OpenAI transcribed over a million hours of YouTube videos for AI development, possibly violating YouTube's rules.
Tech giants, including OpenAI, reportedly cut corners to harvest data for AI development. OpenAI, facing a supply problem, created a speech recognition tool called Whisper to transcribe YouTube videos. This move raised concerns about violating YouTube's rules, as it prohibits use of videos for independent applications. Nonetheless, OpenAI transcribed over a million hours of YouTube videos, which were then fed into its GPT-4 system.
April 06, 2024
3 Articles