人工智能绘画生成图像的可著作权性分析
Analysis of Artificial Intelligence Painting Image’s Copyrightability

人工智能绘画生成图像的可著作权性一直存在较大争议。为了尽可能地减少法学与具体技术之间的隔阂，选取了当前第四阶段的绘画人工智能进行技术解读。深度学习模型的本质是对比余弦相似度匹配图像数据与文本数据。扩撒模型的本质是基于扩散方程的三维形态计算算法。以技术解读的结论为基础，通过比较绘画人工智能与获得公民资格的人工智能、法人，否定了绘画人工智能的主体资格，进而确定绘画人工智能在图像生成中的角色定位应当是工具。在此基础上进一步讨论，人工智能绘画生成图像是否“具有独创性”、“属于人类智力成果”、“表现思想或情感”，皆可获得肯定的答案。在肯定人工智能绘画生成图像可著作权性的同时，提出了应坚持内容决定主义来具体判断图像是否是著作权意义上的作品。
The copyright of artificial intelligence painting image has always been controversial. In order to reduce the gap between law and specific technology as much as possible, the current fourth stage of painting artificial intelligence is selected for technical interpretation. The essence of a deep learning model is to compare cosine similarity to match image data with text data. The essence of diffusion model is a three-dimensional shape calculation algorithm based on diffusion equation. Based on the conclusion of technical interpretation, by comparing the painting artificial intelligence with the artificial intelligence and corporation who obtained citizenship, the subject qualification of painting artificial intelligence is denied, and the role positioning of painting artificial intelligence in image generation should be a tool. Based on this further discussion, whether the image generated by artificial intelligence painting is “original”, “belongs to the achievement of human intelligence”, “expression of thought or emotion”, can get a positive answer. While affirming the copyright of the image generated by artificial intelligence painting, it is proposed that we should adhere to the content determinism to judge whether the image is a work in the sense of copyright.