Is AI-generated Art an Enemy or a Friend to Artist? | AI生成艺术对艺术家是敌是友?

    The essay argues that AI-generated art is not the enemy of artists. The development of AI technology will free artists from repetitive and mechanical labor and encourage creativity, intelligence and humanity implied in man-made art. It contains three sections. The first section introduces Portrait of Edmond Belamya proclaimed AI-generated painting created with an open-source algorithm called GAN (Generative Adversarial Network). It was sold for an unexpected price of $432,500 during a Christie's auction in 2018. The second section reviews the history of AI-generated art by focusing on some key inventions, such as AARON, the first computer program able to produce basic shape in 1968 and Google's Deep Dream Program in 2014. The third section categorizes art-related AI into three types according to impact. The first type includes software that benefit artists in general, such as image recognition, color specification, and auto 3D construction. The second type are programs that generate new image according to selected style, which will eventually outperform the beginner level designer in terms of speed and quality, only preserving art jobs that require true creativity. The third type is represented by GAN algorithm, which generates its own subject by learning from the fed data, removing human’s critical role of consenting the subject. It is now limited by the 'catastrophic forgetting' trait of GAN system.

    本文论证AI生成艺术并非艺术家的敌人。AI技术的发展可以把艺术家从重复而机械的苦力中解放,并激发艺术家创作蕴含创新、智慧及博爱的作品。本文由三部分组成。第一部分介绍一幅被宣称为AI生成艺术的油画《爱德蒙.贝拉米肖像》。这副画由源代码开放的GAN算法生成,在2018年Christie拍卖会上售出了惊人的432,500美元价格(约合三百零五万人民币)。第二部分回顾AI相关艺术的发展历史。介绍1968年第一个能生成基本形状的软件AARON与2014年的谷歌Deep Dream计划等关键发明。第三部分将与艺术相关的AI按照其影响分为三类:第一类是惠及全体艺术家的软件,例如图像识别、颜色细化及3D自动构建等;第二类包含可以依据已选风格生成新图像的程序。这些程序将在生成图像的速度与质量方面超越新手设计师,进而仅保留需要真正创新能力的艺术岗位。第三类是以GAN算法为代表、可通过学习数据自行生成内容的AI。它们移除了人类在创造艺术中决定内容的关键职责,但目前被GAN系统“灾难性遗忘”特质所限制。