人工智慧之治理——以深度偽造為例

The governance of AI–In the aspect of deepfake

隨類神經網絡之運用日漸成熟，人工智慧之技術應用層面日益廣泛。其中，透過自動化編輯器與卷積神經網路技術，仿製影片之技術已非門檻極高之事。2017年11月，美國社群平台Reddit出現將社交名流之臉部疊加於色情影片演員身上之影片，此類影片係透過深度學習演算法對特定社交名流之臉部進行分析，並將其臉部疊加於影片上，使其如同社交名流本人所拍攝之影片。 （本著作為經濟部科技專案研發成果）

With the maturity of neural networks, the application of artificial intelligence is becoming more and more widely used, and through the utilize of Autoencoder and convolutional neural network technology, the threshold of copying films or audios has been easier than before. In November 2017, A user post an App on the community platform called Reddit. This App can superimposed the face of social celebrities on pornographic film actors, and makes the film look like is real. Such technology is generally referred to as “Deepfake＂, and it's believed that these technologies may contribute to the manufacturing and dissemination of false information, which has attracted the attention of national legislators. Because these technologies is using Deep-learning skills to generate fake films, it involves both the prevent of fake news and the governance of artificial intelligence. The Purpose of this study is to observe international policies, legal measures and related guidelines for Deepfake and artificial intelligence, and to examine whether the current legal system can respond to the impact of deepfake and provide feasible recommendations.