NVIDIA
A jelentések szerint az Nvidia több millió videót gyűjtött be a Cosmos AI modell betanításához
Leaked documents obtained by 404 media report Nvidia has been scraping millions of videos daily from YouTube, Netflix, and other sources to train its unreleased foundational AI model.
- Nvidia's project, codenamed Cosmos, aims to process "a human lifetime visual experience worth of training data per day."
- The company used open-source tools and virtual machines to download videos, including full-length movies and TV shows.
- Employees raised concerns about copyright and ethics, but were told there was "umbrella approval" from executives.
- Nvidia claims its practices are "in full compliance with the letter and spirit of copyright law."
Miért fontos?
Project Cosmos appears to be Nvidia’s big move into video-based AI, which could revolutionize everything from 3D world generation to self-driving cars, digital humans, and more. However, this harsh introduction is not a good look for the company, especially as the industry's practices are coming under intense scrutiny.