Nvidia's Cosmos model sparks controversy, allegedly infringes copyright law

In the process of developing the Cosmos model, NVIDIA downloaded 20 million hours of video from the Internet and extracted 100 million 2-60 second clips from it as training data. This behavior has caused controversy under copyright law because NVIDIA has grabbed a large amount of data from platforms such as YouTube and Netflix without permission. Despite the controversy, NVIDIA's move shows its urgent need for high-quality training data.