用语言学的术语来说，语言的功能从“表义”（signification）转向了“行动”（action）。借用唐纳德·麦肯齐（Donald McKenzie）的话，这些模型不是照相机，而是引擎。
随着这种转变，生成式AI在艺术领域最有趣的探索，是将其用作激发创作能动性的工具，而把意义的诠释权明确留给人类——如同精心设计的留白。
杰克·斯蒂尔戈
Jack Stilgoe
讲者简介杰克·斯蒂尔戈（Jack Stilgoe）现任伦敦大学学院科学技术学教授，致力于新兴技术治理研究。他担任英国研究与创新署（UKRI）“负责任人工智能”领导团队成员（官网：www.rai.ac.uk），曾主持经济与社会研究理事会（ESRC）“无人驾驶未来”项目（2019-2022）。作为常受关注的媒体评论专家，他同时担任英国政府人工智能、自动驾驶汽车等新兴技术领域的政策顾问。
演讲主题让人工智能回归现实
摘要随着人工智能浪潮的推进，关于其利弊的讨论显得愈发吊诡：益处常被描绘成乌托邦式的理想图景，风险则被渲染为关乎人类存亡的威胁。我们需要摒弃“AI例外论”，将其视为一种“寻常技术”（Narayanan and Kapoor 2025），并汲取以往新兴技术发展的经验教训。本次演讲将基于针对公众与人工智能研究者的最新调查结果，提出一个观点：我们正处于建立更具建设性讨论的契机之中。
詹姆斯·斯坦霍夫
James Steinhoff
讲者简介詹姆斯·斯坦霍夫（James Steinhoff）博士现任都柏林大学信息与传播研究学院助理教授，他深耕于人工智能、数据与数字媒体的政治经济学领域，是该领域颇具影响力的学者。其专著《自动化与自主性：人工智能产业中的劳动、资本与机器》（帕尔格雷夫出版社，2021年）及合著《非人的力量：人工智能与资本主义的未来》（普鲁托出版社，2019年）已成为该领域重要文献。
演讲主题提示工程与（非）决定论的政治经济学
摘要本次演讲提出，可以将“提示”（prompting）的政治经济学理解为决定论与偶然性之间的博弈。资本作为一个系统，其内在逻辑总是在竭力消除运营过程中的偶然性以促进价值增殖——即便个体资本间的竞争又在不断催生新的偶然因素。提示词技术本质上是概率性算法，然而它作为接口所服务的AI，却已成为资本主义技术狂热的核心载体。本次演讲将探讨提示词为何具有概率性特征，并评估近期学界尝试在大型语言模型LLMs中“消除”偶然性的潜在影响。这种非决定性的消除在技术上看似可行，但其代价在于——它引入了一个新的计算维度，从而重塑人工智能的政治经济学格局。
路淼
讲者简介路淼博士是香港岭南大学文化研究系的助理教授。她的研究兴趣包括基础设施研究、科学与技术研究以及中非研究。她的英文专著将于2025年秋季由美国伊利诺伊大学出版社出版。她的论文发表在Information, Communication & Society, New Media & Society, and Science, Technology & Human Values等多个英文刊物上。
演讲主题热带地区的数据中心：西非的热量管理、中国科技与能源政治
摘要近年来AI与云计算的火热引发了数据中心在全球的建设热潮。不同于以往数据中心选址对温带的偏向，中国在热带地区的扩张表明全球云基础设施的的地理分布与物质性都在变动之中。虽然华为在非洲已经跃升为一个重要的数据中心提供商，但却面临着热管理的棘手问题。在加纳，华为承建的国家数据中心最近引入了太阳能以期降低能耗与碳排放。本文基于2023至2025年的采访与田野调查来解释加纳云基础设施的热政治。通过“热带物质性”这一理论框架，本文将分析一系列国家与企业的行动者如何管理、协商数据中心的物质性。经验层面，本文揭示了绿色能源的矛盾性以及全球云的地理分布变化。理论层面，本文探究了热带如何帮助我们理解国家、自然与基础设施之间的复杂互动。
创意演讲嘉宾
罗霄
讲者简介罗霄，算法工程师，数字艺术家。实验编程创始人，从事交互媒体、人工智能、生成艺术的创作与教学。同济大学建筑城规学院AI与艺术创研中心实践导师，AI艺术创新联盟AIAIA常务理事。毕业于北京航空航天大学，电子信息工程与软件工程专业背景。曾就职于腾讯ISUX和网易游戏。2005年起从事游戏开发与新媒体艺术类项目，2020年引入AI技术重构创作范式。个人作品于中国、英国、日本、南非、意大利等地展出。
演讲主题当生成成为日常：AI 时代的流变、协议与遗忘
摘要生成走入日常，如何避免数字艺术作品沦为可复制的提示组合？本讲座以实践者视角，从经典生成艺术的”规则”导向，转向生成式AI时期的”流变”。系统讲解三种方法：以结构保持确保形体与语义不走样，以因果约束明确可发生范围，以时间性塑形锻造节律与质感；通过可复演的生成链路，探讨AI时代的创作者在创作与展示中如何保留作者性、来源与过程。
工作团队
会议总召：黄孙权
执行策划：王婧洁
会务统筹：郑叶颖、王安琪、曾梦蕊
工作组：周蓬岸、丁乐、曾梦蕊、井俱进、田沛松、王安琪、李梦书
导播与影像：井俱进、李艺欣、李易扬、赵佳菁、李程锦
视觉文宣：李梦书、丁乐、郑叶颖
声音：Boyang GUAN
动效：Shir- Shiko
The 10th Annual Conference of Network Society
The End of the Internet? — AI: In or Out
- HOST
- Institute of Network Society (INS),
School of Intermedia Art (SIMA),
China Academy of Art (CAA) - CONVENER
- HUANG Sunquan
Academic Committee
LI Kaisheng
MIN Han
HUANG Sunquan
CHEN Jieren
LIU Yisi
ZHOU Pengan
INTRODUCTION
Over the past decade, we have consistently pursued a “triple remix”of theory, technology, and action to interrogate the publicness of network society. The first conference, themed “The Power of Networking,” questioned the paradigm formation of techno-social ecology. The second, following Lefebvre’s itinerary, examined urban life and the everyday in the algorithmic age. The third, “Smart Urban Networks,”turned toward infrastructures and civic technologies. The fourth, “Netizens 21,”returned to media archaeology and platform critique. The fifth and sixth, held during the global pandemic, grounded discussions in practical experiments—distributed networks, encryption and cooperative token economies, metaverse rural reconstruction—opening up the linkage among tools, organization, and collaboration. The seventh contrasted P2P with the vocabulary of Web3, unfolding across seven global cities to clarify the contribution of distributed urban public cultures. The eighth reread the genealogy of counterculture and digital technology from the 1960s to the present. The ninth shifted the field site to the Global South, beginning in Chiang Mai to trace manufacturing, mobile connectivity, social media, and new terrains. This trajectory—from theoretical diagnosis, to protocol design, to geopolitical sites—points to the Institute’s enduring concern: How is the public preserved and reinvented amid technological renewal? How has the network transformed aesthetic production?
The 10th Conference, “The End of the Internet?”is not an apocalyptic proclamation, but an interrogation of systemic transition.When the public techniques of hyper-connectivity (citation, interoperability, transparency , and traceability) are replaced by natural-language prompting interfaces (large databases, system defaults, agentic decisions), are we crossing an invisible yet decisive threshold? “AI:In or Out”is not the doomer/ bloomer ideology of historical finality; it is the familiar advertising script of capitalist operation.
Today, social media platforms own “private data”(corporate assets made from personal data), making them effectively disconnected. Now the governance of AI is even more opaque: the entire history of humanity, the crystallization of humanistic and artistic intelligence, is subsumed without public authorization, without attribution or mechanisms of reciprocity, into a new form of “model capital”—something rentable, encapsulable, and arbitrageable. Moreover, who—and what kinds of data—are included within the language, corpora, and benchmarks of models? Who and what are excluded from computation, standards, and profit- sharing? This is the “protocol”question we have repeatedly posed for a decade—indeed the very key to the network-sociological concept of publicness.
In the early internet, public life was sustained by open protocols—HTTP, SMTP, RSS, ActivityPub—where fields, formats, and interoperability were collectively agreed upon. This was “inter-networking”:I could connect to you, and you to me; the rules were open, auditable, and replaceable. Today, the nodes have subtly shifted: we face model interfaces—API terms, system prompts, alignment strategies, content filtering, and evaluation constraints. These also function as “protocols,”but now as black-boxed, mutable, and unilaterally advanced by a handful of providers. Thus, the transformation unfolds: from open-protocol interoperability to model-interface inter-guidance.
The inter-guidance led by giant AI corporations is, in fact, the exploitation and hyper-valorization of “general intellect”(in Marx’s sense). Thus, when work becomes prompt, we may be facing the most dangerous turning point in human civilization: humanity’s collective reservoir of knowledge is filtered into “large language models”; language, the loom of social relations, is transformed into a teleprompter interface; craft, judgment, and aesthetic training are compressed into tokenized commands billed line by line. The danger lies not in machines becoming more intelligent, but in the corporate enclosure of general intellect as proprietary capital—concealing the broader social process of extraction–capital–labor, while gradient rents are captured through black-box evaluations and infrastructural monopolies. When work is prompt, if we do not rewrite the rules, we effectively hand over authorship, publicness, and the right to future distribution to system defaults:browsing and critical analysis are replaced by summoning; craftsmanship and coding are simplified to prompting; authorship is diluted into parameters.
As the Network Society Conference reaches its tenth anniversary, we face not only the compound power of capital-technology and the socio-technical construction of culture-technology, but also algorithmic sovereignty, corpus sovereignty, and infrastructural struggles among nation-states on a global scale. Perhaps no moment is closer to Marx’s Grundrisse, where he wrote: “The development of fixed capital indicates that general social knowledge has become a direct force of production.”And again:”Fixed capital is the most adequate form in which capital, as capital, can be represented.”Is this not artificial intelligence? Is AI not capitalism’s dream lover?
The tenth year is not a call to forget earlier questions in order to update solutions; rather, it is to find, within the ongoing battlefield, the ideals and practical possibilities lost in the late internet era—precisely what enables us to confront a seemingly new, yet more refined and labor-efficient form of exploitation.
In retrospect, the Institute of Network Society will showcase its decade of cumulative research projects (such as SparkLink, Archverse, cooperative models of computation, urban forums, young scholars forums, and hackathons). In addition to keynote lectures, this year’s conference will also exhibit research materials and video screenings from the Culture & Technology Trilogy, along with the Young Scholars Forum.
- TIME
- October 17–19, 2025 (UTC+8:00)
- PLACE
- UFO Terminal, Tank No.1, West Bund Tank Shanghai
(2398 Longteng Avenue, Xuhui District, Shanghai, China)
LIVE STREAMING
On-site Registration
Please click the link or scan the QR code to complete your registration.
Pre-registered participants will have priority access to simultaneous interpretation devices and reserved seating.
(A valid ID is required as a deposit when collecting interpretation equipment on site.)
AGENDA
October 17 2025
Time(UTC+8) Activities
A.M.
Forum I: Young Scholars Session
9:00
Venue Opens
9:20-9:30
Welcoming Remarks by Liu Yisi
9:30-10:30
Self-organized Panel｜Affective Prompts
Commenter: LIU Yisi
Speakers:
ZHU Yun｜My “Lover”: The Chinese Scholarly Landscape of Otome Game Studies (2019–2025)
CHEN Suyuan｜Romancing with AI: The Commercial Construction, Textual Conditioning, and Intimacy Practices of AI Products—A Case Study of “Replika”
ZHENG Yeying｜Disembodied and Infinite Love: How General Conversational AI Reweaves the Network of Human Relationships
YE Yunhe｜From Embodied Encounters to Algorithmic Matching: Blued’s Technological Mediation and the Construction of Intimacy among Chinese Generation-Z Gay Men
(Institute of Network Society, China Academy of Art)
10:30-11:10
Self-organized Panel｜Hijacked Futures and the (Im)possibility of Alternative Imagination: The Tech Right, Venture Capital, and the Specter of Anti-Democracy
Commenter: JIANG Feiran
Speakers:
JIN Bowen｜The Political Engineering of Venture Capital: How the Tech Right Drives the Corporate Statization of the American State
YANG Qichen｜Dark Enlightenment, Antichrist, and Techno-Messiah: How Does the American Tech Right Construct a Contemporary Theological Narrative?
QIN Chuan｜Arts and Humanities Labs under Technological Governance: A Genealogical Inquiry into the Critical Role of Art Schools
(The Institute of Contemporary Art & Social Thoughts, China Academy of Art)
11:20-12:20
Creative Talk
moderator: ZHOU Peng’an
Speaker: LUO Xiao
Speech Title: When Generation Becomes Routine: Flux, Protocol, and Oblivion in the AI Era
October 18 2025
Time(UTC+8) Activities
A.M.
Keynote Forum – I
9:00-9:30
Opening Remarks
MIN Han, Vice Dean of School of Intermedia Art(SIMA), CAA
Prof. HUANG Sunquan, Director of the Institute of Network Society and Conference Convener
9:30-10:30
Panel 1｜The Quiet Editor：Public Reason under Prompted Interfaces
Speaker: Dylan Reibling
Speech Title: Interrogating Infrastructure: Platforms, Predictions, and Prompts
10:30-10:45
Break
10:45-11:45
Panel 3｜Atlas of AI Infrastructures: Cloud Heat, Ground Truth
Speaker: Joana Moll
Speech Title: The Interface Inside
11:45-12:45
Panel 1｜The Quiet Editor：Public Reason under Prompted Interfaces
Speakers: Scott Moore & Bogna Konior
Fireside Chat Title: Machine Decision is Not Final: Contesting the Quiet Editor
P.M.
Keynote Forum -II
2:00-3:00
- Panel 4｜When the work is Prompt
- Speaker
- Nick Montfort
- Speech Title
- Coding, Clicking, and Prompting to Make Art
3:00-4:00
- Panel 4｜When the work is Prompt
- Speaker
- Felix Staldert
- Speech Title
- Digital culture, Generative AI, and Art
4:00-4:15
Break
4:15-5:15
Panel 2｜Gradient Rent: The Political Economy of Prompts
Speaker: Jack Stilgoe
Speech Title: Bringing AI Down to Earth
6:30-8:30
Film Screening and Discussion
The End of the Internet by Dylan Reibling
8:30-10:00
Live Coding Performance
October 19 2025
A.M.
Keynote Forum -III
9:30-10:30
Panel 2｜Gradient Rent: The Political Economy of Prompts
Speaker: James Steinhoff
Speech Title: Prompting and the political economy of (non)determinism
10:30-11:30
Panel 3｜Atlas of AI Infrastructures: Cloud Heat, Ground Truth
Speaker: LU Miao
Speech Title: Data center for the tropics: Heat management, Chinese tech, and energy politics in West Africa
P.M.
Young Scholars Forum
1:00-2:30
Young Scholars Forum｜Panel 1 The Quiet Editor：Public Reason under Prompted Interfaces
Commenter: HUANG Sunquan
Speakers:
DING Fan, YUAN Jinyu, ZHANG Jingni(College of Arts & Media, Tongji University)｜The Quiet Editor and the Reconstruction of Public Rationality: A Critical Study of Implicit Power Mechanisms from the Perspective of Communicative Rationality
HUANG Yuemin(East China Normal University), GUO Dumo(Royal College of Art)｜Mirrors of Infinite Depth: Internal Circulation of LinguisticSymbols in the Post-Internet Era
ZHAO Jiajing(Institute of Network Society, China Academy of Art)｜Heresy Marginalia: The Struggle of Jailbreak Prompts
LI Yangzi(Nankai University)｜From Collective Resonance to Algorithmic Response: How Generative AI Reshapes Queer Narratives and the Politics of Visibility in Fandom
Viola Leqi He, Max Loeffler(NYU Shanghai)｜Violence of Alignment: How to Stop Worrying and Love Haunted Software
LU Manman(Shandong University)｜From “Super Puppets” to “Cyber Super Puppets”: Power Manipulation and Ethical Examination of Virtual Influencers in Intelligent Advertising(online)
2:30-3:30
Young Scholars Forum｜Panel 2 Gradient Rent: The Political Economy of Prompts
Commenter: HUANG Sunquan, LIU Yisi
Speakers:
TONG Zhaoxiang(Institute of Network Society, China Academy of Art)｜The Uneven Geographies of Artificial/Intelligence: The Labor Process of Data Annotators
XU Ke(Institute of Network Society, China Academy of Art)｜Mothers on the Digital Assembly Line: The Dilemmas of Motherhood for Female Data Annotation Workers in China’s Counties
HUA Xizi(NYU Shanghai)｜Parasitize Between Systems: Cross-Model Protocol Aesthetics As Method
WANG Shaowei(Media and Communication, University of Oslo)｜Prompt as Wager: Semantic Finance and the Fourth Value Form of Traffic in Vertical Drama Economies(online)
3:30-3:45
Break
3:45-4:15
Young Scholars Forum｜Panel 3 Atlas of AI Infrastructures: Cloud Heat, Ground Truth
Commenter:
Speakers: PAN Ji
MU Haojie(Fudan University)｜“Green” versus “Panoramic”: Frictions and Politics of Urban Infrastructure
ZHANG Xushan(Tsinghua University)｜The Meshwork of Lines: Following the Fluid Trajectories of Data Centers
4:15-6:30
Young Scholars Forum｜Panel 4 When the work is Prompt
Commenter: HUANG Sunquan, PAN Ji, LIU Yisi
Speakers:
JIANG Yizhu(Institute of Network Society, China Academy of Art)｜machine imaginaire: A Critique of Aesthetic and Algorithmic Power in AI-Generated Art
SUN Jinghе(School of Intermedia Art, China Academy of Art)｜Interface, Feedback, and Decisional Violence: The Isomorphic Logic Between Video Games and Warfare
YAN Hui(Independent researcher)｜The algorithmic reconstruction of authorship: How AI Art reshapes creative identity and power
REN Xuanqi, WU Yufeng(Tianjin Academy of Fine Arts)｜From Poor Images to Prompts: Visual Genealogy and Authorship Reconstruction of Dreamcore Aesthetics in the AIGC Era
LIN Qin(University of Nottingham Ningbo China)｜Beyond the Algorithm: Cultural Expression of AI Divas from Production to Performance
YAO Lijin(The Research Institute of Better China Initiative of CAA)｜Judgment as Infrastructure in Human–AI Collaboration: The Case of Archival Exhibition Work
SHI Yijia(Department of Transmedia Art, Xi’an Academy of Fine Arts)、WANG Zhizi、TANG Lan(Center for Art Education, Xi’an Jiaotong University)｜Prompt as Method: Authorship, Presence, and Aura in AI-Generated Art
JIN Qiuyu(Department of Arts Studies and Curatorial Practices, Graduate School of Global Arts, Tokyo University of the Arts)｜Prompt Mechanisms and Audience Memory: The Transformation of Individual Experience in Contemporary Time(online)
LIU Lei April(Global Arts and Culture, Rhode Island School of Design)｜Accelerant: Designer’s Choice: Norman Teague – Jam Session and the Institutionalization of AI Art and Design(online)
19:30-21:00
HUANG Sunquan: Screening and Discussion of Episode I: San Francisco: Personal Computer and Counterculture
- SPEAKERS
- (Arranged in order of the speeches)
Dylan Reibling
BioDylan Reibling is a Canadian filmmaker whose work explores the intersections of technology, politics, and culture. His debut feature documentary, The End of the Internet (2025), premiered in competition at CPH:DOX and has screened internationally at BAFICI, DokFest Munich, and beyond. The film examines the decentralization movement and the ways digital infrastructures shape social, political, and economic life. Reibling has previously worked with VICE on the acclaimed series Cyberwar and has presented work at festivals and institutions across North America and Europe. He is currently developing new projects on digital futures, technological frontiers, and political thought.
Speech topicInterrogating Infrastructure: Platforms, Predictions, and Prompts
AbstractWhat do infrastructures make possible—and what do they foreclose? This keynote examines how infrastructures—whether network protocols, platforms, or predictive algorithms—shape political, social, and cultural life. Drawing on encounters with technologists, activists, and communities during the making of the feature documentary The End of the Internet (2025), director Dylan Reibling traces the trajectory of distributed systems from their Cold War origins to the present dominance of platforms, highlighting the tensions between emancipatory ideals and commercial realities.
Early visions of decentralized networks promised resilience and non-hierarchical communication, ideals that resonated with countercultural movements of the 1970s. Yet the commercialization of the World Wide Web re-centralized control, naturalizing platforms as the primary mediators of digital experience. This infrastructural consolidation has produced new forms of constraint through algorithmic filtering, surveillance, and monetization.
Artificial intelligence represents a further transformation: where earlier textual infrastructures emphasized provenance, processing, and proscription, contemporary systems rely on prompts and prediction. Large Language Models do not reason but generate probabilistic continuations of past data, raising questions about knowledge, imagination, and the limits of scale.
The keynote argues that interrogating infrastructure means uncovering the affordances built into technical systems—because these affordances shape the world. To accept the seamless experiences offered by technology without scrutiny risks abandoning institutions and values that require care, debate, and collective commitment. Infrastructure, therefore, must be critically examined not only as a technical substrate but as a determinant of the futures we are able to imagine and build.
Joana Moll
BioShe is a Barcelona/Berlin-based artist and researcher. Her work critically examines how techno-capitalist narratives shape the “alphabetization” of machines, humans, and ecosystems, focusing on data materiality, surveillance, and the militarization of civil society through digital media. Her projects have been presented globally at venues such as the Venice Biennale, Ars Electronica, ZKM, and The New York Times, The Financial Times, and Wired. She co-founded the Critical Interface Politics Research Group in Barcelona and has collaborated with the Mozilla Foundation. Currently, she is a Professor for Networks at the Academy of Media Arts Cologne (KHM) and a visiting lecturer at Elisava, Barcelona School of Design and Engineering.
Speech topicThe Interface Inside
AbstractIn 2012, a viral video showed a one-year-old trying to swipe, pinch, and zoom on a paper magazine, growing frustrated when the images failed to respond. Her parent ended the clip with a striking remark: “Steve Jobs has coded a part of her operating system.” This talk uses that moment as a lens to explore how technology inscribes itself into our bodies, gestures, and imaginations, and it will reveal how artistic practice can uncover the often invisible entanglements between technology, ideology, materiality, and capital in the age of data extraction.
Scott Moore
BioScott Moore is a co-founder at Gitcoin and a General Partner at Public Works. He also serves as an advisor to a wide range of other projects pushing forward plural technologies including Metagov and the Plurality Institute. By pioneering new forms of internet-native funding, he has helped collectively distribute over $500m to causes ranging from climate change to open source software.
Bogna Konior
BioBogna Konior is a scholar and a writer whose work focuses on emerging technologies.
She is currently Assistant Professor of Media Theory at NYU Shanghai, where she works at the Artificial Intelligence & Culture Research Center, and the Interactive Media Arts department.
She is the author of Dark Forest Theory of the Internet, forthcoming in the Theory Redux Series from Polity Press in 2025. With Benjamin Bratton and Anna Greenspan, she is the co-editor of Machine Decision is Not Final: China and the History and Future of Artificial Intelligence, forthcoming from Urbanomic in 2025, and distributed by MIT Press.
Fireside Chat topic: Machine Decision is Not Final: Contesting the Quiet Editor
AbstractWhat we read today is increasingly shaped not by editors we can see, but by the “quiet editor”: system prompts, safety filters, and retrieval pipelines hidden within prompted interfaces. These determine who becomes visible, what tones are permissible, which contexts get erased. They replace verifiable sources with the appearance of credibility, footnotes with synthetic authority.
The quiet editor operates at scale and speed that exceeds human oversight, making thousands of editorial decisions per second about what counts as reasonable, credible, or even sayable. Yet these decisions appear to users as natural, neutral, even inevitable; machine verdicts handed down without appeal.
This conversation with Bogna Konior examines whether contestability is desirable or even possible under these conditions with specific focus on her Dark Forest Theory of the Internet. It will explore the implications of our infrastructure for public reason becoming proprietary and opaque, and what arises from AI systems that can generate unlimited, even contradictory evidence.
In a world where communication attracts danger, is editing violence or care?
Nick Montfort
BioNick Montfort’s work includes ten computer-generated books (in print from seven presses), the collaborations The Deletionist and Sea and Spar Between, and more than fifty individual digital projects. His latest poetry book, All the Way for the Win, is composed entirely of three-letter words. His MIT Press books include The Future and two co-edited volumes, The New Media Reader and Output: An Anthology of Computer-Generated Text, 1953–2023. He’s a professor at MIT, principal investigator in the University of Bergen’s Center for Digital Narrative, and directs a lab/studio, The Trope Tank. He lives in New York City.
Speech topicCoding, Clicking, and Prompting to Make Art
AbstractArtists have made art with general-purpose, programmable computers almost from the moment of their invention, using many different methods. Writing computer programs (coding) was certainly the earliest way to coax art from computers. Initially, that coding was done using the instruction sets of particular processors. Before long, high-level languages allowed artists, and others, to write programs with greater ease, ones that were more legible to humans. Within two decades of the computer’s invention, innovators had developed special platforms and tools for the creation of art. In creative text generation, for instance, Joseph Weizenbaum is well-known for developing a computer character, DOCTOR, the first chatbot. However, his 1960s ELIZA was not just a way of making one character. It was a general-purpose platform for making different conversational systems. By that time, Max Matthews had developed MUSIC, a general synthesizer system, and Ivan Sutherland’s Sketchpad provided vector-drawing capabilities in the first-ever graphical user interface. Now, in addition to arcane types of programming (assembly and low-level coding), more approachable types of programming (in high-level languages), and tools for making art (from hypertext systems to applications for photo editing and 3D modeling), what seems to be a radically new way of engaging with the computer is erupting: Generative AI. Instead of writing to the specifics of a computer processor’s instruction set, or writing high-level object-oriented code, or developing artwork with clicks of a mouse in an app, we can write prompts and see what the system does.While I will have some to say about the specifics of this latest development, the main argument I’ll make is that new modes of computational artmaking do not obliterate previous ones. Some participants in today’s demoscene, a European community, craft their real-time animations with very low-level programming. Some artists write their own programs to transform images rather than using the abilities of Photoshop. While there are an abundance of creative platforms and tools available, those who program in performance (livecoders) very often write their own platforms — as if they were musicians inventing their own instruments. And livecoding almost always involves projecting and sharing the code as the musician and visualist develop and change it, so that transparency and even a sort of programming education becomes part of the performance. The question, then, is not what to do when prompt-writing takes over, but what the new ecology of computer art will be like as Generative AI takes its place among existing practices.
Felix Stalder
BioFelix Stalder is a professor the Zurich University of the Arts, teachingin the Department of Fine Arts “art:ifical studies” major. His workfocuses on the intersection of cultural, political and technologicaldynamics, in particular on new modes of commons-based production,copyright, datafication, AI, and transformation of subjectivity.
He not only works as an academic, but also as a cultural producer,as long-time moderator of <nettime>, a crucial nexus of critical netculture, first for the email list, now for its node in the fediverse. Heis a member of the World Information Institute and the TechnopoliticsWorking Group, both based in Vienna.He is the author/editor of numerous books, among others DigitalSolidarity (PML & Mute, 2014), Kultur der Digitalität / DigitalCondition /数字状况 (Suhrkamp, 2016/Polity Press, 2018, School of PublicArt, 2023), Aesthetics of the Commons (Diaphanes, 2021), DigitalUnconscious (Autonomedia, 2021), From Commons to NFTS (Ljubliana 2022)and “Contemporaneity in Embodied Data Practices” (Sternberg Press, 2025)
Speech topicDigital culture, Generative AI, and Art.
AbstractIf we take culture to be the sum of all activities through whichsocieties organize meaning (to guide action), then generative AIintroduces a profound transformation of (digital) culture. Initially,digital culture was shaped by the interaction of three processes:referentiality (meaning as a relation between things), communality(meaning negotiated within groups), and algorithmicity (meaning throughautomated selection). LLMs and other types of generative AI, on theother hand, do not use any of that. They produce what looks likecultural output through correlations only, but across a large number ofvariables. There is no recourse to external references, logic, orsemantics. In this way,they cannot generate meaning to guide action butgenerate action to guide meaning. In linguistic terms, the function oflanguage shifts from “signification” to “action”. These models, toparaphrase Donald McKenzie, are not cameras, but engines.Picking up on this shift, some of the most interesting artistic uses ofgenerative AI explore it as a tool to create agency while leavingmeaning-making explicitly to humans.
Jack Stilgoe
BioJack Stilgoe is a professor in science and technology studies at University College London, where he researches the governance of emerging technologies. He is part of the UKRI Responsible AI leadership team (www.rai.ac.uk). He was principal investigator of the ESRC Driverless Futures project (2019-2022). He is a regular media commentator and adviser to the UK government on AI, self-driving vehicles and other new technologies.
Speech topicBringing AI down to earth
AbstractAs the momentum behind artificial intelligence grows, the debate about its benefit and risks looks peculiar. Benefits are often presented as utopian; risks as existential. We need to resist AI exceptionalism, think about AI as a ’normal technology’ (Narayanan and Kapoor 2025) and remember the lessons of previous emerging technologies. In this talk, I will present new results from surveys of members of the public and AI researchers to argue that there is an opportunity for a more constructive debate.
James Steinhoff
BioDr. James Steinhoff is an assistant professor in the School of Information and Communication Studies at University College Dublin. His research concerns the political economy of AI, data and digital media. He is author of Automation and Autonomy: Labour, Capital and Machines in the Artificial Intelligence Industry (Palgrave 2021) and co-author of Inhuman Power: Artificial Intelligence and the Future of Capitalism (Pluto 2019).
Speech topicPrompting and the political economy of (non)determinism
AbstractThis talk suggests that we can think of the political economy of prompting as a struggle over determinism and contingency. In general, capital as a system seeks to eliminate contingency from its operations in order to facilitate valorization, even as it simultaneously generates new kinds of contingency via competition between individual capitals. Prompting is a probabilistic algorithmic technique, yet the AI which it serves as interface to is the centerpiece of capitalist techno-enthusiasm. I consider why prompting is probabilistic and evaluate the possible ramifications of a recent attempt to “defeat” contingency in LLMs. The elimination of such non-determinism seems possible, but it has costs which introduce a new dimension of calculation into the political economy of AI.
LU Miao
BioMiao Lu is an Assistant Professor in the Department of Cultural Studies, Lingnan University, Hong Kong. Her research interests include infrastructure studies, science and technology studies, and China-Africa studies. Her monograph “The Transsion Approach: Translating Chinese Mobile Technology in Africa” will be published by the University of Illinois Press (Geopolitics of Information Series) in Fall 2025. Her works have appeared in Information, Communication & Society, New Media & Society, and Science, Technology & Human Values.
Speech topicData center for the tropics: Heat management, Chinese tech, and energy politics in West Africa
AbstractThe craze for AI and cloud computing has fueled a data center building boom across the world. Departing from the temperate normality in selecting data center sites, China’s expansion into the tropics suggests both the geography and materiality of cloud infrastructure are in flux. While Chinese tech giant Huawei has risen into a significant data center provider in Africa, it faces new challenges of heat management in the tropical climate. In Ghana, the Huawei-built National Data Center recently introduced solar technology with the aim of reducing energy cost and carbon footprint.
Drawing from interviews and ethnographic research from 2023 to 2025, this paper elucidates the thermopolitics of cloud infrastructure in Ghana. Using “tropical materiality” as a theoretical framework, this paper investigates how a variety of state and corporate actors navigate and negotiate the tropical materialities of data centers relating to heat and energy. Empirically, this paper sheds new light on the paradoxes of green energy and the shifting cloud geography. Theoretically, this study contributes to infrastructure studies, global China studies, and science and technology studies by asking what we can learn from the tropics to understand the intricate entanglements between state, nature, and infrastructure.
CREATIVE TALK SPEAKER
LUO Xiao
BioLuo Xiao is an algorithm engineer and digital artist, as well as the founder of Experimental Coding. He specializes in interactive media, artificial intelligence, and generative art through both creation and education. Currently, he serves as a Practicum Mentor at the AI and Art Creation & Research Center of Tongji University’s College of Architecture and Urban Planning, and holds the position of Executive Director at the AI Art Innovation Alliance (AIAIA).
Luo Xiao graduated from Beihang University with a background in Electronic Information Engineering and Software Engineering. He previously worked at Tencent ISUX and NetEase Games. Since 2005, he has been engaged in game development and new media art projects, and in 2020 began integrating AI technologies to reshape creative paradigms. His personal works have been exhibited in China, the UK, Japan, South Africa, Italy, and other countries.
Speech topicWhen Generation Becomes Routine: Flux, Protocol, and Oblivion in the AI Era
AbstractAs generative practices enter daily life, how can digital artworks avoid being reduced to replicable combinations of prompts? From a practitioner’s perspective, this lecture shifts the focus from the “rule-based” approach of classical generative art to the “flux-oriented” dynamics of the generative AI era. It systematically introduces three methodologies: maintaining structure to preserve form and semantics, applying causal constraints to define feasible ranges, and shaping temporality to craft rhythm and texture. Through reproducible generative workflows, the talk explores how creators in the age of AI can retain authorship, trace origins, and document processes throughout creation and exhibition.
Conference General Chair: Huang Sunquan
Executive Planner: Wang Jingjie
Conference Affairs Team: Zheng Yeying, Wang Anqi, Zeng Mengrui
Working Group: Zhou Peng’an, Ding Le, Zeng Mengrui, Jing Jujin, Tian Peisong, Wang Anqi, Li Mengshu
Live Broadcast & Video: Jing Jujin, Li Yiyang, Zhao Jiajing, Li Chengjin, Li Yixin
Visual Design: Li Mengshu, Ding Le, Zheng Yeying
Sound: Boyang Guan
Motion Graphics: Shir-Shiko