TapTechNews on August 6th reported that documents disclosed by the tech media 404Media showed that Nvidia has collected a large amount of copyrighted content for training artificial intelligence (AI).
Internal mails, emails, Slack conversations, and related documents disclosed by the media showed that Nvidia collected video materials from multiple sources such as YouTube to expand the data set for training AI.
Conversations disclosed by the media showed that employees involved in the project had questioned that using YouTube videos and data sets compiled for research purposes might have legal issues. However, the project manager's reply was that this collection model had obtained the highest level of approval from the company and could use these contents to train AI, and believed that it was 'fully in line with the literal description and spirit of the copyright law'.
One Nvidia employee revealed that the company required employees to collect videos from video sources such as YouTube and Netflix for training Nvidia's Omniverse 3D generator, self-driving system, and 'Digital Human' products.
TapTechNews quoted the media report that the project was internally called the Cosmos project. In order to avoid YouTube's detection, Nvidia used virtual machines with rotating IP addresses to download content to avoid being banned.
Nvidia employees wrote: 'We are using AWS, and restarting an instance will allocate a new public IP, so this has not been a problem so far.'
Related Reading:
'Tech giants exposed for using YouTube content to train AI without authorization, including Apple and Nvidia.'