NVIDIA has found itself at the center of a scandal over the methods of collecting data for its AI systems. The company was caught mass-downloading videos from popular platforms without the consent of copyright holders.
404 Media obtained internal NVIDIA documents that reveal details of the Cosmos project. The company’s employees downloaded tons of videos from YouTube, Netflix, and other services every day. The scale is impressive: every day, content equal to 80 years of viewing was processed.
Why all this? NVIDIA planned to use the collected data for several ambitious projects. These included creating 3D worlds in Omniverse, improving self-driving cars, and developing a “digital human.”
When the information surfaced, NVIDIA tried to justify itself. The company’s representatives said that their actions do not violate copyright law. They say that they only take ideas and information, not the finished product. In addition, in their opinion, AI training falls under “fair use.” However, the owners of the sites from which the content was downloaded do not agree with this position.
