Apple appears to be among the tech giants that have used unauthorized means to train its AI models.
Proof News' findings revealed that companies such as Apple, NVIDIA, Anthropic, and others used scraped YouTube information to train their AI models.
According to the research, these tech companies used subtitles from more than 173,000 YouTube videos taken from over 48,000 channels.
The videos used to train AI came from well-known creators and celebrities such as Marquees Brownlee (MKBHD), MrBeast, PewDiePie, Jimmy Kimmel, Stephen Colbert, and John Oliver.
The research revealed that scraped videos "contain video transcripts from educational and online learning channels like Khan Academy, MIT, and Harvard."
The datasets known as YouTube Subtitles were created by EleutherAI, a non-profit AI research company, reports Wired.
It's important to note that YouTube forbids such data collection without the company's permission. After discovering the vast amount of material used to teach AI, many previously mentioned content creators expressed discontent.
David Pakman, host of "The David Pakman Show," shared that no one asked for permission to use his content with the CEO of Nebula, David Wisku. He called this act "theft" and said it would only harm the creators.
EleutherAI didn't comment on ProofNews's findings. The company has also established a tool to check whether "[...]your YouTube video used to train AI.”
Your email address will not be published. Required fields are markedmarked