邵立昊 SHAO Lihao


人工智能艺术 AI Art :


     抵抗智能图像
     Resisting Intelligent Images  

     在迪福森中见到我
     Meet Me in Diffusion

     识别的欲望_猪笼草Pineapple
     Desire for Recognition_
     Nepenthes Pineapple

     屏幕书写
     On-screen Writing

     西瓜非机_想象对抗算法
     Watermelon Non-Plane_
     Imagination Beats Algorithm

     概念刻度（ongoing）
     Scale of Concepts  


装置艺术 Installation Art :

        
     漩涡
     疫书


关于

邮箱


西瓜非机_想象算法
Watermelon Non-Plane_
Imagination Beats Algorithm



2021 
人工智能影像，打印 
循环视频 
1600 × 900 mm 

StyleGAN 
NFT平台用户行为 
NFT 

      此作品是一套由影子组成的NFT作品,其主体物在图像中是缺席的,需要买家和观众自行想象和构建。通过这一设置,作者提出了一个独特的观点:NFT平台与其用户群体构成了一个包含大量人类行为模式数据的生成对抗网络(GAN)。这个网络虽然尚不可见,但却真实存在,其生成的结果有可能无法从真实的人类行为中区分。

     在当前的NFT市场中,大量相似、元素堆叠的程序化作品充斥其中。NFT平台如同GAN中的生成网络,不断生成这些作品;而买家和用户则扮演着判别网络的角色,通过买卖、评价等行为,以价格来标记和区分这些作品。在这一过程中,双方共同构建并训练了一个拥有大量人类价值观、审美观和社会运作模式的GAN。这个网络所运算出的结果,可能连人类自己都无法分辨。

     为了延缓甚至阻止这个GAN运算出难以区分真伪的结果,作者创作了这套NFT作品。通过要求买家自行想象和构建影子的主体物,作者试图打破将作品内容单向传输给用户和买家的路径,促使他们主动参与到作品的创作中来,从而扰乱甚至逆转GAN的算法。

     补充说明:生成对抗网络(GAN)由一个生成网络和一个判别网络组成。生成网络负责分析、模仿数据集中的样本并生成结果;判别网络则负责分辨生成结果与真实样本。两个网络不断对抗,目的是使判别网络无法判断生成网络的输出结果是否真实。Nvidia的超写实人脸生成器StyleGAN就是一个很好的GAN应用例子。
2021 
AI Video，Print 
Loop 
1600 × 900 mm 

StyleGAN 
NFT platform user behavio 
NFT 

This artwork is a set of NFT pieces composed of shadows, where the main objects are absent from the images and require the buyers and viewers to imagine and construct them. Through this setup, the author proposes a unique perspective: NFT platforms and their user communities constitute a Generative Adversarial Network (GAN) that contains a large amount of human behavioral pattern data. Although this network is not yet visible, it truly exists, and the results it generates may be indistinguishable from real human behavior.

In the current NFT market, a large number of similar, element-stacked, and programmatic works are prevalent. NFT platforms act as the generative network in a GAN, continuously creating these works; while buyers and users play the role of the discriminative network, using behaviors such as buying, selling, and rating to label and differentiate these works by price. In this process, both parties jointly construct and train a GAN that possesses a vast array of human values, aesthetic perceptions, and social operational modes. The results computed by this network may be indistinguishable even to humans themselves.

To delay or even prevent this GAN from producing results that are difficult to distinguish from reality, the author created this set of NFT artworks. By requiring buyers to imagine and construct the main objects of the shadows themselves, the author attempts to break the path of unidirectionally transmitting the content of the artworks to users and buyers, encouraging them to actively participate in the creation of the pieces, thereby disrupting or even reversing the GAN's algorithm.

Supplementary explanation: A Generative Adversarial Network (GAN) consists of a generative network and a discriminative network. The generative network is responsible for analyzing, imitating samples from the dataset, and generating results; while the discriminative network is tasked with distinguishing the generated results from real samples. The two networks constantly confront each other, with the goal of making the discriminative network unable to determine whether the output of the generative network is real. Nvidia's hyperrealistic face generator StyleGAN is an excellent example of a GAN application.