Trends

AI models trained on YouTube videos by Google and Open AI

AuthorBTW Editorial

Reading Time2 min

PublishedApril 8, 2024

Primary DomainN/A

Content TypeN/A

TopicN/A

EntityN/A

RegionN/A

Time HorizonN/A

ImpactN/A

ConfidenceN/A

Headline

Context

Both OpenAI and Google have turned to transcribing YouTube videos to further train their AI models, which could infringe on creators’ copyrights. The two tech giants cut corners on Meta to get as much data as possible to train their AI models. OpenAI used Whisper to transcribe over a million hours of YouTube video, feeding the transcripts into GPT-4, the AI system used for the ChatGPT chatbot. Google, which owns YouTube, also transcribed videos for AI model training.

Evidence

Pending intelligence enrichment.

Analysis

The transcriptions of videos by both businesses can violate the copyrights of the original producers. Lawsuits related to copyright and licence have resulted from other uses of creative content for AI training. OpenAI’s use of YouTube videos may also violate Google’s rules prohibiting the use of its videos for “independent” applications and “automated means (such as bots, botnets, or scrapers)” of accessing its videos. Also read： Google and Stanford researchers launch AI fact-checking tool Also read： Google DeepMind CEO Demis Hassabis receives knighthood for AI technology

Key Points

OpenAI and Google used the speech recognition tool Whisper to transcribe more than 1 million YouTube videos to train their AI models.
OpenAI’s use of YouTube videos may violate Google’s rules , which prohibit the use of its videos for standalone applications as well as access through automated means.