404 Media reports that NVIDIA has been clandestinely extracting data from Youtube and Netflix to train AI, citing evidence from various sources including Slack messages, emails, and other internal documents.
Ming-Yu Liu, Nvidia’s research team deputy head and leader of the mentioned project, revealed via email that each day AI receives a volume of image data equivalent to the lifespan of an individual, around 80 years, for training purposes.
An anonymous employee disclosed that they were tasked with pulling data from Netflix, Youtube, and other online sources to develop various AI products for the company, such as Omniverse and autonomous vehicle systems, requiring a substantial amount of physics-based data.
This is not the first time NVIDIA has faced allegations of data extraction from online sources, as recently NVIDIA has also been questioned for surreptitiously extracting data from Youtube, similar to Apple, Anthropic, and Saleforce.
Reference: digitaltrends, 404 Media
TLDR:
NVIDIA has been accused of covertly mining data from Youtube and Netflix to train AI, using a variety of sources including Slack messages and emails. Ming-Yu Liu, a key figure at Nvidia, revealed that AI receives a vast amount of image data daily for training purposes. Reports suggest that NVIDIA has a history of extracting data from online sources, triggering suspicions similar to other tech giants.
Leave a Comment