Posted 4/7/2024, 3:30:56 PM
OpenAI and Google Face Backlash Over Use of YouTube Videos to Train AI Models Without Consent
- OpenAI and Google used Whisper to transcribe over 1 million hours of YouTube videos without creator consent to train AI models, potentially violating copyright
- Transcribing and using YouTube videos to train models may infringe on video creators' copyrights to their content
- OpenAI's use of YouTube videos likely violates Google's terms of service prohibiting using its videos for independent applications
- People at Google allegedly knew about OpenAI's unauthorized YouTube use but neglected taking action because Google did the same
- In July 2023, Google updated terms of service to allow using public online material like Docs and Maps reviews to further train its AI models