From Ziff Davis: 2024-04-06 13:13:52

OpenAI and Google used YouTube videos to train their AI models, potentially breaching copyright. OpenAI transcribed over a million hours of videos with Whisper to train GPT-4, despite concerns it violated YouTube’s terms. Google knew and allowed it for its own AI training. Licensing all data is financially unrealistic for AI companies.

YouTube CEO addresses concerns about OpenAI using YouTube videos to train Sora, an AI video creation tool. Using YouTube videos would violate terms of service. Sora is set to launch later this year.



Read more at Ziff Davis: OpenAI and Google Reportedly Used YouTube Videos To Train Their AI Models