论人工智能深度学习中著作权的合理使用

On The Fair Use of Copyright in Deep Learning of Artificial Intelligence

自由地分析数据是通过深度学习发展人工智能的前提。如果人工智能分析的对象恰巧是作品,则不可避免地需要对其进行扫描与复制,由此产生著作权的侵权争议。互联网时代,作品以数据的形式进行传播。由于作品在利用上存在非竞争性的特点,人工智能可以借助于机器分析不断从作品中发掘出新的价值,满足不同主体的特定需要。尽管深度学习中对作品的分析行为与原作品息息相关,但机器分析后形成的价值既不是对作品的原样呈现,也不是对作品内容的演绎性表达,而是独立的、新的作品增值。作品增值的结果是对作品思想层面的利用,但是知识增值的过程深度学习则不可避免地要对作品进行转换性使用,可归入合理使用的范畴。将人工智能深度学习纳入合理使用不仅与合理使用规则变迁的历史趋势相一致,也符合著作权法激励创新的目的,更有助于人工智能产业的健康发展。

The freedom to analyze data is a prerequisite for developing artificial intelligence through deep learning. If the object of artificial intelligence analysis happens to be a work, it is inevitable to be scanned and copied, which may result in copyright infringement disputes. In the age of the Internet, works are transmitted in the form of data. Due to the non-competitive features of use of works, artificial intelligence can continuously discover new value from works by means of machine analysis to meet the specific requirements of different subjects. Although the analytical behavior of works in deep learning is closely linked to original works, the value created after machine analysis is neither the simple presentation of original works nor the deductive expression of contents of original works, which reflects independent and new increment of value upon works. The result of value-increment of works is the utilization of the ideological level of wo rks, but the process of the value-increment of knowledge - deep learning - inevitably requires the transformative use of works, which can be classified as the fair use. To incorporate deep learning of artificial intelligence into fair use is not only consistent with the historical trend of change of rules of fair use, but also in line with the purpose of innovation incentive of copyright law and contribute to the healthy development of artificial intelligence industry.